| With the rapid development of our country’s economy and urbanization,critical progress has been made in realizing the Chinese dream and fulfilling the task of great Chinese nation rejuvenation,but environmental problems have become increasingly serious.In 2005,in order to raise awareness of protecting the ecological environment,our country put forward the scientific thesis that "Golden mountains and silver mountains are not as good as green mountains and green mountains,and green mountains and green mountains are the golden mountains and silver mountains".Among them,garbage classification and disposal are the key.Our country still uses assembly line manual sorting method for garbage classification.This traditional method has problems such as low automation,high demand for manpower and labor,harsh environment and low sorting efficiency,which cannot meet the increasing demand for garbage disposal.Now,our country gradual completion of "curving overtaking" in the field of artificial intelligence,the continuous improvement of industrial automation level,the experience and technology of automated factories are constantly improving.It is feasible to propose and complete a system that uses machine vision for non-contact classification and detection of garbage and uses industrial robots to realize automatic and intelligent garbage sorting.At the same time,the research of this system has strong theoretical significance and important application value.First,this article completes the hardware construction of the intelligent waste sorting system,including industrial computer,industrial robot and industrial camera,and builds system control software on the industrial computer to complete the robot control and the scheduling between the various components of the system.Through experimental tests,the precise control of industrial robots is completed.Secondly,this paper completes the research of machine vision algorithms,including deep learning networks for detecting garbage category and location and image processing algorithms for detecting garbage posture.The deep learning in this article uses SSD as the basic network,modifies the skeleton structure and adds an attention mechanism to extract better image features,and uses feature pyramids to perform cross-layer feature fusion.The feature maps used for detection contain multiple information.Use better activation function and loss function to improve the accuracy of detection.Utilizing the advantages of simple processing of traditional digital images,fast calculation speed and low computing power requirements,the posture detection of the garbage that has been located is performed.The experimental results show that the machine vision algorithm proposed in this paper can meet the practical application requirements of garbage sorting,and the robot can greatly improve the accuracy and success rate of garbage sorting through known posture information.Finally,this paper proposes a set of system calibration methods and system cooperative control strategies.In order to unify the world coordinate system,this paper uses the conveyor belt as an intermediate medium to first complete the calibration of the conveyor belt and the industrial camera,and then complete the calibration of the conveyor belt and the industrial robot,thereby completing the calibration of the industrial camera and the industrial robot.Compared with the original algorithm,the improved deep learning algorithm in this paper has an accuracy improvement of 6.5%,and the average error of attitude detection is 1.5°,which improves the level of automated garbage sorting and is of great significance to garbage disposal and environmental protection.The sorting model has high portability and scalability in the industrial assembly line. |
