Rank-Aware Quantile Activation for Motion-Robust Crop Segmentation in UAV Imagery

2026-05-31

AI Summary

Motion blur induced by high-speed drone imaging erases high-frequency texture details, severely degrading semantic segmentation performance for rare crop categories. To address this, this work introduces— for the first time in remote sensing image segmentation—a rank-aware quantile activation mechanism, proposing a dual Quantile Activation (QAct) module that replaces conventional magnitude-based gating with instance-level rank normalization. This design preserves feature ordinal information while enhancing robustness to blur. Combined with zero-shot and blur-supervised training strategies, comprehensive evaluations across multiple blur severities on the Agriculture-Vision 2021 dataset demonstrate consistent and significant improvements in mean Intersection over Union (mIoU) under all blur conditions. Notably, the method excels on rare classes that rely heavily on structural and textural cues, with the Distill-QAct hybrid strategy achieving the best overall performance.

Abstract

Motion blur from high-speed UAV acquisition de-grades semantic segmentation on rare texture-dependent classes with high agronomic value. Standard CNNs rely on high-frequency magnitude features that blur destroys, causing statistical erasure of minority signals. We propose Dual Quantile Activation (QAct), a rank-aware block replacing magnitude gating with instance-level rank normalization. Evaluated onAgriculture-Vision 2021 across zero-shot and blur-supervised regimes at multiple severities, QAct is the dominant architectural factor: it delivers consistent mIoU gains over ReLU across both regimes and all severities, with strongest gains on rare structural and texture-dependent classes. Some dominant classes (water,planter skip) show mixed per-class performance under distillation. At moderate blur, zero-shot QAct outperforms distillation-trained ReLU; across all severities, Distill-QAct achieves best performance, confirming rank aware activation and blur-domain training are complementary robustness sources.

Problem

• motion blur
• semantic segmentation
• UAV imagery
• texture-dependent classes
• rare classes

Innovation

• Quantile Activation
• Rank-aware Normalization
• Motion Blur Robustness
• UAV Imagery Segmentation
• Rare Class Enhancement