摘要
图像识别是当前人工智能发展的1个热点。针对农业果园采集图像光照不均匀，背景复杂多样等特点，提出了一种基于静态背景差分的最大类间方差改进算法用于果园复杂背景图像的分割。通过调节参数y达到不同对比度下果实与背景环境的最佳分割。然后利用距离变换和分水岭算法进行粘连果实的分离，实现果实图像的识别，并且通过实验分析和计算确定了最佳参数y范围在0．2～0．3之间。实验验证表明，图像识别效果良好，能够为果园品质在线检测提供很好的技术基础。
Image recognition is a hotspot in the development of artificial intelligence. Due to theuneven illumination and complex background of imagescollected from orchard agriculture, this paper proposed a new improved OTSU which is based on static background subtraction algorithm. By adjusting the parameters γ to achieve optimal segmented fruit under different contrast with the background environment, the researchers use the distance transform watershed algorithm to separate adhesions fruit, and determine the optimum parameter γ range between 0.2 to 0.3 viaexperimental analysis and calculations. Results show that the image recognition works well, whichcan provide a good foundation for the technical quality of the orchard online testing.
作者
陈亚峰
张晓明
黄亚鸿
徐日华
CHEN Ya-feng ZHANG Xiao-ming HUANG Ya-hong XU Ri-hua(College of Information Science and Technology, Beijing University of Chemical Technology, Beijing 10029, China Department of Computer Science, Beijing Institute of Petrochemical Technology, Beijing 102617, China Department of Automation, Beijing Institute of Petrochemical Technology, Beijing 102617, China)
出处
《北京石油化工学院学报》
2016年第4期61-66,共6页
Journal of Beijing Institute of Petrochemical Technology
基金
北京石油化工学院优秀责任教授资助项目(BIPTPOPME-2015)
2016年北京市大学生科研训练项目(2016J00087)
关键词
图像识别
大津阈值
静态背景差分
分水岭算法
image recognition
Otsu
static background subtraction
watershed algorithm
