基于混合特征的RGB-D数据初始配准方法
|
摘要
针对彩色物体的配准问题,提出一种面向RGB-D数据的初始配准方法.通过几何和颜色的邻域信息构建混合特征,并根据混合特征在源点云中获取物体的特征点;由归一化后的颜色值和混合特征构造特征描述符,通过该特征描述符搜索对应点;再由分块对齐策略,进一步剔除相似性较小的点对,将剩余特征点进行分块配准,选择最优的刚性变换.为了验证该初始配准方法的有效性,通过精确配准算法进一步细化配准结果.实验结果表明,基于混合特征的RGB-D数据初始配准方法对于两片部分重叠点云配准是有效的.
Aiming at the registration problem of colored objects,a coarse registration method for RGB-D data is proposed.The mixed feature is constructed by the neighborhood information of geometry and color.The feature points of the object are obtained from the source point cloud according to the mixed feature.The feature descriptor is constructed from the normalized color value and the mixed feature.The corresponding points are searched by feature descriptors.Then,a strategy of partition alignment is proposed.The point pairs with smaller similarity are further eliminated.The partition alignment with the remaining feature points are performed and the optimal rigid transformation is selected.In order to verify the validity of the coarse registration method,the registration results are further refined by the fine registration algorithm.The experimental results show that the coarse registration method of RGB-D data based on mixed features is effective for two partial overlapping point cloud registration.
Aiming at the registration problem of colored objects,a coarse registration method for RGB-D data is proposed.The mixed feature is constructed by the neighborhood information of geometry and color.The feature points of the object are obtained from the source point cloud according to the mixed feature.The feature descriptor is constructed from the normalized color value and the mixed feature.The corresponding points are searched by feature descriptors.Then,a strategy of partition alignment is proposed.The point pairs with smaller similarity are further eliminated.The partition alignment with the remaining feature points are performed and the optimal rigid transformation is selected.In order to verify the validity of the coarse registration method,the registration results are further refined by the fine registration algorithm.The experimental results show that the coarse registration method of RGB-D data based on mixed features is effective for two partial overlapping point cloud registration.
引文
[1] GUO Yulan,SOHEL F,BENNAMOUN M,et al.A novel local surface feature for 3Dobject recognition under clutter and occlusion[J].Inform Sciences,2015,293:196-213.
[2]赵夫群,周明全,耿国华,等.颅骨点云模型的局部特征配准方法[J].中国图象图形学报,2017,22(8):1120-1127.
[3] MEDAN G,SHAMUL N,JOSKOWICZ L.Sparse 3Dradon space rigid registration of CT scans:method and validation study[J].IEEE Trans Med Imaging,2017,36(2):497-509.
[4] BESL P J,MCKAY N D.A method for registration of 3Dshapes[J].IEEE Trans Patt Anal Mach Int,1992,14(2):239-256.
[5]薛耀红,梁学章,马婷,等.扫描点云的一种自动配准方法[J].计算机辅助设计与图形学学报,2011,23(2):223-231.
[6]韩宝昌,曹俊杰,苏志勋.一种区域层次上的自动点云配准算法[J].计算机辅助设计与图形学学报,2015,27(2):313-319.
[7] FISCHLER M A,BOLLES R C.Random sample consensus:aparadigm for model fitting with applications to image analysis and automated cartography[J].Comm ACM,1981,24(6):381-395.
[8] AIGER D,MITRA N J,COHEN-OR D.4-Points congruent sets for robust pairwise surface registration[J].ACM Trans Graphic,2008,27(3):1-10.
[9] YANG Jiaolong,LI Hongdong,CAMPBELL D,et al.Go-ICP:aglobally optimal solution to 3DICP point-set registration[J].IEEE Trans Patt Anal Mach Int,2016,38(11):2241-2254.
[10] YANG Jiaqi,CAO Zhiguo,ZHANG Qian.A fast and robust local descriptor for 3Dpoint cloud registration[J].Inform Sciences,2016,346:163-179.
[11] ZENG A,SONG Shuran,NIENER M,et al.3D-match:learning local geometric descriptors from RGB-D reconstructions[C]//2017IEEE Conference on Computer Vision and Pattern Recognition,Honolulu,Hawaii,IEEE,2017:199-208.
[12] PAULY M,GROSS M,KOBBELT L P.Efficient simplification of point-sampled surfaces[C]//2002Conference on IEEE Visualization,Boston,Massachusetts:IEEE,2002:163-170.
[2]赵夫群,周明全,耿国华,等.颅骨点云模型的局部特征配准方法[J].中国图象图形学报,2017,22(8):1120-1127.
[3] MEDAN G,SHAMUL N,JOSKOWICZ L.Sparse 3Dradon space rigid registration of CT scans:method and validation study[J].IEEE Trans Med Imaging,2017,36(2):497-509.
[4] BESL P J,MCKAY N D.A method for registration of 3Dshapes[J].IEEE Trans Patt Anal Mach Int,1992,14(2):239-256.
[5]薛耀红,梁学章,马婷,等.扫描点云的一种自动配准方法[J].计算机辅助设计与图形学学报,2011,23(2):223-231.
[6]韩宝昌,曹俊杰,苏志勋.一种区域层次上的自动点云配准算法[J].计算机辅助设计与图形学学报,2015,27(2):313-319.
[7] FISCHLER M A,BOLLES R C.Random sample consensus:aparadigm for model fitting with applications to image analysis and automated cartography[J].Comm ACM,1981,24(6):381-395.
[8] AIGER D,MITRA N J,COHEN-OR D.4-Points congruent sets for robust pairwise surface registration[J].ACM Trans Graphic,2008,27(3):1-10.
[9] YANG Jiaolong,LI Hongdong,CAMPBELL D,et al.Go-ICP:aglobally optimal solution to 3DICP point-set registration[J].IEEE Trans Patt Anal Mach Int,2016,38(11):2241-2254.
[10] YANG Jiaqi,CAO Zhiguo,ZHANG Qian.A fast and robust local descriptor for 3Dpoint cloud registration[J].Inform Sciences,2016,346:163-179.
[11] ZENG A,SONG Shuran,NIENER M,et al.3D-match:learning local geometric descriptors from RGB-D reconstructions[C]//2017IEEE Conference on Computer Vision and Pattern Recognition,Honolulu,Hawaii,IEEE,2017:199-208.
[12] PAULY M,GROSS M,KOBBELT L P.Efficient simplification of point-sampled surfaces[C]//2002Conference on IEEE Visualization,Boston,Massachusetts:IEEE,2002:163-170.