People Counting For Getting In/out Of A Bus Based On Stereo Vision And Implementation

Bus is an important public transport tool for most people, but along with the rapiddevelopment of economy,unreasonable route-planning,low site coverage, and frequenttraffic accidents are growing serious. To solve the problems above, the development ofIntelligent Transportation System(ITS) is of great importance and accurate statistics ofpassenge flow is a very important aspect of ITS.People counting for getting in/out of a busplays an important role in helping us with bus route planning and bus real-timedispatching. In addition, accurate traffic statistical techniques also have important socialand economic benefits.Image contains rich information, the passenger flow detection technology based onthe image has become a active new field today.The technology is divided into twocategories:monocular vision and stereo vision.Monocular vision is vulnerable to theinterference factors such as the light change, shadow in the scene While stereo vision cansolve those problems very well.This paper applied the stereo vision technology topassenger flow detection, put forward a kind of passenger flow detection algorithm basedon stereo vision.After that,the algorithm was transplanted to the DM6446platform andimplemented.Firstly,the paper made an introduction to stereo vision. After a comprehensivecomparison of many stereo matching algorithms, dynamic programming is chosen as thefinal algorithm for disparity map, whose main advantage is that it is a global approach andrelatively smooth dense disparity map can be obtained.Secondly,we integrated depth information into the traditional background subtractionmethod for the target detection,used threshold method for target segmentation and finallymarked the exact position of the target using the connected components labeling method.Finally, the traditional multiple-feature-fusing tracking algorithm is improved.Wemade improvements in three parts, which is the innovation of the paper.The final resultsshow that the improved tracking algorithm is more accurate and quick in tracking movingtargets. We transplanted the algorithm to the DSP platform and set up the people countingsystem based on the stereo vision after the algorithm was verified. After that, the systemwas optimized and the performance of the system was tested, the results show that thesystem can correctly extract and track the target, eventually achieve the correct number ofthe passenger flow.