Research On Pedestrian Detection Based On Edge Symmetry And Uniform Local Binary Patterns Like

With the developing of computer technology and widely used of theory of computer vision, object recognition and tracking has become a hot issue in recent years. Pedestrian detection, a kind of object recognition, plays an important role in auxiliary drive system. Because of its great value in auxiliary drive system, pedestrian detection has become the hottest issue in computer vision and intelligent vehicle. At the same time, due to different shapes, gestures, appearances and dressing of pedestrians, pedestrian detection is easily affect by those factors. In addition, light, weather, and other complicated external environment will also have influence on pedestrian detection. Therefore, pedestrian detection is a challenging task for researchers.Existing pedestrian detection methods are slow and can not detect pedestrian in time. In order to solve those problems, we propose a method based on edge symmetry and uniform local binary patterns like, named ES-ULBPL. Our method scans the images with sliding windows at various scales to identify the locations and scales of the contained human figures. For the purpose of raising the detection speed, we get vertical edge and compute the symmetry of every window to pre-detect pedestrian and determine candidate region of pedestrian. Then we take uniform local binary patterns like into detection, named ULBPL, to get texture features like of pedestrian. And we magnify the values of ULBPL to get better ability to describe texture features of pedestrian. Finally, we combine it with SVM to verify whether there are pedestrians in image or not. Experiment results show that compared to method based on histogram of oriented gradients, our method is more quickly and has a higher accuracy.