自然背景下基于反射模型的树叶彩色图像分割  

Foliage Color Image Segmentation Based on Reflectance Model in Natural Background

作　　者：沈雯雯[1] 白杰云[2] 官俊[1] 任洪娥[1,3] 

机构地区：[1]东北林业大学信息与计算机工程学院,哈尔滨150040 [2]哈尔滨工业大学计算机科学与技术学院,哈尔滨150001 [3]黑龙江省林业智能装备工程研究中心,哈尔滨150040

出　　处：《森林工程》2014年第6期13-16,21,共5页Forest Engineering

基　　金：中央高校基本科研业务费专项资金(DL12C B05);国家林业公益性行业科研专项(2010040071)

摘　　要：为实现固碳释氧测量中叶片图像的高精度分割,提出一种基于反射模型的树叶彩色图像分割算法。首先,在RGB颜色空间中,针对光照对于叶片颜色不一致的影响,采用改进的白平衡方法进行滤波;其次,针对RGB颜色空间的非均匀性、色域窄、亮度分量混合的问题,将其变换到Lab颜色空间,去除亮度分量,消除光照对于叶片分割的影响;然后,依据Lab颜色空间的均匀性,定义颜色特征的欧氏距离为聚类准则;最后采用最小分类器进行聚类分割。实验结果的对比分析表明该方法是有效的。To achieve leaf image segmentation with high precision during measuring the volume of fixed carbon and oxygen re- lease, a method for foliage color image segmentation based on reflectance model was proposed. First, in RGB color space, an im- proved white balance method was used to eliminate the effects of the illumination inconsistencies. Second, the RGB color space was converted to lab uniform color space due to the non-uniformity, a narrow color gamut and mixed illumination, and the luminance com- ponents were removed to eliminate the influence of light on image segmentation. Third, according to the uniformity of the lab color space, the Euclidean distance clustering criterion of color features was defined. Finally, the minimum distance classifier was adopted in image segmentation. Experimental results showed that the proposed method is effective.

关 键 词：图像分割 颜色空间 反射模型 颜色恒常性 

分 类 号：S781[农业科学—木材科学与技术]