A pedestrian detection method based on SVM classifier and optimized Histograms of Oriented Gradients feature

Pedestrian detection by camera sensor is an important function in intelligent vehicle. Histograms of Oriented Gradients (HOG) features is a kind of efficient pedestrian feature. We optimized the HOG features to achieve an accurate human detection system. We don't normalize the input detection windows but resize the cell and block by same ratio. In the processing of calculate the HOG features, the Integral image is used for a better performance. The linear SVM is used as a classifier for the pedestrian detection result. Simulation results showed that the approach method is effective.