Wheat Lodging Segmentation Based on Lstm_PSPNet Deep Learning Network

Lodging is one of the major issues that seriously affects wheat quality and yield. To obtain timely accurate lodging information identify potential factors leading to lodged in breeding programs, we proposed a lodging-detecting model coupled with unmanned aerial vehicle (UAV) image features at multiple plant growth stages. The UAV was used collect canopy images ground area five PSPNet improved by combining convolutional LSTM (ConvLSTM) timing model, inserting attention module (CBAM) Tversky loss function. effect network monitoring under different sizes stages investigated. experimental results show (1) Lstm_PSPNet more effective prediction, precision reached 0.952; (2) choosing an appropriate size could improve segmentation accuracy, optimal this study being 468 × 468; (3) its accuracy sequentially from early flowering late maturity, three evaluation metrics increased 0.932 0.952 for precision, 0.912 0.940 recall, 0.922 0.950 F1-Score, good extraction mid reproductive Therefore, can make full use temporal sequence effectively extract areas provide comprehensive reference technical support crops