AN APPROACH FOR FEATURES MATCHING BETWEEN BILATERAL IMAGES OF STEREO VISION SYSTEM APPLIED FOR AUTOMATED HETEROGENEOUS PLATOON

摘要

Recently, the stereo vision system (SVS) has been developed in measuring tasks. Using SVS in measuring tasks plays an important role in automated highway system (AHS) because the SVS can be used instead of high cost distance sensors. The AHS is being developed to be run in different environments (i.e. unstructured and dynamic environment) and to form different shapes of platoon (i.e. heterogeneous platoon). In this work, a proposed method has been developed to improve the performance of the SVS in terms of automated heterogeneous platoon. The first stage of improvement has been here introduced by proposing a method for matching the left and right image of SVS (i.e. bilateral image). The significant contribution of the proposed method is to localize the points of interest in matching considering the shape of context assigned to the localized object of interest. The idea behind this development is to localize alternative connected points whenever the back view of the preceding vehicle BVPV (i.e. reference object) is influenced by the environment, including but not limited to sunshine reflection. In order to eliminate the surrounding objects of the BVPV, the Histogram of Oriented Gradient (HoG) has been developed by a proposed enhanced procedure. The latter depends on passing a precise knowledge about the position of edges and enhancing the gradient of intensity values. As for feature extraction, the proposed method has been developed to use the smoothed image generated by DoG instead of using the original image. The similarity value between each enhanced blocks is calculated based on the Euclidean Distance. Similarity value of the successful matching is greater than 99% between each enhanced blocks. In comparison with other methods, including SIFT and HoG, the proposed method extracts many corresponding features at different distances (from 3 to 10 meters) for the whole BVPV.