Novel Approach for Crop Weed Detection in Wheat Fields Using Deep Learning Algorithms

Abstract

Agricultural productivity is greatly affected by weeds. To remove these weeds with chemical pesticides is harmful to the ecological environment. Also, with overall level of agricultural production rising, it is becoming more and more crucial to accurately distinguish between crops and weeds in order to perform accurate spraying just on the weeds. For generating precise spraying methods, it is required to identify the crop location and weed location more precisely. In recent years, many weed detection techniques are explored. This approach ranges from conventional to machine learning to deep learning. It is quite necessary to identify the color and texture features from image using image processing techniques for conventional approach. Then these conventional approaches are combined with some classical machine learning techniques. Any classical machine learning method necessitates a limited amount of training time, a low requirement for graphics processing units, and a limited sample size. There are two main approaches to weed detection from images: classification and segmentation.