Crop Seeds Classification Using Traditional Machine Learning and Deep Learning Techniques: A Comprehensive Survey

Seed quality is a critical factor in agricultural production, and variety purity is a crucial measure of seed quality. Ensuring seed quality is essential for optimizing crop yields and maintaining the integrity of farming systems. Traditional methods of seed quality assessment can be time-consuming, labour-intensive, and destructive to the seeds being tested. As a result, there is a growing interest in developing non-destructive approaches to seed quality assessment that are both effective and efficient. Artificial intelligence, particularly Machine Learning (ML) and Deep Learning (DL) techniques, has emerged as a promising solution for ensuring seed quality and variety purity. For many years, classical ML algorithms have been successfully applied to solve seed quality and variety problems using image processing techniques. More recently, DL techniques have also shown great promise in accurately classifying seed varieties using both RGB and hyperspectral imaging data. Given the importance of this topic and the rapid advances in ML and DL techniques, there is a need for a comprehensive review of the various problems related to crop seed variety classification and the practical solutions that have been developed using these techniques. To the best of the authors’ knowledge, this survey paper is the first to provide a comprehensive overview of the use of image processing techniques for crop seed classification and identification. In preparing this paper, the authors have focused on several critical aspects of this topic, including seed-related problems, the basic principles of ML and DL, image data acquisition approaches, pre-processing techniques, hardware and software requirements, experimental setup, and model evaluation parameters. The paper also provides a categorization of crop seed classification tasks, including the identification of damaged seeds, diseased seeds, seed sorting, and single and multiple crop seed varieties. To support their analysis, the authors have compiled summary tables of state-of-the-art works related to ML and DL approaches for crop seed classification, organized by their respective categories. These tables provide a valuable resource for researchers and practitioners working in this field, highlighting the key features and performance of different techniques and identifying areas for future research and development.