Multiframe Detection via Graph Neural Networks: A Link Prediction Approach
Zhihao Lin, Chang Zheng Gao, Junkun Yan, Qingfu Zhang, Bo Chen, Hongwei Liu · IEEE Transactions on Aerospace and Electronic Systems · 2025
Multi-frame detection algorithms can effectively utilize the correlation between consecutive echoes to improve the detection performance of weak targets. Existing efficient multi-frame detection algorithms are typically based on three sequential steps: plot extraction via a relatively low primary threshold, track search, and track detection. However, these three-stage processing algorithms may result in a notable loss of detection performance and do not fully leverage the available echo information across frames. For the application of graph neural networks in multi-frame detection, the algorithms are primarily based on node classification tasks, which do not focus on directly outputting target tracks. In this paper, we reformulate the multi-frame detection problem as a link prediction task in graphs. First, we perform a rough association of multi-frame observations that exceed the primary threshold to construct observation association graphs. Subsequently, the multi-feature link prediction network is designed based on graph neural networks, which integrates multi-dimensional information, such as Doppler, signal structures, and spatio-temporal coupling of observations. By leveraging the principle of link prediction, we unify the process of track search and track detection into one step to reduce performance loss and directly output target tracks. Numerical results indicate that the proposed algorithm improves the detection performance of weak targets while suppressing false alarms, compared with traditional single-frame and multi-frame detection algorithms. Additionally, interpretability analysis shows that the designed network effectively integrates the utilized features, allowing for accurate target associations.