Converting data to decisions is one of the major precision agriculture challenges. This webinar includes presentations on: using cell phone images to estimate soil organic carbon, neural networks to classify weed images, video based deep learning for estimating maize ear height, and automated data correction for plant breeding locations.
Join us on September 10th, 2020 at 13:30 Central Daylight Time (UTC -5).
Soil organic matter (SOM) is considered as the backbone of soil health and soil quality. Thus, its’ estimation is critical to support the development of management decision including precision agriculture. To overcome challenges of laborious, rather expensive, and time-consuming laboratory measurements, recent advances in image acquisition systems provided a new dimension of image-based SOM prediction. However, challenges remain in using soil images taken directly in the field due to variable soil surface conditions including vegetation cover, illumination, and soil moisture. This study quantifies the effects of soil moisture on the relationship between SOM and color parameters derived from cell phone images and establishes suitable SOM prediction models under varying conditions of soil moisture contents (SMCs). This study showed potential of cellular phone to be used as a proximal soil sensor fast, accurate and non-destructive estimation of SOM both in the laboratory and field conditions.
Iftach Klapp - A Deep Learning System for Estimation of Melons Individual Weight and Location Using UAV Images
George Mohler - Automated Corn Ear Height Prediction Using Video-Based Deep Learning
In corn breeding, hand-measurement of ear height is a labor-intensive process, thus limiting scalability. Here we show that it is feasible to automate estimation of the average ear height of a row of corn in experimental fields used for corn breeding. For this purpose we use point pattern analysis on predicted shank-node locations extracted from video captured on uncalibrated cameras moving through a plot at a fixed height from the ground (4 feet and 2 feet). First, a convolutional neural network-based object detection system (YOLOv3) was trained to detect the ear-stalk connection point and applied to the collected videos. Detected ear position and time information from each frame were super-imposed into a point pattern and point-features were then extracted. Using ridge regression to predict the average ear height per plot, we achieved 0.772 concordance, 2.989 inches root mean squared error, and 2.263 inches mean absolute error compared with hand-measured average ear height. This deep learning system can be utilized by mounting cameras onto the plot combine harvester to collect the necessary videos during harvest and could be expanded to quantify other phenotype measurements of interest that are labor-intensive to collect.
Getiria Onsongo - A Tool for Automatically Identifying and Correcting Errors in Plant Breeding Location Data
Advances in big data technologies offer great promise in the area of data-driven plant breeding. To fully realize this promise, disparate sources of data such as genotype, environment, management and socioeconomic data need to be integrated. One of the primary challenges to collectively analyzing these disparate sources is errors in location data. Common errors include flipped latitude and longitude values, missing negative signs and in some cases missing location data. In this presentation, a tool for automatically detecting and correcting errors in location data will be presented. It includes a visualization tool that plots both flagged and corrected locations on a map making it easy for users to validate results.
Francisco R Pereira - Imputation of Missing Parts on Orthomosaic Images Obtained by an UAV Using Data Mining Techniques
In recent years, the emergence of Unmanned Aerial Vehicles (UAV), with high spatial resolution, has broadened the application of remote sensing in agriculture. However, UAV images commonly have specific problems with missing areas due to drone flight restrictions. Data mining techniques for imputing missing data is an activity often demanded in several fields of science. In this context, this research used the same approach to predict missing parts on orthomosaics obtained by UAV using a PlanetScope and Sentinel images as auxiliary data. The spectral bands (blue, green, red and infrared) and the NDVI (Normalized Difference Vegetation Index) derived from the PlanetScope and Sentinel-2 images were used as predictor variables. The prediction accuracies and comparison between original and estimated maps showed the Randon Forest algorithm use is feasible for the proposed objective.
Dharmendra Saraswat - Performance of Convolution Neural Network for Assessing Classification Accuracy of Weed Images Acquired From Publicly Available Datasets
Moderator: Dr. Alex Thomasson, Head, Department of Agricultural and Biological Engineering, Mississippi State University