目的 针对现有钢材缺陷识别算法特征图利用不充分、识别准确率低、参数量大等问题，基于脉冲神经网络，提出一种用于钢材缺陷识别的稠密卷积脉冲神经网络(DCSNN)模型，减少系统消耗和内存占用。方法 首先，采用卷积编码，对输入图片进行特征提取和编码。其次，采用稠密连接算法搭建稠密卷积脉冲神经网络，实现特征重复利用，抑制梯度消失，并通过替代梯度下降算法进行网络训练。最后，在带钢数据集上进行测试，实现带钢缺陷识别。结果 实验结果显示，DCSNN在测试集上的准确率为98.61%，参数量为0.5万，结论 在钢材表面缺陷识别问题上表现出良好效果。

The work aims to propose a dense convolutional spiking neural network (DCSNN) model for steel defect identification based on spiking neural network aiming at the problems of insufficient utilization of feature images, low recognition accuracy and numerous parameters of existing steel defect identification algorithms, so as to reduce system consumption and memory occupation. Firstly, convolutional coding was used to extract and encode the features of the input images. Secondly, the dense convolutional spiking neural network was constructed by the dense connection algorithm to realize the reuse of features and suppress the disappearance of gradients. Then, the network was trained by alternative gradient descent algorithm. Finally, the test was carried out on the strip steel dataset to realize the defect identification of strip steel. The experimental results indicated that the accuracy of DCSNN on the test set was 98.61% and the number of parameters was 5 000. The proposed model shows a good effect on the identification of steel surface defects.