INTELLIGENT JUDGMENT SYSTEM FOR VEHICLE-OVERTAKING BY MOTION DETECTION IN SUBSEQUENT IMAGES

We propose an intelligent judgment system to determine the exact timing for overtaking another car in simulated dynamic circumstances by motion detection and feature analysis in subsequent images based on digital image processing in this paper. The strategic methodology in detection of motion vectors extracted from the source video file can effectively evaluate and predict the behavior of surrounding vehicles and at the same time give an appropriate suggestion whether the driver could overtake another car, which is not only different from the traditional methods in computational feature-only analysis for some specific static image but also provides an innovative idea among the related applications of intelligent transportation system (ITS). Our system makes use of the video-typed files recorded from the rear-view mirror in the vehicle that should be obtained from the digital image capturing devices. It also constantly reevaluates the information of motion vectors in surrounding environments to update the useful information between the driver's vehicle and background. In order to tackle the problem of real-time processing, this paper simplifies the processes of feature selection and analysis for video processing in particular. The crucial features used to give dynamic information, motion vectors, can be obtained from defined consecutive images. We define a variable number of images according to the extent of motion variation in different real-time situations. Our dynamic features are composed of geometrical and statistical characteristics from each processed image in the defined duration. Our scheme can identify the difference between the background and object of interest, which also reveals the dynamic information of the determined number of images extracted from the whole video file. Our experimental results show that the proposed features can give the useful information in a given traffic condition, such as the locations of surrounding vehicles, and the way of vehicles' moving. The real-time problems of ITS are taken into consideration in this paper and the developed feature series are flexible to the changes of occasions. More useful features in dynamic environments as well as our feature series will be applied in our systematical mechanisms, and the improvement on real-time problems by motion vectors should be progressively made in the near future.