In criminal investigations, investigators often have to analyze a vast amount of witness and surveillance footage from a specific incident. We propose an approach to fuse multiple video sources into a common context by reconstructing the crime scene and projecting all available sources spatially and temporally aligned into it. Subsequently, the scene can be interactively explored with visual analytics support.

The pipeline consists of three main steps: The reconstruction of a static environment, the transformation of videos to 4D point clouds and the visualization of all components, including automatic object and person detections.

Input: Video footage from surveillance cameras, drones, mobile cameras recorded after or during an incident.

Output: We create a static 3D scene with all camera locations (geo-registered).

• (A) Reconstructed static surface mesh
• (B) Embedded dynamic object reconstruction
• (C) Movement trajectory of a dynamic object
• (D) Custom user annotations
• (E) Animation path for custom user annotations
• (F) Interactive distance measurement tool
• (G) Interactive replay and temporal navigation
• (H) Supportive visualization for temporal exploration of automatic detections

@conference{pollok:2019:future,
 author    = {Thomas Pollok and 
              Matthias Kraus and 
              Chengchao Qu and 
              Matthias Miller and 
              Tobias Moritz and 
              Timon Kilian and 
              Daniel A. Keim and 
              Wolfgang Jentner},
 booktitle = {{VICTORIA} International Conference: Improving Video Analysis for a Safer Europe as part of ICDP 2019},
 title     = {The Future of Crime Scene Investigation},
 type      = {Poster}, 
 year      = {2019}
}