PhenoStitch: Training-Free Panoptic Crop Mapping from Satellite Image Time Series

2026-08-01

AI Summary

This work addresses the poor generalization of existing approaches to panoptic crop mapping, which often rely on dense annotations and task-specific training. The authors propose the first end-to-end pipeline that requires no gradient-based training: it leverages a frozen Segment Anything Model for oversegmentation, fuses Sentinel-1 radar and NDVI time-series data into dual-harmonic phenological features, merges regions via Potts model energy minimization to delineate fields, and performs nearest-prototype classification with topological closure post-processing using only 20 labeled samples per class. Evaluated on PASTIS-R, the method achieves 20.0 mIoU, 76.2 segmentation quality, and 6.2 panoptic quality with less than 1% of the usual annotation budget, substantially outperforming current baselines, and demonstrates strong cross-regional consistency on ZueriCrop.

Abstract

Panoptic crop mapping requires both delineating individual agricultural parcels and assigning a crop type to each parcel from satellite image time series. Existing approaches typically rely on dense parcel-level annotations and task-specific model training, which limits their applicability to new regions and growing seasons. We introduce PhenoStitch, a panoptic crop-mapping pipeline that requires no task-specific gradient-based training. A frozen Segment Anything model first oversegments each patch into class-agnostic regions. For each region, optical NDVI and Sentinel-1 backscatter series are summarized by an analytic double-harmonic phenological signature. Adjacent regions are then merged into parcels by minimizing a Potts graph energy, and each parcel is classified by nearest-prototype matching using only (k) labeled parcels per class. A final topology-closure step produces the panoptic map. Under a matched budget of (k=20) parcels per class, corresponding to less than 1% of the available labels, PhenoStitch achieves 20.0 crop mIoU, 76.2 segmentation quality, and 6.2 panoptic quality on PASTIS-R under a 5-fold, 3-seed evaluation. It outperforms the evaluated frozen foundation-model, few-shot, and matched-budget supervised baselines under the same protocol, with a consistent ranking also observed on ZueriCrop. Ablation studies show that radar observations contribute the largest performance gain, while the graph-energy merge and compact phenological signature provide further improvements. These results demonstrate the effectiveness of combining label-free parcel delineation with few-shot phenological recognition for panoptic crop mapping under limited supervision.

Problem

• panoptic crop mapping
• satellite image time series
• limited supervision
• agricultural parcel delineation
• crop type classification

Innovation

• training-free
• panoptic crop mapping
• phenological signature
• few-shot learning
• graph-based merging