A Biological Echo Identification Method with Single Polarized Weather Radar Using Super-Pixel Segmentation

Weather radar networks, given their advantages of continental coverage and year-round operation, show great potential in the study of large-scale aerial biological migration. The precise identification and extraction of biological echoes from vast amounts of weather data serve as the essential metrics for assessing the biological monitoring capability of weather radar systems. However, complex weather conditions pose an obstacle to extracting biological information accurately. This study introduces super-pixel segmentation approach in the domain of weather radar-based biological recognition. A cluster-based method is proposed to extract biological information from localized spatial weather radar rendered images. From the perspective of colour space, we confirm the necessity and evident benefits of differentiating them at local spatial level. We collected 750,000 historical records from 108 weather radars nationwide and handpicked 36,000 representative migration cases for labelling, creating a large-scale trainable dataset for studying aerial ecological migration patterns with weather radars. We apply the proposed method to two typical scenes, and the results demonstrate that the accuracy of the proposed method is above 89.38 % in the whole scenes.