Accurate and rapid acquisition of spatial location and change information of buildings from high-resolution remote sensing images is of great significance in urban planning management, military detection, disaster emergency response and other applications. The high-resolution optical images obtained by the Earth monitoring platform can provide more and more detailed information such as color, texture, structure, especially edge, which is the most important data source for building target extraction. Through the research of comparative scholars, it is found that with the continuous development of deep learning technology, using artificial intelligence methods to systematically supervise deep learning and intelligent recognition of cities of different types, sizes and locations can achieve fine identification of urban buildings. Therefore, making full use of high-resolution visible remote sensing images to extract urban building targets and realizing accurate and reliable automatic extraction of building targets has far-reaching academic significance and application value.