Long-Distance and Real-Time Vehicle Detection on Highway Using RCF Network and PPHT Based on Ultraweak FBG Array

Vehicle detection plays an essential role in highway traffic monitoring, management, and intelligent service. However, accurate and real-time vehicle detection is still challenging due to the impact of factors, such as complex traffic and diverse environmental conditions, as well as the unpredictable nature of vehicle movement. To effectively address these issues, in this article, a signal processing method for long-distance and real-time vehicle detection on the highway was proposed using the richer convolutional features (RCFs) network and the progressive probabilistic Hough transform (PPHT) based on ultraweak fiber Bragg grating (UWFBG) array. First, a long-distance vehicle vibration sensing system based on the UWFBG array was developed to achieve continuous traffic monitoring. After that, a dataset was constructed using a time-sliding window based on the spatiotemporal vibration distribution map of vehicle trajectories. Then, the RCF network was employed to reconstruct the line segments representing the vehicle trajectories. Finally, PPHT was executed for line segment detection, and a lightweight line segment nonmaximum suppression (LSNMS) algorithm was proposed to eliminate redundant line segments. From the acquired experimental results, it was demonstrated that the proposed method can detect closely spaced vehicle trajectories in real-time with an average F-measure of 0.96 and a real-time detection resolution of 15 m.