Study On Garbage Detection And Cleanliness Evaluation Method Based On Computer Vision

At present,the automation degree of floor cleaning is low,and most public places still use artificially guided cleaning equipment or full-covered cleaning methods,and the work efficiency is low.In addition,the floor cleanliness still needs to be judged by human eyes,so there is a lack of automatic feedback process,resulting in unnecessary waste of resources.With the continuous development of artificial intelligence,it is possible to use vision technology to realize garbage detection and classification.The image is acquired through the camera,digital image processing or deep learning technology is used for detection,and then the cleanliness level is further assessed,and the cleaning equipment can be guided to complete the cleaning task efficiently in the subsequent.Therefore,the research on garbage detection and cleanliness assessment has certain research significance and application value.In order to improve the autonomy and intelligence of cleaning equipment,this paper studies effective methods of garbage detection and cleanliness assessment,so that the equipment can complete the autonomous detection of garbage by obtaining visual information,which can improve the efficiency of cleaning operations and save resources.The specific research contents are as follows:(1)For the current situation where there is no applicable public data set for ground garbage detection,the two data sets in this article have been established by manual photography.They are the test paper data set for detecting irregular dust garbage and other common ground garbage data set.(2)For the irregularly distributed dust garbage,an indirect detection method based on image processing is proposed.The Grab Cut algorithm combined with the saliency map is used to segment the test paper image,and the segmentation effect is compared with the classic watershed algorithm and the original Grab Cut algorithm.Then,through image graying,binarization and contour detection,the dust stains on the test paper image are quantified,and the cleanliness level can be assessed.Finally,the effectiveness of the method is verified by testing experiments.(3)Using the classic deep learning target detection model for garbage detection and comparison,selecting the YOLO v3 network with both accuracy and speed,and optimizing the self-built garbage data set and data enhancement operations to enrich the data under different conditions.Then during network training,optimization is performed by adding a BN(Batch Normalization)layer to the input of each layer,which improves the robustness of the model.(4)The cleanliness assessment method based on weight distribution is studied.Different weight coefficients are assigned to each type of garbage,and the cleanliness index is weighted with the corresponding amount of garbage to achieve the cleanliness level assessment.Finally,through real-time video detection experiments,the effectiveness of the detection method in this paper is verified.