Portal de Periódicos da CAPES

Research on the Distribution Map of Weeds in Rice Field Based on SegNet

2022; Springer Nature; Linguagem: Inglês

10.1007/978-981-19-2452-1_9

2190-3026

Sheng Zhu, Shihao Li, Ze Yang,

Remote Sensing and LiDAR Applications

Rice is one of the major economic crops in China. Weeds in paddy fields compete with rice for natural resources such as sunlight, water, and soil. At present, most farmers use uniform spraying for weeding. Unreasonable use of herbicides will have a serious impact on the quality of rice. Visible light images of rice paddies taken by an unmanned aerial vehicle (UAV) are used in this paper. Three semantic segmentation models, FCN, U-Net, and SegNet, were used to recognize rice straw images. The pixel accuracy (PA) (PA: retain the first only) of the three models is 88.8%, 89.4% and 84.5%, respectively. The mean intersection over union (MIoU) (MIoU: retain the first only) was 61.4%, 68.8%, and 64.0%, respectively. The experimental results show that the research method of UAV remote sensing images based on the SegNet deep learning model can effectively reflect the difference between rice and weeds. The distribution map of weeds in paddy fields was obtained, and then, the plant protection UAV was guided to apply pesticides accurately.