Research Of Machine Vision Techniques On Sorting System

Vision system is a common component of sorting systems,which can satisfy the needs of the sorting system for identification and positioning.A vision system consists of hardware and software,in which the vision calibration algorithm and the various types of image processing functions constitute the software part of the visual system.As a typical scenario in industrial applications,sorting system has special requirements on the vision system.Based on the practical working environment of the sorting system,this thesis focuses on the calibration and image processing technologies of the sorting system.Aiming at reducing the difficulties faced during the development of image processing applications,the thesis proposes the development of the image processing function block with high auto-configurability regarding parameters,high environmental adaptability and high robustness.Specifically,this thesis mainly addresses the following two issues:1)The rapid calibration method for sorting systems.Establishment of the coordinate mapping between the camera system and the sorting system based on traditional calibration algorithms is too complicated for sorting systems.The achieved positioning accuracy is far more than that required in such scenario.Furthermore,the relationship between the camera pixel coordinate and the sorting work coordinate cannot be directly established.To solve these problems,a fast calibration method is proposed to improve the calibration efficiency by reasonably reducing the positioning accuracy of the sorting system.2)The recognition and location modules for sorting system.Identification and location are the most important functions of vision module in the sorting system.In order to realize the aforementioned objectives,the BLOB analysis method,edge detection and template matching function block are implemented in this thesis.The main contents of this paper are as follows:1)This thesis discusses the current mainstream calibration methods and the practical environment of the sorting system.A fast calibration algorithm adapted from the Zhang Zhengyou calibration method is proposed.2)This paper analyzes the interference factor of image quality in the sorting environment,then a highly automated image preprocessing algorithm is proposed to solve the problems causing image quality reduction in sorting system.This algorithm provides the function of image preprocessing without any manual parameter settings,and has encouraging adaptability and robustness in the experimented sorting application environment.3)According to the characteristics of the targets in the sorting system,three kinds of image processing function blocks were developed,which can recognize and identify positions of most of the targets in the system.The adjacent target segmentation method,the accelerated search algorithm and the sub-pixel processing method for improving positioning accuracy is discussed during the design and implementation of these function blocks.4)Through the experimental way,the actual performance of the functional blocks represented in this thesis are tested and analyzed,and the feasibility of the algorithm is verified.