一个基于计算机视觉的监控系统算法设计

The Design for a Monitoring System Based on Computer Vision

提出了一种新的基于计算机视觉的双目识别算法。该算法利用象素点的深度信息,采用统计的方法给场景建模,通过时间滤波克服光照渐变,依靠深度算法特性克服光照突变,并采用图像深度块融合技术,减少数据量与计算量,快速识别。在DSP平台上实现了该算法。实测证明,该算法与传统方法相比,具有较低的计算复杂度和较低的误检率,在视频监控应用中取得了很好的效果。

A new recognizing and tracing algorithm based on two-eyes computer vision is presented. This algorithm makes use of the depth information of each pixel, then it adopts the statistics method to modeling the scene and overcomes the gradual light change by time filter. It can also overcome the sudden light change by the characteristic of the depth algorithm. It can reduce data and computation by the technique of fusing the depth blocks, then recognize the scene quickly. It is realized the algorithm on the DSP platform. The simulation shows that this algorithm has a lower complexity and lower miss rate. It makes a good effort in the practice of monitoring.