Probabilidad de infracción de velocidad de vehículos utilizando visión artificial en cámaras de campo amplio

We estimate vehicle velocities and detect speeds greater than the speed limit in videos captured with a wide-field fisheye camera (360° × 183°) in the context of transport analysis. To do this: 1. We implemented a calibration method that allows us to map points from the fisheye camera's image to a georeferenced map obtained from a satellite image. 2. blobs of moving objects were detected using a background subtraction algorithm. Feature points were detected inside the blobs and tracked using an algorithm based on SURF and FLANN techniques. 3. We fitted polynomials to the point trajectories, It's coefficients were estimated online by the recursive least squares algorithm. 4. The speed of feature points was estimated. 5. We calculated the probability of the speed being above the allowed limit.