The algorithm of vehicles trajectory extraction based on video image sequences

In allusion to the current traffic monitoring, driver skill examination and other requirements, the algorithm of vehicles trajectory extraction based on video image sequences is proposed. First, based on researching the characteristic of road images, this algorithm has used multi-objective optimization technique to improve feature point matching mechanism based on Euclidean distance, which used after the feature of SIFT(Scale Invariant Feature Transform) was extracted, in order to realize the exact match about SIFT feature points. Secondly, the relationship of slope between the feature points is used to gain the vehicle´s shift direction. At last, according to the connection of coordinate transformation, principle of relative distance was used to extract the actual track of vehicle. Result: the precision of the vehicles trajectory about the algorithm is higher. The experiment shows this algorithm has certain advantages in the applicability of trajectory extraction method and the extraction precision, compared to the common the technology of vehicles trajectory extraction.