A Deep Learning Parking Detection Algorithm Using W-Shape Magnetic Wireless Sensor Networks

Abstract

With the number of motor vehicles in cities rapidly increasing, scheduling and management in large parking lots need to be timelier and more accurate, and vehicle detection technology plays a crucial role in this process. Magnetic sensors have attracted enormous interest in vehicle detection owing to their low cost and easy installation. However, most of the studies extract features manually, and the results are given by the potential difference in magnetic signal which is susceptible to environmental interference. Moreover, each parking space corresponds to a sensor, which is costly to deploy and maintain in a large parking lot. The main challenge lies in the difficulty of eliminating the interferences from vehicles in adjacent parking spaces that decrease the accuracy. In this article, a deep learning parking detection algorithm based on <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$W$ </tex-math></inline-formula> -shape magnetic wireless sensor network is proposed for large parking lots with vertical parking spaces. The number of required sensors is greatly reduced by analyzing the sensor deployment which can greatly reduce the costs accordingly. To improve the adaptability of the algorithm in different environments, the deep learning model is trained with the collaborative information from multiple sensors. Test in the actual scene shows that the algorithm can significantly reduce costs and improve adaptability while ensuring high detection accuracy. Thus, it is easier to implement in large parking lots.