ZHEJIANG UNIVERSITY (20240331165). CROSS-DOMAIN REMOTE SENSING IMAGE SEMANTIC SEGMENTATION METHOD BASED ON ITERATIVE INTRA-DOMAIN ADAPTATION AND SELF-TRAINING simplified abstract

CROSS-DOMAIN REMOTE SENSING IMAGE SEMANTIC SEGMENTATION METHOD BASED ON ITERATIVE INTRA-DOMAIN ADAPTATION AND SELF-TRAINING

Organization Name

ZHEJIANG UNIVERSITY

Inventor(s)

JIANWEI Yin of HANGZHOU, ZHEJIANG PROVINCE (CN)

YUXIANG Cai of HANGZHOU, ZHEJIANG PROVINCE (CN)

YINGCHUN Yang of HANGZHOU, ZHEJIANG PROVINCE (CN)

YONGHENG Shang of HANGZHOU, ZHEJIANG PROVINCE (CN)

ZHENQIAN Chen of HANGZHOU, ZHEJIANG PROVINCE (CN)

ZHENGWEI Shen of HANGZHOU, ZHEJIANG PROVINCE (CN)

CROSS-DOMAIN REMOTE SENSING IMAGE SEMANTIC SEGMENTATION METHOD BASED ON ITERATIVE INTRA-DOMAIN ADAPTATION AND SELF-TRAINING - A simplified explanation of the abstract

This abstract first appeared for US patent application 20240331165 titled 'CROSS-DOMAIN REMOTE SENSING IMAGE SEMANTIC SEGMENTATION METHOD BASED ON ITERATIVE INTRA-DOMAIN ADAPTATION AND SELF-TRAINING

Simplified Explanation: This patent application describes a method for semantic segmentation of remote sensing images across different domains using iterative intra-domain adaptation and self-training.

Key Features and Innovation:

• Training source-target inter-domain adaptation models
• Generating target domain category segmentation probabilities and pseudo labels
• Sorting target domain image segmentation probability credibility scores
• Training target intra-domain iterative domain adaptation models
• Generating target domain segmentation results

Potential Applications: This technology can be applied in various fields such as agriculture, urban planning, environmental monitoring, and disaster management for accurate image segmentation.

Problems Solved: This technology addresses the challenges of semantic segmentation in remote sensing images across different domains by reducing inter-domain and intra-domain differences.

Benefits:

• Improved accuracy in cross-domain remote sensing image semantic segmentation
• Efficient removal of erroneous pseudo labels during self-training
• Enhanced segmentation effects in target domain images

Commercial Applications: The technology can be utilized in industries such as agriculture, urban development, environmental monitoring companies, and disaster response organizations for precise image analysis and decision-making.

Prior Art: Prior research in the field of remote sensing image segmentation and domain adaptation can provide valuable insights into the development of this technology.

Frequently Updated Research: Stay updated on advancements in remote sensing image analysis, domain adaptation techniques, and self-training methods to enhance the performance of this technology.

Questions about Remote Sensing Image Semantic Segmentation: 1. How does this technology improve the accuracy of semantic segmentation in remote sensing images? 2. What are the potential applications of this technology in disaster management scenarios?

By utilizing iterative intra-domain adaptation and self-training, this patent application introduces a novel method for semantic segmentation of remote sensing images across different domains.

Original Abstract Submitted

a cross-domain remote sensing image semantic segmentation method based on iterative intra-domain adaptation and self-training, includes training source-target inter-domain domain adaptation models, generating target domain category segmentation probabilities and pseudo labels, sorting target domain image segmentation probability credibility scores, training target intra-domain iterative domain adaptation models, and generating target domain segmentation results. the invention utilizes source-target domain inter-domain adaptation to reduce source-target domain inter-domain differences, and, utilizes target intra-domain adaptation to reduce target intra-domain differences and improve the accuracy of cross-domain remote sensing image semantic segmentation models; furthermore, by proposed classifying and sorting target domain images based on segmentation probability credibility, prediction results with good segmentation effects are selected as pseudo labels; meanwhile, a new pseudo label screening strategy is proposed to remove pixel points that are highly likely to be erroneous from pseudo labels, thereby avoiding the impact caused by erroneous pseudo labels during self-training within target domains.