Change detection methods based on low-rank sparse representation for multi-temporal remote sensing imagery
- 138 Downloads
With the development of remote sensing image applications, remote sensing imagery is an important technology to make a dynamic detection for changes of lands or coastal zones. Though high resolution remote sensing imagery provides very good performance and large information for the spatial structure variation, the limitation of its spatial structure makes change detection difficult. In this paper, we propose two change detection methods for multi-temporal remote sensing images which are based on low-rank sparse decomposition and based on low-rank matrix representation. An observation matrix is constructed by ordering each band of remote sensing images into a vector. We utilize bilateral random projection method to make low rank decomposition to get a sparse matrix. We then obtain the change map by using nearest neighbor to cluster the change parts from the sparse matrix. On the other hand, by dividing the difference set of multi-temporal remote sensing images into non-overlapping squares with equal size and tiling these squares, an observation matrix is built up. We make up a feature space matrix by low rank matrix representation to build a sparse representation model, and combine nearest neighbor method to make change detection for multi-temporal remote sensing dataset. This change detection method is addressed by iterating between kernel norm minimization and sparsity minimization. The experimental results show that our proposed methods perform better in detecting changes than the other change detection methods.
KeywordsChange detection Sparse representation Low-rank sparse decomposition Low-rank matrix representation
This work was supported by the National Natural Science Foundation of China (Nos. 61163042 and 61663007).
- 3.Jianya, G., Haigang, S., Guorui, M., et al.: A review of multi-temporal remote sensing data change detection algorithms. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 37(B7), 757–762 (2008)Google Scholar
- 14.Chen, Z., Elvidge, C.D., Groeneveld, D.P.: Vegetation change detection using high spectral resolution vegetation indices. Remote Sens. Change Change detect. Techn. 2395 (1998)Google Scholar
- 22.Chen K, Huo C, Zhou Z, et al.: Semi-supervised change detection via Gaussian processes. In: Geoscience and Remote Sensing Symposium, 2009 IEEE International, IGARSS 2009. IEEE, 2009, 2: II-996-II-999Google Scholar
- 26.Zhou, T., Tao, D.: Godec: Randomized low-rank & sparse matrix decomposition in noisy case. In: International Conference on Machine Learning. Omnipress, Madison (2011)Google Scholar
- 30.Liu, G., Lin, Z., Yu, Y.: Robust subspace segmentation by low-rank representation. In: Proceedings of the 27th International Conference on Machine Learning (ICML-10). pp. 663-670 (2010)Google Scholar
- 31.Moody, D.I., Brumby, S.P., Rowland, J.C., et al.: Change detection and classification of land cover in multispectral satellite imagery using clustering of sparse approximations (CoSA) over learned feature dictionaries. In: Applied Imagery Pattern Recognition Workshop (AIPR), 2014 IEEE. IEEE, pp. 1–10 (2014)Google Scholar
- 33.Zhou, J.: Matrix Analysis and Application. Sichuan University Press, Chengdu (2008)Google Scholar
- 34.Roweis, S.T.: EM algorithms for PCA and SPCA. In: Advances in Neural Information Processing Systems, pp. 626–632 (1998)Google Scholar
- 39.Shulei, Wu, Bai, Yong, et al.: The optimal band combination joint WCOMP+BPNN classification method for remote sensing image. J. Comput. Inf. Syst. 11(8), 2873–2884 (2015)Google Scholar