Building Extraction Using Mask Scoring R-CNN Network

Abstract

Extracting buildings from high resolution remotely sensed images is very practical, which can be applied to urban modeling and so on. The development of computer vision has become better, and the accuracy of recognition of convolutional neural networks has exceeded the accuracy of recognition of human eyes. In this paper, we used a deep convolutional neural network in remote sensing to achieve building extraction. The method in this paper is not based on semantic segmentation, but instance segmentation, which considered each building as an independent individual to achieve building extraction. The results showed that the proposed method is able to extract buildings with high accuracy.