| 引用本文： |
-
张海瀛,刘鑫,王明阳,等.采用强制检验-证据增强的跨时空多源异构数据关联方法[J].电讯技术，2026，（4）：647 - 653. [点击复制]
- ZHANG Haiying,LIU Xin,WANG Mingyang,et al.A Method for Cross-temporal and Cross-spatial Multi-source Heterogeneous Data Association Using Mandatory Inspection and Evidence Augmentation[J].，2026，（4）：647 - 653. [点击复制]
|
|
| 摘要: |
| 在跨时空多源异构数据关联问题中，由于大时间跨度导致的目标跨时空情况特别突出，传统的基于预测-判别和新型的基于深度嵌入模型的关联方法均难以适用，是亟待解决的难题。针对大时空转移条件下目标跨时空多源异构数据关联问题，提出一种采用强制检验-证据增强的跨时空多源异构数据关联方法，考虑目标物时空约束，利用多源异构数据之间的共有信息，包括属性、速度、位置、航向等，构建强制检验和证据增强关联特征，实现关联判决。基于仿真数据的实验结果表明，相比于单一维度的关联方法，该方法可以显著提高跨时空多源异构数据关联的命中率和召回率，〩its@1和Hits@5命中率分别提高了至少11%和4%，达到了99%和100%，R@1和R@5召回率分别提高了至少11%和9%，达到了99%和87%。此外，构建的跨时空多源异构数据关联框架，允许用户快速拓展新的关联特征，如语义关联特征、外观关联特征等，具有良好的可拓展性。 |
| 关键词: 数据关联 跨时空数据关联 多源异构数据关联 强制检验 证据增强 |
| DOI：10.20079/j.issn.1001-893x.241115004 |
|
| 基金项目: |
|
| A Method for Cross-temporal and Cross-spatial Multi-source Heterogeneous Data Association Using Mandatory Inspection and Evidence Augmentation |
| ZHANG Haiying,LIU Xin,WANG Mingyang,WANG Chenggang |
| (Southwest China Institute of Electronic Technology，Chengdu 610036，China) |
| Abstract: |
| In the context of cross-spatiotemporal multi-source heterogeneous data association，the cross-spatiotemporal situation of targets arising from large temporal spans poses a particularly prominent challenge.Both traditional prediction-discrimination-based and novel deep embedding model-based association methods struggle to apply effectively，highlighting an urgent need for solutions.Addressing the issue of cross-spatiotemporal and multi-source heterogeneous data association for targets under the condition of large spatiotemporal shifts，a method using mandatory inspection and evidence augmentation-based is proposed.This method considers the spatiotemporal constraints of targets and leverages the shared information across multi-source heterogeneous data，including attributes，speeds，positions，headings，etc.，to construct mandatory inspection and evidence augmentation association features，thereby enabling association decisions.Experimental results based on simulation data demonstrate that，compared with single-dimensional association methods，the proposed method significantly improves the hit rate and recall rate of cross-spatiotemporal multi-source heterogeneous data association，the hit rates of Hits@1 and Hits@5 are increased by at least 11% and 4% respectively，reaching 99% and 100%；the recall rates of R@1 and ㏑@5 are increased by at least 11% and 9% respectively，reaching 99% and 87%.Furthermore，the constructed cross-spatiotemporal multi-source heterogeneous data association framework allows users to rapidly extend new association features，such as semantic association features and appearance association features，showcasing excellent scalability. |
| Key words: data association cross-spatiotemporal data association multi-source heterogeneous data association mandatory inspection evidence augmentation |
《电讯技术》编辑部 