Information Extraction Of Elm Sparse Forest In Otindag Sandy Using GF-2 Satellite

Elm sparse forest widely distributed in the Otindag Sandy Land is a very special vegetation types.It is also an important component of the Otindag Sandy Land ecosystem.It plays an important role in windbreak and sand fixation,climate regulation and grassland ecosystem maintenance.However,it did not cause enough attention,even so far,the query on the elm forest research is very small,and research content is limited to the community structure,species composition and the influence of human disturbance on elm sparse forest.There are few studies on its specific distribution range,density,spatial pattern and so on.The research methods are mostly based on traditional methods such as survey,field visits and historical documents.They are laborious and have great financial resources,and the investigation period is long and difficult to update.It is also difficult to meet the demand of obtaining a wide range of elm spatial distribution.Therefore,Technology to automatically identify the spatial distribution of elm trees is necessary.With the development of remote sensing technology,the spatial resolution of remote sensing images is getting higher and higher.The crown of each tree can be clearly seen in the high resolution remote sensing image.According to the characteristics of the geometric shape,size and spatial pattern displayed on the image,tree crown information can be estimated accurately.The accurate information of elm sparse forest canopy is a prerequisite for other scientific and rational research,and it is also an important reference for decision makers.In this paper,the domestic GF-2 high resolution satellite remote sensing satellites were used as the data source,and the research was implemented on the Zhenglan Banner in Xilingol lengue of Inner Mongolia,using the stratified extraction method to extract tree crown of the elm sparse forest.Firstly,according to the characteristics of elm sparse distribution in the sand,combining NDVI threshold with GEOBIA image analysis technology to extract the elm distribution area,and using local maximum filtering and regional growth method to extract single tree crown,and then analyzing the coverage and forest density of the elm trees with the single wood scale.Finally,the accuracy of the extracted results is evaluated with the measured data of the field.The main conclusions are as follows:(1)Domestic GF-2 data have high spatial resolution and high positioning accuracy characteristics.Using the GF-2 fusion data combined with the NDVI threshold and GEOBIA image analysis technology can extract the elm distribution area well.NDVI threshold can quickly and effectively mask most of the non-elm area,reducing the follow-up elm distribution area accurate extraction of the workload,which is suitable for the extraction of elm trees in sparse forest area.The overall accuracy was 92% and the Kappa coefficient was 0.84.(2)In the detection of elm crown position,this paper proposed to 4 pixels as a unit for local maximum filtering,in order to prevent missing small crown elm,and then a single pixel as a unit for the second filter,reducing multi extremum point can be detected in the same crown.It was verified that the average user accuracy is 0.72 and the average producer accuracy is 0.47.The mean absolute error(MAE)is 24.4 m2,the mean relative error(MRE)is 0.44,the mean bias error(MBE)is 9.2 m2.R2 between the measured crown area and the delineated crown area was 0.6.(3)According to the statistics,there were 2284950 elm trees in the study area.The total coverage area of elm crowns is 224.76 km2,accounting for 2.86% of the whole study area.Most of the elm sparse forest coverage is less than 5%.Most elm sparse forest densities are less than 10 trees per hectare.The results show that the elm forest concentration area is mainly distributed in the middle and northeast of Zhenglan Banner.The others are sporadic.