Marine rescue method based on improved YOLOV4 target detection

The invention discloses a sea rescue method based on improved YOLOV4 target detection, and belongs to the field of sea rescue. The image preprocessing (rain removal and defogging) method based on improved pixel-by-pixel filtering is used for processing an image collected by an unmanned aerial vehicle through a carried high-definition camera, a shot blurred image is converted into a clear image, and effective data is provided for subsequent image detection. An improved YOLOV4-based target detection algorithm is used for accurately identifying people falling into water, and a feature pyramid converter (FPT) is used for improving the image semantic extraction capability of ocean floaters (to-be-rescued people) with small sizes, so that the success rate of search and rescue is improved. A feature pyramid converter (FPT) is added behind a traditional feature pyramid structure, a feature map of small target semantic information is enhanced, and detection of small target objects on the sea is facilitated. 本发明公开了一种基于