Research On Trajectory Extraction Technique Of Swarm

In object recognition and tracking system, gathering the object trajectory and other information is an important part of it. It can obtain the objects'spatial coordinates, time parameters and reliable data, which would be the research base of multi-object motion analysis, scene analysis. At present, for large-scale swarm tracking and trajectory extraction technology are tough, so swarm trajectory extraction system has high research value.This project aims at application scenarios for swarms, through a fixed camera capturing swarm motion video sequence, object detection, object tracking and other processes to extract the moving objects two-dimensional image coordinates of the location in each frame, and associate moving objects of the same continuous sequence in the image sequences according to different feature value. Thus we get the whole trajectory information of moving objects.The main achievements of this paper include:(1) Reserch on Object Detection Technology. This paper first describes and analyzes in-depth of several classical methods on object detection, and then gives the algorithm used in this paper, that is the improved Gaussian mixture model with shadow detection object detection algorithm. This method can remove a series of background and shadow interference pixels in the scene, and it has the ability to adapt to environment changes and a brief suspension of objects. At the same time, this paper studies the related techniques of binary denoising in foreground extraction to geta complete foreground.(2) Reserch on Object Tracking Technology. In this paper, a detailed study of several mainstream tracking algorithms have given, and based on them we analyze the particle filter based on the mean shift algorithm. This method can handle the situation that multiple objects have the occlusion and distinguish targets. Finally, we save the trajectory information to the file after we get the position information at each time which is just the trajectory. (3) Construct Object Tracking and Trajectory Extraction System. In this paper, we sort out the whole process of object tracking and trajectory extraction, design and construct the trajectory extraction system using components to connect. That is, to capture swarm motion video through the camera, and use target recognition and tracking algorithm for tracking separate moving objects in image sequences, to extract the target coordinates and other related trace parameters.This paper tests a single-target as well as multi-target tracking and trajectory extraction, and it could extract most of the object trajectory which is proved that the system can effectively solve the swarm tracking and trajectory extraction problem. In addition, this paper gives experimental analysis of occlusion in multi-object movement and also research on the background interference and other issues, the experimental results shows that algorithms used in this paper can effectively resolved those problems.