Study On Corner Detection And Pose Detection Of Parallel Mechanism Moving Platform Based On Binocular Vision

Binocular vision is an important application of computer vision technology,it is widely used in general automation and intelligent system.Using binocular vision technology to detect the moving platform posture of parallel mechanism has the characteristics of convenience,flexibility and high precision.However,due to the complex environment with ambient light,shadows,unrelated entities and cluttered backgrounds,it is difficult to extract image features accurately and efficiently.Therefore,the detection results of moving platform pose by binocular vision are not ideal,and there is no general solution.Aimed at the problem:(1)this paper design a new self-coding markers,proposed a corner detection algorithm based on coordinate feedback and established a real-time's image corner matching strategy: a double-sorting stereo matching algorithm.In this way,the 3d coordinates of feature corners can be detected quickly,accurately and automatically in complex environments;(2)the solution method of target object's pose was simplified,through the arrangement of self-coding markers,a general detection model of moving platform pose is established.In this way,the pose of moving platform based on binocular vision in complex environment can be detected quickly,accurately and automatically.Firstly,the linear model and nonlinear model of camera are analyzed.Then,Matlab calibration toolbox was used to calibrate the binocular camera based on Zhang's calibration method,and the re-projection error of the calibration result is less than 0.0058%;On the basis of the calibration results,image correction was carried out based on OpenCV,and the corrected original image was highly consistent with the expectation.Then,the feature extraction of image is analyzed and a new algorithm of corner detection based on coordinate feedback is proposed.On this basis,a real-time matching strategy of corner features is established,called: A double-sorting stereo matching algorithm.And then,the 3d reconstruction of corner points is realized by triangulation.Experimental results show that this method can quickly detect the 3d coordinates of the corner points,and the distance error between the corner points is less than 0.24%,and it is robust to light and noise.Finally,the pose detection model of moving platform based on fixed platform is established.On this basis,the detection model based on the initial pose of the moving platform is derived.The two detection models can realize the requirement of comprehensiveness and universality of the moving platform pose detection.In the end,the experiment platform was set up for pose detection.Meanwhile,the laser tracker was also used for the comparative experiment of pose detection.Experimental results show that this method achieves the purpose of rapid,high precision,universal and automatic detection of moving platform pose in complex environment.