Algorithm And Technology Of Image Area Measurement Based On Binocular Stereo Vision

Computer vision is a new and developing technology; this technology used in many fields,such as binocular vision measurement and robot navigation. General measurement method of thearea for the large area such as buildings has the disadvantage of artificial operation. However thetechnology of measurement based on the computer vision has the following advantages:non-contact technique with the advantages of high speed, high accuracy, low cost and convenientfor artificial operation. So it has a very broad application prospects and deep research value. Inthis paper, the parallel binocular stereo vision system is adopted, based on the parallel binocularstereo vision, the three key technology of camera’s calibration, stereo matching, and areaoperation in the process of image area measurement are studied. The main research contents andresults of this thesis are as follows:In camera calibration module, first, the images collected from the camera are pixel cornerdetected using traditional Harris corner detection algorithm; however traditional Harris cornerdetection algorithm can only detect the pixel corner. The pixel value of corners that obtained bydetection has a great influence on the accuracy of the calibration results, so at the beginning, usingtraditional Harris to detect the pixel value of corners, then to calculate the sub-pixel value ofcorners from the observation that any vector from true corner location to its neighborhood isorthogonal to the image gradient. And then classic Zhang Zhengyou camera calibration method isadopted to calibrate the camera and acquire the parameters of the camera. In order to decreasingthe error of calibrating in the experiment, we take ten pairs of images for the chessboard fromdifferent directions and extract corner points of the chessboard in every image. The sub-pixelcorner coordinates as input of camera calibration, then calibrating the camera’s internalparameters, sub-pixel corner detection and camera calibration program are realized in Matlab.In the module of the stereo matching, the arithmetic of feature point detection and featurepoint matching is mainly studied. This paper studies the classical SIFT feature point matchingalgorithm and feature points matching experiments for the different cases of image was carriedout, but given in this paper, we study the image is relatively simple graphics, using the SIFTfeature point matching algorithm seem complicated and match for a long time, so basing on theactual situation, we find a suitable image feature point matching method in this paper. First cornerdetection was carried out on the left and right images respectively, the feature points in the left as a template image, with the feature points of right for template matching, similarity of the biggest,is considered to be the best matching point. This algorithm is not only simple but also match theshort time.Last, in the module of the area calculation, this paper studies the three methods of computingarea and for each of three methods are analyzed and compared. Using the three methods are testedon the same image respectively and compare the experimental results. The spatial geometricalrelation between actual image and image in imaging plane is deduced in this paper. Thencomplete the measurement of the image area. This method is simpler than pixel area method andthe error is small, the condition to realize is easier than the spatial coordinate method.Using two common USB cameras, we compile each of this experiments’code with C++baseon OpenCV and Matlab in VC2006and Matlab.