Cotton pests classification in field-based images using deep residual networks

Abstract Pests in agriculture are a major cause of losses in crops worldwide. Cotton is an important source of textile fiber, and insect pests and their management are the highest variable cost in cotton production. Identifying the most harmful cotton pests in field conditions helps acting for environmentally acceptable and yet economically effective strategies. This research proposes a classification system for major cotton pests (primary and secondary). A new ground-truth dataset containing RGB cotton field images is presented, along with a novel deep residual learning design to classify major pests automatically from given images. Performance is evaluated against other convolutional neural networks and the proposed ResNet34∗ model achieved the highest accuracy with f-score of 0.98.

[1]  Jinyan Li,et al.  Automatic classification for field crop insects via multiple-task sparse representation and multiple-kernel learning , 2015, Comput. Electron. Agric..

[2]  Andreas Kamilaris,et al.  Deep learning in agriculture: A survey , 2018, Comput. Electron. Agric..

[3]  Romain Raveaux,et al.  A survey on image-based insect classification , 2017, Pattern Recognit..

[4]  Jeremy Howard,et al.  fastai: A Layered API for Deep Learning , 2020, Inf..

[5]  Geoffrey E. Hinton,et al.  Deep Learning , 2015, Nature.

[6]  Chris Chatwin,et al.  AUTOMATIC PLANT PEST DETECTION AND RECOGNITION USING k-MEANS CLUSTERING ALGORITHM AND CORRESPONDENCE FILTERS , 2013 .

[7]  Qiang Yang,et al.  A Survey on Transfer Learning , 2010, IEEE Transactions on Knowledge and Data Engineering.

[8]  Li Fei-Fei,et al.  ImageNet: A large-scale hierarchical image database , 2009, CVPR.

[9]  Doyeon Kim,et al.  Performance effect analysis for insect classification using convolutional neural network , 2017, 2017 7th IEEE International Conference on Control System, Computing and Engineering (ICCSCE).

[10]  Saeid Minaei,et al.  Vision-based pest detection based on SVM classification method , 2017, Comput. Electron. Agric..

[11]  Nitish Srivastava,et al.  Dropout: a simple way to prevent neural networks from overfitting , 2014, J. Mach. Learn. Res..

[12]  Michael Mayo,et al.  Automatic species identification of live moths , 2006, Knowl. Based Syst..

[13]  Xi Cheng,et al.  Pest identification via deep residual learning in complex background , 2017, Comput. Electron. Agric..