An Efficient Additive Kolmogorov-Arnold Transformer for Point-Level Maize Localization in Unmanned Aerial Vehicle Imagery

2026-01-12

AI Summary

This study addresses the challenges of point-level maize plant localization in high-resolution UAV imagery, where small targets suffer from low representation, high computational cost, and scene complexity. To this end, the authors propose the Additive Kolmogorov–Arnold Transformer (AKT), which introduces the Kolmogorov–Arnold representation theorem to agricultural remote sensing for the first time. The method incorporates a Pade Kolmogorov–Arnold Network (PKAN) module to enhance functional expressivity and integrates a low-complexity additive attention mechanism to improve small-object feature modeling while substantially reducing computational overhead. Evaluated on a newly curated dataset (PML) comprising 1,928 images with 501,000 point annotations, the approach achieves an F1-score of 62.8%—outperforming the state of the art by 4.2%—with 12.6% fewer FLOPs, 20.7% higher inference throughput, a mean absolute error (MAE) of 7.1 in plant counting, and root mean square errors (RMSE) of 1.95–1.97 cm in inter-plant spacing estimation.

Technology Category

• Machine Learning: Active Learning
• Knowledge Representation and Reasoning: Knowledge Acquisition
• Computer Vision: Remote Sensing / Geospatial AI

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

High-resolution UAV photogrammetry has become a key technology for precision agriculture, enabling centimeter-level crop monitoring and point-level plant localization. However, point-level maize localization in UAV imagery remains challenging due to (1) extremely small object-to-pixel ratios, typically less than 0.1%, (2) prohibitive computational costs of quadratic attention on ultra-high-resolution images larger than 3000 x 4000 pixels, and (3) agricultural scene-specific complexities such as sparse object distribution and environmental variability that are poorly handled by general-purpose vision models. To address these challenges, we propose the Additive Kolmogorov-Arnold Transformer (AKT), which replaces conventional multilayer perceptrons with Pade Kolmogorov-Arnold Network (PKAN) modules to enhance functional expressivity for small-object feature extraction, and introduces PKAN Additive Attention (PAA) to model multiscale spatial dependencies with reduced computational complexity. In addition, we present the Point-based Maize Localization (PML) dataset, consisting of 1,928 high-resolution UAV images with approximately 501,000 point annotations collected under real field conditions. Extensive experiments show that AKT achieves an average F1-score of 62.8%, outperforming state-of-the-art methods by 4.2%, while reducing FLOPs by 12.6% and improving inference throughput by 20.7%. For downstream tasks, AKT attains a mean absolute error of 7.1 in stand counting and a root mean square error of 1.95-1.97 cm in interplant spacing estimation. These results demonstrate that integrating Kolmogorov-Arnold representation theory with efficient attention mechanisms offers an effective framework for high-resolution agricultural remote sensing.

Problem

• point-level localization
• UAV imagery
• small object detection
• computational complexity
• agricultural remote sensing

Innovation

• Kolmogorov-Arnold Network
• Additive Attention
• Point-level Localization
• UAV Imagery
• Efficient Transformer