Clustering Study of Vehicle Behaviors Using License Plate Recognition

Abstract

Ubiquitous computing and artificial intelligence contribute to deploying intelligent environments. Sensor networks in cities generate large amounts of data that can be analyzed to provide relevant information in different fields, such as traffic control. We propose an analysis of vehicular behavior based on license plate recognition (LPR) in a rural region of three small villages. The contribution is twofold. First, we extend an existing taxonomy of the most widely used clustering algorithms in machine learning with additional classes. Second, we compare the performance of algorithms from each class of the taxonomy, extracting behavioral patterns. Partitional and hierarchical algorithms obtain the best results, while density-based algorithms have poor results. The results show four differentiated patterns in vehicular behavior, distinguishing different patterns in both residents and tourists. Our work can help policymakers develop strategies to improve services in rural villages, and developers choose the correct algorithm for a similar study.