High-resolution remote-sensing image ship target detection method based on deep learning

Abstract

Provided is a high-resolution remote-sensing image ship target detection method based on deep learning. The method includes steps: acquiring a high-resolution remote-sensing image including a ship target, and performing artificial marking on the ship in the image by employing a minimum quadrangle capable of surrounding the ship target; setting a structure of a deep learning network model, connecting feature maps with different semantic properties and resolutions, then setting multiple reference quadrangles for positions of the feature maps, and classifying the reference quadrangles by employing a full connection layer; and training the network model, inputting a to-be-detected high-resolution remote-sensing image to the trained network model, detecting a ship area in the image, and constructing a minimum quadrangle surrounding the ship target so that the accurate position of the ship target is indicated. According to the method, compared with the conventional ship target detection method, the interference of environment factors can be more effectively resisted, and ships in different locations and with different dimensions and attitudes in the high-resolution remote-sensing image can be detected more stably and accurately.

Description

technical field [0001] The invention belongs to the field of high-resolution remote sensing image ship target detection, and specifically relates to a high-resolution remote sensing image ship target detection method based on deep learning, which can effectively resist the interference of environmental factors and stably and accurately detect high-resolution remote sensing images. Ships in different scenes, with different sizes and attitudes in the image are detected. Background technique [0002] Ship target detection technology mainly refers to the process of automatically detecting ship targets from images using appropriate target extraction and recognition methods. Efficient and high-precision detection of ships is of great significance in route planning, military target reconnaissance, and marine environment monitoring. Therefore, ship target detection has important research value. [0003] With the launch of many high-resolution satellites and the improvement of the q...

AI Technical Summary

Problems solved by technology

[0005] Whether it is the inherent features of the ship or the features extracted based on the general feature description operator, these features rely on manual design, and are local and low-level features. They have weak semantics and poor robustness, and are difficult to resist different lighting, weather, clouds, The interference brought by environmental factors such as waves to the ship target detection task is prone to false detection and missed detection.

In addition, the existing ship target detection algorithms can usually only detect ship targets within a specific scale range, and ships in remote sensing images include civilian yachts, cargo ships, and military warships, etc. Different types of ship sizes The difference is huge, and the existing methods cannot flexibly adapt to changes in the size of the ship in the image

In addition, the existing ship detection algorithm indicates the position of the ship target by constructing a horizontal rectangular bounding box. Boxes do not give an accurate indication of the actual location of the ship's target

Images