基于混合高斯模型定位的火灾烟雾纹理特征提取  

Smoke Texture Character Extraction Based On Mixture Gaussian Model

作　　者：任厚平[1] 张永明[2] 张维农[1] 袁非牛[2] 余春雨[2] 

机构地区：[1]合肥中科院合肥智能机械研究所,230031 [2]合肥中国科技大学火灾科学重点实验室,230027

出　　处：《微计算机信息》2005年第11S期83-85,共3页Control & Automation

基　　金：中国博士后科学基金的资助(NO.2004036155)

摘　　要：视觉(图像型)火灾探测需要提取较高质量的目标纹理特征用于火灾识别。本文利用混合高斯模型对烟雾目标进行前景提取,并且屏蔽掉非运动的背景图像,然后根据结果图像的灰度分布,计算灰度共现距阵,提取纹理特征。Vision based fire detection has many advantages over traditional methods. In vision based fire detection approaches, it is required to extract better texture character of flame or smoke than other fields. We mainly take a advantage of mixture Gaussian model and frame difference to adaptively extract a variational foreground image from a still background. This paper uses mixture Gaussian model to extract the foreground image and screen the background. Then compute Gray Level Cooccurence Matrix of the foreground image and extract the texture character.

关 键 词：混合高斯模型 火灾区域 纹理提取 前景提取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]