针对传统无损检测技术在罐车内壁裂纹检测中效率低、抗干扰能力差等问题，提出一种基于卷积神经网络的热成像裂纹识别方法。研制了一种滚动式电加热棒作为热激励源，并采用新的激励方式对被检测表面进行热激励；根据热量传输过程中遇到裂纹时温度产生异常的原理，对被检测表面裂纹进行判断；采集热激励后的红外热图像作为训练样本，并搭建 5层卷积神经网络对样本进行训练。实验表明，利用红外热成像与卷积神经网络可以对裂纹进行准确识别；检测效率高、鲁棒性强；并且在测试集上识别准确率达到 96.50%。

To solve the problems of low efficiency and poor anti-jamming ability in the detection of cracks on the inner wall of truck tanks using traditional non-destructive testing techniques, this paper proposes a thermal imaging crack recognition method based on convolutional neural network (CNN). A rolling electric heating rod was developed as a thermal excitation source, and a new excitation method was used to thermally stimulate the surface to be inspected. According to the principle of abnormal temperature generated during the heat transfer process, the surface crack was detected. The thermally excited infrared thermal images are used as training samples, and a six-layer CNN was built to train on the samples. Experiments show that infrared thermal imaging and the CNN can accurately identify the cracks. The detection efficiency is high and the model is robust. Furthermore, the recognition accuracy on the test set reaches 96.50%.