基于高光谱遥感技术的田间地膜识别研究  

Research on Field Membrane Recognition Based on Hyperspectral Remote Sensing Technology

作　　者：黄华成 吴雪梅[1] 张康 张珍 肖远 HUANG Huacheng;WU Xuemei;ZHANG Kang;ZHANG Zhen;XIAO Yuan(School of Mechanical Engineering,Guizhou University,Guiyang 550025)

机构地区：[1]贵州大学机械工程学院,贵阳550025

出　　处：《计算机与数字工程》2024年第2期541-547,共7页Computer & Digital Engineering

基　　金：贵州省科技计划项目(编号:黔科合平台人才[2019]5616)资助。

摘　　要：白色地膜呈透明状,当其覆盖在土壤上时,与土壤颜色相近,难以区分。针对该问题,提出了一种基于无人机高光谱成像技术的识别方法。论文以贵州省毕节市双龙镇为研究区,采集田间地膜低空遥感高光谱图像,然后对高光谱图像进行镜头、反射率、大气校正,建立田间地膜、土壤、植物的感兴趣区(ROI),最后分别利用支持向量机(SVM)、主成分分析法(PCA)、波谱角分类(SAM)和特征光谱段(FSS)四种方法来对高光谱图像中的地膜目标进行识别,结合最大类间方差阈值分割,形态学处理对识别的结果进行优化,得到田间地膜的面积与分布。结果显示:对于土壤中的较亮的部分和地膜中凹陷的部分,SAM和SVM显然比PCA和FSS的识别效果更好;对于植物叶片的边缘,SAM识别效果优于SVM。在识别效果评估中,SAM识别的效果略优于SVM。SAM识别的准确率为95.76%,精准率为90.48%,召回率为99.13%,精准率与召回率的调和平均值为94.61%。论文提出的SAM方法为低空遥感高光谱图像的田间地膜识别提供了新方法。The white film is transparent,and when it is covered with soil,for their color is very similar to the soil,making it difficult to distinguish.To address this problem,this paper proposes a method of identification based on the hyperspectral image from unmanned aerial vehical(UAV).This paper takes Shuanglong town,Bijie city,Guizhou Province as the research area.The high-spectral image of low-altitude remote sensing of field film is collected,then the lens,reflectivity and atmospheric correction of hyperspectral image are performed,and the regions of interest(ROI)of field film,soil and plant are established,finally support vector machine(SVM),principal component analysis(PCA),spectral angle classification(SAM)and feature spectral segment(FSS)are used to identify the film targets in the hyperspectral images,combined with OSTU threshold segmentation and morphological pro-cessing to optimize the identification results,and the area and distribution of the film in the field are obtained.The results show that SAM and SVM are obviously better than PCA and FSS for brighter parts of soil and depressed parts of the film,as for the edges of plant leaves,SAM is better than SVM.In the recognition evaluation,SAM recognition is slightly better than SVM.SAM identification accuracy is 95.76 percent,precision is 90.48 percent,recall rate is 99.13 percent and F-score is 94.61 percent.The SAM method proposed in this paper provides a new method for field film recognition of low-altitude remote sensing hyperspectral images.

关 键 词：田间地膜 高光谱 主成分分析 支持向量机 光谱角 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]