基于Lab颜色距离和GMM的树木图像分割算法

Segmentation algorithm of tree image based on lab color-distance and GMM

针对传统的高斯混合模型(GMM)分割算法在树木图像分割上存在的不足,文中结合Lab色彩模型的颜色和空间的相关性,提出了一种基于Lab颜色距离的GMM的树木图像目标分割算法。该算法首先利用图像颜色信息,将图像的多个颜色通道进行Lab颜色距离的计算,然后基于Lab颜色距离建立GMM,最后根据得到的概率值自适应寻找最佳分割阈值。实验结果表明该方法得到的果树区域更准确,并且对阴影、光照不均匀图像具有很强的鲁棒性,分割的平均准确率为98.2%。

For the shortcomings of traditional GMM image segmentation algorithm on tree using GMM,an effective algorithm based on the GMM of the Lab color distance was proposed in this paper which combined color and spatial correlation in Lab space. First,it makes use of image color information to calculate the Lab color distance through multiple color channels,and then,establishes the GMM based on Lab color distance,finally,finds the best segmentation threshold adaptive based on probability value obtained to get the effective segmentation image. The experiment results show that the algorithm not only can segment the needed fruit object from complex background,but also can reduce the error-detection rate. The tested accuracy of fruit tree images segmentation was 98. 2%.