基于改进的均值漂移的森林火灾图像提取技术

Technology Research on Forest Flame Image Extraction and Recognition Based on Improved Meanshift Algorithm

为了解决呈红色山地路面等对火灾图像提取的干扰问题,考虑林区图像的R通道直方图的分布曲线,采用一种基于均值漂移改进算法,自适应地得到背景与火灾,山地路面的分割阈值T1。通过阈值分割将火灾和山地路面从背景图像中分割出来,再利用改进算法对R通道占比率直方图进行处理,自适应地得到火灾与山地路面的分割阈值T2,从而实现火灾的自适应阈值分割并结合平滑直方图曲线去噪法进一步完善分割效果。在对均值漂移算法进行改进后,阈值的获取自适应性更强,实验表明该方法具有良好的分割效果和运算效率。

In order to solve the flame image extraction problem disturbed by the red land,the distribution curve of R-channel forest image histogram is investigated.Adaptive segment thresholds T1 of background and forest flame and the red land is obtained by using a method based on improved meanshift algorithm.The forest flame and the land with red color are separated from the background area image by the adaptive segment thresholds,and then occupation ratio curve of R channel image histogram is processed by using the same algorithm and adaptive segment thresholds T2 of segmenting forest flame and the red land is obtained.The method of smoothing histogram curve is introduced to get more accurate threshold.Experimental results show that this adaptive threshold segmentation method is rapid and has better segmentation effect and calculation efficiency.