PROJECT TITLE :

An Extended Random Walker Approach for Object Extraction by Integrating VGI Data and VHR Image

ABSTRACT:

Automatic extraction of objects in urban areas from terribly-high-resolution (VHR) images is of great significance to several applications. Existing approaches take into account little data on spatial relationships, backgrounds, and prior data of target objects, leading to that they did not perform well in object extraction. Now free and quick growing volunteered geographic info (VGI) can be accessed easily; so, they'll be used as previous data to improve the performance. This study develops an extended random walker (RW) approach to make a bottom-up and top-down mechanism for extracting target objects by combining VHR images and VGI knowledge. Novel aspects of our approach include: 1) both the shape and spectral previous terms are incorporated into the extended RW algorithm; two) an finish-to-finish framework is proposed to automatically choose each foreground and background seeds with the help of VGI information; and three) the shape prior of VGI information provides prime-down data to select background and foreground seeds and facilitate fuse bottom-up image data (i.e., foreground and background seeds and spatial relationships) to extract target objects. The extended RW approach was validated on building and lake datasets, and its performance is evaluated on each pixel and object levels. Quantitative comparisons with the original RW and Random Forest (RF) algorithm indicate that the proposed approach achieves vital better performance. Besides, it can successfully extract the partly occluded buildings.