The Research And Application Of Binocular Vision Based On The Feature Point For River Model

As the development of computer and image technology, computer vision measurement methodbased on image analysis and relevant technology has been widely used in various fields, all kinds of usefulinformation was successfully extracted and used. Although there are several methods of depth measurementfor river model, but most of them need to contact. For this, in order to achieve measurement of the depth ofthe river model, we attempt to establish binocular vision measurement model by using of computer visiontheory on the basis of the experimental study of the river model.Research of Binocular vision technology is one of the most important branches of computer visiontechnology. Its principle is that by simulating the vision function of the human eye with computer, not onlyinformation can be extracted from image or image sequences, but also formation and motion recognitioncan be applied to3-D scenery and object in the objective world. Its main components are video cameracalibration, image collection, feature extraction, stereo matching, depth measurement and depthinformation interpolation. In the light of the specific purpose of measuring is the depth of informationmeasuring, this paper takes feature extraction and matching as research focus. Rough content of this paperis as below:1. This paper systematically elaborated theory basis and system component of binocular visionmeasurement based on double CCD(charge-coupled device）video camera, analysis of the mostsimple model of binocular vision-the parallel binocular vision model.2. As for image feature-point extraction, this paper put forward applying corner-point testing toriver-model with SUSAN operator, adaptively selecting gray level difference t and by adding lowerlimit geometric threshold of g，which not only effectively eliminates noise and acquires automaticextraction effect of the target feature-point in river-model image, but also has better self-adaption.3. In this paper, the corner points of the target can be as the corresponding feature points. Apoint-to-point matching relationship is build between the two images. this paper used windowbased sparse points Gray-scale matching method to realize the stereo matching of the feature pointin the picture, combined with polar line constraint condition of the corresponding point and disparity constraint condition, It analyzed mainly the impact of the size of window on the testresults.4. After feature extraction and stereo matching, this paper lastly not only built experiment model andspecific method of this measurement, that was parallel binocular vision model and trianglemeasurement method, but also realized the measurement of the depth information and analyzed themeasurement result.