• Acta Optica Sinica
  • Vol. 38, Issue 2, 0210003 (2018)
Fengguang Xiong, Wang Huo, Xie Han*, and Liqun Kuang
Author Affiliations
  • School of Computer Science and Technology, North University of China, Taiyuan, Shanxi 030051, China
  • show less
    DOI: 10.3788/AOS201838.0210003 Cite this Article Set citation alerts
    Fengguang Xiong, Wang Huo, Xie Han, Liqun Kuang. Removal Method of Mismatching Keypoints in 3D Point Cloud[J]. Acta Optica Sinica, 2018, 38(2): 0210003 Copy Citation Text show less
    Keypoint matching process of general method
    Fig. 1. Keypoint matching process of general method
    Keypoint matching process of proposed method
    Fig. 2. Keypoint matching process of proposed method
    Sketch map of intersection angle
    Fig. 3. Sketch map of intersection angle
    Examples of 3D point cloud. (a) Bunny'; (b) Bunny″; (c) Armadillo' ; (d) Armadillo″; (e) chef; (f) rs1; (g) rs4; (h) bun000; (i) bun045; (j) top2; (k) bun180
    Fig. 4. Examples of 3D point cloud. (a) Bunny'; (b) Bunny″; (c) Armadillo' ; (d) Armadillo″; (e) chef; (f) rs1; (g) rs4; (h) bun000; (i) bun045; (j) top2; (k) bun180
    State chart of 3D point cloud
    Fig. 5. State chart of 3D point cloud
    Matching graphs between Bunny' and Bunny″. (a) Matching by NN algorithm; (b) matching by RANSAC algorithm; (c) matching by proposed algorithm
    Fig. 6. Matching graphs between Bunny' and Bunny″. (a) Matching by NN algorithm; (b) matching by RANSAC algorithm; (c) matching by proposed algorithm
    Matching graphs between Armadillo' and Armadillo″. (a) Matching by NN algorithm; (b) matching by RANSAC algorithm; (c) matching by proposed algorithm
    Fig. 7. Matching graphs between Armadillo' and Armadillo″. (a) Matching by NN algorithm; (b) matching by RANSAC algorithm; (c) matching by proposed algorithm
    Matching graphs between chef and rs1. (a) Matching by NN algorithm; (b) matching by RANSAC algorithm; (c) matching by proposed algorithm
    Fig. 8. Matching graphs between chef and rs1. (a) Matching by NN algorithm; (b) matching by RANSAC algorithm; (c) matching by proposed algorithm
    Matching graphs between chef and rs4. (a) Matching by NN algorithm; (b) matching by RANSAC algorithm; (c) matching by proposed algorithm
    Fig. 9. Matching graphs between chef and rs4. (a) Matching by NN algorithm; (b) matching by RANSAC algorithm; (c) matching by proposed algorithm
    Matching graphs of bun045 and bun000. (a) Matching by NN algorithm; (b) matching by RANSAC algorithm; (c) matching by proposed algorithm
    Fig. 10. Matching graphs of bun045 and bun000. (a) Matching by NN algorithm; (b) matching by RANSAC algorithm; (c) matching by proposed algorithm
    Matching graphs of top2 and bun180. (a) Matching by NN algorithm; (b) matching by RANSAC algorithm; (c) matching by proposed algorithm
    Fig. 11. Matching graphs of top2 and bun180. (a) Matching by NN algorithm; (b) matching by RANSAC algorithm; (c) matching by proposed algorithm
    Point cloudNumber of original pointsNumber of edge pointsRate of edge points /%Number of initial keypointsNumber of final keypointsRatio of removal /%Time of removal /msEfficiency of removal /ms
    Bunny'3594710132.822713156442.355770.016
    Bunny″359478252.302508182027.436240.017
    Armadillo'1729748360.4877144342.5431200.018
    Armadillo″1729741880.117026566.5532320.019
    chef17692013140.7483567019.7637080.021
    rs11143641320311.54139429179.1232460.028
    rs4135666131139.67116233171.5133340.026
    bun0004025623455.83113550855.2414120.035
    bun0454009725146.2799334864.9513990.035
    top23829829757.7797533166.0513050.034
    bun1804025123935.95115146459.6913760.034
    Table 1. Information about points in the 3D point cloud and the time of removing keypoints on the edge
    Point cloud 1Point cloud 2AlgorithmNumber of correct matching pairsNumber of all matching pairsError of rotationError of translation
    Bunny'Bunny″NN25316.09814.364
    RANSAC330.1970.344
    Proposed16160.1810.109
    Armadillo'Armadillo″NN961122.1268.574
    RANSAC231.1071.668
    Proposed77770.2850.188
    Table 2. Matching results between complete 3D point cloud and complete 3D point cloud
    Point cloud 1Point cloud 2AlgorithmNumber of correct matching pairsNumber of all matching pairsError of rotationError of translation
    chefrs1NN1558104.0801447.870
    RANSAC0332.585426.013
    Proposed12121.48918.945
    chefrs4NN207288.5191245.380
    RANSAC039.911159.958
    Proposed993.16743.351
    Table 3. Matching results between complete 3D point cloud and 3D point cloud with clutter and occlusion
    Point cloud 1Point cloud 2AlgorithmNumber of correct matching pairsNumber of all matching pairsError of rotationError of translation
    bun045bun000NN49972.55405.385
    RANSAC037.900011.378
    Proposed20200.71600.962
    top2bun180NN126131.909075.135
    RANSAC2395.4290158.701
    Proposed670.52030.875
    Table 4. Matching results between partial 3D point cloud and partial 3D point cloud
    Point cloud 1Point cloud 2AlgorithmTime of keypoint matching /ms
    Bunny'Bunny″NN3343
    RANSAC9294
    Proposed4550
    Armadillo'Armadillo″NN10234
    RANSAC22680
    Proposed16692
    chefrs1NN15369
    RANSAC20570
    Proposed22332
    chefrs4NN15248
    RANSAC23031
    Proposed21298
    bun045bun000NN4699
    RANSAC14203
    Proposed6518
    top2bun180NN4413
    RANSAC10368
    Proposed7097
    Table 5. Comparison of matching time
    Fengguang Xiong, Wang Huo, Xie Han, Liqun Kuang. Removal Method of Mismatching Keypoints in 3D Point Cloud[J]. Acta Optica Sinica, 2018, 38(2): 0210003
    Download Citation