| With the improvement of modern industrial production level and product qualityrequirements in today’s society.Automated and intelligent production has become an inevitabletrend.The development of computer vision technology also has become a prerequisite andguarantee of many automated production systems. As one of the important means of fabricatingautomatically and intelligently,mechanical parts’automatic identification and sorting also hasbecome a pressing problem to be solved. However, most methods currently proposed are carriedaround mechanical parts’category detection, But in the actual production, and some also need theresearch on automatic sorting and outputting mechanical parts with different orientation.To solve this problem, this paper increases the application research on automatic sorting andoutputting mechanical parts with different orientation based on parts’ category recognition,anddesigns a set of system for automatic sorting mechanical parts based on computer visiontechnology. System is divided into three parts: video capture and image conversion module,image processing module and sorting actuator. Systems sets image acquisition,processing andoutputting functions in one to achieve to automatic identification and real-time outputting ofvarious mechanical parts. It shows that the proposed method of the article is feasible andeffective through analysis of experimental results.The main research work and innovation pointsof the system are summarized as follows:(1) The video capture and image conversion module is composed of DFK23G445camera andVT-LT2-HR1260W ring light. Its design implements image real-time inputting,provides thepossibility for the real-time online detection,reduces the image blurring caused byman-made jitter and the influence of the surrounding environment, and can greatly reducethe burden of part of image processing and provide effective help to improve the accuracyand real-time performance of parts detection;(2) Image processing module provides a rigorous algorithm for part detection.The algorithmmakes use of the classical corner detection to find the parts’position with adaptive and thetemplate matching to recognise parts’ orientation,at the same time,puts forward a newparameter analysis method, ROI extraction and normalization to establish template image with adaptive.And for the combination of visual detection and intelligent control closely,puts forward a new method of region labeling, as the basis of sorting operations;(3) Sorting actuator is composed of MC100USB data acquisition card, YL-62infrared sensorand cylinder.Its design provides the implementation method for execution of parts’ detectionand outputting, remains part orientation rather than falling collection based on the realizationof various types parts’automatic recognition and real time sorting in production line, greatlyimproving the collection efficiency and the subsequent assembly of parts;(4) At present, on the basis of the majority of shape recognition in parts recognition andmatching field, this paper increases the application of orientation detection, and adoptsVS2010developing system application software,to complete parts automatic imageacquisition, intelligent recognition and real-time outputting, the design of the softwareprovides a good platform for the system;(5) Adopting the system software, to7parts of samples, shooting the4group of1120picturesto be detected, respectively in the ring light, natural light condition, CCD camera lens8mmand16mm environment, and compared with the artificial detection, it is found that thesoftware can be used to complete the sorting of parts, and the accuracy rate achieves above90%, the speed also reaches the requirement of efficiency. |
