针对传统交流电磁场检测技术中缺陷识别困难、智能化程度不高等问题,提出一种基于单次多盒检测器(SSD)的交流电磁场缺陷智能识别方法。首先通过仿真建立不同类型的缺陷可视化成像数据库,使用数据增强算法对数据库进行扩充,提高数据库的泛化能力;然后基于SSD算法建立交流电磁场缺陷智能识别方法,为缺陷智能评估与缺陷判定奠定基础;最后利用不同类型缺陷检测试验验证了该方法的缺陷识别效果。试验结果表明,基于SSD的交流电磁场缺陷智能识别方法能够正确识别不同类型缺陷,识别准确率达98%,检出缺陷的置信度均在90%以上,可为结构物缺陷的智能识别与智能评估提供方法支撑。

Aiming at the problems of difficulty in defect identification and low level of intelligence in traditional alternating current field measuremen (ACFM), this paper proposes a method for defects intelligent identification of ACFM based on single shot multibox detector (SSD). Different types of defect visualization imaging databases were established through simulation models and experiment, and data enhancement algorithms was used to expand the database to improve the generalization ability of the database. Defects intelligent identification method of ACFM based on SSD may lay a foundation for defects intelligent evaluation and defects judgment. Different types of defect testing experiments were carried out to verify the efficiency of the defects intelligent recognition method of ACFM based on SSD. The experimental results showed that the defects intelligent identification method of ACFM based on SSD could correctly identify different types of defects. The accuracy of the recognition was 98%, and the confidence of detecting the defects was above 90%. The method can provide support for the intelligent identification and evaluation of structural defects.