Computer Vision for Efficient 3D Crime-Accident Scene Reconstruction

This award was competitively made in response to a proposal submitted by Carnegie Mellon University (CMU) to a National Institute of Justice FY 2013 solicitation: 'Applied Technology Research and Development for Criminal Justice Purposes'.

In its application, CMU proposes to develop assistive technologies to reconstruct scenes, such as crime and accident scenes, in 3D using low-cost imaging devices, to quantify the accuracy of the reconstructed scenes, and to study the effect of human factors in the collection and processing of the forensic evidence. A key aspect of CMU's proposal is the implementation of the assisted image capturing system on a cell phone. They propose to develop an approach and algorithms to assist the human (crime scene investigator) to take a sequence of pictures that can then be successfully merged into a single model, thereby avoiding the generation of multiple disconnected sub-models. This structured data acquisition will increase both the completeness and the geometric accuracy of the reconstructed 3D image. In that law enforcement personnel in the field will have a wide range of forensic capabilities; this effort will also produce a complete characterization of crime scene investigator's ability to produce accurate maps and scene reconstructions in 3D on the field, with limited training. This research also had a clear application in courts applications as well. ca/ncf