
I developed this project during my time as a Police Digital Researcher to address the time-consuming task of CCTV footage analysis faced by the Criminal Investigations Department and Police Digital Investigators. The system automatically extracts and saves motion-detected clips for manual review, eliminating the need to watch the entire recording.
Tools and Methodologies
I implemented this solution in C++ using Dear ImGUI for the graphical interface, ensuring efficient interoperability with C++ OpenCV. Specifically, this uses a Mixture of Gaussians background subtraction model to detect motion, and, by using C++, achieves a 2x speed-up compared to OpenCV in python.
Deployment and Reception
This reduced the workload for reviewing CCTV footage significantly. However, a key limitation is that it requires careful tuning of the parameters to ensure it does not miss important footage. However, this is largely mitigated by the motion preview window, which displays the quantity of motion detected for each frame in the video, allowing them to better establish the optimal motion threshold.
Leave a Reply