Research On The Method Of 3D Reconstraction Based On Binocular Stereo Vision

Use visual access to information from the outside world is an important form of human observation and understanding of the world, accounting for most of the sum of access to information. This shows that the higher the utilization of human vision, but also a huge amount of information to obtain through visual, demonstrates the importance of the human visual function.The human visual system can be imaged of the outside world through the eyes, visual description of the system has the ability to acquire three-dimensional information from two-dimensional landscape, which for people's daily production and life is great significance.With the further development of information technology,computers and other advanced equipment, etc. have some intelligent robots, robot conferred human visual function has become one of the objectives pursued by humans may explore. Computer Vision is the use of computers to achieve the visual information processing, the objective three-dimensional scene recognized by the computer vision. Its purpose is to make computers with human visual perception skills, the use of two-dimensional image perceive three-dimensional of the world. This capability includes three-dimensional objects in the scene geometry, shape, position, posture, exercise and so on, and can describe and store this information. The localization of the object based on computer vision has become a hot research field of robotics, and binocular stereo vision is an important branch of computer vision research field, by simulating the human visual system to perceive the objective world, which is one of improtant source of robots get outside information.It is widely used in many fields of industry, agriculture, aerospace, medical and so on.In this paper, research on binocular stereo vision spatial positioning of several key technologies. Six main areas: image acquisition, camera calibration, image pre-processing, feature extraction and stereo matching and three-dimensional reconstruction.Using CCD camera and 1394 image acquisition card study was performed by the camera to capture images, and use VC ++ 6.0 to development and design image acquisition system, to achieve store the data. Calibration of the camera, using a calibration method Zhang Zhengyou board, while use of manual measurement and Harris corner detection method to obtain the image coordinates chessboard template corner by analysis and calculation to obtain internal and external camera parameters. In the image preprocessing, image smoothing, eliminating the noise introduced during imaging causing image distortion, we use median filtering method, and then processed binary image value processing, making it grayscale and extracted edge outline of objects with an enhanced Hough transform. In stereo matching portion for round objects, we focus on the SIFT feature point extraction and matching algorithms to extract center coordinates of the object,according to the final spatial orientation of binocular stereo vision principle, by the center coordinates instead of spatial coordinates of the object, and It causes errors to establish a mechanism for error analysis, and ultimately determine the three-dimensional coordinates of the object, establishment the depth map.Based on Open CV in the VC++6.0development platform, the standard image processing and conversion to match,the eventual establishment the depth map.