Automatic Indoor as-Built Building Information Models Generation by Using Low-Cost RGB-D Sensors

Creating accurate digital representations of existing indoor spaces remains a significant hurdle for modern construction management. Current techniques often struggle to interpret complex room layouts from sparse or noisy spatial data. Previous research has shown that high-end laser scanners provide reliable results but remain prohibitively expensive for many routine projects. That uncertainty drove the need for more accessible hardware solutions that do not sacrifice structural integrity. No prior work had resolved the difficulty of extracting precise architectural relationships from low-quality sensor inputs without human oversight. This gap motivated the development of automated frameworks capable of handling significant data noise. Existing methods frequently rely on manual intervention to correct errors in point cloud segmentation. This study addresses these limitations by leveraging affordable sensor technology to streamline the documentation of built environments.

The aim of this study is to develop an automated framework for generating indoor as-built building information models using affordable sensor technology. This research seeks to overcome the high costs and time-intensive nature of traditional scanning equipment. The authors address the technical challenge of processing noisy spatial data without relying on manual input. They specifically target the difficulty of segmenting complex indoor elements from low-quality point clouds. The motivation stems from the need to make digital documentation more accessible for routine construction and renovation projects. By integrating low-cost hardware, the researchers intend to provide a scalable solution for the industry. This work explores whether autonomous systems can match the accuracy of expensive laser-based alternatives. The study ultimately focuses on streamlining the entire workflow from initial data capture to final model generation.

Review approach involves testing an automated framework designed to process raw spatial data from affordable sensors. The investigation employs a Structure sensor paired with a tablet to capture indoor environments. Researchers evaluated the performance by comparing the output against established Terrestrial Lidar System benchmarks. The design focuses on eliminating human interaction during the segmentation and extraction phases of model creation. Data collection occurred in standard rooms ranging from 45 to 67 square meters to ensure practical applicability. The approach utilizes specific algorithms to filter noise inherent in low-cost hardware inputs. Every stage of the pipeline functions autonomously to transform raw point clouds into finalized digital structures. This methodology emphasizes speed and cost-effectiveness as primary metrics for evaluating the proposed system.

Key findings from the literature reveal that the proposed method achieves an element extraction accuracy of 100%. The system demonstrates a mean dimension reconstruction accuracy of 98.6% across tested indoor spaces. Furthermore, the mean area reconstruction accuracy reached 93.6% during experimental trials. Data collection time for a typical room dropped to 4-6 minutes, compared to 50-60 minutes using traditional laser systems. The automated processing time for generating models is approximately 30 seconds. This represents a substantial improvement over the 10-minute duration required by conventional semi-manual techniques. The results confirm that the framework maintains competitive robustness despite the lower quality of the initial sensor inputs. These metrics highlight the efficiency gains achieved by replacing manual workflows with this autonomous pipeline.

The authors demonstrate that their automated framework successfully converts low-cost sensor data into reliable architectural models. Synthesis and implications suggest that this approach offers a viable alternative to expensive laser scanning systems. The results indicate that high levels of geometric accuracy are achievable despite the inherent noise in affordable hardware. These findings imply that construction professionals can significantly reduce both labor and equipment costs during site documentation. The researchers propose that their method enhances overall workflow efficiency by drastically shortening time requirements. This study confirms that fully autonomous processing is possible for standard indoor environments. The evidence supports the integration of these tools into routine building maintenance and renovation tasks. Future implementation could transform how practitioners manage indoor spatial data across diverse project scales.