Research On Larch Plantation Recognition Based On Multi-source Data

Larix gmelinii is a common cold-temperate and temperate tree species in most forest areas of China.It is not only widely distributed naturally but also has the Characteristic of fast-growing,good material quality,strong resistance and wide adaptability,and has become the main afforestation tree species in northern China.Larch plantation can reduce the pressure on the supply and'demand of natural forest timber to a certain extent.With the steady advancement of China's key forestry projects,the information extraction of larch plantation has become a new demand in the field of larch.How to,use remote sensing to quickly obtain the spatial distribution information of Larch Plantation in China has important theoretical value and practical significance for grasping the spatial distribution of regional larch plantation,promoting rational planting of Larch plantation,improving the cultivation and management level of larch plantation resources and achieving sustainable management.This paper takes Larch plantation in Mengjiagang Forest Farm,Jiamusi City,Heilongjiang Province as the research object,and uses Landsat8,GF-1 remote sensing image and DEM model with combining the ground sample plot and the forest resource investigation data to carry out forest classification experiment for extracting the spatial distribution of Larch plantation with high precision which through the selection of feature space and classifier.It is for the purpose of studying the theory and method of larch plantation information extraction.The main contents and conclusions are as follows:(1)The paper explores the effect of multi-feature fusion data on the extraction of Larch plantation by optimizing temporal features,ratio characteristics,texture features and combining with topographic factors to construct the classification feature database of larch plantations,The results show that the extraction of Larch plantation is influenced by the information of data source.The correct application of temporal features,ratio features,texture measures and topographic features plays a very important role in obtaining high-precision classification results,the extraction results of Larch plantation based on multi-feature fusion data are obviously better than that of single-temporal remote sensing data.(2)The maximum likelihood method,spectral angle method,K nearest neighbor method and random forest method were used to extract forest types in the study area to screen the best classification method for Remote Sensing Extraction of Larch plantation.The results show that among the four classification methods,the random forest method has the best comprehensive effect,the comprehensive evaluation accuracy of the classification results supported by Landsat8 image and GF-1 image are 92.47%and 94.31%respectively,however,it takes a long time,so it is suitable for academic research work of larch plantation extraction in small area.The comprehensive evaluation accuracy of the maximum likelihood method supported by Landsat8 image and GF-1 image is 92.20%and 93.8%respectively.The method is simple,less time-consuming and high precision,so it is suitable for the mapping research of Larch Plantation in large area.The comprehensive evaluation accuracy of the spectral angle method and the K-nearest neighbor method under the support of two data sources has reached more than 85%,which meets the precision requirements of Remote Sensing Extraction of forest types,they can also be used for information extraction of Larch plantation.(3)In this paper,Landsat8 image and GF-1 image were used as the main remote sensing data sources,and used to study the remote sensing data sources suitable for Larch plantation extraction through comparing and analyzing the remote sensing extraction accuracy of Larch which supported by the two images.The results show that Landsat8 image has abundant spectral information,and the data is easy to obtain,the processing speed is fast,the image coverage area is large,so it is suitable for mapping Larch Plantation in large area.GF-1 image has high spatial resolution and high researching ability to temporal characteristics,ratio features.,texture features and topographic factors.,meanwhile it has high extraction accuracy of Larch plantation,but its data volume is large,processing speed is slow and time-consuming,so it is suitable for academic research of Larch plantation extraction in small area.