Weed Species Identification and Detection in Crops Using Deep Learning Technique

Abstract

Agriculture is the backbone of the nation s economy but weed intervention is something that has been impacting the crop yield and quality lately. Weeds are those unwanted plants that grow between cultivated crops, which reduce the purity of the crops. Crops are severely affected by weeds for their quality and yields. Farmers use the traditional methods for weed removal which is time-consuming and also makes it difficult to identify the difference between weed and crop. Therefore, to improve the quality of crop production through weed identification and classification, it is extremely important to integrate the modern technologies prevalent in today s world such as Artificial Intelligence (AI) and Deep Learning (DL). Furthermore, along with the identification and classification of weeds, this work also analyzes weed growth and density based on the different segmentation techniques. This work uses a total of 8400 RGB weed and soya bean crop field images using the Crop Weed Field Image Dataset (CWFID) dataset for the identification and detection of weeds. The weed growth has been identified by the Biologische Bundesanstalt Bundessortenamt and Chemische Industrie (BBCH) coding system. Thereafter, the weed growth, coverage area, and density are identified using vegetation segmentation. This research proposes Hybrid Deep Segmentation-Convolutional Neural Network (HDS-CNN) based architecture approach for identifying weed and crop growth estimation. This work implements a modified U-SegNet based segmentation method. In this work, we propose four different modified pooling layers and reduce the pooling layer of the traditional segmentation model which has recognized the weight of the weed leaf. The proposed algorithms achieved the best accuracy of 98.95%. The evaluation of financial misfortunes and impact due to weeds in farming is a critical perspective of considering which makes a difference in formulating suitable management methodologies against weeds. The proposed approach has been generalized to different weed speci