Recent Developments of Seeds Quality Inspection and Grading Based on Machine Vision

Abstract.As a rapid, nondestructive, cost-effective, objective detection technique, machine vision (MV) has been gradually applied in seeds inspection. In the past, almost all studies focused on varieties identification and quality detection for static seeds in laboratory, but the practical applications such as portable instruments, sorting systems in seed quality detection were not common in industrial uses. For static or dynamic seeds detection, the basic steps including image acquisition, feature extraction, classifier design are the same, the only differences is that the seed is moving for dynamic detection, and the unsatisfactory seeds are automatically rejected. While no related papers have reviewed the development of seed portable instruments and sorting systems. In this work, the feature extraction and classifier design in MV system, the constitution, operational principle, advantage of visual portable instrument were introduced at first. Then performance and characteristic of some commercial instruments were compared. Next, an overview of real-time automatic quality inspection sorting systems was explained in detail. Problems including the imperfect seed singulation, the low processing efficiency below real-time requirements and the detection limitations of MV restricted the fast development of seed sorting systems and research progress for each problems was also explained at the same time. Finally the noteworthy attention of developing seed visual sorters were proposed.