INTELLIGENT IDENTIFICATION METHOD OF MIGRATORY ANIMALS BASED ON WEATHER RADAR

Weather radar echoes encompass precipitation, biological data, and other pertinent information, rendering them valuable tools for aeroecological monitoring. Convolutional Neural Networks (CNNs) can effectively categorize echo images, whereas Fully Convolutional Networks (FCNs) excel in pixel-level classification of organisms and precipitation within these echo images. We leveraged a comprehensive dataset replete with extensive echo data spanning a significant spatiotemporal range to train the FCN-8s network. The empirical findings underscore the remarkable capabilities of FCN-8s in elevating the average accuracy and recall rate for the biological component to levels exceeding 0.9. Achieving a balanced equilibrium between classification accuracy and recall rate holds paramount significance in augmenting the precision of aeroecological monitoring via weather radar.