基于深度学习的液氯罐车射线图像缺陷自动识别研究  

Research on automatic recognition of defects in radiographic images of liquid chlorine tankers based on deep learning

作　　者：李兴红 LI Xinghong(Gansu Zhongtian Chemical Co.,Ltd.,Linxia 731601,China)

机构地区：[1]甘肃中天化工有限责任公司,甘肃临夏731601

出　　处：《中国高新科技》2024年第12期15-17,共3页

摘　　要：由于液氯罐车射线图像可能包含多种不同类型的缺陷,并且每种缺陷在图像中的表现形式也可能存在较大的差异,导致识别各类缺陷个数较低的现象。针对上述现象,提出基于深度学习的液氯罐车射线图像缺陷自动识别研究。对液氯罐车射线图像进行预处理,消除噪声、增强图像特征,对深度学习模型中的卷积层、池化层、全连接层以及分类器进行设计,建立液氯罐车射线图像缺陷识别模型,对模型进行训练,采用前向传播算法计算模型的输出,利用反向传播算法逐层回溯误差,优化模型的权重参数,采用自适应学习策略,迭代训练,实现液氯罐车射线图像缺陷的自动识别。实验结果表明,研究方法能够有效识别出液氯罐车射线图像中的多种缺陷形式,识别各类缺陷个数较高。The Radiographic image of liquid chlorine tanker may contain many different types of defects,and the manifestation of each defect in the image may be quite different,which leads to the phenomenon of low number of defects identified.Aiming at the above phenomenon,the research on automatic recognition of defects in Radiographic images of liquid chlorine tank cars based on deep learning is proposed.The Radiographic image of liquid chlorine tanker was preprocessed to eliminate noise and enhance image characteristics.The convolution layer,pool layer,full connection layer and classifier in the deep learning model were designed,and the defect identification model of liquid chlorine tanker Radiographic image was established,and the model was trained.The output of the model was calculated by forward propagation algorithm,and the error was traced back layer by layer by backward propagation algorithm,and the weight parameters of the model were optimized.The automatic defect identification of liquid chlorine tanker Radiographic image was realized by adopting adaptive learning strategy and iterative training.The experimental results show that the research method can effectively identify various types of defects in the Radiographic image of liquid chlorine tanker,and the number of defects is high.

关 键 词：深度学习 液氯罐车 射线图像 图像缺陷 缺陷自动识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]