摘要
采用一种改进密度空间聚类算法进行巡检图像分割,算法采用简单线性迭代聚类(SLIC)进行预处理,有效降低了内存消耗并提高了运行效率,同时有效改善了图像中目标边界作为背景来处理的问题;算法采用斜交空间距离作为距离度量,充分考虑变量间的相关性,改善了高维聚类不准确的问题.改进算法与DBSCAN对比实验表明:改进算法的聚类结果能有效分离目标和背景,保持边缘完整和连续,运行效率与聚类准确性有很大的改善,可以对巡检图像进行更有效的分割.
An improved density clustering image segmentation algorithm was mentioned in this paper. Through pre-processing by simple linear iterative clustering (SLIC),this algorithm had much better efficiency and made better use of memory.And the border of the targets in the image was more prop-erly processed.The algorithm improved the precision of the high-dimensional clustering by spatial dis-tance measure and considering the relation of variables.Compared with DBSCAN,the new algorithm improves the accuracy of recognizing the target from image′s background and keeps intact of the im-age.And both of the algorithm efficiency and clustering accuracy are improved and get a better result than the traditional way.
出处
《华中科技大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2015年第S1期473-476,共4页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金青年基金资助项目(61105083)
新世纪优秀人才支持计划资助项目(NCET-11-0634)
北京市共建项目(GJ2013005)