Research On Techniques For Binocular Stereo Vision Based On Structured Light

With the development of measurement techniques, three-dimensional visualmeasurement has been widely used in the field of industrial measurement because ofthe characteristics of non-contact measurement、fast speed、good system flexibility, Inrecent years, active vision based on structured light measurement technology hasattracted more and more attention because of its strong anti-interference ability andhigher accuracy advantages. The difficulty is that acquisition of high-precisionthree-dimensional of the surface data is not easy, large amount of measurement dataneeds to be computed and possible error source exist in the structured lightmeasurement technology, therefore, to seek effective solutions for these problems is themain issue of3D measurement field.In this paper, related with the actual projects, we do research on several key issuesof measurement system,technology of line structured light based on binocular vision, onthis basis,we do detection to thread defects, and thus to improve the system’smeasurement performance.The measurement accuracy of structured light measurement system is mainlyaffected by the line of the laser stripe center positioning accuracy、system calibrationaccuracy and structured light stripe matching accuracy, we do depth study of theseaspects. First, given the contradiction between the high positioning accuracy ofstructured light strip center、 high robustness of algorithm between intensivecomputation、bad features of real time, this paper proposes an algorithm of strip centerextraction based on the cross-correlation algorithm, it can automatically split theeffective laser stripe region from the image, and locate the stripes center accurately;Secondly, we do precise positioning of the checkerboard corners based on the planecalibration method through local constraint relations, and use line laser-assistedcalibration method to change the original algorithm to get outside and inner calibrationparameters rely on coordinates of the corner; finally, we summarize the current laserstripe matching method, and put forward a matching method based on Sift-algorithm.After completing structured light strip center extraction、system calibration andstripes matching，we use this system to obtain the3D clouds of target surface, thenrecover the surface morphology to test the validity of this structure light measurementsystem, and analyze the problems of actual measurement process as well as improvingthe measurement system’s inadequacies, finally, do positioning to the system’sresearch direction of future.