Ship target fine detection method with rotation invariance

Abstract

The invention discloses a ship target fine detection method with rotation invariance, and the method comprises the steps: labeling a sample remote sensing image through a three-point labeling method; calculating to obtain position information, category information and confidence of a target in the sample remote sensing image; constructing to obtain a neural network model; performing feature extraction and recognition on the sample remote sensing image through a neural network model; updating parameters in the neural network model through a gradient descent algorithm; after the neural network model is trained for multiple times, obtaining a target refined detection model; and taking a to-be-identified remote sensing image as input of the target fine detection model, and outputting position information, category information and confidence of a target in the to-be-identified remote sensing image. The method has the capability of learning the geometric attitude information of the target, implies the orientation of the target in the triangular frame, can effectively predict the direction and position information of the target, and achieves the judgment of the types and directions of various types of military ships and civil ships.

Description

technical field [0001] The invention belongs to the technical field of remote sensing image processing, and in particular relates to a refined detection method of a ship target with rotation invariance. Background technique [0002] Target detection in remote sensing images is one of the basic tasks of satellite image processing, and its basic purpose is to extract the category and location information of interested targets from remote sensing images. This task has a wide range of applications in many fields. Detecting ships from remote sensing images is an important task, and it is also the basis for high-level applications such as remote sensing image analysis, image content understanding, and scene understanding. [0003] Since the remote sensing image is acquired from top to bottom, the appearance of the ship target is greatly affected by the viewing angle. The direction of the ship can be in any direction from 0 to 360 degrees. How to ensure that the ship can be correct...

AI Technical Summary

Problems solved by technology

Ship targets in remote sensing images often appear from an oblique perspective. Under this perspective, the rectangular frame obtained by direct target detection using the existing deep learning target detection method contains a large amount of background redundant information and overlapping areas, which cannot meet the requirements of the target detection system. Accurate target positioning and fine-grained detection requirements

Images