duy-phamduc68/cctv-inverse-perspective-mapping

This is a calibration tool for aligning traffic camera footage with satellite imagery. The specific technique it uses is called Inverse Perspective Mapping, a way of mathematically transforming a camera's angled, ground-level view into a flat, top-down perspective that matches a satellite or map image. Once calibrated, the tool creates a two-way projection: you can convert a point in the camera frame to its location on the map, and the other way around. The tool handles common camera issues such as lens distortion and parallax, the way nearby objects appear to shift more than distant ones as the camera moves. It outputs the calibration results, including distortion correction data, homography matrices, camera pose, and scale, as reusable JSON files, which the author refers to as G-projection files. These can then be used by other software in the same project pipeline. This module was extracted from a larger project called TrafficLab 3D and published as a standalone repo so that the calibration process can be improved over time. Currently the workflow is largely manual. The tool is written in Python and uses PyQt5 for its graphical interface. It is intended for researchers or developers working on traffic analysis who need to map camera footage onto real-world coordinates. A companion blog post and a guide video covering the broader TrafficLab 3D project are linked in the README.