Research On Passenger Statistics On The Bus Based On Multi-block LBP And AdaBoost

As a part of ITS(Intelligent Traffic System), the passenger statistics on the bus gets a lot of scholar's attention in recent years. The effective passenger statistics on the bus is good to monitor those charging phenomena by bus drivers. It is also help to solve bus-scheduling problem and alleviate traffic jams. Due to the high density of passengers boarding and alighting in our country, and the congestion and disorder when getting on and off, the traditional methods can not meet the demand of passenger statistics on the bus.Based on the comprehensive analysis of current passenger statistics on the bus, this paper raises a method of passenger statistics on the bus, mainly including target detection and target tracking. When the vertically mounted camera captures the video on the bus, we use the Ada Boost algorithm based on multi-block LBP features to detect people's heads in the video, and then set up a target chain, and then use Camshift algorithm based on Kalman prediction to track and count those target. Compares to traditional Haar feature, multi-LBP feature is more fit to head detection for its rotational invariance and gray-scale invariance. It can describe the texture features. the method in this paper improve the precision of the passenger statistics.We do a lot of experiments with videos captured by day and night,and with the images including 2400 head pictures with the size of 24*24 and 3700 pictures without head.The results show that this method can realize the passenger statistics, and basically meet the needs of practical application.