基于SVM的棉田冠层图像分割方法研究
|
摘要
复杂环境下的棉田冠层图像由于成像环境光照不均、冠层阴影部分与土壤背景对比度低等因素给棉田冠层图像的准确分割带来了困难.在对上述问题进行研究后,提出了一种基于支持向量机(SVM)的图像分割方法.该方法首先以图像目标和背景两类像素的分割特征值构建样本数据集;运用训练好的SVM分类器对棉田冠层图像进行分割处理,最后使用数学形态学滤波方法对分割结果进行优化,获得棉田冠层图像的精确分割结果.实验结果表明,该方法可有效分割出棉田冠层区域,分割误分程度以及分割准确程度均优于常用分割方法.
It is very difficult to accuracy segment cotton canopy image under complex environment. There are many factors that cause this problem. Such as imaging environment light, low comparative degree between shadow canopy and soil background so on.In this paper, a segmentation method based on support vector machine(SVM) is proposed. Firstly, the image pixel is divided into two type of target and background, extracts representative segmentation pixel eigenvalues to build a sample dataset. Secondly, combined with SVM algorithm and sample data set for training, use the trained SVM classifier to segment the cotton canopy image. In the end, use mathematical morphology filtering method to optimize the segmentation results to obtain the exact segmentation results of cotton canopy images. The experimental results show that the proposed method can effectively segment the cotton canopy region,and the segmentation error degree and segmentation accuracy are better than the commonly used segmentation methods.
It is very difficult to accuracy segment cotton canopy image under complex environment. There are many factors that cause this problem. Such as imaging environment light, low comparative degree between shadow canopy and soil background so on.In this paper, a segmentation method based on support vector machine(SVM) is proposed. Firstly, the image pixel is divided into two type of target and background, extracts representative segmentation pixel eigenvalues to build a sample dataset. Secondly, combined with SVM algorithm and sample data set for training, use the trained SVM classifier to segment the cotton canopy image. In the end, use mathematical morphology filtering method to optimize the segmentation results to obtain the exact segmentation results of cotton canopy images. The experimental results show that the proposed method can effectively segment the cotton canopy region,and the segmentation error degree and segmentation accuracy are better than the commonly used segmentation methods.
引文
[1]辛海强,冯志敏.棉花种植区卫星遥感监测研究[J].测绘与空间地理信息,2014,37(2):26-27.
[2]刘晓晨.基于Landsat8遥感数据的棉花长势快速监测[J].新疆农业科技,2015,37(6):50.
[3]BAI J H,LI J,LI S K,et al.Monitoring the cotton growth conditions based on lai from remote sensing and expert knowledge[J].Advanced Materials Research,2011,317-319:1050-1057.
[4]冀高.基于数字图像处理的棉花群体特征提取[D].北京:北京邮电大学,2007.
[5]任玉刚,张建,李淼,等.基于分水岭算法的作物病害叶片图像分割方法[J].计算机应用,2012,32(3):752-755.
[6]史晓楠,欧阳震.图像分割常用方法综述[J].科技视界,2014,33(4):33.
[7]GURJEET K S,RAJNEER K.Review on recent image segmentation techniques[J].International on Computer Science and Engineering,2013,5(2):109-112.
[8]宋淑然,洪添胜,王卫星,等.基于马氏距离的荔枝图像分割设计方法[J].沈阳农业大学学报,2005,36(6):655-658.
[9]周俊,王明军,邵乔林.农田图像绿色植物自适应分割方法[J].农业工程学报,2013,29(18):163-170.
[10]王佳男.采用基于区域的自动种子区域生长法的彩色图像分割方法[J].计算机应用,2007,19(6):97-101.
[11]任守纲,陆海飞,袁培森,等.基于显著性检测的黄瓜叶部病害图像分割算法[J].农业机械学报,2016,60(9):11-16.
[12]CHENG MINGMING,LIN WENYAN,ZHENG SHUAI,et al.Efficient salient region detection with soft image abstraction[C]//International Conference on Computer Vision Theory and Applications.IEEE,2013:1529-1536.
[13]王方永,王克如,李少昆,等.利用数字图像估测棉花叶面积指数[J].生态学报,2011,31(11):3090-3100.
[14]王萍,唐江丰,王博,等.基于数学形态学的植物叶片图像分割方法研究[J].浙江农业学报,2012,24(3):509-513.
[2]刘晓晨.基于Landsat8遥感数据的棉花长势快速监测[J].新疆农业科技,2015,37(6):50.
[3]BAI J H,LI J,LI S K,et al.Monitoring the cotton growth conditions based on lai from remote sensing and expert knowledge[J].Advanced Materials Research,2011,317-319:1050-1057.
[4]冀高.基于数字图像处理的棉花群体特征提取[D].北京:北京邮电大学,2007.
[5]任玉刚,张建,李淼,等.基于分水岭算法的作物病害叶片图像分割方法[J].计算机应用,2012,32(3):752-755.
[6]史晓楠,欧阳震.图像分割常用方法综述[J].科技视界,2014,33(4):33.
[7]GURJEET K S,RAJNEER K.Review on recent image segmentation techniques[J].International on Computer Science and Engineering,2013,5(2):109-112.
[8]宋淑然,洪添胜,王卫星,等.基于马氏距离的荔枝图像分割设计方法[J].沈阳农业大学学报,2005,36(6):655-658.
[9]周俊,王明军,邵乔林.农田图像绿色植物自适应分割方法[J].农业工程学报,2013,29(18):163-170.
[10]王佳男.采用基于区域的自动种子区域生长法的彩色图像分割方法[J].计算机应用,2007,19(6):97-101.
[11]任守纲,陆海飞,袁培森,等.基于显著性检测的黄瓜叶部病害图像分割算法[J].农业机械学报,2016,60(9):11-16.
[12]CHENG MINGMING,LIN WENYAN,ZHENG SHUAI,et al.Efficient salient region detection with soft image abstraction[C]//International Conference on Computer Vision Theory and Applications.IEEE,2013:1529-1536.
[13]王方永,王克如,李少昆,等.利用数字图像估测棉花叶面积指数[J].生态学报,2011,31(11):3090-3100.
[14]王萍,唐江丰,王博,等.基于数学形态学的植物叶片图像分割方法研究[J].浙江农业学报,2012,24(3):509-513.