Research And Implementation Of Key Techniques For Estimating Main Stand Factors In Multi-source Remote Sensing Forest Resources Survey

China's forest resources monitoring has entered a rapid development stage,the traditional forest resources survey methods have been difficult to adapt to the new situation of forest production and ecological construction needs.With the rapid development of satellite technology and computer technology,digital forest resources survey and monitoring has become the main development trend.How to use remote sensing data to realize the accurate and rapid extraction of the main forest factors in the second class survey,it has become the focus of forestry investigation and management department.In this paper,the Chaihe Forestry Bureau of the Inner Mongolia Autonomous Region is the research area,and the four main forestation factors of the small-scale storage volume,canopy closure,average tree height and average breast diameter in the second-class survey of forest resources are taken as the main research objects,and Sentinel-2A and GF-1PMS image is the main data source,combining first-class survey sample data of forest resources,second-class survey small class data,DEM data and forest "one map" data,using remote sensing information,terrain information and texture information as independent variable factors,using k-NN algorithm,partial least squares and robust estimation are used to estimate the main stand factors,and the accuracy of the four stand factors estimation modeling by different data sources,different modeling methods,classification modeling and texture information is studied.Law of influence.On the basis of theoretical research on key technologies such as remote sensing information extraction,modeling variable optimization,forest stand factor estimation algorithm,and small class attribute information extraction,C/S architecture is used,C#programming language is used in combination with ArcGIS Engine and plug-in Develop technology,program and implement key technologies in the Visual Studio 2013 development environment,and form a set of software systems dedicated to the estimation of the main stand factors in the second-level survey of forest resources,and use the second-level survey in the research area respectively Small class data and field measured small class data are used for accuracy verification and analysis.The main conclusions and outcomes are as follows:(1)In the case of sufficient sample plots,the precision of forest factor estimation can be improved effectively by establishing multiple forest factor estimation models according to forest land type.(2)The estimation accuracy of the main stand factors based on the Sentinel-2A image is generally better than that based on the GF-1 PMS image.The former is more suitable for the estimation of the main stand factors in large-scale forest resource second-class surveys.(3)When both parameterized and non-parametric methods are used for forest factor estimation,from the perspective of the accuracy of each forest factor estimation model,under the same conditions,the accuracy of the forest factor estimation model based on k-NN method is obvious It is superior to the other two parameterization methods,and has good modeling stability.The robust estimation and partial least squares have similar model accuracy,and are easily affected by the accidental random sampling.From the perspective of the actual estimation accuracy of each stand factor,the k-NN method has achieved good results in the estimation of each stand factor,which fully meets the accuracy requirements of the second-class survey of forest resources,and the parameterized method The accuracy of the estimation is quite different.The estimation accuracy of the parameterized small class accumulation based on the Sentinel-2A image is less than 70%.The partial tree least squares average tree height estimation based on the GF-1PMS image and the parameterized method for the accumulation of the small class The estimation accuracy of volume and average breast diameter is less than 70%.It can be seen that the parameterization method can not fully meet the accuracy requirements of these four stand factors.(4)When texture information is used as an independent variable factor to participate in the estimation of major stand factors,the impact on the estimation and modeling of each stand factor is not consistent.When the k-NN method is used,the texture information has no significant effect on the accuracy and stability of the estimation and modeling of each stand factor;when the parameterized method is used to construct the estimation model,there is a large difference in the impact caused by the texture information.The texture information of the two data sources greatly improves the accuracy of the canopy estimation model,but the impact on other forest factor estimation modeling is quite different.Based on the Sentinel-2A image,the partial least squares accumulation volume The model accuracy of the estimation and the robust estimation of the average breast diameter based on the GF-1PMS image has improved.The modelling accuracy of the stand a-factor estimation under other conditions has remained basically unchanged or declined to varying degrees.(5)On the basis of theoretical research,this paper has successfully developed the main stand factor estimation system for the second-class survey of forest resources.Through the actual application effect test,when the small-class data of the second-class survey is used as the verification standard,the k-NN method Accumulation volume estimation accuracy is up to more than 80%,canopy closure degree estimation accuracy is up to 90%,average tree height estimation accuracy is up to 85%,and average chest diameter estimation accuracy is up to 75%.the above.When using the field-measured small class data as the verification standard,the k-NN method estimates the accuracy of the small class accumulation volume to 81.8%,the canopy closure degree to 88.8%,and the average tree height to 84.3%.The estimation accuracy of the breast diameter is 84.9%,which can meet the practical application requirements of the second-class survey of forest resources.The research results of this paper can improve the efficiency of the second-class survey of forest resources.