Design Of Waste Plastic Bottle Sorting System Based On Vision And Industrial Robot

With the growing national economy,the number of plastic bottles consumed in China every year has gradually increased,and most of the plastic bottles are naturally degraded by burying,causing greater pollution to the environment.Different colors of plastic bottles can replace oil to produce different colors of polyester chemical fiber products after recycling treatment,and the current factory recycling of plastic bottles is still a manual way,the method of sorting is inefficient,in the harsh environment will also cause great harm to the human body.In order to improve the sorting efficiency of plastic bottle recycling,this paper designs a waste plastic bottle sorting system based on vision and industrial robots.This paper focuses on image preprocessing,plastic bottle shape characteristics,color recognition and target positioning,and designs a new end effector,using PLC as the system control core and industrial robot as the sorting mechanism to complete the system design.The research work in this paper is as follows:(1)In the actual situation,plastic bottles may overlap when they enter the conveyor belt through the hopper,and the overlapping of plastic bottles will interfere with image recognition.This paper proposes a shape feature descriptor based on the pixel point set of the contour boundary,using the Sobel edge detection operator and the Moore tracking algorithm to extract the pixel points of the contour boundary of the plastic bottle,and the coordinates of the pixel point set are converted to In the polar coordinate system,it is normalized so that the extracted shape features remain unchanged under rotation,scaling and translation.The shape feature descriptors of individual plastic bottles and overlapping plastic bottles are used to establish a sample library to perform shape matching on the plastic bottles on the conveyor belt,complete the screening of overlapping plastic bottles,and avoid the interference of overlapping plastic bottles on color recognition.(2)In order to reduce the impact of reflection caused by the liquid residue on the bottle body and the uneven surface of the bottle body,this paper uses the K-means clustering algorithm to perform color clustering on the image that has been filtered and denoised,and the colors are quantized into green,blue and transparent 3 colors,the standard value of each color in the palette is obtained,and the color of the plastic bottle is described by the proportion of the number of pixels of each color on the plastic bottle body.(3)Due to the different shapes and sizes of plastic bottles,the characteristics of plastic bottles of different scales are fundamentally different,resulting in low positioning accuracy for small plastic bottles.In this paper,the YOLOv3 target positioning algorithm is used to locate the plastic bottle in the image,and two scales are used to extract the features of the plastic bottle from shallow to deep.The shallow features are used to locate small plastic bottles,and the deep features are used to locate large plastic bottles.After predicting the frame,use the non-maximum value suppression algorithm to remove redundant prediction frames in the figure to obtain the position of the plastic bottle.(4)In order to make the sorting device successfully grab the plastic bottle,an end effector combining "L" finger and suction cup is designed,first using the suction cup to contact the surface of the plastic bottle,and after the plastic bottle is adsorbed,the "L" finger closes to clamp the plastic bottle.In this paper,Solid Works is used to draw the model of the end effector and import it into ADMAS to establish a virtual prototype to verify the feasibility of the end effector.(5)Build a test platform to experiment with shape matching and color recognition.On the test surface,the accuracy rate of plastic bottle shape matching reached 98.3%,and the accuracy of overall color recognition reached 97.6%,which met the actual needs of the factory.