Target Location Technology Based On Binocular Panoramic Vision

With the development of computer vision technology, panoramic vision technology will be an important direction for future development, it has the advantage of a large field of view,you can get a wealth of environmental information, therefore, in areas such as intelligence and military applications play an important effect. In particular,a wide range of targets within the three-dimensional positioning to become a hot spot in recent years, in order to obtain three-dimensional information of the surrounding environment, there has binocular panoramic vision. This paper studies based on binocular panoramic vision targeting technology, including build binocular panoramic vision system, obtain a panoramic image of the correction, match the correction target image and localization algorithms.Constructing problem of binocular panoramic vision system is a prerequisite for targeting,a standard structure of the visual system tends to provide a great convenience for post-processing. This paper selects vertically aligned in the same direction binocular catadioptric panoramic vision system with hyperbolic mirror as a reflection mirror, were constructed up and down catadioptric panoramic vision system. Order to make locate computing more convenien,the most crucial point of building this system is the optical axis of the camera and the symmetry axis of hyperbolic mirror should coincide. In this paper,the method used to solve this problem is to draw a circle centered at the center of the camera,adjust the position of the camera, making panoramic images and circles overlap in the software interface, so basically solve this problem, based on the principle that is: the panoramic image is a standard circle.However,due to the existence of human error is inevitable in the process of constructing the system, making the system has some bias, which will affect the accuracy of the target location. In order to reduce the systematic bias, this paper for panoramic image correction.Correction including, obtaining the center of the panoramic image and calculate the deviation angle on the panoramic image pair. The center of the panoramic image obtained by the Hough circle detection, calculations of the deviation angle, In the calculation of the deviation angle of the panoramic image using a method, which without special markers, using only the vertical line perpendicular to the camera imaging plane in the experimental environment as a useful feature. And verify the feasibility of the correction algorithm in VC + + software platform programming.Stereo matching problem must be resolved before the target position calculation, it is the process of establishing relations parallax, due to the greater distortion in the panoramic image,so choose a high performance SIFT matching algorithm. This article studies the SIFT algorithm, including feature detection, feature point descriptor generation, matching algorithm theory, match results, in accordance with SIFT stereo matching principle, explains the scale invariance and rotation invariant feature points. The part is to make preparations for targeting.Finally, the derivation and implementation of targeting algorithms are under a binocular panoramic vision system. This paper proposes a localization algorithm based on vertical non-single viewpoint hyperbolic catadioptric panoramic stereo vision system, by analyzing the mathematical imaging model of the non-single viewpoint catadioptric panoramic stereo vision system to derive the formula for positioning and to analyze the accuracy of localization algorithm, identify the main factors for affecting the measurement accuracy of the system.Further,the system structure and parameters matching algorithm is the cornerstone of the positioning algorithm, which directly affects the accuracy of target location algorithm. In the laboratory environment, using structural parameters of the binocular panoramic system has been constructed, doing experiments to verify the feasibility of targeting algorithm, the results show that the algorithm is effective.