A Hybrid Deep Learning Approach for Cotton Weed Classification

Main Article Content

Himabindu Cheruku
Vemuri Madhukar

Abstract

Cotton is crucial; effective weed management ensures its yield and quality. Conventional weed control methods, especially herbicides, have led to herbicide resistance and environmental concerns. This study introduces a novel hybrid deep learning network named Shepard Convolutional Fuzzy ShuffleNet (ShCFShuffleNet) for cotton weed classification. The methodology involves preprocessing input images using a Kalman filter, segmenting weeds with Psi-Net, augmenting data with shearing and random erasing, and extracting features for classification. The ShCFShuffleNet integrates Shepard Convolutional Neural Network (ShCNN) and ShuffleNet with fuzzy concepts to improve accuracy. The proposed approach achieves superior performance with precision, Sensitivity, and specificity of 91.3%, 90%, and 92.7%, respectively.

Article Details

Section
Articles