摘要
以数码相机采集的茶叶图像为对象,研究茶叶嫩芽的识别方法。采用基于Lab颜色模型中a分量、b分量信息的K-means聚类法识别彩色图像中的茶叶嫩芽。对不同距离采集的茶叶图像,对比分析Ostu法(最大方差自动取阈法)和3个聚类中心的K-means聚类法的目标识别效果和识别效率。结果表明,Ostu法虽然可以完成嫩芽的识别,平均识别率在89%左右,但不能较好的保证分割后嫩芽的完整度。基Lab颜色模型和K-means聚类法的识别算法能较好的区分嫩芽和背景,平均识别率达到94%左右,且能较好的保证分割后嫩芽的完整度,为智能采摘技术研究提供技术支持和理论基础。
Tea images in Guizhou University and Guizhou Academy of Tea Garden were taken by camera to study the identification methods of the tea buds. The identification algorithm which takes advantage of K-means clustering method based on a and b component of Lab color model was proposed. Using tea images taken with different distances, the target recognition effect and efficiency gotten by Ostu method and K-means clustering with 3 cluster centers method were compared. The Ostu method for tea buds identification was fast but the average recognition rate was low, only up to 89%. Furthermore, the Ostu method could not guarantee the integrity of divided buds. Compared to Ostu method, the algorithm based on Lab color model and K-means clustering could distinguish buds from background well, with average recognition rate 94%. Moreover, the algorithm could guarantee the integrity of divided buds better, which provides technical support and theoretical basis for the future intelligent tea harvesting studies.
出处
《中国农机化学报》
2015年第5期161-164,179,共5页
Journal of Chinese Agricultural Mechanization
基金
贵州省科学技术基金项目(黔科合J字[2011]2199号)
贵州大学引进人才科研项目(贵大人基合字(2013)38号)
