采用轮廓片段空间关系实现遮挡目标识别
|
摘要
为了实现高比例遮挡情况下的目标识别,提出一种基于轮廓片段空间关系的目标识别算法.首先,在采用轮廓的形状上下文特征进行粗匹配的基础上,对模板图像和待识别图像分别进行图像骨架关键接合点的提取和轮廓形状质心的提取.然后,以图像像素中心点为原点建立坐标系,以图像骨架关键接合点和轮廓片段质心在各自图像建立的坐标系内的位置确定空间关系.最后,制定空间关系参数约束标准,筛选满足空间关系约束准则的目标库图像为最后识别结果.与现有遮挡目标匹配算法相比,该算法可以实现高比例遮挡情况下的目标识别,在目标遮挡比例为60%的情况下,识别率可达到78%.
In order to realize the target recognition in the case of high proportion occlusion,a target recognition algorithm based on the spatial relationship of contour segments was proposed.Firstly,based on rough matching using contour context features of the contour,the image skeleton key joints and contour shape centroids were extracted respectively from the template image and the image to be recognized.Then,the coordinate system was established with the image pixel center point as the origin.The spatial relationship was determined by the position of the image skeleton key joint and the contour segment centroid in the coordinate system established by the respective image.Finally,the spatial relation parameter constraint criteria were formulated,and the target library image satisfying the spatial relationship constraint criterion was selected as the final recognition result.Compared with the existing occlusion target matching algorithm,the proposed algorithm can realize the targets recognition with high proportion of occlusion,and when the target occlusion ratio is 60%,the recognition rate can reach 78%.
In order to realize the target recognition in the case of high proportion occlusion,a target recognition algorithm based on the spatial relationship of contour segments was proposed.Firstly,based on rough matching using contour context features of the contour,the image skeleton key joints and contour shape centroids were extracted respectively from the template image and the image to be recognized.Then,the coordinate system was established with the image pixel center point as the origin.The spatial relationship was determined by the position of the image skeleton key joint and the contour segment centroid in the coordinate system established by the respective image.Finally,the spatial relation parameter constraint criteria were formulated,and the target library image satisfying the spatial relationship constraint criterion was selected as the final recognition result.Compared with the existing occlusion target matching algorithm,the proposed algorithm can realize the targets recognition with high proportion of occlusion,and when the target occlusion ratio is 60%,the recognition rate can reach 78%.
引文
[1]史思琦,石光明,齐飞.基于特征完整描述的部分遮挡目标识别算法[J].系统工程与电子技术,2011,33(4):913-918.
[2]JIA Q,FAN X,LIU Y,et al.Hierarchical projective invariant contexts for shape recognition[J].Pattern Recognition,2016,52:358-374.
[3]代沅兴,于天彪,杨建宇,等.一种新的基于轮廓特征的指尖及手指指向检测方法[J].东北大学学报(自然科学版),2018,39(7):937-941.
[4]李平,魏仲慧,何昕,等.采用多形状特征融合的多视点目标识别[J].光学精密工程,2014,22(12):3368-3376.
[5]王波,薛方正,李祖枢.一种基于联合轮廓特征矢量的目标识别方法[J].传感器与微系统,2011,30(7):64-67.
[6]杨亚飞,郑丹晨,韩敏.一种基于多尺度轮廓点空间关系特征的形状匹配方法[J].自动化学报,2015,41(8):1405-1411.
[7]张天行,段锦,祝勇,等.一种改进的特征点匹配局部遮挡目标识别算法[J].计算机系统应用,2014,23(3):158-161.
[8]史思琦,石光明,李甫.基于轮廓特征多层描述和评价的部分遮挡目标匹配[J].光学精密工程,2012,20(12):2804-2811.
[9]BELONGIE S,MALIK J,PUZICHA J.Shape matching and object recognition using shape contexts[J].IEEETransactions on Pattern Analysis&Machine Intelligence,2006,4170(4):483-507
[10]黄伟国,胡大盟,杨剑宇,等.用于遮挡形状匹配的弦角特征描述[J].光学精密工程,2015,23(6):1758-1767.
[11]郑加宽.基于骨架特征的形状识别方法研究[D].南昌航空大学图书馆,2014
[12]KROLUPPER F,FLUSSER J.Polygonal shape descripttion for recognition of partially occluded objects[J].Patte-rn Recognition Letters,2007,28(9):1002-1011.
[13]SUN K B,SUPER B J.Classification of contour shapes using class segment sets[C]//Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition.San Diego:IEEE,2005:727-733.
[14]黄伟国,顾超,尚丽,等.基于轮廓分层描述的目标识别算法研究[J].电子学报,2015,43(5):854-861.
[2]JIA Q,FAN X,LIU Y,et al.Hierarchical projective invariant contexts for shape recognition[J].Pattern Recognition,2016,52:358-374.
[3]代沅兴,于天彪,杨建宇,等.一种新的基于轮廓特征的指尖及手指指向检测方法[J].东北大学学报(自然科学版),2018,39(7):937-941.
[4]李平,魏仲慧,何昕,等.采用多形状特征融合的多视点目标识别[J].光学精密工程,2014,22(12):3368-3376.
[5]王波,薛方正,李祖枢.一种基于联合轮廓特征矢量的目标识别方法[J].传感器与微系统,2011,30(7):64-67.
[6]杨亚飞,郑丹晨,韩敏.一种基于多尺度轮廓点空间关系特征的形状匹配方法[J].自动化学报,2015,41(8):1405-1411.
[7]张天行,段锦,祝勇,等.一种改进的特征点匹配局部遮挡目标识别算法[J].计算机系统应用,2014,23(3):158-161.
[8]史思琦,石光明,李甫.基于轮廓特征多层描述和评价的部分遮挡目标匹配[J].光学精密工程,2012,20(12):2804-2811.
[9]BELONGIE S,MALIK J,PUZICHA J.Shape matching and object recognition using shape contexts[J].IEEETransactions on Pattern Analysis&Machine Intelligence,2006,4170(4):483-507
[10]黄伟国,胡大盟,杨剑宇,等.用于遮挡形状匹配的弦角特征描述[J].光学精密工程,2015,23(6):1758-1767.
[11]郑加宽.基于骨架特征的形状识别方法研究[D].南昌航空大学图书馆,2014
[12]KROLUPPER F,FLUSSER J.Polygonal shape descripttion for recognition of partially occluded objects[J].Patte-rn Recognition Letters,2007,28(9):1002-1011.
[13]SUN K B,SUPER B J.Classification of contour shapes using class segment sets[C]//Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition.San Diego:IEEE,2005:727-733.
[14]黄伟国,顾超,尚丽,等.基于轮廓分层描述的目标识别算法研究[J].电子学报,2015,43(5):854-861.