Identification and Tracking of Takeout Delivery Motorcycles Using Low-Channel Roadside LiDAR

In recent years, food takeout services have grown rapidly in large cities. Likewise, the number of delivery motorcycles has grown with the rise of the takeout delivery market. Therefore, identifying and tracking motorcycles is essential as primary steps toward preventing and predicting delivery riders' crashes using trajectory data. First, kinematic features were proposed to combine with shape and density information as input indicators of random forests (RFs) to classify the traffic objects. Then, a multifeature fusion method was proposed to track the traffic object by constructing a mathematics matrix to present the adjacent point cloud frames. Furthermore, an improved Hungarian algorithm was developed to track the same object in the mathematics matrix. The experiment results showed that the average recognition accuracy of the proposed method is 99.59%, the average tracking accuracy is 92.53%, the tracking accuracy for delivery motorcycles is up to 98%, and the tracking speed stability is 98%. This study contributes to developing proactive strategies to reduce crashes related to takeout delivery motorcycles.

Automated summary

Food takeout services and the number of delivery motorcycles have been rapidly growing in large cities. Therefore, it is crucial to identify and track motorcycles using trajectory data in order to prevent and predict delivery riders' crashes. This study proposes kinematic features combined with shape and density information as input indicators for random forests to classify traffic objects. A multifeature fusion method is developed to track traffic objects by constructing a mathematics matrix, and an improved Hungarian algorithm is used for tracking the same object in the matrix. Experimental results show high recognition and tracking accuracy for delivery motorcycles, contributing to proactive strategies for reducing crashes related to takeout delivery motorcycles.