针对传统火灾探测中灵敏度不高、响应慢的问题,提出一种基于特征融合的图像型火灾探测方法.结合火焰的颜色、运动以及闪烁特征,检测出疑似火灾区域中的火焰像素,排除非火焰像素,并用支持向量机对疑似火焰像素进行验证,采用形态学方法和区域融合判断出火灾区域.实验结果表明,该方法对多种火灾和非火灾场景具有较好的适应性、较强的抗干扰能力以及较高的探测率.%In order to overcome the disadvantages of traditional fire detection, such as low sensitivity and speed, an image fire detection method based on feature fusion is proposed. According to the color, motion and flame flicker features, fire pixel in candidate fire region is detected and non-fire pixels are removed. The candidate flame pixels are verified by support vector machines with a radial basis function kernel. Fire pixels which are verified are merged into fire regions by using morphological closing and region. Experimental results show that the proposed method can recognize fire effectively and has higher accuracy. It is more robust to noise, strong anti-jamming and high detection rate.
展开▼
