Research On Garbage Detection Algorithm For Green Environment

Municipal waste incineration power generation is one of the important methods for the reuse of biomass energy.However,the incineration of recyclable wastes will lead to waste of resources and environmental damage.At present,waste treatment plants mainly rely on manual sorting to screen out recyclable waste from municipal waste.This method has low work efficiency and high cost,which is not coducive to the physical and mental health of personnel.In this paper,the problem of recyclable waste sorting faced by biomass incineration power generation is studied.Object detection technology is used to enable waste sorting before incineration power generation,so as to improve work efficiency.The main work of this paper is summarized as follows:(1)A multi-dimensional recyclable garbage detection dataset(Mi Re G)is constructed.The dataset contains multi-dimensional information such as recyclable garbage images,garbage categories,coordianates and angles.In this paper,data collection is carried out in the actual environment,and a large amout of data preprocessing work is carried out.A total of 6694 images are constructed.The corresponding multi-dimensional information is obtained through the manual annotation method.(2)In order to improve the detection accuracy,this paper introduces the CA attention mechanism into the Yolov5 network to improve the ability of the network to search targets.On this basis,hyperparameters are adjusted according to the measured environment to improve the detection accuracy.The data verification shows that the overall detection accuracy of the improved network has increased from 90.5% to 91.6%,and the detection accuracy of carbon has increased from 85.9% to 90.2%.(3)Garbage detection algorithm based on Center Net and its improvement: In practical applications,for the crawling of recyclable garbage,it is necessary to improve the network in order to reduce the phenomenon of missed detection.And in order to improve the gripping efficiency of the gripper,it is necessary to detect the angle information of the target object in the background.On the basis of Center Net,this paper conducts two researches:(a)Garbage detection algorithm based on Center Net and feature fusion.In view of the phenomenon of missing detection,this paper takes Res Net-50 and Res Net-101 as the backbone network respectively to optimize the network structure of Center Net,and introduces a feature fusion module to improve the netwok to enhance the ability of feature extraction.The results show that the detection accuracy is improved by 1% respectively.(b)In order to achieve the angle detection,this paper improves the end of Center Net and adds an angle detection operator to complete the detection of the angle of the target object.The results show that the angle error is within 0.5 degrees.