A Framework for Weed Detection in Agricultural Fields Using Image Processing and Machine Learning Algorithms

Weeds are a nuisance for farmers, and they're also bad for their crops. Crop growth could be harmed as a result of its presence. As a result, farmers place a high value on weed control. Weeds must be removed from agricultural fields at least once a week, whether they are sprayed with herbicides or removed manually with equipment. The goal of this study is to use the Lego Mindstorm EV3 to develop an automated weed control robot that can be linked to a computer. To distinguish between weeds and crops, an automatic picture classification system has been developed Weedicides will be applied directly to the weeds that have been discovered in or near the robot. The convolutional neural network algorithm is used to process the picture of the item in the image classification approach. Farmers may cut down on the amount of time it takes to monitor their crops by using technology, especially artificial intelligence. The crops can also benefit from this new technique. This is an exciting time to be in the field of artificial intelligence, particularly in the area of deep learning. In one of its many uses, computer vision is used to identify objects. This thesis is based on the integration of these two technologies. As an alternative to the FarmBot Company's system, a system for the identification of various crops and weeds was developed in this paper. FarmBot's API provides access to photos, which are processed using computer vision and transferred to an RCNN that can identify plants on its own via transfer learning.