Semantic change parsing finds detailed land-cover changes from remote images
QSCP: Beyond Class-Name Prompts for Query-Guided Semantic Change Parsing
Computer Vision and Pattern Recognition
Summary
Detecting changes in land cover from satellite images usually just shows where something changed or what the new land type is. The authors created a method that understands more detailed questions like what changed from forest to urban area, not just where. Their system can handle simple categories, synonyms, or even sentence descriptions to return maps showing what exactly changed over time. This helps people get clearer, specific insights into how places have changed based on tailored queries.
What this means in practice
- •For urban planners: Identify specific land-use transitions such as forest to built-up areas from satellite images using natural language queries for detailed planning insights.
- •For environmental monitoring teams: Generate paired temporal maps showing exact land-cover changes requested via category names or descriptive sentences to support environmental assessments.
Authors
Yuan Qian, Jie Ma
Abstract
Traditional change detection (CD) identifies changes between bi-temporal remote sensing images, while semantic change detection (SCD) assigns predefined land-cover classes. However, mapping all changes may not meet a user's specific needs. Referring change detection (RCD) enables selective retrieval through category prompts. However, existing category-prompted RCD uses the queried category to specify the destination of a change and returns only a binary mask of the corresponding regions. Users may instead request a particular transition and paired semantic maps to understand what changed into what. Such requests require explicit source and target reasoning beyond target-class localization. To address these needs, we propose query-guided semantic change parsing (QSCP), which supports category names, synonyms, and intent-bearing sentences and returns a query-specific mask with paired temporal semantic maps. QSCP parses requests into intents and semantic slots, composes bidirectional visual evidence, and predicts both temporal states with a query-conditioned decoder. On SECOND, QSCP outperforms RCDNet on synonym, sentence, and transition queries and improves end-to-end semantic prediction over evaluated semantic baselines. WHU-CDC experiments further assess cross-dataset transfer and consistency across equivalent expressions without target-domain training. Code is available at https://github.com/qianyuancs/QSCP