烟火视频序列的混合高斯模型利弊探讨

Discussion about the advantages and disadvantages of Gaussian mixture model for fire video

烟火视频序列背景模型的质量直接影响到火灾监控的准确性,烟火初起阶段与发展阶段是火灾监控的关键时间环节,目前以视频序列背景模型利弊分析为主题的文章较少。本文首先针对烟火视频序列建立混合高斯模型,然后利用帧间差提取动目标区域,最后,对影响混合高斯模型质量的更新率进行实验分析。实验结果表明,在烟火初起期、发展期固定环境下,混合高斯背景更新与帧间差相结合的方法能有效地提取出动目标区域。实验也表明,更新率值过大或者过小时,直接导致动目标区域提取不准确。若更新率值过小,背景模型更新较慢,造成背景区域误判为动目标,动目标区域出现拖影问题。若更新率过大,背景模型更新较快,运动速度缓慢或暂时停滞的目标会被误检测为背景,动目标区域会出现空洞现象。

The background model for fire video directly affects the accuracy of fire monitoring. Fire is divided into several stages. The early stage and the development stage are the key points of fire control. There are few articles relating to discussion about the advantages and disadvantages of the background model. In this paper, the Gaussian mixture model for fire video is constructed. Then, moving target areas are obtained by frame difference method. Finally, the advantages and disadvantages of the Gaussian mixture model for fire video is discussed by means of the update rate. The experiment results show that the moving target area can be extracted effectively using Gaussian mixture model and frame difference method during the early stage and the development stage of fire. The update rate directly affects the accuracy about the extraction of the moving target area. If the update rate is too small, the update about background model is slow. The background areas are wrongly identified as moving targets. If the update rate is too large, the background model is updated quickly, the target of slow motion or temporary stagnation will be wrongly identified as background.