Sea and land port segmentation method based on spatial and semantic alignment fusion

CN118691827BActive Publication Date: 2024-12-13SANYA SCI & EDUCATION INNOVATION PARK WUHAN UNIV OF TECH
0 Cites 0 Cited by

Patent Information

Application Number
CN202411170116.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-26
Publication Date
2024-12-13
Estimated Expiration
2044-08-26

AI Technical Summary

Technical Problem

The existing remote sensing image land-segmentation technology has missed and misaligned divisions when identifying and dividing ship targets in port areas, and cannot effectively deal with complex port backgrounds, resulting in inaccurate segmentation.

Method used

Using a method based on spatial and semantic alignment and fusion, a deep convolutional neural network and a Transformer feature extraction module combine spatial and semantic feature alignment to perform multi-scale feature fusion and segmentation prediction to improve the accuracy of sea and land port segmentation.

Benefits of technology

The precise segmentation of sea and land boundaries and small targets of port ships has been achieved, and the segmentation effect and model training accuracy of sea and land port segmentation tasks have been improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118691827B_ABST
    Figure CN118691827B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of deep learning and computer vision technology, and discloses a method for segmenting land and sea ports based on spatial and semantic alignment fusion, comprising the following steps: inputting a remote sensing land and sea port segmentation dataset image, extracting features from a deep neural network and a Transformer through a dual-branch feature encoder to obtain multi-scale features, and performing spatial and semantic feature alignment and fusion on the features, and optimizing the dual-branch feature alignment process through back propagation. Multi-scale feature fusion is performed on the fused decoder features, and an auxiliary loss function is used for fusion supervision and back propagation. The fused features are sampled to the original image resolution size for segmentation prediction. Back gradient propagation is performed through a segmentation loss function to train and update network model parameters, improve the accuracy of land and sea port boundary segmentation, and improve the training accuracy of the network model in remote sensing land and sea port image segmentation.
Need to check novelty before this filing date? Find Prior Art