1 November 2019 in IEEE Geoscience and Remote Sensing Letters
This letter proposed an advanced superpixel-based Markov random field (MRF) model for unsupervised change detection. First, simple linear iterative clustering was implemented to the difference image to get a superpixel map. The initial change map is then obtained using fuzzy c-means by clustering the superpixel map. Finally, the initial change map of superpixel is iteratively refined by improved superpixel-based MRF that incorporated the spatial attraction model. It considers the spatial information of adjacent superpixels to complete the miss-detection superpixels and eliminate false-alarm superpixels. The experimental results over two data sets demonstrate the effectiveness of the proposed method.
1 September 2020 in Computers and Electronics in Agriculture
6 April 2020 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Image registration is a basic step for remote sensing image processing, such as classification, change detection and so on. Registration of Unmanned aerial vehicle (UAV) images is a challenging task because of illumination change, obvious rotation, viewpoint change and similar textures. Therefore, the feature points on the buildings are removed by morphological processing when detecting feature points in this study. Considering the similarity of terrain structural properties between adjacent UAV images, the scene shape similarity feature (SSSF) descriptor is utilized to capture the structural similarity. However, SSSF descriptor is influenced by the illumination changes of UAV images. The Phase Congruency algorithm is adopted to replace the Canny operator to detect structural features when calculating the SSSF descriptors because of its robustness for illumination change. Then the normalized correlation coefficient of SSSF descriptors is used as the similarity metric to detect tie points. Finally, the point pairs with large residuals are eliminated by Random Sample Consensus algorithm and the tie points are detected by iteratively removing the mismatched points. The proposed method was compared to some recent related methods using correct matching ratio (CMR), root mean square error (RMSE) and Time. Four experiments including Habitation, Bare-land, Mixed-terrain, Farmland and Buildings are carried out to evaluate the effectiveness of the proposed method. Registration results demonstrate that the proposed method is robust for illumination and viewpoint change of UAV images and usually generates more accurate registration results than some popular registration methods.