An Effective Sequence Image Mosaicing Approach towards Auto-parking System

摘要

For its enormous market demand, Auto-parking system has come to be a hot research topic in Intelligent Vehicle research area recently. To develop the Auto-parking system, this paper presented a new method for parking spaces detection based on the image mosaicing. Considering the characteristics, two pre-processing were presented to improve the results of image mosaicing: firstly, the Adaptive Non-maximal Suppression method (ANMS) was introduced in choosing corners, which improving the distribute characteristic of the corners and the accuracy of the image matching; secondly, the Inverse Perspective Mapping(FMP) is introduced to transform the images, which partially eliminate the perspective effects of the original images and then successfully solve the difficulty of image mosaicing for two images with too different view angles; finally an improved RANSAC method were presented, which effectively improved the accuracy of the registration parameter's estimation from the corrupted matching points. The experiments proved that this proposed method was simple and effective.