Wheat Ear Detection in RGB and Thermal Images Using Deep Neural Networks
| Main Authors: | Grbović Željana, Panić Marko, Marko Oskar, Brdar Sanja, Crnojević Vladimir |
|---|---|
| Format: | Proceeding eJournal |
| Bahasa: | eng |
| Terbitan: |
, 2019
|
| Subjects: | |
| Online Access: |
https://zenodo.org/record/3492023 |
Daftar Isi:
- The number of farmers who use smart phones is increasing rapidly and furthermore RGB and thermal cameras are becoming more and more available either as smart phone gadgets or as integrated parts of the smart phone. Using them, farmers could have early information about the wheat yield. Currently, counting ears on part of a field and extrapolating the values for the whole field requires ears to be counted manually, which is prone to subjective evaluation, takes a lot of time and requires large human resources. In the case of larger fields, samples must be taken from more than one location, which additionally slows down the process. The aim of the research was to develop a system for wheat ear recognition and counting, which is necessary step for further estimation of wheat ear coverage density in the field, and yield prediction. Images of winter wheat were taken at 4 dates during the growing season and segmented manually to acquire the ground truth. Image segmentation was done using deep learning. Namely, convolutional neural networks were applied to RGB and thermal images and the results were compared to the ground truth to assess the system accuracy. Development of a comprehensive system for ear counting and yield prediction has a huge practical value for crop monitoring and optimal decision-making in wheat production.