As remote sensing monitoring of mobile source exhaust can be affected by the complex external environment, it is difficult to establish a correaltion model between vehicle driving conditions and pollution emissions through traditional statistical methods. For this reason, the research on the analysis of influencing factors and emission prediction based on remote sensing monitoring of mobile sources is carried out. Firstly, Spearman correlation is usedto exclude the factors that have no correlation with CO, HC and NO, the main components in emission of mobile source pollution. Secondly, Lasso algorithm is used to choose the principal influencing factors. And after principal components analysis and theselection of algorithm and architecture, the Back-Propagation (BP) neural network model is established as the optimal algorithm. Finally, the validity of the model for predicting the main components of emission of mobile source pollution is verified on the test set. The results of model prediction show that the prediction models based on feature selection and BP has high prediction accuracy, which can reduce the cost of mobile source pollution emission detection and provide theoretical basis for policy making.