摘要
针对现有技术在花生种子筛选过程中分类复杂、准确率低、速度慢的问题,提出了基于卷积神经网络的花生种子的筛选识别算法。根据实际情况将花生种子分为完好花生、破损花生2类研究对象,收集1500张花生照片建立图像库,搭建卷积神经网络结构,提取花生种子图像的颜色特征和纹理特征,优化网络提高筛选的准确率和快速性。试验结果表明优化完成后的卷积神经网络筛选准确率为98.21%,筛选速度为16.4 ms/粒,说明该系统准确率高、筛选速度快,可以满足农业对花生种子的实际筛选要求。
In order to solve the problems of complex classification, low accuracy and slow speed in the peanut screening process, the author proposed a screening and recognition algorithm based on convolutional neural network for peanut seeds in this paper. According to the actual situation, peanut seeds were divided into two categories: intact peanuts and broken peanuts. 1500 peanut photos were collected to build an image database. Convolutional neural network structure was constructed, and the color and texture characteristics of peanut seed images were extracted to optimize the network and improve accuracy and speed for screening. The experimental results showed that the accuracy of convolutional neural network screening after optimization was 98.21%, and the screening speed was 16.4 ms/particle, which indicated that the system had high accuracy and fast screening speed.
作者
张永超
赵录怀
卜光苹
ZHANG Yong-chao;ZHAO Lu-huai;BU Guang-ping(City College,Xi’an Jiaotong University,Xi’an 710018,China;Electrical College,Xi’an Jiaotong University,Xi’an 710048,China)
出处
《江西农业学报》
CAS
2020年第1期77-82,共6页
Acta Agriculturae Jiangxi
关键词
花生
种子
筛选
卷积神经网络
图像处理
特征提取
Peanut
Seed
Screening
Convolutional neural network
Image processing
Feature extraction
