Research On The Division Of Subcompartment In The Western Mountain Forest Of Tianshan

Forest classification is an important work of remote sensing monitoring of forest resources,and it is also one of the important methods to grasp the present situation of forest resources.The rapid development of remote sensing technology provides a convenient technical way for large area forest remote sensing classification,but the small class division of forest resources survey also mainly depends on manual visual interpretation,which inevitably has the subjective factors of technicians,and even the phenomenon that different personnel operation will produce different zoning results.and the results of forest classification based on different spatial resolution remote sensing data still have great uncertainty.although many research results maximize the accuracy of local region classification,it is difficult to apply to other regions.Using different remote sensing data sources to carry out remote sensing classification of large area forest land is still an urgent problem in forestry production practice.Therefore,this paper studies the spruce forest in western Tianshan Mountain and the Qiaxi National Forest Park in Gongliu Sub-Bureau of State-owned Forest Administration of western Tianshan Mountain in Xinjiang.Based on the Landsat 8、GF-1 and GF-2 images,the multi-scale segmentation and canopy density estimation(1)When three different remote sensing data implement scale segmentation,this paper mainly uses OIF value algorithm and ESP evaluation tools to obtain the optimal band combination of each data and the optimal scale parameters of different object objects.Through OIF calculation and research,it is found that the near-infrared band has abundant information and strong independence,and the most outstanding contribution to OIF in the optimal band combination.On this basis,the image segmentation experiment with ESP tool was carried out,and the optimal scale parameters at different levels were finally obtained.(2)Spectral information,texture information and terrain factors were selected as the object feature factors to obtain the identification factors of forested remote sensing region,and the canopy density estimation model was constructed and evaluated.The results showed that GF-2 model with the highest resolution EA 89.93%and RMSE 0.069 5.It can better reflect the actual situation of forest canopy density in Xinjiang.(3)Applying object-oriented classification method and combining HIS transform value,spectral information,vegetation index factor,texture feature and terrain factor to carry out hierarchical remote sensing classification research,the final conclusion is that GF-2 image data with the highest resolution has the best classification effect.Compared with the extraction of forestland area in the category ii survey of forest resources in 2014,the consistency reached 94.47%.Therefore,the use of high-resolution image data can better carry out object-oriented classification and provide a reliable basis for the division of small class in this area.