• Journal of Infrared and Millimeter Waves
  • Vol. 39, Issue 2, 263 (2020)
Liang HUANG1、2, Bing-Xiu YAO1、*, Peng-Di CHEN1, Ai-Ping REN1, and Yan XIA1
Author Affiliations
  • 1Faculty of Land Resource Engineering, Kunming University of Science and Technology, Kunming650093, China
  • 2Surveying and Mapping Geo-Informatics Technology Research Center on Plateau Mountains of Yunnan Higher Education, Kunming650093, China
  • show less
    DOI: 10.11972/j.issn.1001-9014.2020.02.014 Cite this Article
    Liang HUANG, Bing-Xiu YAO, Peng-Di CHEN, Ai-Ping REN, Yan XIA. Superpixel segmentation method of high resolution remote sensing images based on hierarchical clustering[J]. Journal of Infrared and Millimeter Waves, 2020, 39(2): 263 Copy Citation Text show less
    Flow chart of remote sensing image segmentation
    Fig. 1. Flow chart of remote sensing image segmentation
    Gradient distribution of experimental images(a) remote sensing image, (b) sobel gradient image, (c) gradient distribution, (d) remote sensing image, (e) sobel gradient image, (f) gradient distribution
    Fig. 2. Gradient distribution of experimental images(a) remote sensing image, (b) sobel gradient image, (c) gradient distribution, (d) remote sensing image, (e) sobel gradient image, (f) gradient distribution
    Segmentation results using AMR-WT by changing the value of s and m(a) s = 1, m = 1; (b) s = 2, m = 1; (c) s = 3, m = 1; (d) s = 4, m = 1; (e) S1s = 1, m =1; (f) S2s = 1, m = 6; (g) s = 1, m = 12; (h) s = 1, m = 20
    Fig. 3. Segmentation results using AMR-WT by changing the value of s and m(a) s = 1, m = 1; (b) s = 2, m = 1; (c) s = 3, m = 1; (d) s = 4, m = 1; (e) S1s = 1, m =1; (f) S2s = 1, m = 6; (g) s = 1, m = 12; (h) s = 1, m = 20
    Simulation graph of clustering(a)clustering simulation, (b)segmentation results, (c)clustering simulation, (d)segmentation results
    Fig. 4. Simulation graph of clustering(a)clustering simulation, (b)segmentation results, (c)clustering simulation, (d)segmentation results
    Test data sets (a) S1 imag, (b) S2 image, (c) S3 image, (d) S4 image, (e) S1 image, (f) S2 reference image, (g) S3 reference image, (h) S4 reference image
    Fig. 5. Test data sets (a) S1 imag, (b) S2 image, (c) S3 image, (d) S4 image, (e) S1 image, (f) S2 reference image, (g) S3 reference image, (h) S4 reference image
    Experimental results of over-segmentation(a) S1 proposed superpixel algorithm, (b) S1 SLIC algorithm, (c) S1 LSC algorithm, (d) S1 Meanshift algorithm, (e) S2 proposed superpixel algorithm, (f) S2 SLIC algorithm, (g) S2 LSC algorithm, (h) S2 MeanShift algorithm, (i) S3 proposed superpixel algorithm, (j) S3 SLIC algorithm, (k) S3 LSC algorithm, (l) S3 Meanshift algorithm, (m) S4 proposed superpixel algorithm, (n) S4 SLIC algorithm, (o) S4 LSC algorithm, (p) S4 MeanShift algorithm
    Fig. 6. Experimental results of over-segmentation(a) S1 proposed superpixel algorithm, (b) S1 SLIC algorithm, (c) S1 LSC algorithm, (d) S1 Meanshift algorithm, (e) S2 proposed superpixel algorithm, (f) S2 SLIC algorithm, (g) S2 LSC algorithm, (h) S2 MeanShift algorithm, (i) S3 proposed superpixel algorithm, (j) S3 SLIC algorithm, (k) S3 LSC algorithm, (l) S3 Meanshift algorithm, (m) S4 proposed superpixel algorithm, (n) S4 SLIC algorithm, (o) S4 LSC algorithm, (p) S4 MeanShift algorithm
    The time of over-segmentation
    Fig. 7. The time of over-segmentation
    Segmentation results of S1 and S2 (a) S1 FNEA 50, (b) S1 FNEA 100, (c) S1 SH method, (d) result of S1 by proposed method, (e) S2 FNEA 50, (f) S4 FNEA 100, (g) S2 SH method, (h) result of S2 by proposed method
    Fig. 8. Segmentation results of S1 and S2 (a) S1 FNEA 50, (b) S1 FNEA 100, (c) S1 SH method, (d) result of S1 by proposed method, (e) S2 FNEA 50, (f) S4 FNEA 100, (g) S2 SH method, (h) result of S2 by proposed method
    Segmentation results of Test1
    Fig. 9. Segmentation results of Test1
    Segmentation results of S3 and S4(a) S3 FNEA 50, (b) S3 FNEA 100, (c) S3 SH method, (d) S3 result by proposed method, (e) S4 FNEA 50, (f) S4 FNEA 100, (g) S4 SH method, (h) S4 result by proposed method
    Fig. 10. Segmentation results of S3 and S4(a) S3 FNEA 50, (b) S3 FNEA 100, (c) S3 SH method, (d) S3 result by proposed method, (e) S4 FNEA 50, (f) S4 FNEA 100, (g) S4 SH method, (h) S4 result by proposed method
    Segmentation results of Test2
    Fig. 11. Segmentation results of Test2
    序号传感器大小/像素分辨率
    S1QuickBird433×5500.61 m
    S2QuickBird2480×9750.61 m
    S3无人机726×4680.05 m
    S4无人机979×5860.05 m
    Table 1. Information of segmentation data sets
    Liang HUANG, Bing-Xiu YAO, Peng-Di CHEN, Ai-Ping REN, Yan XIA. Superpixel segmentation method of high resolution remote sensing images based on hierarchical clustering[J]. Journal of Infrared and Millimeter Waves, 2020, 39(2): 263
    Download Citation