Abstract: Voids, porosity or pores (Collectively referred to as "porosity") are the defects with microscopic distribution characteristics that are prone to occur during the curing process of resin-matrix composites. Porosity influences mechanical properties of composite structures. Using traditional optical man-observation or pure gray-scale statistical methods, it is difficult to accurately perform a quantitative evaluation of porosity, and the testing efficiency is very low, thus affecting the accuracy of constructing an ultrasonic porosity evaluation model for composite parts, and increasing the risk of missed detection and misjudgment. It is detrimental to the quality control and safety of key composite components. In view of the complex multi-directional lay-up structure characteristics and porosity features in carbon fiber-reinforced resin-matrix composites (CFCs), a cross-modal artificial intelligent (CAI) porosity evaluation method based on the light reflection behavior in their surfaces was studied. The reflection behavior of optical waves on the areas of fibers, matrix resin, and porosity was analyzed. A CFC porosity CAI evaluation experimental system was constructed. Using different CFC porosity samples prepared under the real manufacturing process condition, the variation characterization of optical reflection signals at the surfaces of different directional fiber, resin, porosity, and interfaces were studied. The CAI evaluation model was constructed. The CAI quantitative evaluation results and effects were analyzed. The experimental results show that: Based on the signal variations and their imaging characteristics from the optical reflection behavior at the CFC surface, using the constructed CAI model and method, the porosity quantitative evaluation can be accurately carried out. The accuracy of porosity evaluation from the two frame signals reaches up to 100%. The intelligent recognition, annotation, and quantitative results list of measured porosity are completed within 0.5 second, approximately 0.14 second. The porosity testing efficiency has been greatly improved using the CAI evaluation method. The minimum porosity that can be accurately identified reaches to 5.5 μm × 6.5 μm in size and 4.5 μm in orientation width at 20 μm² porosity discrimination threshold, thus providing a fast, intelligent visual characterization, and quantitative evaluation method for CFC porosity testing.