Multi-pedestrian Detection in Crowded Places on Quadrotor's View

In this paper, detecting multi-pedestrian in crowded areas on quadrotor's view is explored. This paper proposes a new method to combine Kalman filter with detection to reduce missed pedestrians. The detection is based on the supervised machine learning method, which is support vector machine (SVM) with histogram of oriented gradients (HOG) feature. And Kalman filter is used to estimate the location of the missed pedestrians. The experimental result is shown and real-time multi-pedestrian detection are realized on GPU Geforce GTX960M.