基于亮度守恒假设的火灾图像运动特征提取

Motion Feature Extraction of Fire Images based on Brightness Conservation Assumption

论述了火灾火焰和烟雾图像特征分析及提取方法,分析了这些方法尚存在的问题,指出了火灾图像运动特征分析相对于静态特征分析的优势.针对亮度不变假设在提取火灾火焰和烟雾图像运动特征时存在的问题,提出将目前最新的光流计算方法——亮度守恒假设应用于视频火灾探测中,结合全局平滑性假设,推导了亮度守恒方程解的迭代形式,并对基于亮度守恒假设的火灾火焰和烟雾图像运动特征识别方法进行了分析和讨论,为开发更加准确可靠的视频火灾探测识别算法提供了理论支撑.

The fire flame and smoke image feature analyzing and extraction methods are discussed, while the remaining problems of these methods are analyzed and the advantage of fire motion feature compared to static features is pointed out. Aiming at the specific application problems of brightness constancy assumption in fire flame and smoke motion feature extraction, the latest optical flow technique-brightness conservation assumption is put forward and applied in video fire detection. Together with the global smooth assumption, the iteration form of the solution of brightness conservation assumption equation is derived. The recognition method of fire flame and smoke motion feature based on brightness conservation assumption is analyzed and discussed, which can provide the theoretical support for the development of video fire detection recognition algorithm which is more accurate.