论文部分内容阅读
火灾视觉特征的提取是火灾图像探测的关键问题。提出了一种描述火灾火焰轮廓脉动信息的火焰图像质心距脉动模型,即先用质心距来描述火焰的轮廓信息,再利用傅立叶描述子描述火焰轮廓变化的时空信息,采用距离来度量相邻两帧图像轮廓的变化幅度。实验结果表明,对于火灾火焰视频图像而言,其相邻两帧之间的脉动幅度、变化频率要明显高于受控燃烧、光源干扰等非火灾情况,可有效应用于火灾火焰的识别。研究表明,该方法能有效地探测火灾的发生,降低系统的误报率。
The extraction of fire visual features is the key issue of fire image detection. A model of flame image centroid pitch pulsation describing the pulse flare information of fire flames is proposed. The contour information of flames is first described by centroid distance. Then the Fourier descriptor is used to describe the space-time information of flame profile change. The distance is used to measure the adjacent two The extent of the change in the outline of the frame image. The experimental results show that for the fire flame video images, the fluctuation amplitude and frequency of the two adjacent frames are obviously higher than those of non-fire, such as controlled combustion and light source interference, which can be effectively applied to the identification of fire flames. Research shows that this method can effectively detect the occurrence of fire and reduce the false alarm rate of the system.