предыдущий|следующий

Improved ROEWA edge detector for SAR Images

  • role： First author第一作者
  • Affiliation：

    State Key Laboratory of Geohazard Prevention and Geoenvironment Protection (Chengdu University of Technology), Chengdu 610059, China

    Key Laboratory of Geoscience Spatial Information Technology, Ministry of Land and Resources of the P. R. China, Chengdu 610059, China

    College of Earth Sciences, Chengdu University of Technology, Chengdu 610059, China

  • Email：liuhang326@163.com
  • Introduction：刘夯(1990—　)，男，硕士研究生，研究方向为微波遥感应用、计算机视觉。E-mail：liuhang326@163.com
LIU Hang123,  
  • Affiliation：

    State Key Laboratory of Geohazard Prevention and Geoenvironment Protection (Chengdu University of Technology), Chengdu 610059, China

    Key Laboratory of Geoscience Spatial Information Technology, Ministry of Land and Resources of the P. R. China, Chengdu 610059, China

    College of Earth Sciences, Chengdu University of Technology, Chengdu 610059, China

HE Zhengwei123,  
  • Affiliation：

    Key Laboratory of Geoscience Spatial Information Technology, Ministry of Land and Resources of the P. R. China, Chengdu 610059, China

    College of Tourism and Urban-Rural Planning, Chengdu University of Technology, Chengdu 610059, China

ZHAO Yinbing24,  
  • Affiliation：

    Key Laboratory of Geoscience Spatial Information Technology, Ministry of Land and Resources of the P. R. China, Chengdu 610059, China

TENG Jing2,  
  • Affiliation：

    School of Resources and Environment, University of Electronic Science and Technology of China, Chengdu 611731, China

WANG Yong5

реферат

This paper reports on a method to improve the Ratio Of an Exponentially Weighted Averages (ROEWA) edge detector, so that the improved edge detector can accurately determine the positions and directions of edges for Synthetic Aperture Radar (SAR) images. We attempt to build an optimal edge detector for SAR images to obtain better results of edge detection. The edge strength index is redefined as an inverted, signed, and normalized minimum ROEWA (IROEWA), which is utilized to quantitatively describe the phase step of edges. A new method that accurately calculates edge direction is developed based on the edge strength map from IROEWA. We can obtain the possible values of edge directions in this manner, which continuously distributes from 0 degrees to 180 degrees. Therefore, we must improve the Non-Maximum Suppression (NMS) algorithm, so that it can process sub-pixels. Finally, the improved NMS algorithm is also added into the edge detection workflow. This improved edge detection algorithm is called IROEWA & NMS. We conducted two experiments for IROEWA & NMS: one employed nature SAR images, whereas the other adopted a simulation SAR image. Experiment results show that the IROEWA & NMS outperforms the original ROEWA with watershed thresholding. The IROEWA operator is faster than the ROEWA operator under the same conditions. We applied a Receiver Operating Characteristic (ROC) curve to evaluate the IROEWA & NMS and determined that its Area Under the Curve (AUC) is 0.97570; thus, it approximates the ideal optimal detector. The detection rate at the position of the optimal point in the ROC curve of the IROEWA & NMS is as high as 0.95232, whereas the false alarm rate is as low as 0.00214. The IROEWA & NMS exhibits suitable performance on both the detection and false alarm rates. It has significant application value in several fields, such as the segmentation and edge detection for SAR images.

ключеви́че слова́

SAR images;edge detection;ROEWA;Non-Maximum Suppression(NMS);Receiver Operating Characteristic(ROC)

References

  1. 1.
    An C J, Xin Y L and Chen Z P. 2011. Edge detection algorithm for SAR image based on improved ROEWA. Journal of Image and Graphics, 16(8): 1483–1488
  2. 2.
    Bovik A C. 1988. On detecting edges in speckle imagery. IEEE Transactions on Acoustics, Speech, and Signal Processing, 36(10): 1618–1627
  3. 3.
    Canny J F. 1986. A computational approach to edge detection. IEEE Transactions on Pattern Analysis and Machine Intelligence, PAMI-8(6): 679–698
  4. 4.
    Fjørtoft R, Lopès A, Marthon P and Cubero-Castan E. 1998. An optimal multiedge detector for SAR image segmentation. IEEE Transactions on Geoscience and Remote Sensing, 36(3): 793–802
  5. 5.
    Jia C L and Kuang G Y. 2007. An improved edge detection algorithm for SAR images. Journal of Electronics and Information Technology, 29(2): 379–382
  6. 6.
    Lee J S. 1986. Speckle suppression and analysis for synthetic aperture radar images. Optical Engineering, 25(5): 636–643
  7. 7.
    Mason S J and Graham N E. 2002. Areas beneath the relative operating characteristics (ROC) and relative operating levels (ROL) curves: statistical significance and interpretation. Quarterly Journal of the Royal Meteorological Society, 128(584): 2145–2166
  8. 8.
    Paillou P. 1997. Detecting step edges in noisy SAR images: a new linear operator. IEEE Transactions on Geoscience and Remote Sensing, 35(1): 191–196
  9. 9.
    Sun C L. 2006. Study of ATR Algorithm Performance Evaluation Method based on ROC Curve. Changsha: National University of Defense Technology: 28–31
  10. 10.
    Wu Y H, Chen T Z and Shu Y. 2010. Road extraction from high resolution SAR images based on the directional ROEWA detector. Computer Engineering and Science, 32(8): 71–74
  11. 11.
    Yang D L, Lu C H, Zhang J L, Yang K and Chen X T. 2013. Edge detection in SAR images based on ROEWA and Hough transform. Journal of Electronic Measurement and Instrument, 27(6): 543–548
  12. 12.
    Zaman M R and Moloney C R. 1993. A comparison of adaptive filters for edge-preserving smoothing of speckle noise // Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing. Minneapolis, MN, USA: IEEE: 77–80 [DOI: 10.1109/ICASSP.1993.319751]
  13. 13.
    Zhao L J, Jia C L and Kuang G Y. 2007. Overview of edge detection in SAR images. Journal of Image and Graphics, 12(12): 2042–2049

Читать полностью

The above content is generated by Large Model Translation. The translated content is for reference only. We do not assume any commercial or legal responsibilty for any consequences arising from the use of our website