基于频谱特性和自适应虫孔定位的小麦识别  

Wheat recognition method based on spectrum properties and adaptive wormhole positioning

作　　者：王周璞 何小海[1] 吴小强[1] 董德良 李晓亮 Wang Zhoupu;He Xiaohai;Wu Xiaoqiang;Dong Deliang;Li Xiaoliang(College of Electronics and Information Engineering,Sichuan University,Chengdu 610065,China;Sinograin Chengdu Storage Research Institute Co.,Ltd.,Chengdu 610091,China)

机构地区：[1]四川大学电子信息学院,四川成都610065 [2]中储粮成都储藏研究院有限公司,四川成都610091

出　　处：《信息技术与网络安全》2018年第12期35-38,43,共5页Information Technology and Network Security

摘　　要：为了实现对小麦不完善粒批量、快速、准确地识别,提出了一种基于小麦图像特征的小麦不完善粒识别方法。采集不完善粒小麦图像,对每粒小麦图像提取其自适应虫孔特征、频谱特性、周长、最小外接圆面积等89维特征参数。研究结果表明,对于常见小麦不完善粒类别,尤其是发芽粒和生虫粒具有较高的识别率。该识别方法对正常粒、破碎粒、生虫粒、生病粒、发芽粒的识别率分别为98. 75%、97. 50%、93. 02%、99%、96. 25%,平均识别率为96. 90%,相较于传统的图像处理识别方法,识别准确率提高20%左右,表明提出的方法能有效运用于小麦不完善粒检测实际应用场景中。In order to recognize unsound kernels of wheat batchly,rapidly and precisely,an unsound kernels of wheat recognition method which is based on wheat image features is proposed in this paper.We collected the images of unsound kernels of wheat and extracted 89dimensional feature parameters such as adaptive wormhole feature,spectral feature,perimeter,and minimum circumscribed circle area for each grain image.The results show that for common unsound kernels of wheat types,especially germinated grains and insect grains have a higher recognition rate.For this recognition method,the recognition rates of normal,crushed,insect,diseased,and germinated grains were 98.75%,97.50%,93.02%,99%,and 96.25%,the average recognition rate was 96.91%.Compared with the traditional image processing and recognition methods,the recognition accuracy is improved about 20%.It proved that the method which we proposed can be effectively applied to practical recognition application scenarios of unsound kernels.

关 键 词：自适应虫孔特征 频谱特性 小麦不完善粒 发芽粒 生虫粒 识别率 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]