Optimized Convolutional Neural Network for Robust Crop/Weed Classification

Precision farming makes extensive use of information technology, which also aids agronomists in their work. Weeds typically grow alongside crops, lowering the production of those crops. Weeds are eliminated with the aid of herbicides. Without knowing what kind of weed it is, the pesticide may also harm the crop. The weeds from the farms must be categorized and identified in order to be controlled. Automatic control of weeds is essential to enlarge crop production and also to avoid rigorous hand weeding as labor scarcity has led to a surge in food manufacturing costs, especially in the developed countries such as India. On the other hand, the advancement of an intelligent, reliable automatic system for weed control in real time is still challenging. This paper intends to introduce a new crop/weed classification model that includes three main phases like pre-processing, feature extraction and classification. In the first phase, the input image is subjected to pre-processing, which deploys a contrast enhancement process. Subsequent to this, feature extraction takes place, where "the features based on gray-level co-occurrence matrix (GLCM) as well as gray-level run-length matrix (GLRM) " are extracted. Then, these extracted features along with the RGB image (totally five channels) are subjected to classification, where "optimized convolutional neural network " (CNN) is employed. In order to make the classification more accurate, the weight and the activation function of CNN are optimally chosen by a new hybrid model termed as the hybridized whale and sea lion algorithm (HW-SLA) model. Finally, the superiority of the adopted scheme is validated over other conventional models in terms of various measures.