Distance Illumination And A Significant Goal Of Stereo Vision Measurement

Binocular stereo vision is an important research subject of computer vision, and has a wide range of application in the field of military and industrial. This paper studies camera calibration, stereo matching algorithm and binocular passive ranging technology. We lays emphasis on the region-based stereo matching algorithm and the special application of binocular stereo vision under low illumination. We set up the binocular stereo system under normal light and the significant target recognition and distance ranging binocular stereo system under low illumination.In order to solve the problem that the fixed size of support window and the same weight value will influence matching effects in binocular stereo matching under normal light condition, an algorithm which using the adaptively growth method and adaptively distribute weights of the support window based on the color-gray edges is adopted in this paper. Firstly, color edges based on the color difference are added to the gray edges of the image. Then grow the support window size dynamically by using image edge information and set a weight model that combined the geometrical distances with color similarity constrain. Finally, obtain dense disparity map by using the sum of weighted image difference between the two images. The experimental results combine with the Middlebury database and evaluation system show that this algorithm can improve the matching speed and accuracy in edge regions. We build a binocular stereo vision system using this algorithm for stereo matching and achieve the desired effect.In order to solve the problem that there are lots of error matching pixels and target recognition rate is low under low illumination in the binocular stereo vision, the threshold selection method which combines the maximum entropy and change of light intensity is adopted to extract significant object, the recognition rate is improved. A ranging method based on the principle of a non parallel ranging model is presented. The ranging error caused by the non parallel optical axis and the measurement error of optical center distance are decreased. Finally, we build the significant target recognition system. The results show that the binocular stereo vision device, utilizing the ranging method and threshold selection in this paper, has a higher target recognition rate and ranging precision under low illumination.