The gaseous pollutant BC, CO and NO_x in atmosphere are mainly emitted by vehicle exhaust at traffic dense metropolitan areas. It looks like that BC, CO and NO_x are homologous in pollution sources and often used as fingerprints of traffic source at traffic dense metropolitan areas. However, these gaseous pollutants have also other significant pollution sources, such as power production sources, biomass burning, domestic cooking and natural process. Identification in the source categories of these three gaseous pollutants at traffic dense metropolitan areas needs more accurate methods. Here, detrended fluctuation analysis and multifractal methods are applied to homology analysis of BC, CO and NO_x pollution sources based on the observational data from 1 July, 2015 to 30 June, 2016 for a traffic site in Hong Kong. The results show that BC, CO and NO_x pollution are characterized by different long term correlations and multifractality. Long term effects of the past pollution emissions on pollution sources identification are very important, which is always neglected in previous studies. By K-means clustering method, a novel measure is proposed for describing the differences among the BC, CO and NO_x pollution sources quantitatively. The comparison results suggest that fractal parameters can be new quantitative indexes reflecting the differences among BC, CO and NO_x pollution sources. Multifractal parameters of air pollutants can be new benchmarks for testing the simulation results by air pollution sources identification. This work is helpful for the further improvement of air pollution sources identification methods.