An Efficient Method for Seeds Classification and Quality Testing Using Fast R-CNN

Research into agriculture is gaining an increasing attention in recent years. The newest innovations use a variety of computational technology to make agricultural tasks easier. Seed quality is the most important of many factors that determine agricultural output. Further insights into quality manufacturing, seed quality control, and impurity discovery can be gained by proper seed classification. Seeds have traditionally been sorted into different categories based on their appearance (including their color, texture, and shape). This is often done by experts who must visually inspect each sample, which is a laborious and time-consuming process. Seed sorting is a labor-intensive process that can be greatly improved by automating the inspection process. This study presents a classification system for seeds using Fast RCNN, inspired by the success of computational intelligence methods in other image classification challenges. The suggested method includes five stages: wiener filter preprocessing; HE image enhancement; binarization; adaptive watershed segmentation; and Rapid R-CNN model training. The proposed approach outperforms other two models such as CNN and RNN.