High Speed Structure Light Scanning System Based On Stereo Vision

Traditional structured-light measurement is a measurement technique. It emitted linear laser light on the object. This method can detect the formation of the band of light on an object by using a high speed camera. By using an image processing technique for processing with the light, we can extract information about the two-dimensional surface of the object. When the laser and camera moves with respect to the object, you can get a three-dimensional contour size. Because of the limited output frequency of the camera and the processing speed, traditional methods cannot meet the demand for high speed scanning. Therefore, the project proposes a high speed structure light scanning system based on stereo vision to meet the high-speed rail sections board.This paper introduces the laser triangulation method, the concept of binocular stereo vision measurement method. It introduces the overall program structure of the optical scanning system based on three-dimensional machine vision. Then paper introduces the main content, including image processing algorithms and hardware and software design, system calibration theory and experimental verification.Hardware design and image processing algorithms chapter describes the architecture and hardware design and FPGA for five bands of light extraction algorithm. After the code is portable across FPGA, the acquisition hardware resolution is 1280× 1024, frame rate is 300fps and the acquisition section rate is 1500 per section. Software system design relies on DALSA Sapera LT library and MFC development environment, to achieve a set of images, caching, to achieve the function of setting, caching, drawing, and so saving.System calibration includes the calibration of the calibration parameters and structure of light triangulation parameters of binocular stereo vision. The former uses Zhang Zhengyou calibration method. The latter uses least squares to fit parameters. System acquisition rate is 1500 per section. The correct rate of the system can reach 98.26%. Within the calibrated measuring range, the average error is about 3mm. Through experiments, we can see that the system is stable and can achieve the desired requirements in the calibration of the measuring range.