• Acta Optica Sinica
  • Vol. 43, Issue 20, 2028001 (2023)
Qingxuan Zeng1, Qiang Li1,*, and Weizhi Nie2
Author Affiliations
  • 1School of Microelectronics, Tianjin University, Tianjin 300072, China
  • 2School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China
  • show less
    DOI: 10.3788/AOS230839 Cite this Article Set citation alerts
    Qingxuan Zeng, Qiang Li, Weizhi Nie. Lidar SLAM Algorithm Based on Online Point Cloud Removal in Pseudo Occupied Area[J]. Acta Optica Sinica, 2023, 43(20): 2028001 Copy Citation Text show less
    Overall framework of DOR-LOAM system
    Fig. 1. Overall framework of DOR-LOAM system
    Schematic of pseudo occupancy region division
    Fig. 2. Schematic of pseudo occupancy region division
    Schematic of dynamic point cloud memory area formed by point cloud matching. (a) A priori map; (b) current frame
    Fig. 3. Schematic of dynamic point cloud memory area formed by point cloud matching. (a) A priori map; (b) current frame
    Schematic of sliding window mode of a priori map
    Fig. 4. Schematic of sliding window mode of a priori map
    Sequence 00 and sequence 07 dynamic point cloud removal effect for different algorithms. (a) Original iamges; (b) Removert algorithm; (c) ERASOR algorithm; (d) DOR-LOAM algorithm
    Fig. 5. Sequence 00 and sequence 07 dynamic point cloud removal effect for different algorithms. (a) Original iamges; (b) Removert algorithm; (c) ERASOR algorithm; (d) DOR-LOAM algorithm
    Detail of point cloud removal effect of sequence 05 for different algorithms. (a) Ground truth; (b) ERASOR algorithm; (c) DOR-LOAM algorithm
    Fig. 6. Detail of point cloud removal effect of sequence 05 for different algorithms. (a) Ground truth; (b) ERASOR algorithm; (c) DOR-LOAM algorithm
    Comparison of tracks from sequence 00 to sequence 09
    Fig. 7. Comparison of tracks from sequence 00 to sequence 09
    Comparison of dynamic removal rate parameters
    Fig. 8. Comparison of dynamic removal rate parameters
    Comparison of running time of each module
    Fig. 9. Comparison of running time of each module
    SequenceEvaluation indexRemovertERASORDOR-LOAM
    00RP85.5093.9897.29
    RR99.3597.0897.06
    F191.9095.5097.18
    01RP94.2291.4996.45
    RR93.6195.3894.87
    F193.9193.4095.65
    02RP76.3287.7384.27
    RR96.8097.0197.30
    F185.3592.1090.32
    05RP86.9088.7395.33
    RR87.8898.2697.53
    F187.3993.3096.42
    07RP80.6990.6297.30
    RR98.8299.2798.80
    F188.8494.8098.04
    AverageRP84.72690.5194.13
    RR95.29297.4097.11
    F189.47893.8295.52
    Table 1. Comparison of dynamic point cloud removal accuracy for different algorithms
    SequenceE-LOAMT-LOAMNDT-LOAMPSF-LOSuMa++DOR-LOAM
    E¯trans /%

    E¯rot /

    [(°)·m-1

    E¯trans /%

    E¯rot /

    [(°)·m-1

    E¯trans /%E¯trans /%E¯trans /%

    E¯rot /

    [(°)·m-1

    E¯trans /%

    E¯rot/

    [(°)·m-1

    001.170.00470.980.00600.790.640.640.00220.820.0028
    012.920.00622.090.00521.461.321.600.00461.830.0050
    022.360.00821.010.00391.090.871.000.00371.110.0036
    031.160.00681.100.00820.650.750.670.00460.650.0043
    041.390.00500.680.00680.310.660.370.00260.590.0038
    050.820.00360.550.00320.540.450.400.00200.540.0023
    061.340.00570.560.00310.560.470.460.00210.480.0023
    071.210.00660.500.00470.270.460.340.00190.370.0025
    081.630.00630.940.00331.040.941.100.00350.930.0030
    091.360.00570.800.00400.740.560.470.00230.710.0028
    101.840.00631.120.00611.120.540.660.00280.920.0037
    Average1.560.00590.930.00490.900.740.700.00290.810.0033
    Table 2. Accuracy comparison of different algorithms
    AlgorithmBaselineBaseline+DORE-LOAMT-LOAMNDT-LOAM
    E¯trans /%1.280.811.560.930.90

    E¯rot /

    [(°)·m-1

    0.00440.00330.00590.0049-
    Average time /ms76.3387.4877.38100-
    Table 3. Comparison of ablation experiments and different algorithms