摘要
针对水稻害虫自动识别问题,对害虫图像采用基于骨架的形态特征提取方法,自动提取害虫的骨架点的分支数、骨架累计长、横轴纵轴比、圆形性,以此作为BP神经网络的输入因子,并以从20种害虫图像提取的特征值,训练网络,较好地解决了这一问题.
Pest features such as branches, total skeleton length, z-length/y-length, and circularity were automatedby extracted from pest images and loaded into a BP neural network. The network was trained with characteristics from 20 images of rice pests. After such training the network was able to perform automated recognition of rice pests.
出处
《北京师范大学学报（自然科学版）》
CAS
CSCD
北大核心
2008年第2期165-167,共3页
Journal of Beijing Normal University(Natural Science)
基金
国家自然科学基金重点资助项目(60736008/F010207)
国家自然科学基金面上资助项目(606731008/F020106)
关键词
图像识别
特征提取
BP神经网络
骨架
image recognition
feature extraction
BP neutral network
skeleton
