Research On Pedestrian Trajectory Based On Matrix Representation

As the rapid development of image processing,pattern recognition and computer vision,research on pedestrian trajectory from videos attracts more and more attention in social life.Anomaly can be detected by tracking specific person.For example,around office building,video surveillance system can detect unexpected suspect by anomaly detecting.At crossroads,acts against traffic rule such as jaywalking can be discerned.In city service,crowd flow pattern can be retrieved from trajectory and used for crowd management.In urban service,detection of vehicle and crowd density can alleviate traffic congestion.In large shopping malls,public transportation hubs as well as large-scale activities,through analyzing the flow of human and degree of confusion,we can prevent stampede,reduce waiting time and enhance people's visit satisfaction.For both anomaly detection and crowd management,we need to research pedestrian trajectories and dig out information behind the surveillance video.The main work for this paper is creating an algorithm for pedestrian trajectory research;it can be used to identify multi-target trajectories for few people scene.Based on this,small group are detected and behaviors are analyzed.In this paper,a pedestrian trajectory identification algorithm based on matrix representation is created.Our idea is inspired from multi-target trajectories identification algorithm based on inter programming,matrixes are used to describe input and output information,based on matrix operation and transform,multi-target trajectories are identified.The algorithm is applicable to sensor networks,single and multi-camera surveillance camera network.Based on the observation information of every time instant and the structure matrix of regions,we use matrix transformation and matrix operations to find the position movement of multi-target in monitoring network.Experiments on 2010 Shanghai Word Expo show the effect of our algorithm.At the same time,we can prove that in both efficiency and accuracy,our algorithm performs better than the algorithm that is based on inter programming.In this paper,a framework for small group behavior analysis based on trajectory merging is designed and achieved.We obtain input information of our trajectory identification algorithm by pre-processing the surveillance video.After improving our trajectory identification algorithm,pedestrian trajectories can be calculated.According to the identified trajectories,trajectory merging algorithm is designed to detect cluster(small group)from pedestrians.At last,behavior analysis algorithm is proposed to research small group behavior.Experiments on 2010 Shanghai Word Expo show that for a particular scene,the proposed framework is able to obtain the behavior patterns,which are pedestrian structure,flow pattern and scene characteristics.These results can be provided to organizers to arrange crowd management,predict visitors and recommend attractions.