Remote Sensing Image Recognition And Monitoring Of Orchards In Mountainous And Hilly Region Based On Landsat

The fruit economy is an important part of modern agricultural economy.Identifying and monitoring the distribution of orchards is of great significance for grasping the status of rural development,agricultural planting structure and farmers’ living conditions,and to formulating agricultural policies.Landsat remote sensing images provide data support for the identification and monitoring of large area and long period of orchard.Computer technologies such as machine learning and digital image processing technology provide technical means for orchard identification and monitoring.At present,the identification and monitoring of orchards are mainly concentrated in plain areas,but few orchards are extracted from the mountainous and hilly areas,especially the Yunnan-Guizhou plateau area represented by Yunnan Province.The distribution of orchards in hilly and mountainous areas is relatively dispersed,and it is easy to confuse forest land and cultivated land,which makes it difficult to apply the existing research methods directly.To this end,this paper takes Zhaotong City,Yunnan Province as the research area,collects samples of orchards and non-orchards based on Landsat TM/OLI satellite remote sensing images in the past 10 years,constructs orchard identification remote sensing feature sets,and applies decision trees,support vector machines and random forest methods Identify and extract orchard information,apply the best classification method to monitor the temporal and spatial changes of the orchard,and use the principal component analysis model to analyze the changes and its causes in the orchard area from 2010 to 2019.The following contents are the main contents and conclusions of this study:(1)Research on remote sensing recognition of orchards in mountainous and hilly areas based on machine learning.Based on the Google Earth Engine cloud platform,using the classification method of random forest built a remote sensing image recognition model of orchards,and the distribution mapping of orchards in Zhaotong City in 2010,2015 and 2019 was realized.The research results show that compared with support vector machines and decision tree classification methods,random forest classification results are better under the same characteristics.The accuracy of random forest classification to extract orchard area exceeds 96.06%,and the Kappa coefficient is above 0.94.The extraction results are in line with the actual situation,and the classification effect is good.The random forest classification extraction of orchard promotes the classification accuracy greatly,and can be used for orchard extraction.(2)Spatial-temporal change monitoring of orchards in mountainous and hilly areas based on geographic information technology.Using geographic information technologies such as land use transfer matrix,overlay analysis,and regional statistics,the dynamic changes in the area of orchards in Zhaotong City during the study period were described and analyzed.The study results show that the area of orchards in Zhaotong improved by 32.15 km2 from 2010 to 2015,and 4.96 km2 from 2015 to2019.The areas of cultivated land,forest land,and construction land were transformed into orchards.It reflects that Zhaotong City has continuously increased the support and cultivation of orchard industry development since 2010.Fruit farmers have developed a large number of mountain fruit trees to plant fruit trees,which has led to an increase in orchard area.The change of orchard planting area appears to be on the rise from 2010 to 2019.In recent years,changes in the natural environment and agricultural development have affected the changes in the area of orchards.(3)Analysis of related factors of orchard area change based on principal component analysis.Based on the 2010,2015 and 2019 statistical yearbooks of Zhaotong City,six related variables(total population,agricultural output value,net income of rural residents,annual precipitation,annual average temperature,annual sunshine)were selected to construct a principal component analysis model to analyze the related factors of the orchard area change,which affect the dynamic change of the orchard area.Studies have shown that the significant changes in orchard area are factors such as agricultural output value and rural residents’ net income.