Abstract
Abstract. Automatic extraction of buildings from high-resolution remote sensing imagery is very useful in many applications such as city management, mapping, urban planning and geographic information updating. Although extensively studied in the past years, due to the general texture of the building and the complexity of the image background, high-precision building segmentation from high-resolution sensing image is still a challenging task. Repeated pooling and striding operations used in CNNs reduce feature resolutions and cause the loss of detail information. In order to solve this problem, we proposed a deep learning model with a spatial pyramid pooling module based on the LinkNet. The proposed model called P-LinkNet that takes advantage of a spatial pyramid pooling module to capture and aggregate multi-scale contextual information. We tested it on Inria Building dataset. Experimental results show that the proposed P-LinkNet is superior to the LinkNet.
Highlights
Automatic extraction of buildings from remote sensing imagery is used in many applications, including urban planning, navigation, and disaster management [Panboonyuen et al, 2019, Wu et al, 2018, Li et al, 2018, Liu et al, 2019]
We found our model is better than LinkNet34
By adding spatial pyramid pooling, P-LinkNet can obtain larger receptive field and multi-scale information at the same time, and alleviated the global infromation loss occurred in LinkNet34
Summary
Automatic extraction of buildings from remote sensing imagery is used in many applications, including urban planning, navigation, and disaster management [Panboonyuen et al, 2019, Wu et al, 2018, Li et al, 2018, Liu et al, 2019]. Since entering the new world, pixel by pixel prediction has been introduced on the basis of feature extraction by classifier such as Support Vector Machines(SVM) [Inglada, 2007], Adaptive Boosting(AdaBoost) [Aytekin et al, 2013], Random Forests [Dong et al, 2015], K-Means [Cheng et al, 2013], and Conditional Random Fields(CRF) [Li et al, 2015]. These methods rely heavily on manual design and implementation, which change as the application domain changes. They are prone to introduce biases and poor generalization, and are time-consuming and labor-intensive
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 callMore From: The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Disclaimer: 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.
