<p>In the context of urban modernization and the construction of digital twin cities, relying on large-scale high-spatial-resolution data from remote sensing technology, rapid and stable identification of traffic-related targets, and further characterization and evaluation of the spatial pattern and evolution laws of urban cultural functional layout have become key technical requirements for refined urban cultural governance and optimized allocation of cultural space. However, the existing detection algorithms still have shortcomings and are unable to support the high-quality identification and stable analysis of the cultural functional space. Therefore, this paper proposes an improved target detection algorithm MPHN-RT-DETR based on RT-DETR, integrating multi-scale attention fusion, Haar wavelet downsampling, lightweight P-RepConv convolution, and a Normalized Wasserstein Distance loss. The function supply field is formed based on the category-function comprehensive weight matrix and the adaptive bandwidth Gaussian kernel density, enabling rapid quantitative assessment of cultural units. Compared with the baseline model, the accuracy of the improved model is increased by 9.07%, the recall rate is increased by 5.16%, the mAP@0.5 rate is increased by 7.58%. In summary, the MPHN-RT-DETR proposed in this paper not only meets the requirements for efficient and precise detection of remote sensing traffic targets, but also, through coupling with the urban spatial analysis module, realizes the automatic extraction and quantitative assessment of cultural functional layout. It can provide data support and decision-making basis for the optimization of public cultural facility location, the planning of cultural tourism routes, and the crowd guidance during major events.</p>