基于混合颜色空间和双次Otsu的黄瓜靶斑病图像分割

为快速、准确地分割黄瓜叶部病害图像,提出一种基于混合颜色空间的双次Otsu算法。算法根据病害图像各部分的颜色特征,首先选取原始彩色图像的R分量进行初始Otsu分割和形态学相关操作,将R分量图分割为背景类和非背景类;然后选取非背景类图像的Cr分量进行第2次Otsu操作,将非背景区域分割为正常叶子类和病斑区域类,得到最终的分割结果。将该算法应用于黄瓜靶斑病图像的分割中,并与R_Otsu算法、H_Otsu算法以及图切割算法进行比较。试验结果表明:与对比算法相比,本算法在分割精度及处理速度2方面的综合分割性能最优,错分率均值和方差分别为2.12%和0.08%,平均处理时间0.2s,算法对光照变化具有一定的鲁棒性。本研究算法可为自然光照条件下黄瓜病害图像实时、准确分割提供技术参考。

Image segmentation of cucumber target spot disease based on hybrid color space and double Otsu algorithm

In order to segment cucumber disease images quickly and accurately, a method based on double Otsu algorithm using two color components was proposed.Firstly, the R component of the original color image was chosen to conduct the first Otsu segmentation and morphological operation and divided into background class and non-background class.Secondly, after comparing different color components of the diseased area with normal leaf area of the non-background class, the second Otsu operation was carried out on Cr_r component of the non-background image.Finally, the Cr_r component was segmented into normal leaf and spot area.In order to conduct the evaluation of the proposed algorithm, the method was applied to segment cucumber target leaf spot disease images under three light conditions.A total of 240 pictures were selected randomly with 80 pictures of each light condition.The R_Otsu algorithm, H_Otsu algorithm and graph cut algorithm were chosen as contrast means.The results indicated that the proposed method had best comprehensive performance in the segmentation accuracy and the processing speed terms compared with the other three algorithms.The mean and the variance of the error rate were 2.12% and 0.08%, respectively.And the average processing time was less than 0.2 s.Moreover, the algorithm was robust to illumination change.The proposed method is universal, can also be applied to segment for the other cucumber disease images under the natural light condition.