Abstract: Composite materials, with their high specific strength, corrosion resistance, and design flexibility, have seen increasing adoption across aerospace, automotive manufacturing, construction, and medical fields. However, defects such as delamination, pores, and cracks frequently occur during their manufacturing process, significantly compromising material performance. Consequently, defect detection is crucial for ensuring product quality. This paper outlines the composition, performance characteristics, and application scenarios of composite materials, elucidating the necessity of defect detection. It systematically reviews and compares existing detection technologies, including non-destructive testing methods such as ultrasonic testing, laser technology, and infrared thermal imaging, as well as the principles and application limitations of traditional machine vision inspection. The paper focuses on deep learning-based composite material defect detection technology, summarizes its application bottlenecks, and provides an outlook on future development trends. As a core branch of artificial intelligence, deep learning leverages adaptive feature extraction, complex data modeling, and high-precision recognition capabilities to effectively overcome traditional limitations in composite defect detection. By harnessing its inherent advantages, deep learning drives the automation and intelligent upgrading of composite quality inspection, providing both technical support and practical guidance for quality control in the composite materials field.