Identification of Electromagnetic Interference Echoes from Weather Radar Data Using Deep Learning-Based Object Detection

As the electromagnetic environment becomes increasingly complex, electromagnetic interference (EMI) has a more significant impact on the quality of weather radar data. However, identifying EMI echoes is a substantial challenge due to small target sizes and diverse echo types. The challenge frequently results in misclassification and missed detection in operational application. To mitigate this issue, this study utilizes observational data from the Suizhou S-band weather radar to construct a dataset of EMI echo images. Subsequently, the study enhances the YOLOv8n-seg object detection model by integrating Switchable Atrous Convolution (SAC) into the C2f module of the backbone network, thereby improving the model’s ability to capture multi-scale features. The LSKA attention mechanism is combined with the SPPF module to further refine the feature extraction process. Moreover, the CBAM attention mechanism is incorporated into the target segmentation module to emphasize critical features while suppressing irrelevant interference. The dynamic Wise-IoU v2 is utilized as the bounding box regression loss function to enhance the precision of bounding box localization. Through these enhancements, the YOLO-REIE model is developed, specifically designed for identifying EMI echoes in weather radar applications. Comprehensive performance evaluation demonstrates the YOLO-REIE model’s superior detection capability across multiple metrics, including precision, recall, mAP50, and mAP50-95, compared to other models such as YOLOv8n-seg, which has an mAP50 value of 93.3%. The findings offer robust technical solutions for identifying radar EMI echoes in complex environment and establish a strong foundation for ensuring the quality of operational radar data.