Large Vehicle Trajectory Tracing Method Based on Edge Calculation
Jun Hai Guo, Yingying Ma, Dechao Gao, Bangzheng Wang, Xueyuan Liu, Ying Liu · 2018
This paper proposes a large vehicle trajectory tracing method based on edge calculation and feature recognition. Specifically, we use the SVM+HOG feature detection method to acquire a large vehicle in the video captured by the edge device, and then perform feature extraction on the image containing the target vehicle. We use a color recognition algorithm based on building a color dictionary and a vehicle classification algorithm based on migration learning to obtain two features with high recognition accuracy. Finally, the edge device in different spatial locations uploads the feature information of the captured large vehicle to the cloud server, and the cloud generates the trajectory according to the feature matching and the space-time constraint to realize the trajectory tracing of the large vehicle. Experiments show that this method can significantly improve the recognition accuracy of large vehicles and have a good tracing effect.