Abstract: The defect detection for the metal shell of solid rocket motor is essential in the assessment of motor conditions. A machine vision-based surface defect detection method for the motor metal shell was proposed in this paper. The detection was achieved using data enhancement technique based on deep convolutional generative adversarial networks (DCGAN), as well as target detection technique based on convolutional block attention module (CBAM). Firstly, on the basis of the existing small amount of real shell defect data, the dataset was enhanced by data enhancement technique. Then, the deep learning algorithm was applied to identify and classify defective targets. Finally, the defect detection for the motor shell was realized. The feasibility of the method was verified through testing various types of shell defects. The results demonstrate a 5.7% enhancement in identification accuracy compared to conventional detection methods, while maintaining the detection efficiency. The model also shows good robustness and generalization. The proposed method may have a good application prospect in practical engineering.