Abstract
A grid method to detect the existence of vegetable seeds into the tray-cells after the seeding operation was developed. Its features include generation of windows, noise suppression and grid operators. Windows are generated and processed at the cavities of the tray instead of processing the whole image of the tray. The image noises are suppressed by using local average filter kernel operation. Then the seed blobs are detected using grid operators. Judgement to whether the blob represented a seed or noise was carried out according to continuity of the grids that represent an object in the 8-or 16-neighborhood. To test the performance of the developed grid method, morphological features of the seeds and noises were extracted and analyzed. The results showed that the grid method suppressed the image noises by 76.9% and successfully detected the seed blobs. The detection accuracy was 100% for cucumber, melon, lettuce, eggplant and green pepper seeds, and was 99.2% for tomato seeds. Hence, the usefulness of the method to detect the existence or non-existence of seeds into the tray-cells was confirmed.
