基于视觉双通路与贝叶斯模型的烟雾检测方法

A Smoke Detection Algorithm Based on the Theory of Two Visual Pathways and Bayesian Model

该文提出了开放环境下基于贝叶斯模型与视觉双通路融合的烟雾检测方法。首先利用Itti视觉注意模型自下而上生成灰度显著性图;然后通过被测图像纹理特征直方图与烟雾样本纹理特征直方图的匹配,自上而下获得纹理显著性图;最后根据贝叶斯模型融合灰度与纹理显著性图,生成最终的烟雾概率显著性图。实验结果表明,该文方法能准确提取图像中的疑似烟雾区域,具有抗光照变化的优势,适用于开放环境下的实时烟雾检测。

An algorithm based on the Bayesian model and the theory of visual bi-pathways to detect smoke in an open environment is proposed.Firstly,a gray saliency map is generated on a bottom-up way based on Itti＇s visual attention model.Secondly,a texture saliency map is generated on a top-down way through matching the texture characteristic histogram of the test picture with that of the smoke samples.At last,a smoke probabilistic saliency map is generated by fusing the mentioned two saliency maps using the Bayesian model.The experimental results show that this algorithm can extract the suspected smoke area accurately in an image and being adapt for the smoke detection in an open environment.