Extracting Montanic Apple Orchard Information Based On Multi-temporal High Resolution Remote Sensing Image

Remote sensing technology has been widely applied in the field of agriculture,with the development of information technology.Many multi-scale spatial resolution data provides huge amounts of information for extracting crop planting area.The resolution of the implemented image data has been improved,so agricultural remote sensing opens the era of high resolution remote sensing step by step.Montanic remote sensing image classification is a conundrum in the field of remote sensing research.It is difficult to classify,because the fluctuation in the topography is big and crop planting is scattered.It is important to promote the sustainable development of apple industry in China,through the dynamic monitoring against apple planting regions and gaining the apple orchard overall area and distribution.This paper takes Yiyuan county as the study area to extract the apple orchard information by using multi-temporal Landsat8 OLI and GF-1 satellite image data combined with the spectral characteristics of main features and DEM data,vegetation index to determine optimum temporal for apple orchard information extraction.Meanwhile,extract the apple orchard information by using a variety of classification method and carry on the comparison to the precision analysis of different classification method.The main results and conclusions are as follows:(1)Three times spectrometric experiments were conducted by using portable hyperspectral feature spectrum instrument in April,May and October 2016.According to the distribution of local vegetation,the spectral characteristics of main features like apple,cherry,peach and pine were collected and made an analysis.The research showed that: apple orchard and other vegetation types had the distinction,meanwhile the ideal bands were visible band from 0.450 to 0.680 micrometer and near infrared band from 0.680 to 0.750 micrometer.(2)Based on the reflectivity extracted from the image,normalized difference vegetation index(NDVI)and enhanced vegetation index(EVI)of the apple orchard,plough,other orchard were calculated respectively by using the apple growth period of four phase(germination period,florescence,young fruit period,fruit ripe period)Landsat8 OLI images.The features which combined with spectral characteristics of the image and spectral characteristics of field survey were analyzed.The research showed that: the difference of the apple orchard and other features' spectral characteristics and characteristics of vegetation index was most obvious.Therefore confirming the optimum temporal was apple florescence in April for extracting apple orchard information.(3)GF-1 image in apple florescence was selected by combining with the DEM data,vegetation index and spectral characteristics of features through using supervised classification,decision tree classification,and object-oriented classification to conduct apple orchard information extraction.The research showed that: the area precision and the overall classification accuracy of supervised classification for the extraction of apple orchard were high with 97.19% and 91.95% respectively.The area precision and the overall classification accuracy of decision tree classification which joined the DEM and slope information also were higher with 94.52% and 90.64%.And the classification accuracy of object-oriented classification was the highest with area accuracy of 98.53% and the overall classification precision of 95.88%.