Abstract: The traditional train operation at stations is mainly based on human inspection, with high labor intensity and hidden dangers of personal safety, as well as false positive and false negative errors, which makes it difficult to effectively guarantee the quality of train operation. With the application of image recognition, data mining, intelligent analysis and system integration technology, the scheme of in-station train running status monitoring system is proposed, which can automatically collect and intelligently analyze the information of train operation at the station including image, video, wheel tread temperature, vehicle running sound, etc., and provide the recognition and alarm function of vehicle abnormal status. This system can facilitate the transformation of train operation from "human inspection" to "machine inspection", "outdoor inspection" to "indoor inspection", and "static inspection" to "dynamic inspection", which is of great significance to improve the efficiency and quality of train operation at the station and ensure the safety of train operation.