A real-time 3D modeling method for buildings driven by IMU and RGB-D fusion

With the rapid development of cutting-edge technologies such as autonomous driving, augmented reality, and intelligent robotics, real-time 3D scene reconstruction is increasingly demonstrating its crucial value in diverse application domains. However, 3D reconstruction faces challenges such as redundant data acquisition, inaccurate pose estimation, and insufficient reconstruction precision, making it difficult to support rapid visualization and measurement analysis. Therefore, this paper proposes an innovative real-time 3D modeling method for buildings driven by IMU and RGB-D fusion. The method incorporates an adaptive sampling strategy for RGB-D cameras based on IMU data calibration, introduces a pose estimation optimization algorithm that combines dynamic feature point cluster centroid prediction with bias detection, establishes a progressive 3D modeling approach constrained by structural features, and develops a prototype system for in-depth case study analysis. Experimental results demonstrate that, compared to existing methods, our approach achieves significant improvements across multiple metrics: data redundancy is reduced by 31%, dynamic object detection precision reaches 93.2%, computational latency decreases by 13.9%, and spatial accuracy improves by 29.0%. This approach enables rapid 3D reconstruction of building environments, automating structural measurements and analysis while significantly reducing manual effort, making it beneficial for applications in smart city development and digital twin implementation.