Aiming at the key steps of hydraulic pump fault diagnosis such as feature extraction and pattern recognition, a fault diagnosis method of hydraulic pump based on symmetrized dot pattern（SDP） image feature and adaptive nearest neighbor classifier（ANNC） is introduced in this paper. Firstly, the time waveform was transformed to snowflake image in polar coordinate space by symmetrized dot pattern（SDP） method. And then, image shape feature which used as fault feature was extracted on the basis of analyzing the characteristic of the image. Finally, the ANNC was employed to classify and evaluate the fault feature. The experiment results of hydraulic pump fault diagnosis shown that the image shape feature can represent the fault states, and the ANNC increased the classification accuracy and validated the effectiveness of the proposed method.