Multi-spectral Remote Sensing Recognition And Application For Easily Confused Tree Species In Mountain Area Based On Cloud Model

The classification based on remote sensing data is an indispensable content of forest resource investigation and monitoring,and the accuracy of classification directly affects the application level and practical value of remote sensing data.How to identify tree species to satisfy certain precision is a key problem in the research of remote sensing image,which is of great significance.At the same time,the theory of cloud model has gradually been applied to the interpretation of remote sensing image,its method is simple and the computation is simple,it can obtain higher precision of the interpretation classification than the traditional method,and has good development foreground.However,there were few researches on forest recognition by combining remote sensing data with cloud model.The study area was sited in mountain Taishan in Shandong Province,the easily confused tree species of Pinus tabulaeformis and Platycladus orientalis in Taishan were selected as the object of study.The spectral reflectance of two easily confused species were extracted based on ZY-3 and ZY-1 02 C multi spectral remote sensing image.With the support of Erdas image,ENVI,ArcGIS,etc.The sensitive spectral index and texture features were screened to set up the recognition model between Pinus tabulaeformis and Platycladus orientaliscombining with the cloud mode.The higher species identification accuracy was obtained,and it could realize the species recognition rapidly.The main contents and conclusions are as follows:(1)Screening out the sensitive spectral indices of Pinus tabulaeformis and Platycladus orientalis.The 158 spectral indices were constructed based on the canopy reflectance of Pinus tabulaeformis and Platycladus orientalis,and the sensitive spectral indices were screened out with the principle of high correlation.It was found that the sensitivity of Pinus tabulaeformis and Platycladus orientalis was mainly in red bands,which is consistent with vegetation characteristics that is mainly reflected in the bands.(2)Establishing the cloud model based on spectral characteristics.Ten sensitive spectral indices of Pinus tabulaeformis and Platycladus orientalis were screened,and one-dimensional cloud model were established.Then compared with the recognition accuracy of ten one-dimensional cloud models,and three sensitive spectral indices of Pinus tabulaeformis and Platycladus orientalis were selected to establish threedimensional cloud models.Finally,the recognition accuracy of Pinus tabulaeformis reached 90.56%,Platycladus orientalis reached 88.92% and the overall precision reached 90.06%.It can provide the scientific reference for other easily confused species recognition with the multispectral remote sensing.(3)Establishing the cloud model based on texture features.Based on the first principal component,the texture features in different sliding window scales were extracted,then these texture features were used to establish the cloud model.It showed that the accuracy of the cloud model established by the texture feature of 3?3 window was the highest,the sensitive texture parameters chosen for Pinus tabulaeformis were mean,entropy and second moment,and the sensitive texture parameters chosen for Platycladus orientalis were mean,variance and contrast.These parameters were used to build up 3-dimentional cloud model,but the recognition accuracy of the easily confused tree species is not high,the classification accuracy of Pinus tabulaeformis was 55.26%,and Platycladus orientalis reached 66.52%.(4)The spatial inversion of tree species in Taishan.Because the recognition accuracy of cloud model that based on texture features was not high,this study mainly focuses on spectral characteristics,and the spatial distribution of Pinus tabulaeformis and Platycladus orientalis in Taishan was retrieved.Better classification accuracy was obtained,the total classification accuracy reached 87.62%,and Kappa coefficient was 0.8146.In summary,combining the cloud model with the spectral characteristics and texture features of multi spectral remote sensing data,this study provides a feasible method and process for tree species identification model,which has positive significance for fast identification and precise management of tree species,and also provides reference for multispectral remote sensing identification of other confusing tree species.