基于视频序列的目标检测与跟踪技术研究

A Study of Object Detecting and Tracking Based on Video Sequences

针对计算机智能监控环境,文中提出一种改进的基于像素灰度出现概率最大值的背景建立方法,该方法克服了光照变化对背景重建的影响,使得背景建立的时间大大缩短。并采用一种新的自适应背景更新算法获得背景图像以进行目标检测,这种方法较好地克服了IIR法更新速度难以取值的缺点,使得更新速率可以达到自适应的效果;在目标跟踪阶段,使用基于卡尔曼滤波的方法对检测出的运动目标进行跟踪,由于卡尔曼预测可以大大减小特征匹配的搜索范围,因此提高了跟踪的实时性。实验结果表明,该文的算法能够快速有效地获得、更新背景,并且能够实时地对运动目标进行跟踪。

This thesis was focused on intelligent surveillance. It proposed an improved method of building background, which is based on the maximum probability value of the pixel. This method overcame the light change on the background of the impact of the reconstruction, making the background of the establishment of the time significantly shortened. Then, adopted a new self- adaptive background updating algorithm to gain a background image for the purpose of target detecting, the new method was proved better than IIR method , for it could obtain an adaptive effect on updating rates; In order to track the target detected, the Kalman filtering algorithm was used. The results showed that this algorithm could obtain and update background image in a short time. It also could track the detected target in real - time.