基于Seed Identification软件的棉籽机器视觉快速精选

Quickly selection for cotton seed based on Seed Identification Software

为了研究Seed Identification软件在棉花种子加工工艺和精选参数选择上应用的可行性,以鲁棉28酸脱绒棉籽为材料,通过扫描仪获取400粒棉籽的PNG图像,利用Seed Identification软件快速提取图像中棉籽的RGB、Lab、HSB、灰度、长度、宽度和投影面积等物理指标,通过卷纸发芽获得每颗幼苗鲜质量作为种子的活力指标,种子物理指标与种子活力的相关性分析表明:幼苗鲜质量与R、S、B(HSB)、b、宽度、长度、投影面积的相关系数均达到0.05显著水平。按R<90、S≤18、B(HSB)≤36、b≤4、宽度>4 mm、长度>7.2 mm、投影面积≥25 mm2对种子进行精选,发芽率可由原来的89%分别提高到96.1%、95.1%、95.1%、95.3%、93.1%、93.5%、94.4%,获选率分别为96.6%、99.2%、98.9%、97.8%、98.6%、97%、94.7%。验证试验将种子按以上指标精选后,发芽率分别为95.1%、95.1%、94.8%、94.8%、94.4%、94.4%、94.8%。该研究为基于机器视觉技术对脱绒棉种实施快速、有效精选提供了理论依据。

In this paper, the correlation between the physical traits and seed vigor of delinted cotton seed（Lu Mian 28） was analyzed. The PNG format images of 400 cotton seeds were acquired with falatbed scanner, and the color features of cotton seed such as RGB, HSB, Lab, gray scale, and width, length and projected area were extracted automatically and quickly using seed identification software developed by our lab. Our seed identification software can identify the image and record related seed physical information, and then output all the information into an Excel file automatically. The identifying results were achieved in 1 second, very quickly, with errors lower than 2%. The germination experiment was performed to get seedling fresh weight as seed vigor. Data analysis showed that R, H, S, B（HSB）, b, width, length, and projected area had a relatively high coefficient of variation（more than 0.1） during the sample. Correlation analysis showed that R, S, B（HSB）, b, width, length, and projected area were all significantly correlated with a seedling＇s fresh weight. The correlation coefficients（R） were-0.128,-0.143,-0.121,-0.151, 0.283, 0.173, and 0.346 respectively. Cotton seeds of R﹤90, S≤18, B（HSB）≤36, b≤4, width 4mm, length 7.2 mm, seed projected area≥25 mm2were selected respectively, and the seed germination rate was improved from 89% to 96.1%, 95.1%, 95.1%, 95.3%, 93.1%, 93.5% and 94.4%, and the selected rates of high quality seeds were 96.6%, 99.2%, 98.9%, 97.8%, 98.6%, 97%, and 94.7%, respectively. The verification test selected cotton seeds based on the physical traits and selected parameters described above, and the germination rate of seeds with R﹤90, S≤18, B（HSB）≤36, b≤4, width 4 mm, length 7.2 mm, seed projected area≥25 mm2reached 95.1%, 95.1%, 94.8%, 94.8%, 94.4%, 94.4% and 94.8%, respectively. Therefore, we could deduce that this seed identification software could be applied to the selection of seed processing technology and the parameter determination of single delinted cotton seed according to seed vigor. The result has great importance for improving the seed processing level of China. In addition, our experiment confirmed that the seed projected area had a higher correlation with seed vigor compared to seed width and length. The reason may be that the projected area combined the information of seed width and seed length. But until now, there have been no related seed processing machines which could select seeds according to the seed projected area. This kind of seed machine is suggested to be developed as soon as possible.