基于模型切换的自适应背景建模方法

Model Switching Based Adaptive Background Modeling Approach

提出了一种基于模型切换的背景建模方法(MSBM)．该方法以熵图像为纽带,实现了不同精细程度的背景模型在空间上的自适应选取和在时间上的自适应切换．对于亮度分布复杂度高的背景区域采用精细的模型以保证运动目标检测的精度,反之采用简单的模型以降低计算量．通过模型结构自适应结合参数自适应,很好地兼顾了检测精度和计算代价．基于高斯混合模型和时间平均模型的双模型切换式运动目标检测算法被用于实验研究,结果表明这种算法的检测效果和单独采用高斯混合模型的检测效果相当,而计算速度却比后者提高很多．

A model switching based background modeling approach （MSBM） has been proposed. This approach uses entropy image as ligament to realize adaptive switching between background models with different elaborations in a spatiotemporal domain. For background regions with high complexity of pixel＇s value distribution, we adopt an elaborate model to guarantee the accuracy of moving object detection; otherwise, we adopt a coarse model to reduce the computational load. Combining adaptive model structure with adaptive model parameter, MSBM can improve the processing speed greatly without sacrificing the accuracy. A double-model-switching moving object detection algorithm based on Gaussian mixture model and temporal average model has been used in the experiment and the results show that it can possesses almost the same detection accuracy and much higher image processing frame rate than Gaussian mixture model.