High-resolution remote sensing image building extraction based on Unet network under satellite platform
星载平台下基于 Unet 网络的高分遥感影像建筑物提取

摘要

城市建设规模逐步扩张让城市内部建筑物数量增多，在城市规划、资源管控、安全管理等多个方面的工作开展期间，都需要对城市内部建筑物信息进行参考，为此精准、快速地获取城市区域内部建筑物信息，也是未来城市建设与发展期间的核心内容。随着现阶段遥感技术的使用，高分辨率影像作为促进城镇化建设的重要数据，在高分辨率的遥感影像下，可以提供更加丰富和精细的建筑物空间信息。同时，在深度学习技术的研发和使用期间，都能为遥感技术的创新应用提供必然的选择，解决传统遥感图像在解析中存在的多种问题。在硬件性能提升的过程中，星载平台的发展让遥感技术的应用范围不断扩展，深度学习在星载平台中的应用，可以实时对遥感影像进行处理，节约大量的时间和资源。本文主要针对星载平台下，Unet网络的高分遥感影像建筑物提取技术进行阐述，期望能为今后技术的创新研发提供参考。

Abstract

The continuous expansion of urban construction has led to a significant increase in the number of buildings within cities. During urban planning, resource management, and safety administration, accurate and rapid acquisition of building information has become crucial for urban development. With the application of remote sensing technology, high-resolution imagery serves as vital data for urbanization projects, providing richer and more detailed spatial information about buildings. Meanwhile, advancements in deep learning technology have become essential for innovative applications of remote sensing, effectively addressing issues in traditional image analysis. As hardware performance improves, the development of spaceborne platforms has expanded remote sensing applications. The integration of deep learning into these platforms enables real-time processing of remote sensing images, significantly reducing time and resources required. This paper focuses on Unet network-based high-resolution remote sensing image building extraction technology under spaceborne platforms, aiming to provide insights for future technological innovation.