激光超声表面缺陷检测的过程中，缺陷的定量表征通常依赖于操作者的判断，易受到人为因素干扰，致使检测结果不稳定。针对这一问题，提出一种基于图像识别的二维卷积神经网络（2D-CNN）的缺陷自动分类检测方法。利用有限元方法模拟激光超声检测过程，并采集超声信号数据用于训练分类模型；使用连续小变换（CWT）处理超声信号得到小波时频图，以小波时频图作为输入训练卷积神经网络（CNN）分类模型，实现对表面缺陷深度的自动分类。验证结果表明：提出的检测方法能够对不同深度的缺陷准确分类，测试的平均准确率达到97.3%；构建的CNN分类模型能够自主学习输入图像的缺陷特征并完成分类，提高了检测结果稳定性，为激光超声缺陷检测的自动化分析处理提供了新的思路。

In the process of the laser ultrasonic surface defects detection, the quantitative characterization of the defects mainly depends on the operator's judgment, and it is easily interfered by the human factors, which leads to the unstable detection results. To solve this problem, an defects automatic classification detection method based on the two-dimensional convolutional neural network (2D-CNN) for image recognition was proposed. The finite element method was used to simulate the laser ultrasonic detection process, and the ultrasonic signal data was collected for training the classification model; the continuous wavelet transformation (CWT) was used to process the ultrasonic signal to obtain the wavelet time-frequency images, and the images were used as inputs to train the convolutional neural network (CNN) classification model to realize the automatic classification of the surface defects depth. The verification results show that the proposed detection method can accurately classify the defects of different depths, and the average accuracy rate of the test reaches 97.3%; the constructed CNN classification model can independently learn the defects features of the input images and complete the classification, which improves the stability of the test results, and provides a new idea for the automatic analysis and processing of laser ultrasonic defects detection.