• Laser & Optoelectronics Progress
  • Vol. 57, Issue 4, 041017 (2020)
Jianfeng Wang1、*, Hongwei Wang1、2、**, and Xueqin Yan1、***
Author Affiliations
  • 1School of Electrical Engineering, Xinjiang University, Urumqi, Xinjiang 830047, China
  • 2School of Control Science and Engineering, Dalian University of Technology, Dalian, Liaoning 116024, China
  • show less
    DOI: 10.3788/LOP57.041017 Cite this Article Set citation alerts
    Jianfeng Wang, Hongwei Wang, Xueqin Yan. Fundamental Matrix Estimation Based on Multiple Kernel Learning-Density Peak Clustering[J]. Laser & Optoelectronics Progress, 2020, 57(4): 041017 Copy Citation Text show less
    Epipolar geometry
    Fig. 1. Epipolar geometry
    Experimental results of basic density peak algorithm(dc=1.8). (a) Decision graph; (b) clustering results
    Fig. 2. Experimental results of basic density peak algorithm(dc=1.8). (a) Decision graph; (b) clustering results
    Experimental results of multi-kernel learning-density peak algorithm. (a) Decision graph; (b) clustering results
    Fig. 3. Experimental results of multi-kernel learning-density peak algorithm. (a) Decision graph; (b) clustering results
    Distribution map of γ
    Fig. 4. Distribution map of γ
    Raster division diagram
    Fig. 5. Raster division diagram
    Flow chart of basic matrix estimation
    Fig. 6. Flow chart of basic matrix estimation
    Average epipolar distance of four methods under different Gaussian variance noise
    Fig. 7. Average epipolar distance of four methods under different Gaussian variance noise
    Average epipolar distance of four methods at different external point rates
    Fig. 8. Average epipolar distance of four methods at different external point rates
    Experimental images. (a) Graffiti; (b) tree; (c) UBC; (d) bike; (e) fruit; (f) INRIA;(g) building; (h) snow-tree
    Fig. 9. Experimental images. (a) Graffiti; (b) tree; (c) UBC; (d) bike; (e) fruit; (f) INRIA;(g) building; (h) snow-tree
    PerformanceindicatorAlgorithmGraffitiTreeUBCBikeFruitINRIABuildingSnow-tree
    Total numberof matchedpointsSURF84722311216700289345576264
    Number of matchedpoints forestimatingfundamental matrixLMedS4241116608350146173288132
    R-RANSAC18567027157313711535
    LO-RANSAC712311722111125020
    Proposed1322281783920254926
    AverageresidualerrorLMedS8.66211.37382.31533.63932.61011.80342.66501.6481
    R-RANSAC7.27441.04931.21012.24661.04661.16323.96131.2221
    LO-RANSAC1.38970.30080.82000.49270.28260.45190.10521.3596
    Proposed1.05130.25380.70170.38590.21830.39720.08441.1466
    Averageepipolardistance /pixelLMedS0.38090.24660.35860.36510.29810.31520.39850.6759
    R-RANSAC0.15110.13620.14610.17120.10260.14810.13240.1192
    LO-RANSAC0.05880.09610.07810.09020.07180.06600.07520.0956
    Proposed0.04140.08300.06910.07820.06730.05920.06440.0868
    Computationaltime /sLMedS0.14810.21760.28170.17740.10530.12280.17360.1378
    R-RANSAC0.16590.23510.41600.14050.08120.13690.16930.1596
    LO-RANSAC0.13020.17700.09060.12350.04430.10430.13960.1206
    Proposed0.14930.20600.09750.10290.04270.08710.09060.1415
    Table 1. Comparison of four algorithms performance
    Jianfeng Wang, Hongwei Wang, Xueqin Yan. Fundamental Matrix Estimation Based on Multiple Kernel Learning-Density Peak Clustering[J]. Laser & Optoelectronics Progress, 2020, 57(4): 041017
    Download Citation