Pedestrian detection based on combinational holistic and partial features

Pedestrian detection has been widely used in many applications, however, it is a challenging task and there are many problems unsolved to be handled. Althougth Histograms of Oriented Gradients (HOG) plus Support Vector Machine (SVM) is the most successful pedestrian detection algorithm, the detection rate is becoming worse when the portions of human partial ocllusions are increasing. We propose an approach of adding the head features based on HOG for improving pedestrian detection rates in the case of body partial occlusions. The experiment demonstrates that our approach is robust to the occlusions.