Assessment of 3D surveying methods applied to indoor fire scenes: A point cloud comparison

When generating virtual reconstructions of crime scenes for use as either investigative or training tools, the accuracy of the spatial data underpinning these reconstructions must be fully understood. This research aims to assess the accuracy and completeness of the spatial data collected at the crime scene, recorded using specialised and off-the-shelf mapping technologies available to law enforcement and forensic investigators. A simulated fire scene in a domestic dwelling in the Guldborgsund municipality, Denmark, was selected as a case study. First, a multimodal dataset of point clouds was assembled using a Leica RTC360 terrestrial laser system (TLS), a Leica BLK2GO handheld mobile laser scanner (HMLS), a Matterport Pro2 3D Camera structured-light scanner (SL), and a Canon EOS 5D Mark III digital single-lens reflex camera (DSLR). Due to its nominal precision and resolution, we identified the point cloud obtained using the TLS as the baseline reference system for the comparative analysis of the selected survey methods. The present analysis included the evaluation of the accuracy and quality of point clouds by comparison with the baseline reference data using visual inspection, inter-cloud distance estimation, measurement uncertainty of selected features and odometry drift analysis. Due to the different nature of the methods, it is not our intention to identify the best surveying method overall. In contrast, we are interested in benchmarking the performance of the selected capture methods both qualitatively and quantitatively in the context of incident documentation.