摘要：In order to realize fast and accurate identification of crops and weeds in complex field background,a method based on improved Faster R-CNN was proposed.Firstly,the collected images of crops and weeds are preprocessed to reduce the image size and expand the data set of crops and associated weeds to enhance the generalization ability of the proposed method.Then,Faster R-CNN deep network is improved,including ZF network with deeper level and two residual learning frameworks is designed to extract image features,RPN uses sliding window mechanism to generate target candidate frame,and ROI Align is used to replace ROI Pooling layer.Finally,the improved Faster R-CNN was used to identify weeds in complex environment.Based on the open source framework of Python and Mmdetec-tion,the experimental demonstration of the proposed method is carried out.The results show that the improved Faster R-CNN training effect is good,and the accuracy and recall rate of the proposed method are significantly better than other comparison methods,which can provide reference for precise weeding.
关键词：Weed identification in farmland;Faster R-CNN network;image preprocessing;RPN;ROI Align;ZF network
