A modified GIP method based on atom searching for selecting outlier samples in STAP
Yunfei Hou, Yingnan Zhang, Wenzhu Gui, Minghai Wang, Wei Dong · Remote Sensing Letters · 2025
Reliable training samples are crucial for space-time adaptive processing (STAP), and the presence of outliers in samples can significantly reduce clutter suppression performance. Additionally, when multiple outliers are present in the training samples, the detection performance of traditional generalized inner product (GIP) methods declines markedly. To enhance the GIP’s ability to detect outliers, a GIP outlier detection method based on atom searching (AS) is proposed. First, a set of suitable atoms is searched using a metaheuristic algorithm based on a fitness function. Second, a fitting residual threshold is set according to the characteristics of clutter atom distribution to eliminate irrelevant atoms, thereby constructing an accurate clutter subspace. Finally, based on the obtained clutter subspace, a GIP detector is constructed. Simulation results show that when the clutter subspace is accurately obtained, AS-GIP is more effective and sensitive than other methods.