基于帧间差分法的动体特征速度聚类分析

发布时间：2018-05-10 18:12

  本文选题：帧间差分法 + 众数聚类分析　； 参考：《计算机应用研究》2016年10期

【摘要】：针对智能视频监控中快速、准确地检测和识别运动物体的问题,提出了一种依据运动物体特征速度来检测识别动体以及解读其语义含义的算法。该方法以相对帧间差分法为基础,通过对预处理后的二值斑块图像的标记,计算斑块的像素长度作为其特征速度,并依据斑块特征速度的众数进行聚类分析,从斑块特征速度得到运动物体的特征速度语义解读和运动物体的检测识别。实验结果表明,斑块的特征速度不仅可以实现对运动物体的检测,而且通过聚类分析可以准确地得出动体特征的语义解读。用特征速度和众数聚类分析方法实现对运动物体的检测识别和语义解读,相对于其他统计算法简单有效,便于智能摄像机的嵌入式开发。
[Abstract]:To solve the problem of fast and accurate detection and recognition of moving objects in intelligent video surveillance, an algorithm is proposed to detect and recognize moving objects and interpret their semantic meanings according to the characteristic velocity of moving objects. Based on the relative inter-frame difference method, the pixel length of the pre-processed binary patch image is calculated as the feature speed, and the clustering analysis is carried out according to the mode number of the plaque feature velocity. The semantic interpretation of feature velocity of moving object and the detection and recognition of moving object are obtained from the feature velocity of plaque. The experimental results show that the feature velocity of patches can not only detect moving objects, but also accurately interpret the semantic features of moving objects by cluster analysis. The method of feature speed and mode cluster analysis is used to detect and recognize moving objects and to interpret the semantics. Compared with other statistical algorithms, it is simple and effective, and is convenient for the embedded development of intelligent camera.
【作者单位】： 大连理工大学管理与经济学部;
【基金】：国家“十二五”资助项目子课题(2013BAK02B06-03)
【分类号】：TP391.41;TN948.6
，

本文编号：1870348