Design and application of a multi-source sensor data fusion system based on a robot phenotype platform

Abstract

The compact, high-throughput phenotyping platform, characterized by its portability and small size, is well-suited for crop phenotyping across diverse environments. However, integrating multi-source sensors to achieve synchronized data acquisition and analysis poses significant challenges due to constraints in load capacity and available space. To address these issues, we developed a robotic platform specifically designed for phenotyping greenhouse strawberries. This system integrates an RGB-D camera, a multispectral camera, a thermal camera, and a LiDAR sensor, enabling the unified analysis of data from these sources. The platform accurately extracted key phenotypic parameters, including canopy width (R² = 0.9864, RMSE = 0.0185 m) and average temperature (R2 = 0.8056, RMSE = 0.1732 °C), with errors maintained below 5%. Furthermore, it effectively distinguished between different strawberry varieties, achieving an Adjusted Rand Index of 0.94, underscoring the value of detailed phenotyping in variety differentiation. Compared to conventional UGV-LiDAR systems, the proposed platform is more cost-effective, efficient, and scalable, with enhanced data consistency, making it a promising solution for agricultural applications.

Plant phenotyping relevance

複数センサーを統合したロボット植物フェノタイピング基盤を開発し、キャノピー幅と平均温度の抽出精度を検証しているため、フェノタイピング手法が中心である。

Code and data availability

The supplied blocks describe a robotic strawberry phenotyping platform and its measurements, but contain no public phenotype dataset, sensor data deposit, author analysis code, or trained model with an availability statement or URL. The only URLs present are the article DOI and the Creative Commons license notice; no '

No evidence-backed public reproduction asset is currently recorded.