The Research Of Obstacles Detection Technology Based On Binocular Vison And Grating Projection

The acquirement of obstacle information is the chief problem to solve fortechnology of keeping car away from colliding. With respect to detection methods oflaser, radar and ultrasonic wave, the method based on vision will be more advantageous,which is featured by the broad, complete description of the target information, inaccordance with the human cognitive habit. Obstacle detection based on stereoscopicvision mainly need to solve two problems: how to extract obstacle target and how toobtain the deeper information of obstacle.On the basis of deeper analysis of the obstacle detection methods, this thesis focouson research of detecting obstacle by vision measuring system built with compromise ofgrating projection and stereoscopic vision, and develops the relative arithmetic as well asmakes experiments accordingly.The thesis introduces the theory of using grating projection to make thethree-dimentional measurement in detail, analyses its systematic optics structure; it alsointroduces the dealing with grating strip by four-step phase shift method and phaseresolving technology.On the basis of measuring model of classic grating projection method, it improvesthe measuring system by considering the correspondence of the phase-height inthree-dimentional space, to get the mapping relation among the three dimentionalcoordinate of phases, image points and object points, in the camera coordinate system.On the basis of the construction of vision measuring system, the thesis also detectsand identifies the obstacles, to develop the relative arithmetic. We use the median filter tomake the preprocessing to the images, then use Sobel operator to make the edgesegmentation to the images and use Hough conversion and road constrained condition todraw the road borderline to establish the Area of Interested(AOI) and reduce the obstacledetection range. Tracking the obstacles with a recursive template matching methodincreases tremendously the instantaneity and reliability of obstacle detection.