Research On Design And Detection Algorithm Of Screws Disassembly Equipment For Electronic Products

Screws are one of the most common fasteners udsed in electronic products.A scientific and efficient screw removal method is necessary for non-destructive disassembly of electronic products.At present,there are few researches on the design of disassembly equipment for screws of electronic waste,and the level of automatic disassembly is low.The combination between disassembly equipment and automatic methods is inadequate,advanced detection algorithm and special diamantling equipment in existing researches are also insufficient.In response to these key issues,this paper conducts the research on the equipment designed for screw removal of electronic products and screw detection algorithm.The main work is as follows:(1)According to the image characteristics--larg electronic products and small screws--the screw detection method of "rough detection+secondary detection" was proposed.The detection module,the movement mechanism and the actuator mechanism of screw removal equipment were designed.Camera calibration was performed,the relationship between 3D world coordinates and 2D pixel coordinates was derived.Finally,the complete disassembly process of screws was introduced.(2)The principle of image detection by YOLOv3 was introduced.A dataset for the rough detection of screws in electronic products was established.From the perspectives of feature extraction network,multi-scale feature fusion network and loss function,ablation experiments of various models based on YOLOv3 were conducted.Finally,a method for rough detection of electronic products screws based on improved YOLOv3 was proposed.(3)The principles of Hough Transform and the common image processing methods used in Hough Transform were introduced.A dataset for the secondary detection of screws in electronic products was made.Fitness functions and performance indicators that meet the multiple detection results were designed.With the help of genetic algorithm,the best image processing methods and parameters combination were found.A secondary detection model of screws based on Hough Transform and genetic algorithm was proposed.(4)The YOLOv3 model was applied to the secondary detection of screws,then the characteristics of Hough Transform and convolutional neural network were compared.Different combination methods of Hough Transform and YOLOv3 from the perspectives of data form,network structure and fusion method were explored,the best fusion model for screws secondary detection was built.