Forensic science proves case in murder trial
Computer scientists, Astrophysicists and Ecologists at LJMU have helped in preparing the prosecution case in a murder.
The application of artificial intelligence was used in the disclosure preparation at the murder trial early this year.
The case is the latest instance of LJMU forensic science being applied to crime-fighting, adding to our work on victim identification, war graves, animal poaching, people trafficking, the illegal drug trade and more.
High-profile work on Deep Learning and Computer Vision by Professor Paul Fergus, Dr Carl Chalmers, Professor Steven Longmore, and Professor Serge Wich led to an approach used by the Metropolitan Police to assist in the preparation of a murder case they were working on.
The police knew they had their suspect but leading up to the trial they needed to discount any other potential suspects.
Prof Fergus picks up the story: “The police were keen to use deep learning and computer vision to analyse CCTV footage because manually looking through large volumes of CCTV footage was not a viable option for them.”
Dr Chalmers said: The Met wanted us to use our AI object detection platform to pull out the images of people and cars to significantly reduce the number of images they would need to look through.”
Paul and Carl had 28 days of CCTV to process - that’s an astonishing 25 images per second for four weeks!
Setting up eight GPUs, side-loaded with 120 Faster RCNN models using Nvidia’s Triton server, they processed the images in seven days and identified all the individual people and cars for the MET to cross-reference with other evidence.
They then developed a Power BI dashboard that allowed the MET to quickly analyse the detections and filter for time, day, object, and camera location.
The accused not only murdered the victim but was he was also using his victim’s personal data to obtain money goods. For these crimes he was sentenced to 35 years in prison.
Added Paul: “In crime and security, AI detection systems are increasingly used to detect and track objects of interest, such as people, vehicles, and suspicious packages, providing valuable data to policing and security services to prevent crime and improve public safety.
“More broadly, they are being employed in everything from healthcare to retail to agriculture.”
The principal use of the teams technology is via Conservation AI, a research project servicing wildlife conservation organisations globally and works with images from the visual spectrum and thermal infrared cameras that are used on drones or in camera traps. Professor Wich said: “It is fascinating to see that the deep learning platform we developed has such varied applications from conservation to a murder case.”
