A digital image analysis algorithm was developed to inspect mildewed rice seeds quickly and accurately based on color feature. Three methods of color feature extracting were tested and the results showed that the blocks method was best. The algorithm was applied to a 5×600 images set,which included black background, white background and both sides images of rice seeds. The algorithm achieved an accuracy of 92% for normal seeds, 95% for spot mildewed seeds and 83% for severe mildewed seeds. Error analysis provided suggestions for increasing the accuracy further.
Transactions of the Chinese Society of Agricultural Machinery