摘要
为了提高马铃薯高光谱图像的滤波效果,降低马铃薯外部损伤检测模型的错误率和误判率,采用小波递推最小二乘滤波算法对马铃薯高光谱图像进行去噪.该算法先对经过主成分分析的马铃薯高光谱图像进行4尺度的小波变换,并将得到的二维小波变换系数集成为一个单一的一维重构向量,利用此向量的系数重构分辨率更高的图像,再将经过小波逆变换的图像使用递推最小二乘滤波算法进行滤波,得到结果图像;同时与改变噪声类型的滤波效果、采用维纳滤波和递推最小二乘滤波算法对马铃薯高光谱图像的滤波效果进行对比.结果表明:该滤波算法对高斯和椒盐噪声均具有良好的滤波效果;与采用维纳滤波算法和递推最小二乘滤波算法相比,滤波效果明显,对提高马铃薯外部损伤检测模型的正确率和识别率具有促进作用.
In order to improve the filtering effect of potato hyperspectral image,and reduce potato ex-ternal damage model error rate and erroneous recognition,thus using wavelet and recursive least squares filtering algorithm for potato hyperspectral image to de-noising.The fundamental principle is that potato hyperspectral image through principal component analysis make use of 4 scale wavelet transform to process,and then the resulting of two-dimensional wavelet transform coefficients integrate into a single one-dimensional reconstruction vectors,and get higher resolution image.Finally,the inverse wavelet trans-form image use recursive least squares algorithm to filter.At the same time compared filtering effect with different noise,and wiener and recursive of least square filter algorithm for potato hyperspectral image.Ex-perimental results showed that it had good filtering effect for the added Gaussian and salt and pepper noise of potato hyperspectral image.Compared filtering effect with wiener filter algorithm and recursive of least square filter algorithm,the filtering effect of the algorithm was obvious,which improve potato external damage detection model accuracy and recognition rate.
出处
《甘肃农业大学学报》
CAS
CSCD
北大核心
2014年第2期170-175,180,共7页
Journal of Gansu Agricultural University
基金
国家自然科学基金项目(61261044)
关键词
马铃薯
高光谱成像技术
小波
递推最小二乘法
去噪
potato
hyperspectral image technology
wavelet
recursive of least square
de-noising
