一种基于DSP的运动目标检测与判别算法

本文提出一种基于DSP、静止摄像机条件下运动目标的实时检测与判别算法。它针对CCD摄像机广泛使用的YUV图像格式，对自适应混合高斯模型进行改进，并根据光亮变化的快慢来更新背景模型中的更新率。；在判别算法中，利用特殊的高宽测量方法和离散度来消除阴影对判别的影响。在多次实验中，本算法显示了良好的性能。

Moving Object Detection and Classification Algorithms Based on DSP

This paper describes an algorithm for real-time moving object detection and classification with a slatic camera based on DSP. It improves the adaptive  hybrid Gausslan model with the YUV image format which is widely used by CCD cameras. And it updates the learning coefficient a depending on the illumin ation changing speed. In the classification algorithm it uses especial height/width measurement and disperse degree to eliminate shadow effect. In many   experiments, the algorithm shows good performance.