Robust people counting in crowded environment

This paper describes a new learning-based method for people counting task in crowded environments from a single camera. The main difference between this method and traditional ones is that it adopts separated blobs as the input of the people number estimator. Firstly, the blobs are selected according to their features after background estimation and calibration by tracking. Then, each selected blob in the scene is trained to predict the number of persons in the blob. Finally, the people number estimator is formed by combining trained sub-estimators according to a predefined rule. Experimental results are shown to demonstrate the functions.