Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FarmMind: Reasoning-Query-Driven Dynamic Segmentation for Farmland Remote Sensing Images

Haiyang Wu, Weiliang Mu, Jipeng Zhang, Zhong Dandan, Zhuofei Du, Haifeng Li, Tao Chao

TL;DR

FarmMind tackles the rigidity of static farmland segmentation by introducing a reasoning-query-driven dynamic framework that actively fetches auxiliary imagery on demand. It leverages a multimodal large language model to analyze segmentation ambiguities, trigger targeted data retrieval (multi-temporal or enlarged imagery), and perform collaborative reasoning to refine results. The approach is validated on a large, multi-region FRSI database, showing superior accuracy and generalization over traditional label-driven and language-driven static methods. This dynamic, open-world retrieval mechanism enhances robustness to phenological variation and boundary ambiguity, with practical implications for scalable, spatiotemporal farmland monitoring.

Abstract

Existing methods for farmland remote sensing image (FRSI) segmentation generally follow a static segmentation paradigm, where analysis relies solely on the limited information contained within a single input patch. Consequently, their reasoning capability is limited when dealing with complex scenes characterized by ambiguity and visual uncertainty. In contrast, human experts, when interpreting remote sensing images in such ambiguous cases, tend to actively query auxiliary images (such as higher-resolution, larger-scale, or temporally adjacent data) to conduct cross-verification and achieve more comprehensive reasoning. Inspired by this, we propose a reasoning-query-driven dynamic segmentation framework for FRSIs, named FarmMind. This framework breaks through the limitations of the static segmentation paradigm by introducing a reasoning-query mechanism, which dynamically and on-demand queries external auxiliary images to compensate for the insufficient information in a single input image. Unlike direct queries, this mechanism simulates the thinking process of human experts when faced with segmentation ambiguity: it first analyzes the root causes of segmentation ambiguities through reasoning, and then determines what type of auxiliary image needs to be queried based on this analysis. Extensive experiments demonstrate that FarmMind achieves superior segmentation performance and stronger generalization ability compared with existing methods. The source code and dataset used in this work are publicly available at: https://github.com/WithoutOcean/FarmMind.

FarmMind: Reasoning-Query-Driven Dynamic Segmentation for Farmland Remote Sensing Images

TL;DR

FarmMind tackles the rigidity of static farmland segmentation by introducing a reasoning-query-driven dynamic framework that actively fetches auxiliary imagery on demand. It leverages a multimodal large language model to analyze segmentation ambiguities, trigger targeted data retrieval (multi-temporal or enlarged imagery), and perform collaborative reasoning to refine results. The approach is validated on a large, multi-region FRSI database, showing superior accuracy and generalization over traditional label-driven and language-driven static methods. This dynamic, open-world retrieval mechanism enhances robustness to phenological variation and boundary ambiguity, with practical implications for scalable, spatiotemporal farmland monitoring.

Abstract

Existing methods for farmland remote sensing image (FRSI) segmentation generally follow a static segmentation paradigm, where analysis relies solely on the limited information contained within a single input patch. Consequently, their reasoning capability is limited when dealing with complex scenes characterized by ambiguity and visual uncertainty. In contrast, human experts, when interpreting remote sensing images in such ambiguous cases, tend to actively query auxiliary images (such as higher-resolution, larger-scale, or temporally adjacent data) to conduct cross-verification and achieve more comprehensive reasoning. Inspired by this, we propose a reasoning-query-driven dynamic segmentation framework for FRSIs, named FarmMind. This framework breaks through the limitations of the static segmentation paradigm by introducing a reasoning-query mechanism, which dynamically and on-demand queries external auxiliary images to compensate for the insufficient information in a single input image. Unlike direct queries, this mechanism simulates the thinking process of human experts when faced with segmentation ambiguity: it first analyzes the root causes of segmentation ambiguities through reasoning, and then determines what type of auxiliary image needs to be queried based on this analysis. Extensive experiments demonstrate that FarmMind achieves superior segmentation performance and stronger generalization ability compared with existing methods. The source code and dataset used in this work are publicly available at: https://github.com/WithoutOcean/FarmMind.
Paper Structure (21 sections, 3 equations, 14 figures, 6 tables)

This paper contains 21 sections, 3 equations, 14 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Label-driven FRSI segmentation
  4. Language-driven FRSI segmentation
  5. Image retrieval-augmented generation
  6. Methodological
  7. Overview
  8. Basic perception
  9. Reasoning query
  10. Collaborative segmentation
  11. A FRSI database collection and organization
  12. Experiments and results
  13. Experimental setup
  14. Method comparison
  15. Ablation and variant analysis
  16. ...and 6 more sections

Figures (14)

  • Figure 1: Different farmland remote sensing image segmentation paradigms: (a) Static Segmentation; (b) Dynamic Segmentation (Our)
  • Figure 2: Overview of dynamic segmentation for farmland remote sensing images. (FSM stands for foundation segmentation model; MLLM stands for multimodal large language model )
  • Figure 3: The complete workflow example of FarmMind. (Conditional mask correction (CMC) strategy and binary remapping are detailed in Section \ref{['subsec:3.4']})
  • Figure 4: Examples of partial data from the FRSI database
  • Figure 5: Segmentation maps of different methods
  • ...and 9 more figures