Detection of cotton waterlogging stress based on hyperspectral images and convolutional neural network

Waterlogging in the early stage of cotton will reduce number bolls and do harm to yield. Early detection waterlogging help farmers adjust management save loss. To evaluate application deep learning for waterlogging, this study applied a convolutional neural network (CNN) classify different durations stress (0, 2, 4, 6, 8, 10 d) based on hyperspectral images (HSIs) leaves. An experiment was designed simulate situation under collect HSIs visible near-infrared (VNIR 450-950 nm) spectra with 126 bands 66 d after sowing (66 DAS). It found spectral curve reflectance higher than that non-waterlogging cotton. Especially near 550 nm 750 nm, increased there were ‘blue shift’ phenomena position red edge spectra. The first principal components band randomly discarding component analysis (PCA) used build dataset. GoogLeNet Inception-v3 (GLNI-v3) VGG-16 models selected detect results showed average time round training GLNI-v3 13.337 s, classification accuracy 96.95% loss value 0.09431. 21.470 97.00% 0.17912. Though these two had similar value, achieved high fewer iterations. short-term can be detected by leaves CNN are suitable HSIs, method provide support yield estimation assessment waterlogging. Keywords: cotton, image, DOI: 10.25165/j.ijabe.20211402.6023 Citation: Zhao J, Pan F Li Z M, Lan Y B, Lu L Q, Yang D et al. Detection network. Int J Agric & Biol Eng, 2021; 14(2): 167–174.