一种基于单张图像的烟雾检测算法

为了提高烟雾检测的准确性,有效排除图像中类似烟雾的区域,降低由于环境和光照不同而产生的误检率,提出一种基于颜色矩并结合邻近算法进行分类学习的方法,对图像中的烟雾进行检测,该算法能克服光照不同、背景复杂的干扰,有效消除疑似烟雾区域的存在。算法由两部分构成:烟雾颜色矩特征的提取和利用邻近算法对特征数据集进行分类学习,自动找出并正确标记新的测试图像中烟雾区域的位置。

A New Image Based Smoke Detection Algorithm

In order to improve the accuracy of the image based on smoke detection algorithms, and eliminate the similar object of smoke region, reduce the false detection rate caused by environment light changes, a new smoke detection algorithm based on colormoment feature and k-Nearest Neighbor algorithm was presented in this paper. This algorithm can effectively reduce the interference of scene illumination detected and complex background. The algorithm in this paper consists of two parts ： extracting the colormoment characters of smoke image and using k-Nearest Neighbor algorithm to learn these characters, then mark the smoke region in the new testing image automatically.