Pedestrians Tracking Based On Least Squares And Intelligent Collision Avoidance Model

Object tracking is one of the most important technology of intelligent video analysis system.The basic principle of object tracking is statistical the characteristics of the targets through a certain way and determine the target location in each video frame. First,in Initial video frame,we will give the location and area of target,and put it as tracking targets.Then,we will collect the targets' location in each subsequent video frame.Tracking algorithm should meet certain robustness to ensure that the targets has also been tracked and avoid lost targets.A novel target detection algorithm based on YUV color space is proposed in this paper.Aim at the existing issues of lacking of adapting illumination changes and background changes in RGB color space,we select YUV color space to establish background model.This method uses the independent feature of luminance and chrominance information, which can effectively improve detection accuracy, and has a good effect on the hand of removing shadows.In addition, in this paper, we focus to propose a pedestrian tracking algorithm that least squares method combined with the adaptive obstacle avoidance algorithm. Firstly, in view of the weakness Kalman algorithm, we use least squares method to predict pedestrian's trajectory. In the mean time,we also combine this algorithm with adaptive obstacle avoidance algorithm. Compared with the traditional pedestrian tracking method, The proposed method in this paper is more in line with the human habit of thinking, more intelligent.The article main research content is as follows:(1) Firstly, the RGB color forms are converted to the forms of YUV, and further respectively construct a codebook model for the every pixel of video frame; Then codebook models are trained so as to achieve the background modeling;(2) Then,judging the results according to whether the current pixels value will match with this pixel codebook, and detecting the foreground target;(3) Using the traditional Kalman algorithm to achieve the initial target tracking and take some approaches to improve it;(4) For the high error rate and high lost rate of pedestrian tracking using Kalman algorithm,we propose to use the least square to predict pedestrian track.Due to the particularity of the pedestrian, the psychological and exercise habits will affect the movement. According to this characteristic, adaptive obstacle avoidance algorithm for pedestrian tracking is propose. While meeting the obstacle, the trajectory of the pedestrian will be more intelligent.