Three-dimensional Object Detection System Based On OpenCV

Acquiring object's Three-dimensional information from multiple images is one important research topic in the domain of machine vision. Use of multiple cameras shooting the same object from different locations can be detected in the three-dimensional contours of the object. With the development of Science and technology, three-dimensional object detection based on machine vision in modern industrial production has become more and more important, in the industrial products line on the automatic on-line detection, quality control has a very wide range of application prospects.This paper analyzes and summarizes the current domestic and international the machine vision in industrial inspection application status and Difficulties, using visual C + + and OpenCV (open source computer vision library) for the development of tools, for which the camera calibration, image preprocessing, image feature points matching and three-dimensional reconstruction of some key technologies such as research, the main work is as follows:1. A method based on OpenCV and multi-view registration combining calibration algorithm is proposed. Using a Zhang's calibration plate and three industrial-grade CCD to be 9 group images shooting from different angles, using OpenCV to calibrate the parameters in the camera. Secondly, in this based on the corresponding relationship between each camera view, afterwards an iterative method is utilized to get the entire coordinate transformation of pair-wise views, thus the precise multi-view registration can be conveniently achieved and then can get the relative positions in them(the camera outside the parameters).Calibration process is simple, easy to operate. The experimental results show that the method is practical in multi-camera calibration.2. To improve the traditional filtering algorithms, an effective addition to image capture to the noise generated in the process, balance the brightness difference between images, enhance the image edges and details, to lay the foundation for follow-up work.3. Proposed based on the epipolar geometry constraint of the image matching algorithm, all the matching points on a highly constrained in the same line, effectively reducing the error matching. The first to use Harris corner detection operator to determine the approximate location of corner points, and then use window-based Sparse points matching to be a large number of rough matching points, and then use the principle of epipolar constraint to remove the wrong match point, to obtain more precise match point location,through OpenGL spatial reconstruction of discrete points.