Automated Physical Quality Assessment of Harvested Seeds: A Critical Review of 2D and 3D Computer Vision Systems

Abstract

Purpose: This paper investigated automated physical quality assessment of harvested seeds. Design/Methodology/Approach: This study provides an extensive review of computer vision-based two-dimensional (2D) and three-dimensional (3D) deployments for the physical inspection of harvested seeds and grains. For this purpose, a total of 75 peer-reviewed articles published between 2022 and 2025 were identified from scientific databases, including Scopus, Web of Science, IEEE Xplore, and ScienceDirect. These articles were based on seed quality assessment, image processing, and artificial intelligence. The selected articles were systematically analysed according to different stages of the processing pipeline, including data acquisition, preprocessing, segmentation, feature extraction, and classification. Research Limitation: This review is restricted to physical quality assessment of harvested seeds, excluding chemical, biochemical, and nutritional parameters. It references 75 peer-reviewed articles published between 2022 and 2025. Findings: This study identified technical problems related to variations in seed samples, hardware setups, segmentation, feature selection, and classification. These problems significantly affect the performance of automated systems. Based on a critical examination of the present automated systems, this paper highlighted the scope for future research. Practical Implication: An advanced, future-ready system can address the need for integrated imaging methods and effective data processing. Social Implication: The adoption of automated seed inspection systems provides assurance of food security. Originality/ Value: This paper identified critical gaps such as the absence of a unified processing framework, the lack of cross-species generalisation, and the limited adoption of explainable AI.

Plant phenotyping relevance

収穫種子の物理品質を画像から評価する2D/3Dコンピュータビジョン手法を体系的にレビューしており、植物形質取得法が中心である。

Code and data availability

The supplied blocks are from a review article on seed quality computer vision; the visible content consists of the article header, a figure caption, and reference lists. No public phenotype datasets, seed images, author code, models, or supplements specific to this paper's own measurements are mentioned, and no data or

No evidence-backed public reproduction asset is currently recorded.