Research on pedestrian counting based on millimeter wave

The thriving development of big data has provided powerful techniques for pedestrian counting and monitoring in public places, which has attracted great attention. Current pedestrian counting solutions based on computer vision and sensors have privacy leakage issues and cost issues. As a wireless technique, millimeter wave has diverse application scenarios because of its advantages of safety, cheapness and stability. In this paper, we propose a novel system to realize pedestrian counting leveraging millimeter wave. Based on the point cloud data collected by radar, we design an efficient denoising algorithm based on the point cloud distribution and transform the problem into a multi-classification task, and then extract features from pedestrian trajectories to form a series of density maps describing pedestrians' movement information. We also propose a classifier model mmCountNet to accurately predict the pedestrian and build a data set containing 2000 samples for model training. The experiment results show that the accuracy of the proposed system is 95.92% in the Single-Crowd testing, and 91.35% in the Multi-Crowds testing, which basically achieves the purpose of pedestrian counting.