论文部分内容阅读
针对内蒙古荒漠化草原牧草监测与数字化程度较低的问题,实现了三种典型牧草的特征提取与图像识别,为多牧草种类识别与草业管理提供依据.运用MATLAB图像处理技术,对糙苏、羊草及鹅绒萎菱菜进行图像预处理,提取了12种颜色矩特征和4种形状特征,利用BP神经网络法实现了三种牧草的图像识别,识别正确率为86.9%.试验结果表明,基于颜色矩特征和形状特征的BP(back propagation,BP)神经网络图像识别方法能够有效地实现典型牧草的图像分类研究.自动识别牧草是草业数字化的重要组成部分,可为监测植被物种多样性、草种退化及病虫草害的控制提供科学依据.
In view of the problems of pasture monitoring and digitization in desert steppe grassland in Inner Mongolia, the three typical forage grass feature extraction and image recognition were realized, which provided the basis for multi-forage species identification and grassland management.Using MATLAB image processing technology, Grass and goose down witch-hazel were preprocessed, 12 color moments and 4 shape features were extracted, and the image recognition of the three species of forages was realized by using BP neural network, and the recognition rate was 86.9% BP (back propagation, BP) neural network image recognition method can effectively study the image classification of typical forage grass.It is an important part of digitalization of grassland that automatic recognition of forage grass can be used to monitor the diversity of vegetation species, Grass species degeneration and pest control provide scientific basis.