Detection of Damaged Cottonseeds Using Machine Vision

Damaged cottonseeds has a disadvantageous influence on cotton yields. The traditional detection of cottonseeds depends on just labor, which is tedious and variant with different operator. An automatic detection system based on machine vision was designed to distinguish the sound cottonseeds from the damaged ones. The objective of this study is to develop image processing algorithms to finish picking out damaged cottonseeds. During the development of the algorithm, three statistical characteristics, mean, variance and the ratio of mean to variance (RMV), were used. Different sizes of detection window were tested. It is proved that 9times9 detection window can perform well. Image algorithm testing on a validation data showed that damaged cottonseeds could be distinguished from sound ones with accuracy of up to 93%.