Uncrewed Systems Technology 049 - April/May 2023

10 April/May 2023 | Uncrewed Systems Technology Researchers in China and Singapore have developed an automated method of counting rice plants using UAVs (writes Nick Flaherty). Rice is cultivated on nearly 162 million hectares of land worldwide. One of the most common ways to quantify its production is to count the number of rice plants to estimate yield, diagnose growth and assess losses in paddy fields. “The new technique uses UAVs to capture RGB images of a paddy field,” said Professor Jianguo Yao from Nanjing University of Posts and Telecommunications in China, who led the study. “The images are then processed using a deep learning network we have developed, called RiceNet, which can accurately identify the density of rice plants in a field, as well as provide higher-level semantic features such as crop location and size.” The RiceNet machine learning framework consists of one feature extractor, at the front end, that analyses the input images, and three feature decoder modules that are responsible for estimating the density of plants in a paddy field, the location of the plants and their size. The last two features are particularly important for future research into automated crop management techniques, such as spraying fertilisers. As a part of the study, the researchers deployed a camera-equipped UAV over rice fields in the Chinese city of Nanchang, capturing images measuring 5472 x 3648 pixels. They then used some of the images as a data set as a reference to train the system, and the rest as a test data set to validate the analytical findings. Out of the 355 images with 257,793 manually labelled points, 246 were randomly selected and used as training images, while the remaining 109 were used as test images. Each image contained an average of 726 rice plants. The RiceNet technique’s signal- to-noise ratio enables it to efficiently distinguish rice plants from their background. The results of the study showed that RiceNet’s mean absolute error and root mean square error were 8.6 and 11.2 respectively, which is comparable to the data generated using manual methods. The research threw up some key tips for effective automation of the counting process. For instance, the team does not recommend acquiring images on rainy days. It also suggests collecting UAV-based images within 4 hours after sunrise to minimise fog time as well as the occurrence of rice leaf curls, both of which adversely affect the output quality. “In addition, we further validated the performance of our technique using two other popular crop data sets,” said Prof Yao. “The results showed that our method significantly outperforms other state-of-the-art techniques, which underscores the potential of RiceNet to replace the traditional manual method.” The RiceNet framework can also be used for other UAV- and deep learning- based crop analysis techniques, to help improve the production of food and cash crops worldwide. Airborne vehicles Rice counter crops up Tests showed that the RiceNet technique outperformed other counting methods The RiceNet system can accurately identify the density of rice plants in a paddy field, as well as provide features such as crop location and size