The Study Of High-temperature Targets Identification And Mixed Pixel Decomposition Method In Remote Sensing

Remote sensing identification of high-temperature targets(such as forest fires,straw burning,volcanic magma,coke-coking,coal spontaneous combustion,oil well torches,etc.)is of great significance and practical value for environmental monitoring,ecological balance,disaster early warning,and resource protection.As an important part of high-temperature target remote sensing identification and attribute inversion,its method research has become a hot issue in remote sensing research.At present,there are various remote sensing methods for the identification and extraction of high temperature targets,including threshold method,principal component analysis method,factor analysis method,Mahalanobis distance method and remote sensing index method.The remote sensing index proposed based on the spectral features of land features lacks a quantitative evaluation index to some extent.In terms of spatial characteristics,mixed pixels lead to different degrees of error in recognition and attribute inversion of high temperature targets.This paper selects forest fires in Qigan Forest Farm in Daxinganling,Inner Mongolia,straw burning in Gongzhuling City,Jilin Province,and Kilauea Volcano in Hawaii.Using Landsat8 OLI remote sensing image data,we focus on high-temperature targets and room temperature based on variance analysis and mixed pixel decomposition methods.The differences in the spectral characteristics of the objects,the construction and screening methods of the high-temperature target remote sensing identification index,the distribution characteristics of the mixed pixels of the high temperature target,and the spectral characteristics,the main conclusions are as follows:(1)According to the basic principle of high temperature target remote sensing recognition and the characteristics of electromagnetic spectrum of surface features,comparative analysis methods are used to study the spectral characteristics and differences of high temperature objects and normal temperature features,which shows that there are significant differences in the spectral characteristics of high temperature objects and normal temperature objects and are mainly reflected in In the near-infrared region,the radiance of 0.845~0.885μm high-temperature target is lower than that of most normal temperature objects,while the radiance of 1.560~1.660μm and 2.100~2.300μm high-temperature targets is much higher than that of normal temperature objects.(2)Based on the difference in spectral characteristics between high-temperature objects and normal-temperature objects,the “separation measure” indicator in the principle of variance analysis was used to quantitatively screen the optimal bands for distinguishing between high-temperature targets and normal-temperature features,and to quantitatively evaluate the optimal band combinations.The high-temperature target recognition index identifies and extracts high-temperature targets through discriminant functions and uses confusion matrix to evaluate the accuracy.It concludes that each band of remote sensing images has different discrimination ability between high-temperature targets and different normal temperature features,and is constructed from the best band.The high temperature target recognition index has different classification suitability for different normal temperature objects,and the suitable identification index for different high temperature targets is also different.As a whole,the identification accuracy of multi-band composite index is better than that of other form indices,and the identification of actual high temperature targets is realized.In the band filtering and combination,the accuracy and efficiency of high-temperature target recognition can be improved.(3)Through the high temperature target mixed pixel decomposition study,the mixed pixel of the high temperature target is decomposed by the full-constrained least square method linear spectral mixing model,and the distribution characteristics and spectral characteristics of mixed pixels of the high temperature target are studied by combining the extraction results of the high temperature target recognition index.The area where the high temperature target is in the mixed pixel is related to the nature and state of the high temperature target,and the spectral characteristics of the mixed pixel of the high temperature target are commonly influenced by other endmembers.The larger the area ratio of the high temperature target is,the larger the pixels are.Near the center of the high-temperature target coverage area,the closer the spectral characteristics of the pixels are to the spectral characteristics of the pure pixel high-temperature target,the smaller the area of the high-temperature target within the mixed pixel at the edge position,and the spectral features are mainly affected by other features in the pixel.End effect.