Research And Development Of Robotic Waste Sorting System Based On Image Processing

With the gradual development of the economy and the continuous improvement of living standards,the problem of garbage disposal is plaguing the development of many cities.Garbage sorting and recycling is an effective way to solve the problem of garbage disposal,and the first task is garbage sorting.The use of robots for garbage sorting can improve efficiency,save manpower,and achieve the purpose of garbage classification and treatment.Aiming at the characteristics of robot garbage sorting task,this dissertation proposes and demonstrates the design scheme of garbage automatic sorting system based on deep learning image recognition,studies and determines the hardware design scheme,especially the selection of key components,discusses the software design scheme,studies and improves the key algorithm(with the modification of the loss function as the breakthrough),builds an experimental system and conducts full experimental verification,which proves the feasibility and advanced nature of the proposed improved model in the experimental system.The specific research content of this dissertation includes the following parts:(1)Based on the characteristics and requirements of the robot garbage sorting system,the scope of application,advantages and disadvantages of the target detection model based on deep learning are studied and analyzed,and the YOLOv8 model with the latest time and the best effect is selected as the basic model of the robot garbage sorting task.(2)Design the overall scheme of the robot garbage sorting system.According to the requirements of sorting tasks and the relationship between equipment,the hardware design and selection were carried out,the software design scheme and improvement points of the garbage sorting system were analyzed,and the target detection algorithm in the garbage sorting system was studied and improved,and the effectiveness of the improvement was verified by comparative experiments.(3)Build a garbage target detection experimental system,shoot garbage images and videos that simulate the production environment,and construct a garbage detection dataset together with the collected garbage images.Experiments comparing the improved model with the previous model and the original model verify that the improved garbage target detection system has better detection performance and improves m AP by about 3%.At the same time,the speed of detecting pictures reaches 50-70 FPS,which meets the real-time requirements of the robot garbage sorting system and can be effectively applied in the robot garbage sorting system.The garbage target detection system constructed in this dissertation has good performance in garbage identification and classification tasks,and has high portability and scalability in actual industrial scenarios,which is of great significance for garbage classification and environmental protection...