logo

SCIENCE CHINA Information Sciences, Volume 61, Issue 10: 102301(2018) https://doi.org/10.1007/s11432-017-9277-x

Multi-scale rock detection on Mars

More info
  • ReceivedMay 31, 2017
  • AcceptedOct 10, 2017
  • PublishedMay 18, 2018

Abstract

In this paper, we propose a novel autonomous Martian rock detection framework via superpixel segmentation. Different from current state-of-the-art pixel-level rock segmenting methods, the proposed method deals with this issue in region level. Image is splitted into homogeneous regions based on intensity information and spatial layout. The heart of proposed framework is to enhance such region contrast. Then, rocks can be simply segmented from the resulting contrast-map by an adaptive threshold. Our method is efficient in dealing with large image and only few parameters need to set. Preliminary experimental results show that our algorithm outperforms edge-based methods in various grayscale rover images.


Acknowledgment

This work was supported by National Natural Science Foundation of China (NSFC) (Grant Nos. 61503102, 61701225).


References

[1] Castano R, Judd M, Estlin T, et al. Current results from a rover science data analysis system. In: Proceedings of IEEE Aerospace Conference, Big Sky, 2005. Google Scholar

[2] Thompson D R, Castano R. Performance comparison of rock detection algorithms for autonomous planetary geology. In: Proceedings of IEEE Aerospace Conference, Big Sky, 2007. Google Scholar

[3] Burl M C, Thompson D R, deGranville C. Rockster: onboard rock segmentation through edge regrouping. J Aerosp Inf Syst, 2016, 13: 329-342 CrossRef Google Scholar

[4] Spencer D A, Adams D S, Bonfiglio E. Phoenix landing site hazard assessment and selection. J Spacecr Rocket, 2009, 46: 1196-1201 CrossRef ADS Google Scholar

[5] Dunlop H, Thompson D R, Wettergreen D. Multi-scale features for detection and segmentation of rocks in mars images. In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Minneapolis, 2007. Google Scholar

[6] Bonfiglio E P, Adams D, Craig L. Landing-site dispersion analysis and statistical assessment for the Mars Phoenix lander. J Spacecr Rocket, 2011, 48: 784-797 CrossRef ADS Google Scholar

[7] Gulick V C, Morris R L, Ruzon M A. Autonomous image analyses during the 1999 Marsokhod rover field test. J Geophys Res Planet, 2001, 106: 7745-7763 CrossRef ADS Google Scholar

[8] Huertas A, Cheng Y, Madison R. Passive imaging based multi-cue hazard detection for spacecraft safe landing. In: Proceedings of IEEE Aerospace Conference, Big Sky, 2006. Google Scholar

[9] Matthies L, Huertas A, Cheng Y, et al. Landing hazard detection with stereo vision and shadow analysis. In: Proceedings of IEEE International Conference on Robotics and Automation, Pasadena, 2007. Google Scholar

[10] Gor V, Mjolsness E, Manduchi R, et al. Autonomous rock detection for Mars terrain. In: Proceedings of AIAA Space Conference and Exposition, Albuquerque, 2001. Google Scholar

[11] Fox J, Castano R, Anderson R C. Onboard autonomous rock shape analysis for Mars rovers. In: Proceedings of IEEE Aerospace Conference, Big Sky, 2002. Google Scholar

[12] Di K C, Yue Z Y, Liu Z Q. Automated rock detection and shape analysis from Mars rover imagery and 3D point cloud data. J Earth Sci, 2013, 24: 125-135 CrossRef Google Scholar

[13] Xiao X M, Cui H T, Tian Y. Robust plane fitting algorithm for landing hazard detection. IEEE Trans Aerosp Electron Syst, 2015, 51: 2864-2875 CrossRef ADS Google Scholar

[14] Gong X J, Liu J L. Rock detection via superpixel graph cuts. In: Proceedings of the 19th IEEE International Conference on Image Processing, Orlando, 2012. 2149--2152. Google Scholar

[15] Ma F, Meng D Y, Xie Q, et al. Self-paced co-training. In: Proceedings of International Conference on Machine Learning, Sydney, 2017. 2275--2284. Google Scholar

[16] Zhang D W, Meng D Y, Han J W. Co-saliency detection via a self-paced multiple-instance learning framework. IEEE Trans Pattern Anal Mach Intel, 2017, 39: 865-878 CrossRef PubMed Google Scholar

[17] Achanta R, Shaji A, Smith K. SLIC superpixels compared to state-of-the-art superpixel methods. IEEE Trans Pattern Anal Mach Intel, 2012, 34: 2274-2282 CrossRef PubMed Google Scholar

[18] Xiao X M, Cui H T, Yao M B. Autonomous rock detection on Mars through region contrast. Adv Space Res, 2017, 60: 626-635 CrossRef ADS Google Scholar

  • Figure 1

    (Color online) Some of the notable rocks on Mars that edge detection algorithms failed to attach contours or generated false boundaries. From top to bottom: grayscale images taken by Spirit Rover's pancam; edge maps obtained by Sobel detector; results from Canny operator.

  • Figure 2

    (Color online) Superpixel of Mars images.

  • Figure 3

    (Color online) Statistical performance on RC* and RC, respectively. (a) PRC curves; (b) AUC and Max-$~F_{\beta}$.

  • Figure 4

    (Color online) Visual performance of proposed algorithm on real Martian images. (a) Original image; (b) and (d) are contrast maps resulting from RC and RC*, respectively; (c) and (e) are corresponding cutting maps; (f) the detecting results by Canny operator; (g) ground truth.

  • Figure 5

    (Color online) Statistical performance on tests with parameters changing. (a) PRC curves; (b) AUC and Max-$F_{\beta}$.

  • Table 1   Parameters setting for multi-scale superpixels segmentation
    Scale 1 2 3 4 5 6 7 8 9 10 11 12
    $~N^i~$ 250 300 350 400 450 500 550 600 650 700 750 800
    $~m^i~$ 5.5 5.5 7 7 8.5 8.5 10 10 11.5 11.5 13 13

Copyright 2019 Science China Press Co., Ltd. 《中国科学》杂志社有限责任公司 版权所有

京ICP备18024590号-1