Urban Change Detection Based on Dempster–Shafer Theory for Multitemporal Very High-Resolution Imagery

Hui Luo,Chong Liu,Chen Wu,Xian Guo

doi:10.3390/rs10070980

Abstract

Fusing multiple change detection results has great potentials in dealing with the spectral variability in multitemporal very high-resolution (VHR) remote sensing images. However, it is difficult to solve the problem of uncertainty, which mainly includes the inaccuracy of each candidate change map and the conflicts between different results. Dempster–Shafer theory (D–S) is an effective method to model uncertainties and combine multiple evidences. Therefore, in this paper, we proposed an urban change detection method for VHR images by fusing multiple change detection methods with D–S evidence theory. Change vector analysis (CVA), iteratively reweighted multivariate alteration detection (IRMAD), and iterative slow feature analysis (ISFA) were utilized to obtain the candidate change maps. The final change detection result is generated by fusing the three evidences with D–S evidence theory and a segmentation object map. The experiment indicates that the proposed method can obtain the best performance in detection rate, false alarm rate, and comprehensive indicators.

Highlights

Remote sensing imagery has been one of the major data sources for studying human development and monitoring ecosystems [1,2,3,4,5,6]
Numerous studies have been proposed for change detection, which can be categorized into three classes [19]: (1) direct comparison, which calculates the difference between the original spectral features, such as image difference, image ratio, and change vector analysis [9,20,21,22]; (2) image transformation, which obtains the change information by transforming the original data into new discriminant features, such as principle component analysis, multivariate alteration detection (MAD), and slow feature analysis (SFA) [23,24,25]; (3) classification based method, which analyzes the “from-to” change types by comparing the classification maps or classifying the multitemporal combinations, such as post-classification method and direct classification [26,27]
We elaborate how to fuse the results of different change detection methods with Dempster–Shafer theory (D–S) theory to improve the performance for multitemporal very high-resolution (VHR) images

Summary

Introduction

Remote sensing imagery has been one of the major data sources for studying human development and monitoring ecosystems [1,2,3,4,5,6]. When multitemporal remote sensing images covering the same areas are able to be obtained, it is a matter to analyze where the landscape conditions have changed and what types the changes are. Dynamic analysis of landscape variation by change detection is extremely useful in many applications, such land-use/land-cover change, urban expansion, ecosystem monitoring, and resource management [10,11,12,13,14,15,16,17,18]. With the development of very high-resolution (VHR) remote sensing technology, multitemporal VHR images are much easier to obtain and are widely used in building change detection, disaster assessment, and detailed land-cover change monitoring [28,29,30,31,32,33]. We can group them into two categories: (1) object-oriented change detection, which utilizes the object as the process unit to improve the completeness and accuracy of the final result [19,37,38]; (2) change detection fusing spatial features, which takes into consideration the texture, shape, and topology features in the process of change analysis [28,39,40]

Methods

Discussion

Conclusion