Aerial View River Landform Video segmentation: A Weakly Supervised Context-aware Temporal Consistency Distillation Approach

2025-11-20

AI Summary

Addressing the challenges of label scarcity, poor temporal consistency, and heavy reliance on fully supervised data in river geomorphology segmentation from UAV remote sensing video, this paper proposes a weakly supervised, context-aware temporal consistency knowledge distillation framework. The method introduces a teacher–student architecture driven by dynamic key-frame selection and updating—the first to incorporate temporal consistency knowledge distillation into weakly supervised aerial image segmentation. It jointly models cross-frame semantic coherence and motion consistency by integrating a context-aware module with explicit temporal constraints. Using only 30% labeled data, the approach achieves significant improvements over state-of-the-art weakly supervised methods in both mean Intersection-over-Union (mIoU) and Temporal Consistency (TC) metrics. This enables robust, continuous geomorphological localization and segmentation in complex aerial scenes.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Semi-Supervised Learning
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

The study of terrain and landform classification through UAV remote sensing diverges significantly from ground vehicle patrol tasks. Besides grappling with the complexity of data annotation and ensuring temporal consistency, it also confronts the scarcity of relevant data and the limitations imposed by the effective range of many technologies. This research substantiates that, in aerial positioning tasks, both the mean Intersection over Union (mIoU) and temporal consistency (TC) metrics are of paramount importance. It is demonstrated that fully labeled data is not the optimal choice, as selecting only key data lacks the enhancement in TC, leading to failures. Hence, a teacher-student architecture, coupled with key frame selection and key frame updating algorithms, is proposed. This framework successfully performs weakly supervised learning and TC knowledge distillation, overcoming the deficiencies of traditional TC training in aerial tasks. The experimental results reveal that our method utilizing merely 30% of labeled data, concurrently elevates mIoU and temporal consistency ensuring stable localization of terrain objects. Result demo : https://gitlab.com/prophet.ai.inc/drone-based-riverbed-inspection

Problem

• Weakly supervised segmentation of aerial river landform videos
• Addressing temporal consistency in UAV terrain classification
• Reducing annotation needs while maintaining localization accuracy

Innovation

• Teacher-student architecture for knowledge distillation
• Key frame selection and updating algorithms
• Weakly supervised learning with temporal consistency