Design Of Vehicle Panoramic Viewer Based On Image Mosaics

With the rapid development of the automobile industry, the problem of the safe driving becomes the hot topic in the world, and many kinds of the vehicle monitoring system emerge at the right moment. In the vehicle monitoring system, the supervisor has to see the wider scene, even the 360 degree panorama scene. Therefore, a vehicle panoramic viewer based on image mosaics is designed, and its experimental facility is displayed in this paper.This paper proposes a vehicle monitoring system based on the feature points image mosaics, in order to get a wider field-of-view scene. Harris corner and SIFT are the two stable and effective feature points algorithm. Harris corner is easy to achieve and invariant to rotation, but the lack of the property of scale invariant is an obvious deficiency, and it is easliy influenced by noise. SIFT algorithm is invariant to rotation, scale variation, partially robust to affine distortion, illumination variance and noise. But SIFT algorithm is complex, and has large calculations, so it is not apt to realize the real-time system.In this paper, the extraction of SIFT feature points is improved by taking advantage of the SIFT extensible property, which resolves the problem of Harris scale variation sensitivity. After the scale-space extreme detection, the Harris corner is detected as the feature point in SIFT scale space, which not only has the advantages of them, but also improves the speed of the points matching. Meanwhile, SIFT descriptor is improved to reduce the length of descriptor on the premise of the feature stability.The main contributions in this paper are that (1) Research the extraction feature points of Harris and SIFT deeply, improve the extraction method of SIFT feature points, and extract the more stable image feature points; (2) Propose a SIFT descriptor calculation method which can reduce the length of descriptors; (3) Blend the moving objects in the overlaping area with the method of moving objects detection and remove the image fusion ghosting effectively; (4) Design a vehicle monitoring system based on image mosaics, and build up its experimental equipment; (5) Optimize the software of system in order to meet the demands of system real-time.