Proceedings
Year
| Filter results109 paper(s) found. |
|---|
1. A Low-cost Multi-view Image to 3d Reconstruction for Plant PhenotypingCurrent 3D plant phenotyping approaches often rely on LiDAR or multi-camera systems, which are costly, require complex calibration, and lack scalability. This study introduces a simple and cost-effective pipeline for 3D plant reconstruction using Hunyuan3D-2.5, a multi-view generative model. Plant samples were photographed directly using a mobile phone, and raw images were processed with a custom Python background-removal pipeline that enhanced plant contours and removed environmental noise. ... C. Huang |
2. A New Paradigm of Datadriven Agrifood Systems“Data-driven agrifood systems” is issued as a new standard terminology of smart farming from the international organization for standardization (ISO), and it has also focused on the needs of small/medium enterprises of farming. Data management scheme has changed the context of decision making on received style of good agricultural practices. Farmers and stakeholders should re-watch the system changes with emerging technologies. Farm management sustainable and community-based shoul... S. Shibusawa |
3. A Physics-informed Neural Network Approach for Simulating Laminar FlowEfficient and accurate modeling in agricultural fields is critical for advancing precision agriculture. These simulations, often involving the prediction of airflow, temperature, and humidity distributions, directly support decisions related to crop management, greenhouse climate control, and irrigation strategies. Computational Fluid Dynamics (CFD) has been a primary tool for decades, offering reliable and high-fidelity simulations through established numerical methods such as the finite-dif... C. Huang |
4. A Simulation-based Matching System for Utilizing Clean Energy from Agri-livestock WasteIn order to mitigate greenhouse gas emissions and air pollution derived from agricultural and livestock waste and to enhance the resilience of the clean energy supply chain, a simulation-based matching system for utilizing clean energy from agri-livestock waste as developed. Building upon a prior research entitled " An Inventory of Greenhouse Gases and a Map of Biomass Energy Utilization in Agriculture and Animal Husbandry Biomass Waste," the system is designed to evaluate the effic... J. Jiang |
5. A Vision-guided Gantry Robot for Efficient Orchid Basket Reorganization in GreenhousesProper alignment of orchid baskets in greenhouses is important to maintain visual uniformity, maximize space usage, and ensure consistent light exposure during flowering. Manual arrangement is time-consuming and labor-intensive, underscoring the need for automation. To address this challenge, we propose an integrated robotic system for automated basket organization. The system combines a cartesian gantry robot, a rotation-aware clamping gripper, dual Intel RealSense D435i cameras, and a light... W. Lin |
6. Adoption of Precision Agriculture in JapanJapan is a country facing global challenges in terms of a declining and aging agricultural population, making the establishment of a sustainable production system a matter of urgency from the perspective of food security. While respecting Japan's traditional knowledge, the author believes that precision agriculture is an effective solution to resolve this situation. We argue that data-driven agriculture presents a higher degree of affinity with Japanese farmers, providing a more viable pa... E. Morimoto |
7. AI for Genomic Agriculture — from Sequence to Field ImpactGenomics offers powerful opportunities to enhance crop yield, resilience, and nutritional value, yet the complexity and scale of genomic, transcriptomic, and epigenomic data pose significant challenges for interpretation and application. Artificial intelligence (AI), particularly machine learning and deep learning, provides powerful approaches to decode this complexity and accelerate precision agriculture. I will present AI-based methods developed in my laboratory for annotating pla... C. Chen |
8. An Intelligent Blade Balancing Control System for Steep-terrain Tea Cutting ApplicationsTea is a famous and valuable beverage in Taiwan. Tea is mainly grown in steep or mountainous areas. The terrain is a challenge for harvesting automation. Manual labor in harvesting tea in complex terrain is time-consuming and affects the economic efficiency of the product. This study proposes a tea-cutting blade balancing control system integrating image processing and fuzzy logic control. A unique mechanism is developed to adapt to the slope of the terrain. The limiting angle is 12 degrees r... W. Lin |
9. Analysis Of Internal Abnormalities Of Tilapia Flesh Using Hyperspectral Imaging And Machine Learning MethodTilapia, the most produced aquaculture species in Taiwan, has experienced significant production loss due to internal abnormalities, notably streptococcosis, which remains undetectable until fillets are cut. The absence of visible external symptoms frequently leads to quality reduction and economic loss. To address this, hyperspectral imaging, capable of capturing subtle spatial and spectral differences, was employed. The objective of this study was divided into two phases: firstly, identific... S. Chen |
10. Apple Weight Prediction Based on Lifecycle Growth InformationIt is essential for determining optimal harvest timing, the grade of quality, and fresh maintenance, all of which directly impact growers’ economic returns, to accurately predict individual apple fruit weight. This study aims to predict the fruit weight of Fuji apples at the main branch level (n = 126) using growth data collected throughout the growing season. Fuji apples were monitored at 23 orchards in 2022 and 2023, and at 24 orchards in 2024. Growth data were col... C. Ryu |
11. Application of Image Processing and Artificial Inteligence (AI) for Cabbage Cultivation MonitoringCabbage requires precise monitoring for during cultivation, e.g., transplanting performance, water stress, growth status, and yield estimation. This study presents image processing and artificial intelligence (AI) techniques to enhance automation for cabbage production operations. High-resolution multispectral and thermal images were acquired using UAVs and ground-based platforms. Seedling detection during transplanting operation was implemented using a YOLOv8 model with a CSPDarknet53 backbo... S. Chung |
12. Applying Retrieval-augmented-generation to Support Farmers in Pest and Disease DiagnosisAccording to the Ministry of Agriculture, crop production in Taiwan reached a value of $275 billion NTD in 2023, highlighting the economic importance of agriculture. However, the industry is now facing serious challenges, particularly in pest and disease identification and crop protection. Due to global ecological challenges, the manifestations of local pests and diseases have changed, making it difficult for farmers to rely on past experiences to identify and manage them effectively. Farmers... Y. Kuo |
13. Automated Identification of Tomato Diseases, Pests, and Disorders Using Ai Models and Smartphone ApplicationsTomato is one of the most important economic crops in many countries, with a substantial global production volume. However, tomato growth is often affected by diseases, pests, and physiological disorders (DPD), which typically manifest as symptoms on leaves, such as specks, yellowing, necrosis, or leaf deformation. These issues significantly reduce tomato yield and quality. Therefore, accurately identifying these symptoms and implementing corresponding management strategies have become crucia... Y. Kuo |
14. Automated Selection of Taiwan Native Breeding Chickens Using Machine Vision and Deep LearningChicken is a primary global source of protein. In Taiwan, the poultry sector is a cornerstone of the domestic food supply. A significant part of this sector is the Taiwan Native Chicken (TNC), a collection of indigenous breeds prized for their unique flavor and cultural value, generating over 26 billion New Taiwan Dollars in 2023. Maintaining the quality of TNC relies on the effective selection of superior breeders. Conventionally, this selection is performed through manual inspection of phen... Y. Kuo |
15. Automatic Calibration of Crop Growth Models for Predicting Corn Economic Optimum Nitrogen RatesThe objectives of this study were to 1) evaluate an automatic model calibration strategy; and 2) compare the performance of DSSAT and APEX models for simulation of maize (Zea mays L.) growth, plant nitrogen (N) uptake, yield in response to different N application rates and the estimation of the economic optimum N rate (EONR). Detailed data collected from eight site- years of N experiments conducted from 2014 to 2016 in Minnesota and Wisconsin, USA were used in this research. The results indic... Y. Miao |
16. Automatic Counting of Chickens Around Feeders Using Convolutional Neural NetworksIn 2023, Taiwan’s chicken industry generated about NTD 93.6 billion, or 43.5% of the livestock production value, underscoring its central role in the sector. Nonetheless, monitoring flock health and housing remains labor-intensive, and adjustments to feeding regimes frequently depend on subjective judgment, limiting standardization and scalability. Because feeding behavior is a key indicator of health and welfare, we present a vision-based system that continuously detects feeders and co... Y. Kuo |
17. Biosynthesis of Silver Nanoparticles from Platostoma Palustre for Agricultural ApplicationsNanoparticle synthesis using natural resources offers a cost-effective and eco-friendly strategy. In this study, silver nanoparticles (AgNPs) were synthesized using Platostoma palustre extract (PPE), rich in polysaccharides and bioactive compounds. Characterization by XRD, SEM, and TEM confirmed successful synthesis. TEM revealed oval-shaped nanoparticles (7-80 nm) with PPE forming a stabilizing layer to prevent agglomeration, while XRD indicated a crystallite size of approximately ... W. Lin |
18. Cabbage Yield Estimation Using Multispectral UAV Imagery and Deep Neural SegmentationAccurate and efficient yield estimation is essential of optimizing crop management, resource allocation, and harvest planning in precision agriculture. Traditional manual methods are time-consuming, labor-intensive, and often lack spatial accuracy. Recent advances in remote sensing and deep learning offer scalable, non-destructive alternatives for yield monitoring. This study proposed a cabbage yield estimation based on an enhanced unity networking (U-Net) segmentation model utilizing multisp... S. Chung |
19. Cfd Evaluation of Uvc Air-cleaning Integration in Greenhouse Hvac SystemsGreenhouse crops in Taiwan are highly vulnerable to airborne pathogens due to the humid climate and poor ventilation. This study evaluated the integration of UVC air- cleaning devices with the greenhouse HVAC system to reduce pathogen concentrations. A SolidWorks model of the NTU smart greenhouse was constructed, and CFD simulations were conducted to compare three configurations in which four UVC units were placed at the upper, middle, and lower regions of the wet pad. Results showed that the... C. Huang |
20. Close-range Remote Sensing Data for Optimizing Horticultural Production ProcessesPlant sensors have been explored over the last three decades, resulting in various non-destructive sensor systems, feasible for usage along the entire horticultural supply chain. This review will show examples of sensor applications, pointing out benefits and challenges of different measuring principles. Particular emphasis is given on recent developments on analyzing plants directly in the field, aiming precise, data-driven production measures. ... M. Zude-sasse |
21. Color Identification and Texture Features of Phalaenopsis Using Deep LearningAs one of the most economically important and widely traded ornamental plants worldwide, Phalaenopsis hold a significant position in the global floriculture industry. The breeding process is traditionally labor-intensive, requiring careful visual assessment of numerous floral traits to select desirable varieties, which underscores the need for scalable, automated solutions. To enhance the efficiency of Phalaenopsis breeding and accelerate phenotypic comparison across varie... Y. Kuo |
22. Current Status and Potential of Digital Agriculture in IndiaIndian agriculture is facing multiple challenges, including climate change, resource depletion, low productivity growth, high post-harvest losses etc. To address these, a strong push toward digital and smart farming is underway. The Government of India has launched major initiatives such as the Digital Agriculture Mission, Digital India Programme, and targeted funding for AI, Machine Learning, and cyber security to support agricultural innovation. The focus is on building Digital Pu... M. Singh |
23. Deep Learning-based Insect Detection on Sticky Traps Captured Via Mobile Phones Under Field Lighting ConditionsInsect pests pose a major threat to agricultural production, requiring effective integrated pest management (IPM) strategies that depend on accurate identification and counting of pests captured on sticky traps. However, mobile phone images taken under natural field lighting often suffer from inconsistent illumination, shadow interference, and low visibility of small insect targets, which significantly reduce the reliability of automated monitoring systems. To address these challenges, this s... T. Lin |
24. Deep Reinforcement Learning Based Robotic Arm Control for Autonomous HarvestingInverse Kinematics (IK) is a traditional method used for robotic arm manipulation, relying heavily on precise calibration and huge computational demands for arms with higher Degrees of Freedom (DoF). In contrast, Deep Reinforcement Learning (DRL) is an innovative approach to manipulation that exhibits greater tolerance for calibration inaccuracies. It trains using noise added to joint angles, allowing it to learn how to compute accurate trajectories even with inaccuracies in the joint angles.... C. Huang |
25. Design and Development of UECS-based Environmental Monitoring and Control Platform Without CodingData-driven agriculture has been increasingly adopted to achieve labor-saving, energy efficiency, and resource optimization in agricultural operations. Among small- and medium-scale horticul- tures, the Ubiquitous Environment Control System (UECS) proposed in 2004 is attracting again due to low cost of introduction. The UECS is an autonomous and distributed open-source en- vironmental monitoring and control platform for greenhouse horticulture. A computer called a node is used in each environ... T. Okayasu |
26. Design of a Collision Avoidance Algorithm for Autonomous Tractors with ImplementsOver the past decade, autonomous tractors have emerged as a key technology in agricultural automation. Global Navigation Satellite System (GNSS)-based navigation is widely used in autonomous tractors. However, since the GNSS cannot perceive the surroundings, an additional perception system is required to ensure the safety of the operation. Paddy ridges, one of the major obstacles in paddy fields, are typically higher than farmland to facilitate water storage. These height differences can lead... H. Kim |
27. Design of a Garlic Seeding Monitoring and Mapping System Using GNSS and Vision SensorsSeeding monitoring serves as the first step in precision agriculture, playing a crucial role in collecting and managing data across the entire agricultural process. While several international companies have recently developed precision agriculture solutions that monitor seeding rate, missing rate, and more, the agricultural environment in Korea presents unique challenges. For instance, in the case of Korean garlic planters, an average missing rate of approximately 10% is observed. When these... H. Kim |
28. Detecting and Removing Defective Carcasses of Taiwanese Native Chickens Using Convolutional Neural NetworksPoultry is one of the most important sources of meat worldwide. In 2023, the production value of poultry in Taiwan reached 59.8 billion NTD, accounting for 27.8% of the economic value of the animal husbandry industry. Among various chicken breeds, Taiwanese native chickens (TNC) are highly favored by consumers for their meat quality and flavor. As the demand for chicken increases, providing high quality meat to the market has become crucial. Unlike broilers, Taiwanese native chickens have div... Y. Kuo |
29. Development of a Lorawan Wireless Node for Monitoring Smart GreenhousesThe adoption of Internet of Things (IoT) technologies in the smart greenhouse domain is rapidly advancing. Greenhouse planting improves quality and yield by controlling factors affecting crop production. Temperature, humidity, and light intensity in greenhouses are important factors affecting crops. Monitoring and regulating these parameters is conducive to improving the quality and yield of crops. Traditional greenhouse monitoring systems that use wired connections often have problems with c... S. Chung |
30. Development of a Low-power Wireless Communication System Using Lora for Structural Monitoring in Greenhouse FoundationsPlastic greenhouses dominate protected cultivation in South Korea but are vulnerable to extreme weather and foundation instability. To address this issue, a low-power, low-cost monitoring system was developed to estimate foundation attitude and detect anomalies such as uplift. The system integrates an IMU (Inertial Measurement Unit)-based sensor node, LoRa (Long Range) communication, and a gateway in a star topology. Field tests, including pipe uplift and natural conditions, confirmed compara... J. Park |
31. Development of a Measurement and Analysis System for Tillage Operations in Paddy FieldsThis study developed a foundational technology for real-time tillage depth measurement using Inertial Measurement Units (IMUs). The ultimate goal is to enable variable-rate tillage operations tailored to spatial variations in topsoil depth. The system consisted of an RTK-GNSS module and two IMUs to measure the respective pitch angles of the tractor and implement. Tillage depth was estimated using a model derived from the geometric relationship between the implement’s pitch angle and its... E. Morimoto |
32. Development of a Mobile Inspection Robot for Stacked-cage Layers Houses in TaiwanIn stacked-cage layers houses, it is essential to know the eggs produced in each cage per day and their distributions for evaluating egg-laying performance and the health status of the layers. A two-wheel-drive mobile inspection robot for egg-counting was thus designed, assembled, and on-site experiments were performed and evaluated in this paper. The path of the mobile robot was pre-designated according to the site floor layout, so the robot can move autonomously aisle by aisle. Multiple cam... A. Cherng |
33. Development of a Small-scale Weeding Robot for Inter-plant Areas Using Vision and Rake MechanismIn low-herbicide or herbicide-free farming systems such as those used for medicinal and herbal crops, weed management remains one of the most labor-intensive tasks. Intra-row weeds, which grow between closely spaced crop plants, are particularly difficult to remove using traditional mechanical methods. Manual weeding, although effective, still poses a significant labor burden and limits the scalability despite the high market value of the crops. To address this challenge, we have de... K. Imaoka |
34. Development of a Small-Scale Weeding Robot for Inter-Plant Areas Using Vision and Rake MechanismIn low-herbicide or herbicide-free farming systems such as those used for medicinal and herbal crops, weed management remains one of the most labor-intensive tasks. Intra-row weeds, which grow between closely spaced crop plants, are particularly difficult to remove using traditional mechanical methods. Manual weeding, although effective, still poses a significant labor burden and limits the scalability despite the high market value of the crops. To address this challenge, we have de... K. Imaoka |
35. Development of a Smart Agriculture Platform for Modern Management of Longan OrchardsSmart agriculture has emerged as a critical approach in modern agricultural systems. This study aimed to develop a smart agriculture platform for longan orchards by integrating Internet of Things (IoT) technologies and digital systems for precision farming. The study population comprised 100 large-scale agricultural producers located in the provinces of Chiang Mai and Lamphun, Thailand. The developed platform incorporated six core technologies: IoT-based smart irrigation, weather monitoring, ... C. Kanjanaphachoat |
36. Development of a Smart Low-Carbon Greenhouse Integrated Plasma- Activated Water and Second-Life EV BatteriesGlobal warming and the excessive use of nitrogen fertilizers pose significant challenges to sustainable agriculture. This study presents a smart low-carbon greenhouse system powered by a solar photovoltaic microgrid, integrated with second-life electric vehicle batteries and a real-time energy management system. Plasma-activated water (PAW) technology is employed to reduce dependence on chemical nitrogen fertilizers while enhancing crop productivity and reducing nitrous oxide emissions. The s... W. Sean |
37. Development of Ai-based Energy Management Strategy in Seawater Desalination Plant Based on Physical ModelingGlobal water scarcity is becoming increasingly severe, and seawater reverse osmosis (SWRO) has become a major technology for freshwater production due to its high efficiency. However, membrane fouling during long-term operation increases transmembrane pressure, reduces flux, and raises energy demand, ultimately lowering efficiency and shortening membrane lifetime. Traditional control and prediction methods struggle with the nonlinear and dynamic nature of these processes. To address this, we ... W. Sean |
38. Development of an Electric-assisted Handling System for Pig FarmIn swine farming, manure management is a critical yet labor-intensive task. With increasing agricultural labor shortages, optimizing farm infrastructure to reduce manual workload has become essential. Many pig farms in Taiwan utilize elevated slatted floors (concrete or cast iron) to separate pigs from their waste, allowing excrement to fall through gaps for later disposal. While this design improves hygiene by reducing direct contact, the heavy and bulky slatted panels pose significant chall... W. Chen |
39. Development of an Integrated Harvesting Machine for Taro FieldsTaiwan cultivates a diverse range of agricultural products, among which taro (Colocasia esculenta) is an important root vegetable. Although several harvesters exist for root crops, their applicability remains limited due to crop-specific requirements, and no dedicated integrated harvesting machine is currently available for taro in Taiwan. Farmers still rely heavily on manual labor, using knives or spades to loosen the soil around taro plants before uprooting them individually—a time-co... W. Chen |
40. Development of Automated Rose Monitoring System with Deep Learning-based Growth Stage ClassificationIn cut-flower cultivation, effective production planning is essential to accommodate seasonal fluctuations in market demand. Precise rose growth stage monitoring is critical for cultivation schedule, environmental control, and harvest timing, yet current practices rely on manual observations, which are time-consuming and prone to subjectivity, limiting consistency and scalability. This study presents an automated monitoring system integrating computer vision and deep learning for ob... S. Chen |
41. Development of Cultivar-optimized Nir Spectroscopy Model for Cherry Tomato Maturity and Sweetness Assessment"Yunu" cherry tomato cultivars hold substantial commercial value in Taiwan’s premium markets, where sweetness serves as a key quality attribute. To enhance cultivar-specific quality assessment, this study evaluates tomato quality in both pre-harvest and post-harvest stages.In the pre-harvest stage, image data were used to establish a Red Ripeness Index (RRI) for evaluating tomato maturity. Color calibration techniques were applied to improve consistency, and the stability and ... S. Chen |
42. Development of Rgb and Lidar Fusion Based Pear Fruit Quantification and Mapping SystemThis study presents a system for accurate fruit quantification using LiDAR-RGB sensor fusion. The system projects 2D fruit detections from a YOLO model onto a 3D map generated via SLAM, assigning a unique coordinate to each fruit to prevent double-counting. This approach achieved an aggregate accuracy of 98.5%, with a predicted total of 535 fruits compared to the 527 observed. The resulting data revealed significant fruit density variations (3.2 to 12.6 fruits/m²), establishing the syste... E. Morimoto |
43. Development of Temperature and Humidity Sensor Calibration Procedure for Multifunctional Orchid Greenhouse Monitoring SystemBacterial soft rot and bacterial brown spot are primary diseases that threaten orchid cultivation, often resulting in substantial economic losses. To address labor shortages and environmental challenges in recent years, the orchid industry is increasingly adopting intelligent disease management systems that combine sensing technologies and data analytics as part of its transformation strategy. The multifunctional monitoring system was developed as an economical, integrating sensors for temper... C. Haung |
44. Development of Vision-guided Autonomous Robot for Phenotypic Monitoring in Tomato BreedingPhenotypic monitoring in crop breeding requires continuous data collection throughout growth cycles, yet traditional manual methods are both labor-intensive and time-consuming. Individual plant tracking over extended periods poses particular challenges due to field scale and measurement frequency requirements across diverse agricultural environments. This study presents an autonomous robotic platform integrating computer vision and precision positioning technologies for automated phenotypic d... S. Chen |
45. Development, Design, and Integration of an Egg Tray System with Unmanned Ground Vehicle for Robotic Poultry AutomationEggs are among the most extensively consumed foods, valued for their nutritional and health benefits. While generally studied, the exact fine representation of a raspberry’s egg shape remains complex. Egg forms are usually classified as globular, ellipsoidal, elliptical, and pyriform- the last of which still lacks a definitive equation. This study presents a new system for modeling egg figures and calculating volume, based on crucial parameters, including the major axis, the m... B. Gonzales |
46. Disease Symptom Recognition and Severity Assessment for Phalaenopsis OrchidsTraditional disease assessment relies on manual visual inspection, which is subjective and often leads to inconsistent results due to variations in human judgment. To address these challenges, this study proposes an automated approach for disease classification and severity grading in Phalaenopsis orchids using the YOLOv8-seg deep learning model. The system integrates instance segmentation with Lab color space analysis, which was found to outperform HSV in distinguishing healthy and diseased ... C. Huang |
47. Dual-channel Imaging and Two-stage Deep Learning for Fertility Detection of Duck EggsIn Taiwan, the waterfowl industry generates a production value of NT$11.2 billion, of which meat ducks contribute about 80% (≈NT$8.9 billion). As the upstream segment of the duck meat industry, the hatching process of duck eggs plays a critical role in duck production. Fertilized eggs require a clean incubation environment to develop properly. To protect this environment, unfertilized eggs need to be removed at an early stage, which makes fertility detection essential. However, conventi... Y. Kuo |
48. Edge-AI-based Dairy Calf Behavior Monitoring System Using Computer Vision and Iot TechnologiesWe present an edge-AI, IoT system for real-time monitoring of dairy calf behavior that runs on embedded system and streams only compact results to the cloud. A lightweight, quantized MoViNet-A2 model deployed on a Raspberry Pi 4 classifies seven behaviors (non-active/active lying, non-active/active standing, feeding, drinking, ruminating) from 4-s clips captured once per minute, and publishes JSON outputs to AWS for dashboards. Field trials on three Holstein calves at the National Taiwan Univ... T. Lin |
49. Embodied Agentic Artificial Intelligence for Precision Agriculture: Cross-domain Experience from Multimodal Generative AIMy team develops inclusive, responsible, and multimodal AI technology across education, healthcare, and digital services grounded in our research in embodied agentic intelligence and large language models. I will share deployed examples from these domains and draw parallels to agriculture, where similar technical challenges persist, ranging from multimodal fusion for contextual reasoning, explainable AI for actionable insights, and data-efficient learning for adaptation and localization. Whil... N. Chen |
50. Enhancing Rice Disease Management: Estimating Pathogen Damage Through Multispectral Imaging AnalysisThis study investigates the application of multispectral imaging (MSI) in conjunction with machine learning algorithms for the early detection and estimation of pathogen damage in rice crops, with a specific focus on Bacterial Leaf Blight (BLB) and Blast diseases. Rice plays a crucial role in global food security, yet these diseases significantly compromise its production. Traditional diagnostic methods are often labor-intensive and time-consuming, necessitating the adoption of innovative tec... I. Sutrisna wijaya |
51. Enhancing Sustainable Farming of Nh: Mechanization of Planting and Post Harvest CleaningNymphoides hydrophylla (NH), commonly known as white water snowflake, is a culturally and nutritionally important aquatic vegetable, particularly valued in Taiwan's Hakka communities. However, its commercial scalability remains limited due to labor-intensive practices in both planting and post-harvest cleaning. This study introduces an integrated mechanized system that combines a seedling planting tool and a cleaning machine, designed to enhance overall production efficiency, reduce ... W. Lin |
52. Estimating Rice Canopy Height Using a Ground-based Slam Lidar SystemThis study evaluates the application of a ground-based LiDAR system, integrated with a Simultaneous Localization and Mapping (SLAM) algorithm, to estimate rice crop canopy height (CH). Using the Velodyne VLP-16 LiDAR sensor, point cloud data were collected and processed to map the rice field. The experimental area covered approximately 600 m² during the crop’s vegetative stage. LiDAR-derived canopy height (LCH) was extracted using percentile-based metrics and compared with manual m... E. Morimoto |
53. Estimation of Crop Coefficient in Malaysian Durian Using Satellite Data and Machine LearningDurian (Durio zibethinus) is a popular fruit and key crop in Southeast Asia, known as the “King of Fruits” for its thorny exterior and distinctive aroma. The crop coefficient (Kc), based on crop evapotranspiration (ETc) and reference evapotranspiration (ETo), is crucial for water efficiency. Currently, there is no Kc value for Malaysian durian. This study introduces a machine learning method utilizing remote sensing data from Sentinel-1, Sentinel-3, and MODIS ET, combined wit... S.K. Balasundram |
54. Evaluating Flight Path Strategy for Uav-based Phenotyping of Individual Muskmelon Plant in Greenhouse EnvironmentsUnmanned Aerial Vehicle (UAV)-based phenotyping is an emerging non-invasive method for high-throughput trait measurement in controlled environments. This study examines how UAV flight trajectory affects reconstruction fidelity and trait accuracy for muskmelon and grape plants in a GPS-denied greenhouse. Two strategies - circular loop and vertical hop - were flown using a UAV with RGB-D SLAM navigation, capturing data with a RunCam Thumb Pro. Data were processed through a GLOMAP structure-from... T. Lin |
55. Evaluation of High-throughput 3d Reconstruction Method for Plants and Its Application to Traits Feature Extraction2D images are widely utilized to monitor and evaluate plant growth, capturing the dynamic and multi-directional nature of plant canopies remains difficult, emphasizing the need for 3D monitoring integrated with plant phenotyping systems.This study aims to introduce a high-throughput plant phenotyping system using 3D plant shape model reconstructed from a dataset of 2D plant images from multiple camera poses. A robot autonomously gathered data by recording video footage of plants from various ... T. Okayasu |
56. Evaluation of Planting Accuracy and Early Growth Uniformity of Spring Cabbage in GreenhousesMechanized transplanting reduces labor and time in greenhouse cabbage production, yet misplacement, over burial, and missing seedlings still compromise uniform stands This study evaluated transplant quality and early growth uniformity with two stages during transplanting and harvesting image and machine learning workflow at plot scale. Two transplanters, automatic and semi-automatic, were tested under ridge widths of 60, 70 and 80 cm and seedling ages of 30 and 35 days. In February after tran... S. Chung |
57. Field Testing of a Laboratory-made Portable Hydroponic Nutrient Analyzer with Ion-selective ElectrodesAs a strategy to address climate change and declining agricultural productivity, hydroponic systems have gained increasing attention. In particular, precise control of nutrient ion composition in nutrient solutions is essential for ensuring stable crop growth and improving product quality. However, most hydroponic farms currently rely on pH and electrical conductivity (EC) sensors for nutrient solution management. While EC reflects the overall ionic strength, it does not provide quantitative ... H. Kim |
58. Fusing Deep Learning and Control Theory for Optimized Sugar Beet Yield PredictionAccurate yield prediction is a vital field of research in precision agriculture, enabling optimal resource allocation and enhanced food security under growing climatic uncertainty. Traditional models struggle to capture complex, non-linear interactions between environmental drivers and crop growth. To address this, we present our approach, a multi-stage method for sugar beet yield prediction and management that integrates deep learning with control-theoretic techniques and mathematical langua... A. Tabbassi |
59. High-reliability Navigation for Multi-functional Robots Using Rfid Triggers and 3d Slam in a Protected HorticultureProtected horticulture in Japan is facing a serious labor shortage, yet existing robots have not achieved sufficient return on investment, and their adoption remains limited. To support the deployment of multi-functional robots, we developed a high-reliability autonomous navigation system that integrates RFID-based event-triggered state transitions with LiDAR-based simultaneous localization and mapping (SLAM).The developed mobile platform was built on an omnidirectional robot equipped with fo... T. Okayasu |
60. Identification of Citrus Diseases, Pests, and Disorders Using Deep LearningTaiwan’s warm climate offers favorable conditions for citrus production, making it the most economically valuable fruit crop in the country. Citrus trees are perennial and mainly propagated asexually. Long-term exposure and limited genetic diversity make them more susceptible to infection by various pathogens. In practice, diagnosis often relies on farmers’ experience, which can be subjective despite their familiarity with local conditions. Microscopic examination by plant patholo... Y. Kuo |
61. Identification of Cucumber Pests, Diseases, and Disorders Using Deep LearningCucumber is an essential economic crop worldwide, which is typically cultivated in summer. The hot and humid conditions make them highly susceptible to various pests, diseases, and physiological disorders, which hinder their growth and lead to significant yield losses. Early and accurate detection is vital to limiting the spread of diseases or pests. However, traditional diagnostic approaches rely heavily on visual inspection by experienced farmers or microscopic examination by specialists, w... Y. Kuo |
62. Improving Depth Accuracy by Using a Real-time Monitoring System for Traditional Tillage MachineryTillage depth has a great influence on soil quality, fuel consumption, and equipment durability in mechanized farming. However, traditional methods often maintain a fixed depth, lacking the ability to adjust in real time. This study proposes a real-time monitoring system that significantly improves the depth measurement accuracy of traditional tillage machinery. The system is equipped with a soil contact wheel combined with an angle sensor, which converts the rotation angle into a depth value... W. Lin |
63. Innovating Irrigation: Affordable Smart Solutions for Water SustainabilityAgriculture accounts for 70–80% of global freshwater use, a level increasingly unsustainable under climate change. This study reports the development and field validation of a low-cost smart irrigation system for tomato and melon in Tuscany (2021–2023). The system integrates evapotranspiration-based models, wireless sensor networks, and adaptive control algorithms. In 2023 it achieved up to 50% water savings compared to traditional practices, without yield reduction, at a total co... A. Matese |
64. Integration of a Real-time Dairy Cow Eye Temperature Monitoring System Based on Deep Learning and Thermal ImagingEarly detection of heat stress and illness in dairy cows is critical for maintaining herd health and optimizing milk production. Among various physiological signals, body temperature is a key indicator of health status. In this study, we present a real-time, non-contact monitoring system that integrates dual-channel thermal imaging and deep learning for precise and automated surveillance. The system processes RGB and thermal video streams in parallel: in the RGB channel, YOLO detects faces, B... T. Lin |
65. Investigation of Seed Monitoring Potential Using Light Dependent Resistor (Ldr) for Cell Type Precision SeedersPrecision seeding is an important operation in modern agriculture, ensuring accurate seed placement at defined rates and intervals to optimize crop performance. Despite their critical importance, conventional seed metering devices often require frequent manual calibration, making them labor-intensive, inefficient, and impractical for both smallholder and large-scale farming operations. Existing seed monitoring technologies are often costly and lack real-time adaptability to varying field cond... S. Chung |
66. Laser- Induced Enhancement of Seed Germination and Early Growth in LegumesLaser technologies are emerging as promising tools in precision agriculture for enhancing plant development and productivity. This study investigates the effects of low- power laser irradiation (532 nm, 1W) on the seed germination and early growth of mung beans (Vigna radiata). Seeds were exposed to laser light prior to planting, and their germination performance, leaf expansion, chlorophyll content, and shoot length were measured and compared to untreated control seeds. The laser-treated see... C. Ding |
67. Lauraceae Timber Identification Using Vision TransformerThe forest coverage in Taiwan exceeds 60%, yet over 99% of annual timber consumption relies on imports. This significant dependence, coupled with frequent incidents of wood misidentification and fraud, highlights the need for accurate and efficient wood species identification systems. Conventional approaches, such as microscopic analysis and sensory- based macroscopic inspection, are labor-intensive, subjective, and require domain expertise, making them unsuitable for large-scale or real-time... Y. Kuo |
68. Low-code Development Environment and Middleware for Ubiquitous Environment Control SystemsThis work presents a low-code development environment that enables non-engineers to construct a customized software for UECS devices automating horticultural facilities as well as a middleware that provides a uniform application executing environment on different platforms for the UECS software. ... T. Nakanishi |
69. Machine Learning Prediction Models for Dual-Horizon Egg Production ForecastingEgg production forecasting presents significant challenges in agricultural supply chain management due to complex seasonal patterns, disease outbreaks, and market volatility. Although various forecasting models have been developed for agricultural production, limited research has systematically compared model performance across different temporal horizons or developed integrated frameworks optimized for diverse planning needs. This study develops a comparative dual-horizon machine learning fr... S. Chen |
70. Measure, Model, Manage: the Unfinished Revolution in AgricultureOver the last 40 years, the paradigm of Measure, Model, Manage has promised an agricultural revolution through data-informed precision management. This shift remains largely incomplete, lagging concurrent innovations in genetics and pesticides. Significant barriers persist in achieving breakthrough innovations for crop data collection and the development of data analysis/decision-making systems. These hurdles include a decades-old "Sensor Crisis" (a lack of appropriate too... A. Werner, A. Holmes |
71. Mobile-based Automated Phenotyping System for Accessible Tomato BreedingTomato breeding programs require extensive phenotypic data collection including fruit development stages and critical timing parameters, yet manual monitoring is labor- intensive and limits breeding program scalability, particularly in resource-limited environments. This study presents a cost-effective automated phenotyping system that requires only smartphone video recording combined with pre-assigned plot numbers, eliminating the need for expensive mobile platforms and making advanced breed... S. Chen |
72. Modeling and Characterization of Unimodal and Bimodal Diurnal Pollen Foraging Patterns in Honeybee ColoniesPollen foraging patterns in honeybee colonies provide essential information on their ecological adaptation strategies. This study proposes a statistical modeling framework to characterize diurnal pollen foraging patterns in honeybee colonies. To support this, data were collected from healthy honeybee colonies during controlled experimental period. The raw pollen harvest data were then segmented into daily time series and converted into hourly histograms to capture foraging rhythms more effect... T. Lin |
73. Modeling the Effects of Greenhouse Environmental Factors on Soft Rot Incidence in PhalaenopsisPhalaenopsis spp. is one of Taiwan’s most important ornamental crops for export. However, during greenhouse cultivation, Phalaenopsis is frequently threatened by bacterial soft rot (Erwinia spp.), particularly under high-temperature and high-humidity conditions that accelerate pathogen spread and cause severe losses in seedlings. This study was conducted in a Phalaenopsis greenhouse located in Houbi District, Tainan, Taiwan. The greenhouse contained 21 planting beds, which wer... C. Huang |
74. Monitoring Chicken Houses with AI Surveillance SystemIn Taiwan, the need of chicken meat accounts for approximately 30% of total livestockvproduction. In order to maintaining animal welfare, floor-rearing chicken farming approaches are widely used in Taiwan. However, traditional poultry management is often labor-intensive which increases the risk of disease transmission. To improve monitoring efficiency, we proposed a smart rail surveillance system to automatically monitor chickens for real-time chicken health assessment. The system comprised a... Y. Kuo |
75. Multivariate Linear Regression Modeling for Predicting Chicken Body Weight Using Age, Uniformity, and Growth RateAccurate estimation of chicken body weight is critical for optimizing feed management, harvesting schedules, and animal welfare in commercial poultry systems. This study proposes a robust predictive framework using multivariate linear regression to estimate the average weight of native broiler chickens based on three explanatory variables: age, uniformity, and daily growth rate. After rigorous data cleaning and outlier removal, the model was trained and validated on 43 field observations coll... H. Lin |
76. Nighttime Piglet Detection Using Deep LearningIn 2023, Taiwan’s pig industry was valued at over NT$85.1 billion, representing nearly 40% of total livestock production. However, effective piglet management remains a challenge due to environmental variability, frequent aggressive behaviors, and labor shortages—especially during nighttime. Traditional monitoring methods rely on manual observation, which is time-consuming, subjective, and impractical for continuous surveillance. To address this, we propose an automated nighttime ... Y. Kuo |
77. Non-destructive Tilapia Quality Determination Using Near-infrared SpectroscopyTilapia represents a significant economic asset in the aquaculture industry due to its high nutritional value and commercial importance. However, internal abnormalities are frequently detected during processing operations, particularly those caused by Streptococcosis, which is among the most prevalent diseases affecting tilapia quality. These quality defects often lead to commercial disputes between aquaculture farmers and fillet processors, highlighting the critical need for non-destructive ... S. Chen |
78. Null Dataset-Based Detection Enhances Robotic Vision in Greenhouse Cherry Tomato HarvestingCluttered cherry tomato greenhouse environments with visually similar distractors often trigger False Positives (FPs) in robotic vision, misguiding the robot’s motion and reducing harvesting success. We introduce a null-dataset strategy that integrates unannotated distractor images into YOLOv8l training, with their proportion tuned through loop refinement to suppress FPs while preserving precision. Optimal null proportions were identified as 12.3% for tomato detection and 8.3% for pedic... P. Yen |
79. Optimizing Power Delivery in Electric Farm Machinery Using a Hybrid Battery and Ultracapacitor SystemAgriculture plays a significant role in global greenhouse gas emissions, contributing notably to climate change. Integrating renewable energy into agricultural operations has become increasingly vital in addressing this challenge. This study investigates the potential of electrifying agricultural machinery using a hybrid energy storage system that combines batteries and ultracapacitors. While batteries offer high energy density, they face limitations such as slow charging and reduced lifespan... S. Wu-yang |
80. Performance Evaluation of Agricultural Spray Nozzle Under Different Pressure Conditions by Image AnalysisSpray nozzles are critical components in agricultural equipment used for pest control, pollination, and so on. The liquid ejected from the nozzle is broken down into droplets due to friction with the air and pressure changes. Consequently, the nozzle performance is often defined by alternative parameters to estimate the actual operating conditions. This study aims to determine the operating parameters of spray injection by photographing the movement of droplets ejected from a nozzle under dif... T. Okayasu |
81. Pest and Disease Image-text Identification System of Leafy Vegetables in Urban Community FarmingUrban community farming has been integrated into education for sustainable food and agriculture. However, the participants are primarily students and novice farmers with limited background knowledge. Managing pests and diseases becomes challenging for these growers as diverse vegetable crops attract various pest and disease species, requiring accurate identification and treatment expertise. There is a need to develop timely identification services and guidance on control measures. In the... S. Chen |
82. Phalaenopsis Seedling Assessment Using Leaf Contour Detection with YOLOIn this study, we propose a vision-based approach for automatically measuring the morphological traits of Phalaenopsis seedlings. By utilizing top-view and side-view images, our method automatically extracts leaf contours to replace traditional manual measurements. A YOLOv8n-seg model was employed to segment the seedlings, and further correction strategies were introduced to improve accuracy. Experimental results demonstrate the potential of our approach to support large-scale seedling classi... Y. Kuo |
83. Plantsaga: Integrating Segment Anything Model with Gaussian Splatting for Plant Organ-level 3d SegmentationOrgan-level 3D phenotyping is essential for crop breeding but remains limited by the high cost of manual annotations. To address this challenge, PlantSAGA (Plant Segment Anything Gaussian Splatting) is introduced as a reference-based framework that enables accurate organ segmentation with minimal annotation. Multi-view muskmelon plants were reconstructed using COLMAP for camera pose estimation and Gaussian Splatting for 3D modeling, while 1~10 reference masks guided organ-level discrimination... T. Lin |
84. Portable DNA Detection Tool for Halal Monitoring Using Spectral SensingPork and its derivatives are non-halal in Islam, raising concerns about cross- contamination in food. With the growing number of Muslim tourists and Taiwan’s efforts to expand its halal F&B exports, strict halal compliance and reliable detection methods are essential. Conventional techniques like PCR offer high accuracy but are limited by long processing times and the need for advanced laboratories. Recombinase Polymerase Amplification (RPA) presents a faster alternative, operating ... J. Chen |
85. Potential of Plant Phenotyping for Data-driven Greenhouse HorticultureWe are trying to investigate the use of various features extracted from plant images for the purpose of environmental control in greenhouses according to the growth conditions of plants. A measurement robot was utilized in order to collect plant images. Plant growth features (apical buds, flowers, fruits, etc.) were extracted by using a deep learning-based detector. In addition, we also introduced a 3D reconstruction technology to obtain the plant shape features such as plant height, internod... T. Okayasu |
86. Power Consumption Signal Characterization of Bldc-based Agricultural Fans for Malfunction Detection for Smart GreenhousesEffective management of environmental parameters, notably temperature and humidity, is critical for ensuring optimal plant growth and productivity in smart greenhouses. Brushless (BLDC) fans are commonly utilized for controlling greenhouse ventilation and humidity levels. The primary aim of this study was to characterize the power consumption of BLDC agricultural fans to identify operational anomalies and facilitate predictive maintenance strategies. An experimental setup was devised, involvi... S. Chung |
87. Precision Nutrient Management in the USA: Current Trends and Future OpportunitiesPrecision nutrient management (PNM) has become integral to modern U.S. agriculture, particularly in optimizing fertilizer use efficiency, reducing environmental impacts, and sustaining profitability. As detailed in recent analyses, the adoption of precision technologies for nutrient management in the U.S. is advanced, especially among large- scale operations in the Midwest Corn Belt. Key technologies facilitating PNM include variable rate technology (VRT), remote and proximal sensing, soil an... S. Phillips |
88. Preliminary Tests for Potato Yield Monitoring Using a Controlled Test BenchAccurate yield estimation is a critical aspect of precision agriculture, particularly for root crops such as potatoes, where direct measurement during harvest can be challenging and labor-intensive. Developing precise and automated methods to enhance the efficiency and accuracy of yield assessments is thus imperative. This study explores the potential of integrating vision-based imaging and non-contact sensing technologies to achieve accurate potato mass estimation under controlled laboratory... S. Chung |
89. Quantitative Assessment of Discharge Depth Effects on Lithium-Based Batteries: LTO, LFP, and NCMThis study explores the impact of depth of discharge (DoD) on the performance degradation of three lithium-based battery chemistries: lithium titanate (LTO), lithium iron phosphate (LFP), and nickel cobalt manganese oxide (NCM). The objective is to establish a standardized methodology for evaluating battery health under partial cycling and to quantify the degradation behavior across three DoD ranges: 0–33%, 34–66%, and 67–100%. LFP and NCM cells were cycled at 1C under room ... C. Huang |
90. Reducing Ground Losses Using a Leaf Segmentation-based Autonomous Sprayer for Papaya GreenhousesPapaya plants have irregular canopy structures, making traditional spraying methods highly labour-intensive and prone to chemical waste due to non-selective application. In precision agriculture, delivering pesticides accurately to target areas is crucial for reducing labour requirements, costs, and environmental impact. Therefore, the integration of smart agricultural machinery and machine vision is necessary to optimise pesticide application. In this study, a low-cost autonomous spraying sy... W. Lin |
91. Regression Model for Estimating Branch Number of Soybean Using Uav-based Multispectral ImagesSoybean (Glycine max (L.) Merr.) is a protein-rich crop, and the number of branches is a significant trait associated with yield. This study aims to estimate the branch number of soybeans using vegetation indices (VIs) extracted from multispectral images mounted on a UAV. The study was conducted on the soybean cultivar Seonpung, sown on June 20, 2022, and June 24, 2023. Vegetation growth was investigated on 20 control and 30 treatment samples on August 20 and September 20, 2022, August 21 and... C. Ryu |
92. Revolutionizing Poultry Health: AI-Powered Real-Time Disease Detection Using YOLO v7 and IQR for Enhanced Farm ProductivityPrompt and accurate detection of poultry diseases is crucial to prevent outbreaks and reduce economic losses. Conventional monitoring systems based on manual inspections are inefficient and prone to error, delaying timely interventions. This study proposes an AI-driven early warning system that integrates YOLO v7 for real-time image detection with Hampel Filters for anomaly recognition. The model specifically targets two critical health indicators: rooster combs and eyes. Over a period of 53 ... A. Santosa |
93. Rgb-based Soil Water Content Prediction Enhanced by Hyperspectral CalibrationWhile hyperspectral imaging (HSI) cameras demonstrate high accuracy for detecting soil water content (SWC)-related spectral variations, their field deployment remains constrained by prohibitive costs and operational complexity. This study investigates utilizing low-cost RGB cameras through HSI-guided calibration for SWC estimation. 210 paired HSI-RGB measurements were acquired across five soil texture classes (0-40% fine particles), fourteen moisture levels (0-39% SWC), and three illumination... J. Park |
94. Robotic Arm Tomato Harvesting System and Next Best View Algorithm DevelopmentReplacing human labor with robots is a trend for future agriculture due to its efficiency and consistency. However, in automatic fruit harvesting tasks, leaf occlusion and the dynamic orientation of fruit make it difficult for robots to directly observe the picking point. To address this problem, this research focuses on tomato harvesting, and proposes a next-best-view (NBV) algorithm based on two main structures: “tomato pose prediction” and a “target-hit-gain function&rdqu... P. Yen |
95. Signal Characterization for Actuator Operation Status Monitoring in Smart Vertical FarmsVertical farming presents a sustainable solution for high-yield crop production in space- constrained environments by enabling precise control over environmental parameters. However, effective implementation depends not only on environmental monitoring but also on the reliable operation of actuators that regulate system condition. The objective of this study was to characterize power consumption signals from actuators within smart vertical farms to facilitate precise monitoring, assessment of... S. Chung |
96. Signal Characterization of Environmental Sensors for Abnormality Detection in Hot Temperature GreenhousesMaintaining optimal microclimatic conditions is critical for crop productivity in greenhouse cultivation. High-temperature environments can induce subtle but critical deviations in environmental parameters, often resulting in reduced crop growth, quality, and yield. This study aimed to characterize the raw signal behavior of environmental sensors to enable early detection of abnormal conditions in hot-temperature greenhouses. An internet of things (IoT)-based sensor network comprising tempera... S. Chung |
97. Signal Characterization of Ict Components for Malfunction Detection for Open-field Irrigation SystemsAgricultural practices in open fields increasingly rely on automated irrigation technologies and ICT components, whose operational status impacts their reliability and efficiency. This study aimed to develop a malfunction detection pattern for sensors and actuators through signal characterization in an open-field irrigation setup. The experiment included environmental sensors and actuators, interfaced with a programmed microcontroller, operating in cycles (On/Off) or alternatively. Signals we... S. Chung |
98. Signal Characterization of Sensors for Operational Status Monitoring in Smart Vertical FarmsVertical farming represents an advanced agricultural practice capable of efficiently producing high-quality crops through precise environmental management, optimal spatial utilization, and consistent production outcomes. Ensuring reliable and accurate performance of environmental sensors is essential for sustaining ideal growth conditions within these advanced agricultural systems. This study aimed to characterize signals from environmental sensors to enhance real-time operational status moni... S. Chung |
99. Smartflow: Ai Optimization of Desalination for Sustainable Agricultural Water ManagementLimited access to reliable freshwater sources is a persistent barrier to agricultural productivity, particularly in coastal and arid regions where rivers, lakes, and groundwater reserves are rapidly declining. Farmers in these areas often struggle to meet irrigation demands, resulting in reduced yields and heightened vulnerability to climate variability. Although seawater desalination provides a potential alternative, conventional reverse osmosis (RO) systems are typically too energy-intensiv... M. Jamaludin |
100. Smartphone Application for Real-time Environment Monitoring of Smart GreenhousesSmart greenhouse technologies significantly enhance agricultural productivity, sustainability, and resource efficiency, yet existing solutions often face limitations regarding affordability, real-time responsiveness, and scalability, especially for small- and medium-sized farms. This research introduces a cost-effective, scalable smartphone- based application designed for real-time monitoring and precise control of essential greenhouse environmental parameters, including temperature, relative... S. Chung |
101. Study on Contect Sensor-based Ridge Tracking Technology for Precision Garlic SeedingRidges are an important part of field operations in agriculture. From soil tillage and sowing to harvesting, ridges serve as the foundation throughout the entire crop production cycle. However, in practical application, ridges are often irregular and poorly maintained. Irregular ridge can disrupt consistent seeding which can result in uneven crop growth and a decline in overall productivity. In the case of garlic, seeding uniformity is directly related to yield. Therefore, addressing the unev... H. Kim |
102. Synthetic Data-driven Validation of Multi-stage Fruit Detection Systems in Controlled Virtual EnvironmentsAccurate fruit counting across development stage is critical for tomato breeding decisions. Yet, the ground truth validation in real field remains challenging where partially occluded fruits cannot be reliably counted manually due to complex environmental factors. To address this need, this study presents a photorealistic simulation approach that complements real field data collection. A virtual environment enables controlled evaluation across three distinct fruit growth stages: green stage f... S. Chen |
103. Theoretical Analysis of Deflection in Deformed Silicone Components for Dried Longan PeelingIn traditional manual processing of dried longan, the fruit is typically peeled by cutting from the stem end with a knife and tearing along the seed axis to separate the flesh. However, to enhance operational efficiency and realize production automation, the development of dried longan processing machinery with automatic peeling capabilities has become an inevitable trend in the industry. The most critical component of such machines is the peeling module, whose geometry and dimensions directl... C. Cheng |
104. Towards in Situ Monitoring of Root Growth Traits: Combining Spectral Imaging with Transparent Bed HydroponicsWe developed a novel method that enables non-laboratory monitoring of the growth characteristics of crop root systems by combining spectral imaging with a transparent bed hydroponics. Root systems of spinach grown were observed through the transparent bottom plate using a hyperspectral camera daily. An optimal index for the classification of root ages (days after emergence) was determined as the ratio of reflectance at 498 and 601 nm. Additionally, the distribution of root age was visualized ... D. Yasutake |
105. Unlocking Canopy Dynamics: Uav-lidar-based Biomass Estimation in Ocimum BasilicumUAV-LiDAR offers a high-throughput route to phenotyping and biomass estimation in basil (Ocimum basilicum L.). Over three crops seasons (2021–2023), we evaluated three commercial varieties across 96 plots under different irrigation regimes and sowing densities. Multi-temporal LiDAR acquisitions quantified canopy height, LAI and volume and were validated against ground truth. Canopy volume strongly predicted fresh biomass (R² = 0.93; mean error < 8%). Across years, fresh bio... P. Toscano |
106. Unsupervised Anomaly Detection of Tipburn in Leafy Vegetables Using Denoising AutoencoderTipburn, a common physiological disorder in leafy vegetables, presents as marginal necrosis but its fuzzy boundaries make annotation costly and inconsistent. We present a label-free pipeline that combines CIE Lab–based preprocessing with a chroma-only denoising autoencoder (DAE) trained solely on healthy samples for real-time, pixel-level anomaly mapping. Lettuce images were acquired under controlled lighting, segmented in CIE Lab space, and reduced to the a channel and a/b chromatic ra... M. Yang |
107. Unsupervised Hyperspectral Image Segmentation Using Deep Global ClusteringHyperspectral imaging (HSI) combines rich spectral and spatial information, supporting field monitoring and crop assessment in precision agriculture. HSI scenes from one dataset usually share the same background and foreground classes, yet spectra from one region differ from those in another. Pixels that describe the same object therefore cluster together in spectral space; mapping these clusters back onto the image yields pseudo-segmentations that can stand in for class labels. However, proc... S. Chen |
108. Using Floral Bract Withering to Identify Green-ripe Pineapples with Deep LearningGreen-ripe pineapples are ideal for extended transportation and storage during summer but are challenging to identify during on-site harvesting. This study introduces a deep learning-based approach using the YOLO-NAS algorithm to detect green-ripe pineapples by analyzing the withering rate of floral bracts at the fruit's base. A high- mounted tracked vehicle, equipped with an Intel D405 depth camera, captures images at a distance of 300–400 mm as it navigates pineapple ridges. The s... S. Chen |
109. Yolo Strawberry Maturity Classification and Harvest Priority with 3d CameraAccurate harvesting timing is essential to improve crop quality and productivity, and recent advances in agricultural automation have led to the emergence of fruit maturity classification and harvest optimization algorithms for agricultural robots as major technical challenges. This study proposes a pipeline for strawberry object detection, maturity classification, distance estimation, and harvest priority. We train a YOLOv8 detector on an open RGB dataset, and estimate the camera-fruit dista... M. Yang |