摘要
为了克服由于烟雾稀薄、远景以及风速带来的干扰,实现火灾的及时准确检测,提出了一种有效的基于小波纹理特征分析的视频烟雾检测算法。该算法首先利用混合高斯模型提取运动区域,对运动区域进行网格化和二维离散小波变换以获取局部信息;然后利用灰度共生矩阵提取每个网格的纹理特征;最后利用自适应神经模糊推理系统(ANFIS)和联合判别准则进行训练和火灾判断。实验表明,该算法检测率达到了94.8%,误检率为1.1%,证明该算法可以充分挖掘图像的局部信息,并提高了检测烟雾的空间分辨率,适宜多种场景,可靠性较高。
In order to overcome interference with thin smoke, distance view, wind speed and realize real-time and accurate fire detection, an valid video smoke detection algorithm based on wavelet texture feature is proposed. Firstly, Gaussian Mixture Model is applied for extracting the motion region. Then the motion region is meshed and transformed by Two-dimension discrete wavelet to obtain local massage. Utilizing gray level co-occurrence matrix, each cellular’s textural features are obtained. Finally, the adaptive neuro-fuzzy inference system and joint criterion are used for training and fire judgment. The Experimental results show that the algorithm detection rate reaches 94.8% and error detection rate is 1.1%, the algorithm can fully excavate local information of the image and improve the spatial resolution of the smoke detection, it is suitable for a variety of scene and has high reliability.
出处
《光学技术》
CAS
CSCD
北大核心
2013年第4期348-353,共6页
Optical Technique
基金
河北省自然科学基金资助项目(F2010001268)
关键词
纹理特征
二维离散小波变换
灰度共生矩阵
自适应神经模糊推理系统
textural features
two-dimension discrete wavelet transform
gray level co-occurrence matrix
adaptiveneuro-fuzzy inference system