Automatic Counting And Grading Of Sunflower Seeds Using Image Processing Techniques.

Manual seed counting and grading are labor-intensive and error-prone processes in agricultural production. This study proposes an HSV-based image processing system for sunflower seed counting and grading. Images were captured with a camera positioned 14 cm above seeds placed on a white background. Preprocessing employed HSV color filtering (Lower: 94, 86, 14; Upper: 120, 254, 128) to segment seeds effectively. Morphological features such as area, length, and width were extracted. Seeds were categorized into three grades: Grade A (≥ 5000 pixels), Grade B (2900–4999 pixels), and Grade C (< 2900 pixels). Experimental results showed counting accuracy above 95% and grading accuracy of 92% compared with manual inspection. Average processing time was 1.8 seconds per image. Although limitations remain in dataset size and preprocessing (painting seeds for color uniformity), the study demonstrates a cost-effective proof of concept for agricultural automation. Future work should incorporate larger datasets and machine learning models to improve robustness in industrial applications.