Multiple Seed Segregation Using Image Processing

Agricultural production is based on seeds and large populations are dependent on agriculture for sustenance. Seed properties are generally assessed by human inspection, which has many limitations. Previously, seed segregation hardware setup was developed, but the limitation of the setup is it can visualize a single seed at a time and no internal features were considered. This work aims at separating multiple different types of seeds as good and bad using object detection analysis. Using a Webcam, multiple images of different seeds are clicked. These seeds are kept in a tray. The images were fed into pre-trained Object detection YOLO5 model. This model classifies the seeds into two categories, good and bad and also classifies the types of seed using Image processing. It is an affordable solution. With this setup, multiple good and bad seeds are detected accurately which is also time efficient. This hardware setup can be used to segregate any number of seeds with a classification accuracy of 97%. To the best of the authors' knowledge, this is a first attempt to develop a dataset and also perform classification of different types of seeds including both good and bad types.