Research And Application Of Pedestrian Detection In Campus

Pedestrian detection is a hot spot in computer vision, and it has great application requirements, such as intelligent video surveillance, vehicle auxiliary driving, human-computer interaction etc. Pedestrian Detection Based on Hough Transform under fixed camera has two important parts.The first part is the detection of foreground. For the background clutter, we utilize the foreground detection technology to remove the background without useless information. We compare and analyze several background detection algorithms by experiment, and finally choose ViBe(Visual Background extractor) algorithm to extract the foreground. After that, we filter some areas that do not satisfy the conditions. We will describe the detail of method as follow. First, using the ViBe(Visual Background extractor) algorithm to extract the moving foreground in videos which were taken by the fixed camera, so that we can get rid of the impact of the background in pedestrian detection. The next step is to find the area of the detected motion and calculate the area of the region, and then get rid of some of the area which is relatively large or relatively small that cannot be the pedestrian area. This can reduce the scope for detecting pedestrians and can improve the operational efficiency also it can improve the accuracy of detect the pedestrians.The second part is pedestrian detection. The head skeleton outline shape in the movement area is similar to a circle(or including the arc) and other vehicles(such as: automotive) that does not contain the key information circular. We get that pedestrian detection can be completed by circle or arc detection and Hough transform has the powerful characteristic to detect the circle. According to this feature we can use the Hough transform to detect pedestrians of the head. Before the detection of the circle in pedestrian’s head we need to do some pretreatment to the binary image which is obtained by the algorithm runs after ViBe like morphological filtering, noise processing and refine the image. It can make a person’s head like a round which is good for us to detect the head of the pedestrians.This method can be used for complex traffic to detect a pedestrian. In addition, this method allows for a better pedestrian detection when pedestrian is blocked and overlapped. It can reduce the impact on the pedestrian which is blocked and overlapped. Besides, the method can get rid of background clutter, such as motion fuzzy, background mess etc. Experiments show that our method can achieve effective result for pedestrian detection.