Abstract
In this paper, the change detection of Multi-Spectral (MS) remote sensing images is treated as an image segmentation issue. An unsupervised method integrating histogram-based thresholding and image segmentation techniques is proposed. In order to overcome the poor performance of thresholding techniques for strongly overlapped change/non-change signals, a Gaussian Mixture Model (GMM) with three components, including non-change, non-labeling and change, is adopted to model the statistical characteristics of the different images between two multi-temporal MS images. The non-labeling represents the pixels that are difficult to be classified. A random walk based segmentation method is applied to solve this problem, in which the different images are modeled as graphs and the classification results of GMM are imported as the labeling seeds. The experimental results of three remote sensing image pairs acquired by different sensors suggest a superiority of the proposed approach comparing with the existing unsupervised change detection methods.
Highlights
In remote sensing, change detection is to identify the difference between two images of the same scene acquired at different times, which is a critical technique in many remote sensing applications, such as land use change monitoring [1], disaster damage assessment [2,3], urban expansion study [4]and map updating [5], etc
In order to assess effectiveness of the proposed unsupervised change detection method, three multi-temporal image pairs are selected as test data sets as shown in Figures 5–7, which were acquired by different sensors and contain various land cover changes
Based thresholding method [21]; (2) Markov Random Field (MRF)-based method [21], in which Besag’s iterated conditional modes (ICM) algorithm [38] is used to minimize the MRF energy function initialized by EM algorithm; (3) Block-PCA method (BPCA) [26], in which PCA is applied to h × h non-overlapping block set to detect the change area; (4) and multi-resolution level set method (MLSK) [24], in which the Chan–Vese algorithm [39] is applied on the pyramid of difference images to obtain the change map
Summary
Change detection is to identify the difference between two images of the same scene acquired at different times, which is a critical technique in many remote sensing applications, such as land use change monitoring [1], disaster damage assessment [2,3], urban expansion study [4]. To guarantee a better result, a large number of training samples with human labeled ground truth need to be fed into the classifiers after feature extraction These methods are known as pre-classification change detection methods. When change and non-change classes are strongly overlapped in feature space, or when their statistical distribution cannot be modeled accurately, the thresholding techniques can not provide satisfactory results To tackle this problem, some researchers proposed improved approaches. Bovolo et al [23] proposed a two-phase detection method, in which the point set derived by a selective Bayesian thresholding is treated as a pseudo-training set of a binary semi-supervised SVM classifier to generate change map.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
