Small target detection method and system for images captured by unmanned aerial vehicle based on YOLO

The invention discloses a small target detection method and system for an unmanned aerial vehicle captured image based on YOLO, and the method comprises the steps: downloading an image target detection data set from a VisDrone2019 challenge competition, and carrying out the preprocessing of the data set; constructing a reference model and loading a pre-training weight of the model; a multi-level feature aggregation center module is added to the neck of the reference network model, the multi-level feature aggregation center module comprises a high-level feature aggregation branch, a middle-level feature aggregation branch and a low-level feature aggregation branch, and each branch comprises a feature alignment module, a weighted feature fusion module and a global information extraction module; a refined feature detection head is added to improve the detection performance of the model on a small target; and taking the training set as input, training the improved network model, and carrying out small target dete