A Traffic Event Detection Method Based on Random Forest and Permutation Importance

Although the video surveillance system plays an important role in intelligent transportation, limited camera views make it difficult to observe many traffic events. In this paper, we collect and combine flow variables from multi-source sensors, propose a PITED method based on Random Forest (RF) Permutation importance (PI) for event detection. This model selects suitable by means of permutation arrangement importance, establishes whole process acquisition, preprocessing, quantization, modeling evaluation. Moreover, real data are collected tested paper evaluating experiment performance, including miss/false rate event, average detection time. The experimental results show that is more than 85% false alarm less 3%. It effective efficient practical application regardless both workdays holidays.