Detection of small targets in aerial imagery plays a crucial role in fields like intelligent traffic management and emergency rescue operations. However, the high-density distribution of small scale targets and complex background interference in aerial imagery severely degrade the performance of existing detection models. To address these problems, we propose the Edge Guided Spatial Frequency Interactive Detection Transformer (ESFI-DETR), a robust transformer framework for small target detection with high precision. Firstly, the Small Target Feature Enhancement Block (STFE) enhances the multi scale feature extraction capability and inference efficiency by leveraging reparameterized partial convolution and contextual anchor attention mechanisms, while simultaneously reducing the parameter count. Secondly, we design the Edge guide Multi Scale Enhancer (EMSE) model, which can extract the edge information in the P2 layer and project it to various parts of the network, enhancing the detection ability of the model in occlusion and complex environments. Thirdly, The Micro Scale Feature Pyramid Network (MS-FPN) is developed by integrating the SPD algorithm with a novel Hierarchical Spatial-Frequency Adaptive Fusion module for spatial frequency dual-domain interactions, enhancing multi-scale feature fusion capabilities while maintaining real-time computational efficiency. Finally, the Inner-MPD IOU mechanism effectively addresses gradient vanishing and background ambiguity in small target detection through joint optimization of primary/auxiliary bounding boxes and contextual feature interactions. Experimental results on public benchmarks demonstrate that ESFI-DETR significantly outperforms existing methods, achieving substantial reductions in false positives and missed detections for occluded small targets under complex aerial scenarios.