Location of Vehicles Using Edge and Surface Image Block Features

The paper presents a novel method for vehicle location in a sequence of images. In the method, a processing of input images taken from the video camera placed above a road is applied. Input images in bitmap format are converted to binary images using analysis of small image gradients. Binary images consist of edge and surface elements. The input images are divided into square blocks of the same size and, for each block, a sum of edge elements and an average of values of the pixels that correspond to surface elements are calculated. The image blocks are classified as background and foreground on the basis of the sums of edge elements and the averages of pixel values. Located vehicles are represented by regions that include blocks classified to the foreground. The region coordinates describe the location of vehicles in individual images. Results of experiments using images obtained in good and poor illumination conditions are provided.