• Laser & Optoelectronics Progress
  • Vol. 59, Issue 14, 1415016 (2022)
Junhui Ge1, Jian Wang1, Yiping Peng1, Jiexuan Li1..., Changyan Xiao1,** and Yong Liu2,*|Show fewer author(s)
Author Affiliations
  • 1College of Electrical and Information Engineering, Hunan University, Changsha 410082, Hunan , China
  • 2Zhejiang Tongji Vocational College of Science and Technology, Hangzhou 311231, Zhejiang , China
  • show less
    DOI: 10.3788/LOP202259.1415016 Cite this Article Set citation alerts
    Junhui Ge, Jian Wang, Yiping Peng, Jiexuan Li, Changyan Xiao, Yong Liu. Recognition Method for Spray-Painted Workpieces Based on Mask R-CNN and Fast Point Feature Histogram Feature Pairing[J]. Laser & Optoelectronics Progress, 2022, 59(14): 1415016 Copy Citation Text show less
    Algorithm framework of workpiece recognition combining Mask R-CNN and 3D feature paring
    Fig. 1. Algorithm framework of workpiece recognition combining Mask R-CNN and 3D feature paring
    Results of Mask R-CNN instance segmentation. (a) Example of correct identification; (b) example of misidentification
    Fig. 2. Results of Mask R-CNN instance segmentation. (a) Example of correct identification; (b) example of misidentification
    Topological structure of affine consistence
    Fig. 3. Topological structure of affine consistence
    Workpiece samples
    Fig. 4. Workpiece samples
    Evaluation index
    Fig. 5. Evaluation index
    Recognition results with different workpieces
    Fig. 6. Recognition results with different workpieces
    Comparison with different methods
    Fig. 7. Comparison with different methods
    ProcessDetonating cover 1Upper plate 1Upper plate 2
    Original image
    Instance segmentation
    3D point cloud
    Filtering preprocessing
    Extraction of ISS key points
    Recognition based on featurepairs verification
    Table 1. Process and result of recognition with different workpiece
    ParameterPrRcF1
    Value99.1099.4399.26
    Table 2. Recognition results
    WorkpieceDimension /(m×m×m)Num /104Coarse /msFine /msTime /ms
    Detonating cover 10.85×0.82×0.5011.903158871202
    Weld holder2.40×0.65×0.0317.6533210641396
    Upper plate 20.75×0.75×1.355.92310524834
    Table 3. Running time
    Junhui Ge, Jian Wang, Yiping Peng, Jiexuan Li, Changyan Xiao, Yong Liu. Recognition Method for Spray-Painted Workpieces Based on Mask R-CNN and Fast Point Feature Histogram Feature Pairing[J]. Laser & Optoelectronics Progress, 2022, 59(14): 1415016
    Download Citation