Threshold Segmentation Based on Information Fusion for Object Shadow Detection in Remote Sensing Images
- School of Information and Communication Engineering, Harbin Engineering University
Harbin 150001 China
yslin@hit.edu.cn - College of Information and Communications Engineering, Dalian Minzu University
Dalian, 116600, China
wangliguo@hrbeu.edu.cn
Abstract
In the shadow detection task, the shadow model is usually consistent with the approximate contour of ontology semantics, it is difficult to extract the features of land covered objects or ground pixels, and easy to be confused into foreground objects in gray scale. Therefore, we present to formulate and apply one new threshold segmentation method based on information fusion for object shadow detection in remote sensing images. Firstly, object shadow pixels are screened using intensity and chromaticity information in HSI color space. Secondly, the remote sensing image is carried out by principal component analysis (PCA) to obtain the first principal component. A new shadow index is constructed using the results obtained from HSI and the first principal component. Thirdly, based on the results of the above two information fusion, a threshold segmentation model is established using the improved threshold segmentation algorithm between the maximum and the minimum threshold segmentation algorithm, so as to obtain the final object shadow detection results. Finally, affluent experiments are conducted on the datasets collected from Google Earth. The results show that the proposed object shadow detection algorithm in remote sensing images can achieve better segmentation and detection (more than 95%) effect compared with state-of-the-art methods.
Key words
object shadow detection, threshold segmentation, information fusion, remote sensing images, HSI color space, PCA
Digital Object Identifier (DOI)
https://doi.org/10.2298/CSIS231230023Y
How to cite
Yin, S., Wang, L., Teng, L.: Threshold Segmentation Based on Information Fusion for Object Shadow Detection in Remote Sensing Images. Computer Science and Information Systems, https://doi.org/10.2298/CSIS231230023Y