基于改进VGG网络的单体热电池X光图像无损检测方法研究

为解决单体热电池生产中出现的安装错误、人工检测耗时耗力的问题,提出一个结合迁移学习和卷积 神 经网络(convolutional neural network,CNN) 的单体热电池缺陷检测模型。首先,对数据集图像进行裁剪、加噪等预处理,以VGG16(visual geometry group 16) 网络作为 模型的骨干架构,在瓶颈层后增添选择性核(selective kernel,SK) 卷积；然后,增添全局平均池化(global average pooling,GAP) 层, 增加Dropout层及添加 L2 正则化等微调操作,得到单体热电池缺陷检测模型Q-VGGNet；最后,在大型公开数据集ImageNet上进 行预训练学习,将获得的权重参数迁移到单体热电池图像识别模型Q-VGGNet上。测试实验表明:6种 网络模型对数据集缺陷图像的总体识别准确率分别达到了98.39%、94.44%、97.27%、96.34%、93.71%、 95.61%,Q-VGGNet网 络模型 对合格图像和 漏装负极、极耳断裂、漏装集流片3种缺陷图像 识别准确率 分别达到了99.6%,95.9%,99.6%和98.4%。检测结果表明:该方法能够更准确、快速地检测热电池缺陷, 拥有良好的缺陷诊断能力,较传统方法提高近3%,为人工检测单体热电池缺陷提供了良好的解决途径。

Research on nondestructive testing method for X-ray image of single thermal battery based on improved VGG network

To solve the problems of installation errors,time and labor consuming of manual detection,an image recognition model for single thermal battery defects based on transfer learning and convolutional neural network (CNN) is proposed.First,the images of the dataset are preprocessed by cropping and adding noise,etc.The visual geometry group 16 (VGG16) network is used as the backbone architecture of the model,and a selective kernel (SK) convolution is used after the bottleneck layer.Then, global average pooling (GAP) layer and Dropout layer are added,and L2 regularization and other fine-tuning operations are also added,an defect recognition model Q-VGGNet for single thermal battery is got.Finally,pre-training learning is performed on the dataset ImageNet,and the learned weight parameters are transferred to the model Q-VGGNet.The testing results show that the overall recognition accuracy of the six net- work models for the defect images on the dataset can reach 98.39%,94.44%,97.27%,96.34%,93.71% and 95.61%,respectively.The recognition accuracy rates of the Q-VGGNet network model for qualified images and the three types of defective images (negative electrode missing,tab broken,and current plate missing) can reach 99.6%,95.9%,99.6% and 98.4%,respectively.The results show that this method can detect thermal battery defects more accurately and quickly, and has good defect diagnosis ability.The accuracy is improved nearly 3% higher than the traditional method,and a good solution for manual detection of single thermal battery defects is provided.