On-road high-emitting vehicle identification by an automatic hyperparameter optimization model based on a remote sensing system

Optical remote sensing systems (RSSs) are ideal for monitoring and identifying high-emitting vehicles on roads, as they can be installed on any road for non-contact measurements. In general, an on-road vehicle is considered a high-emitting vehicle when its instantaneous emissions exceed the specified cut-points, as monitored by RSSs. However, RSS measurements of vehicle emissions are easily influenced by transient operating conditions of passing vehicles and multiple environmental factors, resulting in variable results for the same vehicle, further interfering with the screening of high-emitting vehicles. In this paper, an automatic hyperparametric optimization model is established in an RSS to identify high-emitting vehicles by fusing multi-feature data on environmental factors and vehicle operating conditions obtained from the RSS with the chassis and engine dynamometer test results provided by vehicle inspection stations (VISs). Qualitative and quantitative experimental results show that our model exhibits better recognition performance for high-emitting vehicles in RSSs of different times and sites, which reflects the good self-adaptability of the model. Moreover, the hyperparameters of the model do not need to be manually adjusted, so the model can be automatically trained to meet the requirements of real-time recognition scenarios for high-emitting vehicles on the road.