Fast human detection using a cascade of histograms of oriented gradients

We integrate the cascade-of-rejectors approach with Histograms of Oriented Gradients (HoG) features to achieve fast and accurate human detection system. Our features are HoGs of variablesize blocks that capture salient features of humans automatically. We find the appropriate set of blocks, from a large set of possible blocks, using AdaBoost as a feature selection. Finally, we use the integral image representation and a rejection cascade to greatly accelerate the performance of our system. Taken together our system can process frames at rates of between 5 to 30 frames per second, depending on the density in which we scan the image, while maintaining an accuracy level similar to previously reported methods