Feature Extraction And Location Method Research Based On Binocular Stereo Vision

With the great development of artificial intelligence and machine vision as well as themarvelous progress of human technology, people are requiring for higher standard of living.The quality and comfort of life has become the new criterion to evaluate the living.Consequently, intelligentialize and automation of household life has also become humandesire. As a result, the research on feature extraction and location method based onBinocular stereo vision is becoming a hot topic.This dissertation makes some researches and does lots of experiments which focus onthe household intelligent home service robot that is base on the extraction binocular stereovision feature and location method. Being different from the indoor positioning featurewhich is especially vulnerable to tables and chairs change, the locating information usingthe indoor corner is more stable, which is regarded as robot locating characteristic. What’smore, this method does lots of experiments and comparative analysis about the cameracalibration combining with three common kinds of light and comes up with a stereomatching method on the conditions of the improvement of SIFT algorithm. It can alsoachieve independent development and operation of the software program as well as the finalfeature extraction and location according to Traveler II Robot platform.1) This thesis introduces the research significance, background and the status at homeand abroad of this method. It discusses the measuring principle of binocular stereo vision,coordinate systems and the conversion relations between them. It grays and smoothing theimage acquired by binocular camera, which paves the way of the follow-up imageprocessing.2) This dissertation discusses linear and nonlinear model of video camera and utilizesZhang Zhengyou’s camera calibration method to conduct plenty of experiments analyzingthe influences of three common light conditions (hard light, dark light, light conditions) onthe camera calibration parameter to conclude that according to application prospect of thisarticle, camera calibration does not require to consider light conditions. Feature extractionand location error has mainly to do with feature point extraction accuracy of test board.3) This thesis studies feature extraction and stereo matching based on the Harrisalgorithm and SIFT algorithm as well as raise the improvement of SIFT algorithm. As aresult, this method increases precision, accuracy and matching efficiency.Finally, this thesis, depending on OpenCV and Microsoft Visual C++6.0software,applies MFC modules to develop an application independently and realize image preprocessing, camera calibration algorithm, stereo matching algorithm and locating featurepoints and three-dimensional coordinate as well as distance algorithm program. It functionson the Traveler II Robot platform successfully.