Urban management issues have become a significant challenge for urban administrators, and urban management must be based on modern computer technology and network communication environments. This paper investigates objectoriented building extraction methods, proposes a semi-global filtering method based on region growing for filtering remote sensing data, and then presents building 3D modeling methods and rendering methods in 3D remote sensing image modeling scenarios. In error comparison experiments between the classical TIN filtering algorithm and the proposed filtering method, the improved algorithm demonstrates superior performance, accurately identifying ground points at terrain steep changes, complex buildings, and complex roads. Its Type I error and total error are significantly lower than those of the classical TIN point cloud filtering algorithm, thereby demonstrating the superiority of the proposed improved algorithm.