Night Time Vehicle Detection Using Machine Learning

Monitoring traffic using camera networks is crucial, especially at night when visibility decreases and accident risks rise. Existing methods often struggle with the erratic nature of vehicle lights in low -light conditions, where lights appear as flashes or complex patterns across disconnected image regions. This study introduces a real -time vehicle detection algorithm designed for night time scenarios, leveraging machine learning with a grid of foveal classifiers. These classifiers use a single global image descriptor to predict vehicle locations based on their positions within the grid and ground -truth data. By requiring only point -based annotations for training, the algorithm accelerates database creation. Experimental validation on a new nighttime dataset demonstrates the effectiveness of this approach in accurately detecting vehicles under challenging lighting conditions.