Database Research Summaries2018 Imaging for Crop Stress Diagnosis in Soybeans
Research Focus
The focus of this project is to create a sensing strategy that improves scouting effectiveness and accuracy at a reduced cost.
Objectives
- Soybean Intra- Canopy Imaging SSCHS – Preliminary work has been completed on the SSCHS and image acquisition system. Several stock cameras are linked and controlled from a single, synchronized “main” camera. A second SSCHS will be developed for specifically for soybean crop canopies.
- Reference Library Development for Soybean Crop Health Assessment – To build the Reference Library, a network of crop scouts, growers, and state extensions specialists will be contacted. The library will initially be limited only to crop health problems of importance to Ohio soybean producers.
- Computational Process/Algorithm Development for Soybean Crop Health Classification – Post image acquisition data are transferred from the camera to the image processing, feature extraction and classification software environment.
- Commercial Scale Field Testing – In examining past research in the field, another notable weakness is that the test plots/fields are usually limited to one or two locations.
Results
- It was noted that there was an issue with the ability of the imaging sensors to compensate for intense direct sunlight. Various algorithm approaches have been tested to adjust the functioning of the cameras to account for this. However, given the time of year, there has yet to be an opportunity to test these programming changes. Further, as a secondary plan we have looked at using neutral density filters to help control the amount of light entering the camera and prevent over-saturation.
- From the 2018 cropping season testing, we have designed a Graphical User Interface [GUI] to enable intuitive control of the cameras as well as viewing of the results. We will continue to develop and test this GUI with relevant stakeholders during the second year of the project. We have also continued with Convolutional Neural Net (CNN) training. As the group learns more about artificial intelligence, we continue to improve the functioning of these algorithms.
Importance
- The benefits of an expanded crop health reference library, through the addition of soybean imagery, will include: a quick turnaround time for soybean crop health diagnostics, potential use as an educational tool during grower meetings showing digital images of insect, disease, and nutrient damage; and expanded in-service training opportunities for county educators.
For more information about this research project, please visit the National Soybean Checkoff Research Database.
Funded in part by the soybean checkoff.
