Closed scorpeeon closed 10 years ago
An algorithm for on-board vision vehicle detection using cascade of boosted classifiers. Combines histograms of oriented gradient (HoG) (generative features to eliminate "easy" negative examples) with Haar-like features (discriminative classifiers for fine decisions).
Obviously needs trainin with positive and negative samples. In test the measured detection rate was 94%. Does not discuss tracking, but might be useful for vehicle detection.
ACascadeOfBoostedGenerativeAndDiscriminativeClassifiersForVehicleDetection.pdf