Composite materials are indispensable in the contemporary aerospace sector, with their quality inspection being a critical component. Phased array C-scan, a form of ultrasonic non-destructive testing, is adept at identifying flaws within these materials. Despite its effectiveness, the large surface area of aerospace composites makes manual evaluation of the C-scan images both time-consuming and prone to oversight. To address this, the present study introduces a novel dual-modality image fusion model that leverages both amplitude and depth C-scan images. This model independently extracts features from each image modality and subsequently fuses these features through a weighted summation approach. Our findings indicate that this model boasts an impressive accuracy rate of 97.8%, outperforming single-modality intelligent recognition models by a significant margin. Additionally, when compared to alternative feature fusion techniques, our method demonstrates superior precision, thereby enhancing the reliability of defect detection in composite materials.