The barrel had some features such as small diameter, thin wall and arc trough on inner wall etc，which made it difficult for ultrasonic detecting. And the waves reflected from the inside rifling have badly interfered the identification of echo signal. Compared with the defect echo and rifling wave, the defects were identified through processing method of extracting double characteristic values of time and amplitude of echo signal, it is used to eliminate the interference of rifling wave. For overcoming the disadvantage of single probe data, the local detection results of the two probes were synthesized by data fusion technology, so that the accuracy of defect detection was improved.