ASENet: A Deep Learning Approach to Predict Traffic Speed Compliance During Automated Speed Enforcement

Speed enforcement is one of the most widely used countermeasures to curb speed violations. Automated Speed Enforcement (ASE) not only supplements traditional traffic enforcement but also provides valuable data on traffic violations for safety research. Numerous evaluations have been performed to assess the effectiveness of ASE programs in reducing speed violations. However, there is a significant gap in the research community regarding the prediction of traffic speed compliance during different ASE enforcement periods. To address this gap, this paper presents the first study to predict traffic speed compliance during various automated enforcement periods using a novel deep learning architecture. We collected traffic data from different ASE sites in the Durham Region of Ontario, Canada, covering the warning period, the active-ASE period, and the post-ASE period across multiple rotations. We categorized the features influencing changes in traffic speed compliance using Analysis of Variance and Ordinary Least Squares Regression. Performance evaluations demonstrate that the proposed deep learning approach can predict traffic speed compliance during different periods of automated speed enforcement with a very high degree of confidence.