Abstract:
The intricate geometric parameters of seeds play a pivotal role in unveiling the intricate genetic blueprints, facilitating precise variety identification, and tracking the intricate growth trajectories of higher plants. This understanding holds immense significance within the vanguard of research endeavors, specifically in digital seed evaluation, phenomics, and digital agriculture simulation, where the pursuit of precise and efficient data acquisition methods is paramount. Building upon a meticulous examination of the diverse requirements for accurately measuring plant seed geometric parameters, this paper innovatively introduces a high-throughput measurement approach, leveraging the advancements in handheld scanning technology. This novel methodology meticulously addresses two central challenges encountered in conventional measurement processes, namely missing point cloud data and automatic screening of typical samples. Firstly, confronted with the pervasive issue of incomplete point cloud data during the batch scanning of diverse plant seeds, a sophisticated point cloud completion method is devised. This innovative approach employs ellipse fitting and smooth interpolation techniques tailored to the longitudinal profile contours of seeds, accurately bridging the gaps in scanned point clouds. This ensures the preservation of data integrity and fidelity, forming a robust foundation for downstream analytical endeavors. Secondly, the inefficiencies and laborious nature of traditional manual selection of typical samples are mitigated through the introduction of an automated sample selection algorithm grounded in statistical models derived from principal component analysis (PCA) method. This algorithm proficiently identifies and extracts representative seed samples, dramatically enhancing research efficiency while simultaneously improving data quality and consistency. Based on the completed point cloud data and the selected typical samples, this paper integrates geometric parameter estimation algorithms, enabling automatic measurement of 11 geometric parameters, including volume, surface area, length, width, height, as well as the perimeter and area of three principal component profiles. In the empirical validation of this methodology, a high-precision handheld scanner, the RigelScan Elite, was employed to capture the point cloud data of eight morphologically disparate plant seed types: broad beans, peanuts, kidney beans, soybeans, peas, black beans, red beans, and mung beans, collectively encompassing 3,400 samples. The geometry of broad beans, peanuts and kidney beans is regular, while that of soybeans, peas, black beans, red beans and mung beans is irregular. The obtained 3D point cloud data of plants has a resolution of 0.01 mm and does not include color information. Subsequently, the captured data underwent algorithmic processing for missing data completion, automated screening of representative samples, and automatic measurement of various geometric parameters for individual seeds. The results demonstrate the effectiveness of the proposed point cloud completion method, successfully restoring missing data with an average error of 0.017 mm, irrespective of the seed's geometric regularity. The sample selection algorithm accurately identifies typical samples, boasting an average extraction error of 0.80%, thereby guaranteeing the representativeness and precision of the selected samples. Furthermore, the integrated automatic geometric parameter extraction algorithm achieves an average measurement error below 0.41%. This remarkable feat underscores the successful realization of batch, automatic, and high-precision measurement of seed geometric parameters. In conclusion, this study presents a groundbreaking high-throughput automatic measurement approach for seed geometric parameters, offering a reliable and efficient tool for fundamental data acquisition in digital seed evaluation, phenomics, smart agriculture, and associated disciplines. The implementation of this methodology represents a significant advancement in the precision and scalability of seed characterization, paving the way for future research endeavors in these vibrant and rapidly evolving fields.