Deep learning remote sensing image vessel target identification method based on threshold constraint

Abstract

The invention discloses a deep learning remote sensing image vessel target identification method based on threshold constraint, and the method comprises the following steps: (1) carrying out the threshold segmentation of a remote sensing image through an OTSU threshold segmentation method, and achieving the sea-land separation; (2) extracting shape features of the remote sensing image; (3) on the basis of a deep learning YOLOv5 algorithm, carrying out multi-scale connection fusion on the pyramid network structure at the bottom layer; (4) aiming at the characteristics of the remote sensing image ship target, designing an anchor frame according to the shape characteristics of the remote sensing image ship target; (5) introducing focusing classification loss as a loss function of a YOLOv5 algorithm to carry out regression convergence; (6) on the basis of an improved YOLOv5 algorithm, inputting a vessel sample for training to obtain a model; and (7) carrying out vessel target identification on the remote sensing image according to the trained model. The method optimizes the anchor frame and the loss function according to the shape features of the vessel, improves the generalization performance of the model, and improves the recognition precision of the vessel.

Description

technical field [0001] The invention belongs to the technical field of target recognition, and in particular relates to a deep learning remote sensing image ship target recognition method based on threshold constraints. Background technique [0002] With the rapid development of remote sensing technology, the use of deep learning to quickly and accurately identify targets from satellite images, on the one hand, can replace people in repetitive and tedious work, and free people from the heavy work of remote sensing image interpretation; On the one hand, establishing an end-to-end model structure can not only improve the processing rate of remote sensing data, but also achieve higher recognition accuracy. Introducing deep learning into target detection tasks can get rid of the constraints of manually designing detection features in traditional detection algorithms, extract relevant features through autonomous learning of the model network, and improve the detection efficiency ...

AI Technical Summary

Problems solved by technology

So far, the latest YOLO method has been iterated to YOLOv5, but because its essence is still a single-stage algorithm based on regression, it sacrifices recognition accuracy while prioritizing efficiency.

Images