摘要
以单倍体育种产生的经遗传标记后的玉米品种1050-37为研究对象,研究种子图像的颜色模式类别,将单个玉米种子划分为紫色标记区域、黄色区域和白色区域进行分析.通过分析图像在归一化rgb,HSV模型下的不同颜色特征,选取其中7个作为输入特征参数,构建了一种3层BP神经网络模型,从而实现玉米单倍体种子图像的有效分割.试验表明:该模型对紫色标记区域、黄色区域和白色区域的准确识别率分别为97.61%,93.34%和94.09%;所提取的紫色标记区域对单倍体与杂合体的识别是有效且可靠的.
Based on BP neural network of maize haploid seeds,an image segmentation method was proposed to research 1050-37 corn with genetic marks.According to color features,corn seed images were divided into three color patterns of purple area,yellow area and white area.Different color features of normalized rgb and HSV color space were analyzed,and 7 features were chosen as input parameters to establish a BP neural network model with 3 layers to achieve effective image segmentation of maize haploid seeds.The experiments show that the classification accuracies of the model are 97.61% for purple marks area,93.34% for yellow area and 94.09% for white area,respectively.The purple marks area acquired by BP NN is effective and reliable for the identification of haploid kernels and hybrid kernels.
出处
《江苏大学学报（自然科学版）》
EI
CAS
北大核心
2011年第6期621-625,共5页
Journal of Jiangsu University：Natural Science Edition
基金
国家"863"高技术研究发展计划项目(2010AA101401)
国家自然科学基金资助项目(31071320)
