Building Height Extraction Based on Joint Optimal Selection of Regions and Multi-Index Evaluation Mechanism

Building height is one of the important data for understanding urban development and changes. Building height estimation using a digital surface model (DSM) based on the difference between the roof elevation and the ground elevation of the building is commonly utilized. However, owing to the limitations of existing DSM techniques, invalid values may exist in the DSM. Existing DSM-based methods for estimating building heights typically use the interpolated DSM; however, when there are many invalid values, there may be errors in the interpolation results, which can mislead the selection of ground elevation values. Therefore, we propose a building-height extraction method that combines an optimal selection region and a multiindex evaluation mechanism to reduce the impact of invalid values and complex terrains. First, the optimal area for the ground elevation search was obtained based on the spatial relationship between the target building and surrounding buildings. Second, a joint multiindicator weighted evaluation mechanism was used to obtain the optimal ground elevation value. Finally, the building height was determined based on the difference between the roof and the ground elevations. Four build-up areas were used to test the effectiveness of the proposed method. The results exhibit high accuracy in complex areas with variable ground elevations, with a mean absolute error (MAE) of 1.16 m in building height. In areas with many invalid values and large shadow coverage of the surface areas, the MAE in building height is 0.92 m. Additionally, we verified the accuracy of the ground elevation estimated after interpolation. It is evident that the performance of the original DSM is satisfactory, with a high tolerance for input data and ability to be used in different building scenarios, providing new ideas for studying building height estimates.