A multi-scale ship target detection method based on spatial channel serial attention

By proposing a multi-scale ship target detection method based on spatial channel serial attention, the problems of low accuracy and feature degradation in multi-scale ship target detection are solved, achieving efficient and accurate multi-scale ship target detection and improving detection accuracy and model adaptability.

CN122415994APending Publication Date: 2026-07-17DALIAN MARITIME UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610618064.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-05-07
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Existing technologies have low accuracy in multi-scale ship target detection, which cannot meet the needs of real-time detection. Furthermore, the similarity in color between ships and the ocean leads to features such as low contrast and blurred edges in the images, which degrades the features. Water surface fog and special geometric limits further weaken the feature representation of target detection.

Method used

A multi-scale ship target detection method based on spatial channel serial attention is adopted. By establishing a heavy parameter feature extraction block, a divide-and-conquer feature fusion block and spatial channel serial attention, a multi-scale ship target detection network is built. Pixel threshold constraints and lightweight convolution are introduced to enhance the model's detection capability in different scenarios.

Benefits of technology

It improves the accuracy of multi-scale ship target detection, enhances the model's deployment adaptability in embedded devices with limited storage space, improves detection capabilities in different scenarios, and enhances accuracy and AP metrics.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122415994A_ABST
    Figure CN122415994A_ABST
Patent Text Reader

Abstract

本发明提供一种基于空间通道串行注意力的多尺度船舶目标检测方法,包括:对公开数据集进行处理,建立存在多尺度船舶目标的图像数据集,并对数据集进行划分;设计重参数特征提取块,使用轻量卷积重塑卷积模块为重参数特征提取块,搭建应用卷积模块、重参数特征提取块与金字塔池化层的骨干网络;使用空间通道串行注意力重塑跨阶段局部融合块为分治式特征融合块,搭建应用分治式特征融合块的颈部网络;整合网络框架,搭建基于空间通道串行注意力的多尺度船舶目标检测网络;将检测网络在所述数据集进行训练得到多尺度船舶目标检测模型,部署训练所得模型并调整超参,用于多尺度船舶目标检测任务。本发明能够有效提高模型对多尺度船舶图像的特征提取能力及回归能力以提升检测精度。
Need to check novelty before this filing date? Find Prior Art