摘要
【目的】分析不同玉米病害图像特征,提取病害图像特征参数,探讨准确、快速的病害图像特征数据提取方法。【方法】采用多重分形的分析方法和提升格式的多小波变换对玉米病害图像进行预处理,去除图像噪声。利用多重分形谱理论对去噪后的玉米病害图像进行局部边缘提取,并采用玉米病害图像多重分形谱的特征值作为玉米病害的形状特征。【结果】该方法可以获得玉米病害图像对应的多重分形谱曲线,并采集到8个特征值作为病害图像的特征参数。发现不同玉米病害图像的特征参数有较大差异,而同类玉米病害图像的形状特征参数有一定的规律性。【结论】基于多重分形理论的玉米病害特征参数的提取方法能快速、有效地提取反映病害图像特征的参数,可为玉米病害图像智能识别的进一步研究提供参考。
[Objective]On the basis of image feature analysis, maize disease feature parameters were extracted to explore the way of accurately and swiftly extracting feature data from disease image. [Method]The maize disease image was pre-treated using muhi-fractal analysis method and the lifting scheme of multi-wavelet transform to eliminate the noise. Edge extraction was conducted to the noise-free maize disease image, and the feature value was used as shape feature of maize disease. [Result]The corresponding multi-fractal spectrum curve of different disease images and eight ideal feature values were obtained as feature parameters. The eight ideal feature values were significantly different and certain regularity existed in the feature values from the similar maize disease images. [ Conclusion ]As the parameters, quickly and effectively extracted by multi-fractal analysis method, reflected maize disease image characteristics, it could provide references for follow-up research on intelligent identifi- cation of maize disease images.
出处
《南方农业学报》
CAS
CSCD
北大核心
2013年第5期871-874,共4页
Journal of Southern Agriculture
基金
河南省教育厅科学技术研究重点项目(12A510021)
关键词
玉米病害
多重分形
多小波变换
特征值
maize disease
muhi-fractal
multi-wavelet transform
feature value
