Infrared and Laser Engineering, Volume. 54, Issue 6, 20240478(2025)

Complex infrared background motion compensation algorithm based on LGB descriptor

Jinheng JING, Xiangbin LING, Yuhao XIONG, Wenyuan QI*, and Jian LI
Author Affiliations
  • Shanghai Aerospace Control Technology Research Institute, Shanghai 201109, China
  • show less
    Figures & Tables(19)
    Flowchart of the algorithm in this paper
    Schematic diagram of the construction of circular regions by the FAST algorithm
    Schematic diagram of the principle of the quadtree algorithm
    Schematic diagram of the LGB descriptor generation process
    Schematic diagram of location descriptor generation process
    Schematic diagram of grayscale descriptor generation process
    Schematic diagram of BEBLID descriptor generation process[16]
    Region partitioning in block matching
    Six sequences of experiment in this paper
    (a) Distribution of feature points without any manipulation; (b) The first 100 feature points with the largest response values based on (a); (c) Distribution of feature points after manipulation using the quadtree algorithm based on (a) (The number of retained feature points is 100)
    Difference map after background motion compensation for six image sequences with different algorithms
    Schematic diagram of different blocking strategies
    Performance efficiency analysis of different matching strategies
    (a) Original image; (b) Using MPCM algorithm on the original image; (c) Using MPCM algorithm on the frame difference image without any processing; (d) Using background motion compensation algorithm on the frame difference image after background motion compensation
    • Table 1. Details of each image sequence

      View table
      View in Article

      Table 1. Details of each image sequence

      SequenceNumber of framesResolutionScene descriptionNumber of targets per frameAverage target speed/pixel·frame–1
      1500256×256Complex ground clutter10.3
      2500256×256Complex hill and valley10.3
      3399256×256Complex ground interference11.2
      41599256×256Complex ground-to-air background10.2
      5500256×256Complex forest ravines10.2
      6751256×256Complex building interference10.5
    • Table 2. Individual algorithm details

      View table
      View in Article

      Table 2. Individual algorithm details

      AlgorithmFeature point detection algorithmFeature point rejection algorithmDescriptorMatching methodDescriptor sizeMismatch rejection algorithm
      OursFASTQuadtreeLGDBlock matching64 bitRANSAC
      Ours*FASTQuadtreeLGDPoint-by-point matching64 bitRANSAC
      ORBFASTHarris response value sortingBRIEFPoint-by-point matching64 bitRANSAC
      FREAKFASTQuadtreeFREAKPoint-by-point matching64 bitRANSAC
      BRIEFFASTQuadtreeBRIEFPoint-by-point matching64 bitRANSAC
      BEBLIDFASTQuadtreeBEBLIDPoint-by-point matching64 bitRANSAC
    • Table 3. Correct matching rate under different feature point distribution conditions in Fig.10

      View table
      View in Article

      Table 3. Correct matching rate under different feature point distribution conditions in Fig.10

      FigureSequence
      Sequence1Sequence2Sequence3Sequence4Sequence5Sequence6
      Fig.10(a)85%75%83%64%81%62%
      Fig.10(b)89%82%87%72%85%79%
      Fig.10(c)92%88%91%93%89%87%
    • Table 4. Comparison of results for each indicator across different algorithms

      View table
      View in Article

      Table 4. Comparison of results for each indicator across different algorithms

      SequencesIndicatorsAlgorithms
      OursOurs*ORBFREAKBIREFBEBLID
      Sequence 1OLE3.64473.64715.37483.77413.91985.8128
      NFAM0.21260.20210.38320.29210.27950.3547
      Sequence 2OLE2.85072.84834.13212.98183.20113.7001
      NFAM0.67470.72790.91470.76110.81530.7747
      Sequence 3OLE3.65593.62475.09713.64013.79815.2493
      NFAM0.04810.04790.11370.05470.07530.1468
      Sequence 4OLE2.33612.34354.41612.79072.98092.6958
      NFAM0.11020.11050.32050.15010.15470.1684
      Sequence 5OLE4.04384.13654.29914.46134.39514.3506
      NFAM0.23620.24190.40110.31870.28560.3514
      Sequence 6OLE3.91314.01584.07344.26154.22664.1692
      NFAM0.53620.55810.74610.66970.61310.6966
      Average correct matching rate of the six sequences94%90%76%86%83%78%
    • Table 5. Comparison of time taken of different algorithms

      View table
      View in Article

      Table 5. Comparison of time taken of different algorithms

      Time consumed/msAlgorithms
      OursOurs*ORBFREAKBRIEFBEBLID
      Feature point extraction time6.1566.15616.3526.1566.1566.156
      Descriptor generation time2.2092.2093.8615.5863.8613.535
      Matching time8.01519.20119.31519.17619.20519.215
      Total time16.38027.56639.52830.91829.22228.906
    Tools

    Get Citation

    Copy Citation Text

    Jinheng JING, Xiangbin LING, Yuhao XIONG, Wenyuan QI, Jian LI. Complex infrared background motion compensation algorithm based on LGB descriptor[J]. Infrared and Laser Engineering, 2025, 54(6): 20240478

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Optical imaging, display and information processing

    Received: Oct. 22, 2024

    Accepted: --

    Published Online: Jul. 1, 2025

    The Author Email: Wenyuan QI (qwy77687@163.com)

    DOI:10.3788/IRLA20240478

    Topics