Multi-sensor information fusion to characterise 3D spatial distribution of water stress in strawberries

Abstract

Moisture plays a critical role in crop growth and development, making accurate, efficient, and non-destructive detection and monitoring of crop water stress essential for advancing crop science research and optimizing production management. Traditional non-destructive methods for monitoring water stress primarily rely on color imaging or partial 2D spectral analysis. However, these methods are limited to two-dimensional features and fail to capture the spatial variability of water stress within the three-dimensional canopy structure of crops. To address this limitation, this study integrates RGB-D cameras and thermal infrared cameras and introduces a method for calculating the 3D spatial distribution characteristics of crop water stress using RGB-D-T fusion analysis. This approach enables high-precision detection and analysis of water stress in strawberry plants. An RGB-D-T acquisition system was designed and implemented to collect RGB images, depth images, and thermal infrared images of strawberries subjected to different moisture gradient treatments. Using the YOLOv8-seg deep learning model, semantic segmentation of the crop canopy and the wet reference surface was performed. The segmentation results were fused with 3D point cloud data to generate a 3D dataset incorporating temperature, color, and semantic information. Subsequently, the three-dimensional distribution characteristics and dynamic changes in the canopy water stress index (CWSI) of strawberry plants were analyzed under varying moisture conditions. The results demonstrated that under low moisture gradients (15%–30%), the CWSI value increased significantly and exhibited a concentrated distribution, indicating severe water stress. Conversely, under high moisture gradients (75%–90%), the CWSI value approached zero, reflecting sufficient water supply and complete stress alleviation. Additionally, the study highlighted the variation in the temperature difference between strawberry leaves and the surrounding air, confirming the sensitivity of strawberries to water stress across different reproductive stages. The response to water deficit was most pronounced during the growth phase. By fusing multi-source data, this study achieves 3D visualization and precise quantification of water stress in strawberries, providing innovative insights and technical support for precision irrigation and crop phenotyping research.

Plant phenotyping relevance

RGB-D・熱赤外センサーの融合、3D点群化、深層学習セグメンテーションにより、イチゴの水ストレスを3D定量化する取得・解析手法が研究の中心である。

Code and data availability

公開論文であることは確認できましたが、現在の公式API・許可済み取得経路では本文を自動取得できませんでした。

No evidence-backed public reproduction asset is currently recorded.