Fast part identification method based on vision

Abstract

The invention discloses a fast part identification method based on vision. The method comprises steps that S1, template images of multiple parts are acquired, and a part database is established; S2, to-be-detected part images are acquired, and part characteristics are extracted; S3, the to-be-detected images are identified; and S4, the parts of the to-be-detected images are positioned. The methodis advantaged in that through three aspects of the contour number of the parts, geometric characteristics of each contour and shape characteristics of each contour, the to-be-detected part images andtemplate part images of the part database are compared and identified, and the planar parts can be accurately identified, classified and positioned in real time.

Description

technical field [0001] The invention belongs to the field of robot vision positioning, and relates to a vision-based fast identification method for parts. Background technique [0002] Since the 21st century, with the continuous maturity of robot technology, robot technology has been more widely used. From industrial robots to service robots, robots have gradually entered people's daily lives, bringing us a lot of convenience. With the improvement of human material living standards and the increasing enrichment of spiritual life, robots will be more closely linked with human beings in the future. [0003] At present, the industrial robots used in production generally adopt the method of teaching or offline programming, and simply repeatedly execute the pre-programmed programs to complete the actions set by the program, which is subject to many restrictions in industrial production. [0004] In an industrial production line, there will be a large number of parts recognition...

AI Technical Summary

Problems solved by technology

Traditional manual classification cannot be accurately and accurately identified, and there are the following problems: ① slow recognition speed; ② missed detection

[0006] The paper "Target recognition and positioning of binocular stereo vision, Journal of Intelligent Systems, 2011, 6(4): 303-311, Shang Qian, Ruan Qiuqi, Li Xiaoli", uses binocular stereo vision to realize target recognition and positioning; the binocular stereo The vision system mainly includes four modules: camera calibration, image segmentation, stereo matching, and 3D ranging. Stereo matching is the most critical step in binocular vision positioning, but it is difficult to achieve accurate stereo matching in the target area, and stereo matching Inaccuracy will directly lead to deviations in the acquired depth information, and its real-time performance is the biggest challenge faced by binocular and multi-eye positioning vision systems

Images