Improvement of the identification model for vehicles with high emissions by employing vehicle specific power
机动车比功率在高排污车辆鉴别中的应用

As a useful supplement to the inspection and maintenance (I/M) system, remote sensing is mainly for the detection of vehicles with high emissions. From experimental data in Guangzhou in 2004, we found that vehicles with high emissions, which account for 10% of the total vehicles in the city, contribute over 36.81% of the carbon monoxide (CO), 41.80% of the hydrocarbons (HC), and 48.52% of the nitrogen oxides (NOx) in the total daily emissions. This indicates that vehicles with high emissions constitute the main vehicular pollution sources. The concept of vehicle specific power is introduced and its relation with pollution emissions is analyzed. Through a comparison of 15 different measurements, vehicle emissions show great uniformity in various measurements when vehicle specific power is used as the basic metric. The distribution also indicates that the high emissions of CO or HC occur momentarily in vehicles with a high level of specific power. Thus, those vehicles should not be considered vehicles with high emissions. By applying these conclusions to improve our previously developed artificial neural network model for identifying vehicles with high emissions, the percentage of hits reached 95%.