基于视频的火焰检测方法

Flame detection method based on video

在分析视频监控的火焰检测技术的基础上,针对传统的帧间差分法,提出一种结合分块处理和帧间差分的分块帧组差分的运动检测方法,提高相邻像素间的关联性,降低噪声干扰的影响,通过色调、色饱和度和亮度来分析视频中火焰像素的静态颜色特征,降低亮度对颜色检测的影响。充分利用燃烧火焰的动态频谱特征和空间的能量信息,对不同时刻火焰燃烧的图像时域和空域进行小波分析;分析后的疑似火焰区域利用连通邻域像素的信息对零散的非火焰像素点进行滤除。实验结果表明,该检测方法更具目的性,且颜色检测更具可靠性,降低了闪烁光和类似火焰物体等造成的影响。

During the main research of flame detection techniques based on video, aiming at the traditional method of frame difference, a new approach combined blocks with flame difference on motion detection was introduced. This approach improves the relevance of every pixel in blocks and reduces the noise. Furthermore, static color features of the flame were analyzed by hue, saturation and value in HSV, which reduced the influence of intensity. Then, according to spectrum characters in temporal and energy information in spatial, the wavelet transform was used to detect whether flaming or not. Lastly, the 8 neighbor pixels were used to eliminate the suspected flame area＇s pixels. The results of the experiment prove that these blocks are useful, and the color detection is reliable to detect flames, which reduces the influence of flicker and some objects similar to flames.