Vehicle image mosaics technology based on improved SIFT algorithm

Aimed at the problem of large vehicle outline detect in the field of intelligent traffic, this paper proposes a vehicle stitching algorithm based on improved SIFT feature matching to reduce the requirements of the camera configuration and choose camera position easily. Firstly, the improved SIFT algorithm is used to extract and match feature point quickly. So the superior characteristics of SIFT algorithm is utilized to not only deal with the image's rotation, translation and illumination changes, the stability of visual change and affine transformation, but also ensure the processing speed of algorithm by reducing the complexity of the SIFT algorithm. In this paper, mismatched points are simply and quickly removed by limiting the matching line slope, which is proved to be a good result. After obtaining the stable matching points, we calculate the transformation matrix between two images. Finally, the image pyramid fusion is performed to perfect our stitching image and this pixel level processing has a good ability to eliminate fusion mosaic slot caused by camera position deviation, different light and other causes. Experiment results show that this algorithm reduces the rate of mismatching and improves the low efficiency in traditional image mosaic. The improved SIFT algorithm has distinct advantage in increasing speed, obtaining high resolution and completing vehicle images after mosaic.