People Counting in Videos Based on Machine Learning.

People counting plays a crucial role in various senses, such as concerts, scenic spots, etc. In this paper, we propose a people counting method applied in public areas. We combine Histogram of Oriented Gradient feature extraction and Support Vector Machine classification and applied them to crowd counting. Firstly, images are captured by the camera and HOG feature extraction is applied to generate gradient histograms. Then, Support Vector Machine is used for feature classification and after that, detection frames are generated around each person detected. Finally, number of frames is calculated. The method that combines HOG feature extraction and SVM classification is easy to achieve. Experiments show that it is fast and reaches high accuracy in most occasions.