Building extraction method based on lightweight network

• Main Authors: JIA HAITAO, CHANG LE, SONG YIYUN, LENG GENG, LUO XIN, WEI ZUQI, XU WENBO, REGAZA ADUGNA TESFAYE
• Format: Patent
• Language: Chinese; English
• Published: 03.09.2024
• Subjects: CALCULATING; COMPUTING; COUNTING; PHYSICS

The invention discloses a building extraction method based on a lightweight network. With the high demand for urbanization and the availability of high-resolution remote sensing images, building extraction has become an important subject for intensive research in the research field. According to the network, local and global connection is carried out according to a 1 * 1 convolutional layer to connect different blocks, so that medium-high frequency information is accessed at the same time, a feature extraction stage of the network is realized, the speed of extracting the network by a building is improved, and the light weight of the network is achieved. The multi-level connection can help the proposed method to learn multi-level features and help the network to quickly disperse information from low to high, and back propagation is carried out at the same time. In this case, the network shares information among all blocks, thereby obtaining a better reconstruction result. 本发明公开了一种基于轻量化网络的建筑物提取方法。随着对城市化的高需求和高分辨