摘 要:自适应K-means++ 被用于提取线束连接器的主色特征值。首先对相机采集的图像进行去噪和增强对比度的操作;然后进行图像灰度化处理,通过大津阈值法分离线束主体与背景;再根据Canny 算子提取每根导线的轮廓,由轮廓位置获取增强图像对应的图像块;最后通过线宽选取聚类区域并利用自适应K-means++ 提取主色特征值。实验通过中位切分法、K-means及自适应K-means++ 分别提取特征值,并与人眼视觉观测的特征值进行色差对比。实验表明自适应K-means++ 方法提取的特征值较准确。
关键词:线束连接器;主色特征值;自适应K-means++;色差
中图分类号:TP391.41 文献标识码:A 文章编号:2096-4706(2021)01-0071-06
Research on Line Order Eigenvalue Extraction Method Based on Adaptive K-means++
HUI Wanyu,WU Yuxiu,ZHANG Wenzhong
(School of Electrical and Information Engineering,Anhui University of Techology,Ma’anshan 243032,China)
Abstract:Adaptive K-means++ is used to extract the dominant color eigenvalue of the harness connector. First,denoise and enhance the contrast of the image collected by the camera;then the image is grayed,and the main body and background are separated by Otsu threshold method;then,the contour of each wire is extracted according to Canny operator,and the image block corresponding to the enhanced image is obtained from the contour position;finally,the clustering region is selected by linewidth and the dominant color eigenvalue is extracted by adaptive K-means++. The experiment uses median segmentation,K-means and adaptive K-means++ to extract eigenvalue,and compare them with the eigenvalue observed by human vision. Experiments show that the eigenvalue extracted by the adaptive K-means++ method are more accurate.
Keywords:harness connector;dominant color eigenvalue;adaptive K-means++;chromatic aberration
参考文献:
[1] 范冬冬.基于机器视觉的双排线束检测仪的研究与设计 [D].马鞍山:安徽工业大学,2019.
[2] 吴晖辉,曾宪荣,赖燕君,等. 基于机器视觉的线束连接器字符检测 [J]. 测试技术学报,2018,32(2):174-179.
[3] SHI G M,JIAN W. Wiring harness assembly detectionsystem based on image processing technology [C]//2011 InternationalConference on Electronics,Communications and Control (ICECC).Ningbo:IEEE,2011:2397-2400.
[4] 吴宗胜,薛茹. 基于颜色聚合向量的线序检测方法 [J]. 计算机测量与控制,2019,27(6):182-185.
[5] 石坤泉,魏文国. 采用双边滤波的冷轧铝板表面缺陷图像去噪方法的研究 [J]. 表面技术,2018,47(9):317-323.
[6] 丰远远,高贤君,杨元维,等. 改进对数变换联合局部增强的高分遥感影像阴影自动补偿方法 [J]. 激光与光电子学进展,2020,57(20):67-75.
[7] 刘美,薛新松,刘广文,等. 对比度增强的彩色图像灰度化算法 [J]. 长春理工大学学报(自然科学版),2018,41(5):70-74.
[8] 陈雪鑫,卜庆凯. 基于改进的最大类间方差法的水果图像识别研究 [J]. 青岛大学学报(工程技术版),2019,34(2):33-38+62.
[9] 郑健,王继,宋世铭.Canny 双阈值算子在边缘提取中的优势 [J]. 地理空间信息,2019,17(11):128-130+12.
[10] CHEN G L,WANG G W,JU J,et al. Research on theInfluence of Kmeans Cluster Preprocessing on Adversarial Images [C]//Proceedings of the 2019 7th International Conference on InformationTechnology:IoT and SmartCity.New York:Association forComputing Machinery,2019:248–252.
[11] KUBO Y,NII M,MUTO T,et al. Artificial humeral headmodeling using Kmeans++ clustering and PCA [C]//2020 IEEE 2ndGlobal Conference on Life Sciences and Technologies (LifeTech).Kyoto:IEEE,2020:5-7.
[12] 朱连江,马炳先,赵学泉. 基于轮廓系数的聚类有效性分析 [J]. 计算机应用,2010,30(S2):139-141+198.
[13] 范鹏飞,孙俊. 基于机器视觉的色差检测算法 [J]. 传感器与微系统,2016,35(10):120-122+126.
作者简介:
惠婉玉(1996—),女,汉族,安徽宿州人,硕士研究生在读,研究方向:机器视觉与图像处理;
吴玉秀(1982—),男,汉族,河南安阳人,讲师,博士,研究方向:机器视觉与图像检测;
张文忠(1997—),男,汉族,安徽桐城人,硕士研究生在读,研究方向:机器人调度。