Since the pollution gases produced by on-road high-emitters are significantly harmful to the environment, it is of great significance to identify on-road high-emitters accurately. However, there is still a relatively large promotion space for identifying high-emitters in model selection, evaluation metrics, recognition performance, and other aspects, for both traditional cutpoint-based methods and emerging artificial intelligence-based methods. Therefore, to address the above issues, a method for on-road high-emitters identification is proposed based on mixed kernel extreme learning machine. The method uses mobile source telemetry data obtained from on-road remote sensing detection equipment, and integrates different kernel functions on the basis of kernel extreme learning machine, which can improve the robustness of the model and the recognition performance of on-road high-emitters. The experimental results on remote sensing data collected from the traffic network in Shushan District of Hefei City, China, show that, compared with the other methods, the proposed method has a higher F1-Score, lower missing alarm rate and false alarm rate, which confirms the effectiveness of the method in high-emitter identification. It is indicated that the proposed method can help to identify high-emitters in the traffic road network efficiently and provide basic support for further improving urban air quality.