基于改进的均值漂移算法的高分辨率遥感影像分割方法研究
|
摘要
影像分割是面向对象信息提取的关键,其分割结果的精度高低将对信息提取结果精度产生重要影响。本文在研究均值漂移(Mean shift)算法的基础上,对图像的多维特征向量加以改进。该改进算法先通过改进颜色特征距离公式来对高分辨率遥感影像进行预处理,然后加入纹理特征向量,并设置纹理特征的参数值和迭代次数,进而对高分辨率遥感影像进行分割。文章采用该改进的均值漂移算法对遥感影像进行分割实验,并与业界公认的ecognition软件的分割结果进行对比,实验结果表明:该改进算法得到的分割精度更高,能得到更好的分割结果。
Image segmentation is the key of object-oriented information extraction,and the accuracy of the segmentation results will have an important influence on the accuracy of information extraction. Based on the research of mean shift algorithm,the multidimensional feature vector of image was improved. Firstly,the improved algorithm adopted the improving color feature distance formula to the pretreatment of high-resolution remote sensing image,then it adopted adding the texture feature vector,and setting the parameter values of texture feature and the number of iteration to segmentation of high-resolution remote sensing images. The improved mean shift algorithm was used to divide remote sensing images and compare the segmentation results of recognition software. The experimental results show that the improved algorithm can get higher segmentation precision and better segmentation results.
Image segmentation is the key of object-oriented information extraction,and the accuracy of the segmentation results will have an important influence on the accuracy of information extraction. Based on the research of mean shift algorithm,the multidimensional feature vector of image was improved. Firstly,the improved algorithm adopted the improving color feature distance formula to the pretreatment of high-resolution remote sensing image,then it adopted adding the texture feature vector,and setting the parameter values of texture feature and the number of iteration to segmentation of high-resolution remote sensing images. The improved mean shift algorithm was used to divide remote sensing images and compare the segmentation results of recognition software. The experimental results show that the improved algorithm can get higher segmentation precision and better segmentation results.
引文
[1]COMANICIU D,MEER P.Mean Shift:a Robust Approach toward Feature Space Analysis[C].UNITED STATES:IEEE Transactions on Pattern Analysis and Machine Intelligence,2002,24(5):603-619.
[2]周芳芳,樊晓平,叶榛.均值漂移算法的研究与应用[J].控制与决策,2007,22(8):841-847.
[3]林小平,毛政元,刘建华.纹理特征遥感影像分割研究[J].测绘科学,2010,35(5):31-34.
[4]周家香,朱建军,马慧云,等.改进的Mean-Shift遥感影像分割方法[J].中南大学学报:自然科学版,2012,43(2):620-625.
[5]CHENG Y.Mean shift,Mode Seeking,and Clustering[C].UNITED STATES:IEEE Transactions on Pattern Ananlysis and Machine Intelligence,1995,17(8):790-799.
[6]OJALA T,PIETIKAINEN M,MAENPAA T.Multiresolution Grayscale and Rotation Invariant Texture Classification with Local Binary Patterns[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2002,24(7):971-987.
[7]沈德海,侯建,鄂旭,等.一种抑制高斯噪声的加权滤波算法[J].渤海大学学报(自然科学版),2016,37(04):356-360.
[8]孙悦.基于e Cognition的卫星遥感影像分析技术[J].无线电工程,2014,44(03):35-39.
[9]Polak M,ZHANG Hong,PI Ming-hong.An evaluation metric for image segmentation of multiple objects[J].Image and Vision Computing,2009,27(8):1223-1227.
[2]周芳芳,樊晓平,叶榛.均值漂移算法的研究与应用[J].控制与决策,2007,22(8):841-847.
[3]林小平,毛政元,刘建华.纹理特征遥感影像分割研究[J].测绘科学,2010,35(5):31-34.
[4]周家香,朱建军,马慧云,等.改进的Mean-Shift遥感影像分割方法[J].中南大学学报:自然科学版,2012,43(2):620-625.
[5]CHENG Y.Mean shift,Mode Seeking,and Clustering[C].UNITED STATES:IEEE Transactions on Pattern Ananlysis and Machine Intelligence,1995,17(8):790-799.
[6]OJALA T,PIETIKAINEN M,MAENPAA T.Multiresolution Grayscale and Rotation Invariant Texture Classification with Local Binary Patterns[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2002,24(7):971-987.
[7]沈德海,侯建,鄂旭,等.一种抑制高斯噪声的加权滤波算法[J].渤海大学学报(自然科学版),2016,37(04):356-360.
[8]孙悦.基于e Cognition的卫星遥感影像分析技术[J].无线电工程,2014,44(03):35-39.
[9]Polak M,ZHANG Hong,PI Ming-hong.An evaluation metric for image segmentation of multiple objects[J].Image and Vision Computing,2009,27(8):1223-1227.