中文摘要：

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。

英文摘要：

The paper attempts to present an effective technique for prompt identification of the active infrared smoke based on the texture features consisting of a MOXA infrared camera and the infrared emission light,which can help to distinguish between the smoke image in the dark（ namely,low identifying light or the background with zero illumination）,in which it would be impossible to identify any image by using ordinary video cameras. The image-identifying process by way of the active infrared smoke based on the texture features consisting of a MOXA infrared camera can basically be made up of four steps for the new technique under study. The first step is to use the medium filter method to remove the fuzzy parts of the gray image,which turn to be a bit fuzzy as it has been shown at the first minutes. Secondly,it would be possible to use the three frame difference method to remove the smoke foreground frame,trying to obtain the local and statistics or digital texture features by means of the local binary patterns and the gray level con-current matrix, respectively.And,finally,the texture feature can be put into the Fisher classifier to distinguish between the smoke and non-smoke（ dust and mist） images via the detection signals given to an alarm. For the said purpose,we have designed the identification experiments on the smoke and mist while taking into full account the background wind and the interference and disturbance factors made by the human movement and the background irregularities. Human disturbance considered,we have also taken into more account the texture features by examining the irregular and diffusing features of the smoke while paying attention to the Gabor feature and LBP scale of the image. What is more,the texture feature method is supposed to be of more importance and more implicative than other smoke characteristics-described ones in the dark environment by checking their proper classification rates and response time-lengths. And,one more point to be noted,the best installation position of the