摘要
为了提高玉米单产,在播种之前对种子活力进行检测十分必要。提出一种将传统的软X-射线检测方法与计算机智能识别相结合的新方法。首先建立corn_pixel结构,然后通过确定种子尖端位置和形心建立种胚区域的椭圆不等式对种胚区域进行标识。以椭圆短半轴b及种胚区域渗钡像素比率M1/M和非种胚区域渗钡像素比率K1/K为输入特征,以标准发芽试验结果为输出,建立BP神经网络单粒种子活力识别模型。结果表明,当b=2CD/5时,识别的准确率最高,以该b值为依据进行分组试验的平均准确率可达95%以上。
In order to improve the per unit yield of maize, it is necessary to detect seed vigor before sowing. A new method was put forward by combining traditional soft X-ray seed vigor detection method with computer intelligence. First of all, corn_pixel structure was defined, the elliptic inequality of embryo region was set up by determining the position of the tip and centroid of the seed so the embryo region was marked. The short axis of ellipse（b） ,barium infiltrated percentage（K1/K） of non-embryo region and barium infiltrated percentage（M1/M） of embryo region were set as input feature and result of standard germination test was set as output, a BP neural network single maize identification model was built. The results showed：when b = 2CD/5 ,the accuracy rate was the highest. The average accuracy rate of group tests based on the value of b was above 95 %.
出处
《作物杂志》
CAS
CSCD
北大核心
2013年第3期136-140,共5页
Crops
基金
农业部引进国际先进农业科学技术计划"948计划"(2008-Z24)
黑龙江省教育厅科学技术研究项目(12531468)
