Research On Pedestrian Tracking Algorithm Based On Monocular Vision

With the growing number of vehicles, road traffic accidents,such as caraccidents,does great harm to people’s life safety,which has become a hot socialissue.As a result,developing an intelligent vehicle safety assistant driving systembecome a hot point in computer vision research.The road traffic specific situation in our country is that pedestrians are importantparticipants.While in many cases,the collision accidents between pedestrians andvehicles was due to the driver’s carelessness,wrong judgment or fatigue driving.Inorder to protect the safety of pedestrians in the street more timely and effectively,thekey point in intelligent vehicle safety assistant driving system research is detectingand following the trace of pedestrian forward,as well as giving an alarm to the driverin time if necessary.It also has great importance to reduce the probability of collisionaccidents between pedestrians and vehicles.Road information in front of vehicles can be obtained through cameras installedin vehicles, for intelligent safety and driver assistance system, in particular its focus:moving objects tracking was studied in image sequences. Improved trackingalgorithm based on template matching: a weighted mean normalized related measureis proposed; for postures and size changing of moving objects in image sequences,an improved adaptive template updating policies with combination of coefficient andtarget size variation has proposed; To reduce unnecessary calculations amount in thetracking, kalman filter is raised to forecast regional scope where pedestrians mightoccur in the next frame, and to make relevant matches reducing from the entire imageto the area that pedestrians may appear, to dramatically reduce calculation amount.To a certain extent these proposed improvements make moving object trackingperformance and tracking efficiency by correlation matching algorithm better in theimage sequence.After researching the particle filter theory,propose the particle filtertracking algorithms based on the histogram,which can help tracking more reliable andmore stable.This paper researches on intelligent vehicle safety assistant drivingsystem.Using simplified monocular vision algorithms, estimate the distance betweenpedestrians in front of the road and vehicles, and on this basis, develop early warningmechanisms and early warning strategies for preventing pedestrian collision, whenthe distance between vehicles on the road in front of pedestrians and vehicles is tooclose, drivers are alerted and reminded to take timely and effective protectionmeasures to avoid traffic accidents.