Abstract: In order to detect the losing fault of the bogie block key(BBK) in trouble of moving freight car detection system (TFDS), an fault recognition method for BBK losing was proposed based on Hough transform. Firstly, the improved randomized Hough transform is used to extract the feature of the axle and the endpoint. Based on the geometric relationship between the axle, the endpoint and the BBK, the mathematical model is established to calibrate the region of interest (ROI). Then the relative gradient histogram features of ROI are calculated, and the recognition of the losing fault of the BBK is realized by using the angular distance similarity measure. Experiment results demonstrate that the proposed method can recognize the fault of BBK losing on different types of bogie, and the average fault recognition accuracy reaches 94%.