论文部分内容阅读
视频图像火灾探测技术因其探测效率高、响应速度快等优点而被广泛应用,但普通摄像机无法拍摄零照度(黑暗)甚至低照度环境下的烟雾图像。而火灾发生时往往先产生烟雾,若能实现火灾早期的烟气探测将有利于预防火灾规模的扩大和人员财产的损失。利用红外摄像机结合红外发射灯来成像以完成对火灾的探测,具体探测过程为:首先利用中值滤波去除原始灰度图像中的噪声;其次利用三帧差分法提取烟雾前景;然后通过Local Binary Patterns与灰度共生矩阵分别获取烟雾纹理的局部和统计特征;最后将烟雾纹理特征输入Fisher分类器以识别烟雾与非烟雾(如水汽、飞尘等),并及时发出报警信号。设计有无环境风和人为干扰环境下的烟雾与水汽试验,利用烟雾运动的不规则性与扩散性、图像的Gabor特征、LBP尺度验证所用纹理特征。结果表明,从分类正确率和响应时间看,在低照度环境下所用纹理特征优于其他烟雾特征法。另外,通过调节相机与火源之间的距离,可得到应用于实际工程的红外摄像机最佳安装距离。
Video image fire detection technology is widely used because of its high detection efficiency and fast response. However, ordinary cameras can not photograph smoke in zero illumination (darkness) or even low light environment. In the event of a fire, smog often occurs first. If the detection of flue gas in the early stage of a fire can help prevent the fire from expanding in scale and the loss of property of personnel, Firstly, the noise in the original grayscale image is removed by using the median filter; second, the smoke foreground is extracted by the three-frame difference method; then, the smoke is forecasted by Local Binary Patterns And gray-level co-occurrence matrix to get the local and statistical features of the smoke texture respectively. Finally, the smoke texture features are input into the Fisher classifier to identify the smoke and non-smoke (such as water vapor, fly ash, etc.) and send the alarm signal in time. Design the smoke and water vapor test under environment-free and man-made interference, use irregularity and diffusivity of smoke movement, Gabor feature of image, texture feature of LBP scale verification. The results show that the texture features used in low light environment are superior to other smoke features methods in terms of classification accuracy and response time. In addition, by adjusting the distance between the camera and the fire source, you can get the best installation distance of the infrared camera used in the actual project.