| 摘 要: 针对安检领域历史集装箱扫描图像量越来越大，分析手段单一、图像资源未能充分利用等问题，设计一种融合非结构化图像特征与关联单证结构化信息的图像智能分析系统；依托向量数据库与结构化数据库，实现海量历史扫描图像信息存储；应用孤立森林（iForest）方法进行智能分析，实现异常检测和离群点检测。实验结果表明，该系统能够对海量机检扫描图像进行智能分析和可视化展示，为事后审计和风险追溯提供数据支撑，具有实际意义。 |
| 关键词: 机检扫描 特征提取 图像智能分析 孤立森林 离群点检测 风险追溯 |
|
中图分类号:
文献标识码:
|
|
| Design and Development of an Image Intelligent Analysis System Fusing Unstructured Features with Associated Structured Information |
|
LIWEI
|
NUCTECH COMPANY LIMITED
|
| Abstract: To address challenges in the security inspection field—including the growing volume of historical container scan images, limited analytical methods, and underutilized image resources—this study designs an intelligent image analysis system that integrates unstructured image features with structured information from associated documents. Leveraging vector databases and structured databases, it enables storage of massive historical scan image data. Support Isolation Forest (iForest) methods are applied for intelligent analysis, achieving anomaly detection and outlier identification. Experimental results demonstrate that the system can intelligently analyze and visually present massive machine-inspected scan images, providing data support for post-event auditing and risk tracing, thereby holding practical significance. |
| Keywords: Machine scanning, Feature extraction, Image intelligent analysis, iForest, Outlier detection, Risk tracing |
