• Laser & Optoelectronics Progress
  • Vol. 57, Issue 6, 061005 (2020)
Zongjun Zhang and Fengbao Yang*
Author Affiliations
  • College of Information and Communication Engineering, North University of China, Taiyuan, Shanxi 030051, China
  • show less
    DOI: 10.3788/LOP57.061005 Cite this Article Set citation alerts
    Zongjun Zhang, Fengbao Yang. Road Extraction Algorithm for Remote Sensing Images Based on Improved Expectation-Maximization Clustering[J]. Laser & Optoelectronics Progress, 2020, 57(6): 061005 Copy Citation Text show less
    High-resolution remote sensing image
    Fig. 1. High-resolution remote sensing image
    Road extraction process. (a) Image mathematical morphology processing result; (b) image EM clustering result; (c) image contour detection result; (d) road extraction result
    Fig. 2. Road extraction process. (a) Image mathematical morphology processing result; (b) image EM clustering result; (c) image contour detection result; (d) road extraction result
    Processing results of experimental data 1. (a) Original picture; (b) image contour detection result; (c) road extraction result; (d) extraction results are superimposed on original image
    Fig. 3. Processing results of experimental data 1. (a) Original picture; (b) image contour detection result; (c) road extraction result; (d) extraction results are superimposed on original image
    Processing results of experimental data 2. (a) Original picture; (b) image contour detection result; (c) road extraction result; (d) extraction results are superimposed on original image
    Fig. 4. Processing results of experimental data 2. (a) Original picture; (b) image contour detection result; (c) road extraction result; (d) extraction results are superimposed on original image
    Extraction effects of traditional method. Extraction results of (a)data 1 and (b) data 2 obtained by Hongh transformation based method; extraction results of (c)data 1 and (d) data 2 obtained by watershed threshold segmentation combined with shape features
    Fig. 5. Extraction effects of traditional method. Extraction results of (a)data 1 and (b) data 2 obtained by Hongh transformation based method; extraction results of (c)data 1 and (d) data 2 obtained by watershed threshold segmentation combined with shape features
    Test set data and extraction results. (a) Test set data 1; (b) test set data 2; (c) test set data 3; (d) extraction result of test set data 1; (e) extraction result of test set data 2; (f) extraction result of test set data 3
    Fig. 6. Test set data and extraction results. (a) Test set data 1; (b) test set data 2; (c) test set data 3; (d) extraction result of test set data 1; (e) extraction result of test set data 2; (f) extraction result of test set data 3
    MethodP /%R /%D /%
    Hough transform57.60.8442.4
    Watershed threshold segmentation90.420.609.6
    Proposed method92.63.607.4
    Table 1. Evaluation results for data 1
    MethodP /%R /%D /%
    Hough transform62.40.9237.6
    Watershed threshold segmentation91.815.208.2
    Method of this paper92.54.307.5
    Table 2. Evaluation results for data 2
    Zongjun Zhang, Fengbao Yang. Road Extraction Algorithm for Remote Sensing Images Based on Improved Expectation-Maximization Clustering[J]. Laser & Optoelectronics Progress, 2020, 57(6): 061005
    Download Citation