基于斑点检测的钢管堆垛自动计数方法  

Automated Pipe Counting Methods Based on Blob Detection

作　　者：曹潇 郁云[2] Cao Xiao;Yu Yun(New Energy Research Center,China Electric Power Research Institute Co.,Ltd,Beijing 100192,China;School of Digital Commerce,Nanjing Vocational College of Information Technology,Nanjing 210023,China)

机构地区：[1]中国电力科学研究院有限公司新能源研究中心,北京100192 [2]南京信息职业技术学院数字商务学院,南京210023

出　　处：《信息化研究》2021年第4期31-35,共5页INFORMATIZATION RESEARCH

摘　　要：针对待检测的钢管数量大、背景杂乱、拍摄角度和光照条件不一致、钢管形状和堆垛不整齐等复杂场景,在研究分析现有方法不足的基础上,文章首先提出了利用支持向量机(SVM)算法进行图像阈值估计,然后利用斑点检测对目标图像中的钢管进行识别的方法。现场图像进行识别实验的结果证明,对于复杂场景下的钢管识别,此算法具有很高的精度和鲁棒性。相对于现有的方法,本算法具有更好的识别效果和工程实用性,具有很高的推广意义。Aiming to work in complicated scenes such as large amount of circle containing in a picture,background clutter,uneven illumination,different shooting angles,a synthesized algorithm was proposed based on the former researches.In the algorithm,the support vector machines(SVM)was exploited to estimate detection thresholds which were then input into blob detection algorithm to complete pipe counting.The results of detection tests for on-site picture proved that,for the pipe detection in complicated scenes,this algorithm showed its high accuracy and robustness.Compared with existing methods,this algorithm tends to provide better detecting results and is more applicable in practices,so it can be popularized.

关 键 词：图像处理 阈值估计 斑点检测 轮廓提取 支持向量机 钢管计数 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]