基于机器视觉的水稻种子精选技术

Rice Seeds Selection Based on Machine Vision

以面积和宽长比作为稻种类型的特征参数,建立了稻种类型数据库;以等价矩形长、宽的差值最小为标准,进行了未知稻种类型的判断;以扫描线上黑白像素的变化次数和扫描线数来判断稻种的破裂;以不同阈值提取的稻种面积差来判断稻种是否霉变。选取丰源优299等10种稻种进行实验,分别进行了种子类型判断、工位有无种子判断、几何参数判断以及发霉与破损情况判断,检测正确率分别为100%、91.4%、88.9%和76.8%。

The area and wide length ratio was chosen as the characteristic parameters for different kinds of rice seeds,and a database of rice seeds＇ type was constructed.The equivalence rectangle＇s length and the least wide difference were used as standard to identify unknown rice seeds＇ type.The disrepair of the rice seeds was identified based on the times of the black pixels change to the white pixels on one scan line and the number of scan lines.The mildew of rice seeds was identified based on the difference of area obtained by two different thresholds.Ten kinds of rice seed samples like Fengyuanyou 299 were chosen for the experiments.During the experiments,inspection of seeds＇ type,identification of seeds in work positions,geometry parameters of seeds,mildew and disrepair situations of seeds were done respectively.The accuracy was 100%,91.4%,88.9% and 76.8% respectively.