基于多特征匹配的视频图像火灾识别方法

Video image fire recognition method based on multi-feature matching

针对改进传统火灾识别算法难以有效消除噪声和保护图像边缘细节的问题,提出了一种基于多特征匹配的火灾识别方法。使用改进的自适应中值滤波算法识别检测,通过分离色彩通道建立颜色模型,进一步判别火焰区域,采用帧间差分法对火焰部分轮廓进行分割和提取,获取火焰实时面积。实验结果表明,改进后的算法体系对色彩识别度由原来的87%提升至98.9%,对火灾识别率提升至92.7%,高于90%国际通用水平。相对于传统中值滤波算法和传统混合高斯算法体系,该方法能更好地保护图像细节,避免漏检误检。

This paper proposes a fire identification method based on multi-feature matching designed to address the problem that the improved traditional fire identification algorithm is difficult to remove noise and protect image edge details.The method works by identifying and detecting the flame area by using improved adaptive median filter algorithm;establishing a color model by separating the color channel to distinguish flame area further;segmenting and extracting the flame part contour to obtain the real-time flame area using the frame difference method.The experimental results show that the color recognition of the improved algorithm system is improved from 87%to 98.9%.The fire identification rate increases up to 92.7%,which is higher than the international standard by 90%.Compared with the traditional median filtering algorithm and the traditional mixed Gaussian algorithm system,this method can better protect the image details and avoid missing detection and misdetection.