Static 3D Detection Technology Based On Single Camera

Computer stereo vision technology imitates the imaging process of human eyes, getting object's surface depth information by calculates parallax under different perspectives. In recent years, thanks to computer hardware performance boost as well as the continuous improvement of image processing algorithm, stereo vision technology developing rapidly and widely used in human-computer interaction, virtual reality, military patrols, medical diagnosis, industrial detection, entertainment, and many other areas of life and production.The theoretical basis and key technologies of computer stereo vision system were studied in this work. For the goal of static object's surface 3D reconstruction, corresponding hardware and software environment were designed and built, which can measure the depth information of object, and display it in the form of a three-dimensional. The main work in this paper include:1. The three coordinates involved in stereo vision system and their relationships were analyzed, the camera imaging models and the method of correct distortion were given. For the more, I removed the lens distortion by Open CV and get the undistortion images, and then internal and external parameters of cameras were calibrated by Zhang's calibration algorithm.2. Some necessary enhancement processing on original images like gray, filtering, histogram equalization, etc were been done to improve the quality of the image. Positioning the checkerboard corners accurately by using Harris corner detection algorithm combine of subpixel corner point positioning method. Many kinds of edge detection methods were compared, and their corresponding test results were provided.3. To improved the matching efficiency and veracity, proposed a matching search method based on the geometric constraints. The calculation of the polar and the method of three-dimensional correction were given, at the same time cameras were adjusted to forward parallel alignment structure in math by Open CV. BM matching algorithms are implemented by programming, and then get the parallax image of figure object.4. Three-dimensional measurement system based on single camera was been set up, baseline distance of system can be adjusted according to actual requirements to adapt to different scenarios flexibility. The Open CV environment was been configured on VS system, calibration, image preprocessing, polar correction, stereo matching and other steps were completed combining with Matlab. A complete stereo vision system framework was achieved, get the expected effect of 3D reconstruction, and detailed error analysis was carried out.