Accessibility navigation


A hierarchical Dempster-Shafer evidence combination framework for urban area land cover classification

Yang, F., Wei, H. and Feng, P. (2020) A hierarchical Dempster-Shafer evidence combination framework for urban area land cover classification. Measurement, 151. 105916. ISSN 0263-2241

[img]
Preview
Text - Accepted Version
· Available under License Creative Commons Attribution Non-commercial No Derivatives.
· Please see our End User Agreement before downloading.

952kB

It is advisable to refer to the publisher's version if you intend to cite from this work. See Guidance on citing.

To link to this item DOI: 10.1016/j.measurement.2018.09.058

Abstract/Summary

This paper presents a novel evidence combination framework for urban area land cover classification by using Light Detection And Ranging (LIDAR) data fused with co-registered near infrared and color images. The newly developed combination framework is built with a hierarchical structure involving an improved Dempster-Shafer (DS) theory of evidence for decision making. In the framework, a fuzzy basic probability assignment (BPA) function with fuzzy classes is firstly established based on the DS theory of evidence, and a probability is then assigned to each data source, that is derived from the original airborne LIDAR and the co-registered images. Secondly, an interesting approach is to introduce noise removal in an interim stage at the output of the probability distribution, and then the probability assigned to each data source is redistributed with a designated rule. Finally, a decision is made based on a “maximum normal support” rule, leading to the classification results. The proposed framework has been tested on two datasets. The testing results have shown that it can dramatically reduce the computational time in the classification process, and significantly improve the classification accuracy, i.e. 8.22% on Test 1 and 5.76% on Test 2 compared to the basic DS method. Due to its non-iterative and unsupervised nature, the proposed method is fast in computation, does not require training samples, and has achieved high classification accuracy.

Item Type:Article
Refereed:Yes
Divisions:Faculty of Science > School of Mathematical, Physical and Computational Sciences > Department of Computer Science
ID Code:80424
Publisher:Elsevier

Downloads

Downloads per month over past year

University Staff: Request a correction | Centaur Editors: Update this record

Page navigation