摘要
单高斯模型森林火灾烟图像目标检测方法是根据烟目标内部灰度分布不均匀,且形状和面积会随着时间不停变化等特点,首先,通过火灾烟的动态数据累积图像得到烟目标出现的大致区域,再通过构造单高斯背景统计模型,对目标区域内部进行逐帧分割,最后,统计烟目标面积的变化趋势。经过对森林火灾烟序列图像进行实验和分析,结果表明这种方法对早期火灾产生的烟能够实时检测和识别,并有效提高了火灾烟图像检测的可靠性。
Gaussian statistical model for early forest fire smoke detection method is based on the inner pixels grey scales in a very large range, and the smoke target is always changed with its contour. Firstly, by accumulated dynamic data areas of interest, the rough area of the smoke target is indicated. Then by using Gaussian background statistical model, the rough area of the smoke target is divided. Lastly, the area of the smoke target is counted. In the experiments, we use this method to analyze the early forest fire smoke. The results show that this method can reahimely detect early smoke region, and it improves the detection reliability of fire smoke.
出处
《计算机与现代化》
2009年第2期18-20,53,共4页
Computer and Modernization
关键词
火灾烟图像
目标检测
动态数据累积
高斯背景统计模型
fire smoke image
target detection
accumulated dynamic data
Gaussian background statistical model
