为解决了火焰闪烁频率计算量大、计算结果不准确的问题,首先基于火焰的闪烁特性,选取连续帧图像中可疑区域的相似度作为火焰和干扰源的判别依据,利用正负大样本数据确定了相似度阈值的范围大小;然后,根据火焰的运动特性设计了一种连续帧图像中可疑区域的区域跟踪方法。实验证明:该方法能够跟踪出连续帧图像中的可疑区域,并可排除常见干扰源的影响,准确地进行火灾预警。
A new flame recognition method referring to the similarity of consecutive frames is proposed and the threshold is determined by a large number of sample videos.Besides,a tracking algorithm of suspected areas between consecutive frames is proposed based on the fire moving characteristics.A series of experiment results show that the proposed method can track the suspected areas between consecutive frames exactly and prompt fire warning precisely by eliminating the influence of common interference.