Swarm Fuzzy Feature Lightweight Network Algorithm for Flying Birds Recognition Application

According to the problems of long distance, unobvious features of the flying birds in the air of the airport, recognition for these flying birds is difficult. A swarm fuzzy features lightweight network algorithm is supposed to be applied for flying birds recognition. Firstly, the dynamic K-Means + + algorithm is used to cluster the target samples in the data, which is set to obtain anchor frames that meet different target scales. Then, a full-dimensional dynamic convolution module is added to the backbone network of YOLOv5 target detection, and recognition to dynamically adjust the size and shape of the convolution kernel. Moreover, different features are separated and filtered by association integral method, these feature channels with no obvious recognition features are cut off, that compressed computational complexity and optimize the detection accuracy by reducing the amount of calculated information.