摘要
提出了混合颜色特征下双层K-means聚类分割方法,首先在I分量将图像像素采用K-means聚类分割方法聚为4类;取聚类中心最大的2种像素的a*b*分量,进行第2次K-means聚类,得到病斑图像。对采集的21幅病害图片的试验结果表明,该方法分割结果的平均重合系数为97.53%,平均假阳性率为1.22%,平均假阴性率为3.52%。该研究可为进一步病害特征提取识别与病害程度诊断研究提供技术参考。
Double-layer K-means clustering segmentation method with mixed color features were proposed. Firstly,the image pixels of component I were clustered into 4 types by using k-means clustering method. a* b* components with two largest pixel in cluster center were selected to make the second K-means clustering,and the disease spots images were obtained. The test results of 21 sampled disease images showed that the average overlap coefficient of the segment results was 97. 53%,the average false positive rate was 1. 22%,and the average negative positive rate was 3. 52%. The research could provide technique references for further study on the extraction,identification of disease features,and the diagnosis of disease degree.
出处
《安徽农业科学》
CAS
2018年第3期169-170,198,共3页
Journal of Anhui Agricultural Sciences
基金
陕西省农业科技创新与攻关项目(2015NY034)