基于特征点检测与光流法的运动目标跟踪算法

Moving object tracking algorithm based on feature point detection and optical flow

为了解决当前运动目标跟踪算法在背景模型复杂和目标特征不明显的情况下,导致算法跟踪能力不足的问题,本文分别从特征点检测与光流法分析的角度出发,提出了基于特征点检测与光流法的运动目标跟踪算法。首先,根据图像梯度矩阵最小特征值,通过仿射变换,精确化特征点帧间匹配,排除伪特征点,达到精准检测运动目标特征点的目的。然后,基于图像像素守恒原理,进行2幅图像间变形评估,建立图像约束方程,进一步精确跟踪运动目标。最后,基于软件开发环境QTCreator实现算法,并系统集成。实验测试结果显示:与当前运动目标跟踪技术相比,本文算法拥有更高的准确性与稳定性。

In order to solve the current moving object tracking algorithm in complex background model and target characteristics is not obvious,lead to the problem of insufficient algorithm tracking ability,this paper respectively from the feature point detection and the perspective of optical flow method,is proposed based on feature point detection and optical flow method of moving object tracking algorithm.First of all,according to the minimum image gradient matrix eigenvalue,by affine transformation,accurate feature points matching between frames,eliminate the false feature points and achieve the purpose of accurately detecting moving target feature points.Then,based on image pixel conservation principle,the deformation between two image assessments,establish image constraint equations,further precise tracking moving targets.Finally,based on the software development environment QTCreator algorithm,and system integration.Test results show that compared with the current motion target tracking technology,the algorithm has higher accuracy and stability.