Research on Field Membrane Recognition Based on Hyperspectral Remote Sensing Technology
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 vec-tor 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.
国家科技期刊平台
NSTL
万方数据
