The Research On Binocular Stereo Vision Image Matchingand Target Positioning Method

Stereovision measurement perceives the three-dimentional information of spacescene by the two-dimensioned image information, which is similar to human vision.Stereovision system is increasingly widely used in real life, and it’s accuracy andspeed have already achieved a high requirement. This article makes an intensivestudy to the image matching of binocular stereovision and the target positioningproblem of space scene.Firstly, this article introduces the application status of stereovision, imagingmodel of binocular stereovision and the experiment of binocular camera calibration.Then, this paper focuses on the key point extraction and feature matching process ofSIFT algorithm and SUDF algorithm.In the stage of feature extraction for SIFT algorithm, this article presents abetter multiple scales. The combination of SIFT algorithm and Harris algorithmmakes the detected key point has both corner information and invariance forchanging in scale.In the stage of feature extraction for SURT algorithm, edge detection andmorphological algorithm were added. The detecting of feature points near the edgeof the image decreases the number of feature points, witch no only overcome theinfluence of unstable feature points but also improve the accuracy of subsequentmatches.In the stage of feature matching for SURT algorithm, the dimensions of vectorwas reduced to40dimensions.40dimensions vector whose standard deviation islarger are elected as the new description vector according to the statistics of describephasor of the key points, due to the64-dimensioned describe phasor which derivesfrom image neighborhood has some specific and repeated attributes. This methodguarantees the matching precision and at the same time enhances the matching speedeffectively.As for the target positioning problem of space scene, we have designed thebinocular vision decoding algorithm. Combining the internal and externalparameters of two cameras with image pixel coordinates which bases on theimproved matching algorithm, we solve the three-dimensional information of thetarget in space scene，thereby proving the improved algorithm in line with therequirements of stereovision measurement.