基于运动累积和半透明的视频烟雾探测模型

通过分析早期火灾烟雾运动规律，提出了一种适用于普通可见光视频的运动累积和半透明的视频烟雾探测模型。由于烟雾通常从阴燃点持续冒出，因而通过累积模型度量运动像素的累积程度，能够很好地捕获这种早期火灾的时空视觉特征，同时有效地抑制噪声的干扰。根据烟雾的模糊和部分遮挡背景特性，提出了一种基于高通滤波的半透明遮挡快速模型。半透明模型能有效地表征烟雾遮挡的半透明特性。实验结果表明，累积和半透明模型相结合提高了鲁棒性、增强了抗干扰性，明显地提高了探测的准确率。

Motion Accumulation and Translucence Based Model for Video Smoke Detection

Video smoke detection has many advantages over traditional methods, such as fast response and non-contact, ect. But most of current methods for video smoke detection have high rates of false alarms. By analyzing the characteristics of smoke motion, a motion accumulation and translucence combined model is proposed for video smoke detection. Because smoke often emerges continually from the place of smouldering, an accumulation model is presented to extract these temporal-spatial visual features of early fire over a time window. The model synthesizes information from several frames, so it can mostly suppress noises. Smoke can blur edges of objects and partially obscure the objects of background. After observing the phenomena, a translucence model is presented based on the high pass filter and the model efficiently represents translucence of the smoke. Experiments show that the combined accumulation and translucence model is robust and significant for improving the accuracy of the smoke detection.