Investigation on capability of digital aerial images in recognizing of different tree species in the Hyrcanian mixed forest (Case study: Shastkalateh forest, Gorgan)

Identify tree species and Mapping compound of trees play an important role in making optimal decisions for the management forests ecosystem of large areas. evaluation different sources, of remote sensing such as, aerial digital images of forest resources as an Solutions, Replace Ground methods is Considered in recent years. Remote sensing data, especially digital aerial images with high spatial and radiometric resolution could be a useful tool to identify tree species. By conventional methods pixel-based, classification the pixels of the images can be done the different algorithms. Digital classification conventional methods such as maximum likelihood algorithm the most common methods based on pixel-based classification. The use of modern methods of classification Such as parametric algorithm support vector machine is essential compare the performance.
According to a few studies to examine the ability of these images of urban forests and forestation north of the country, And lack of research evaluating the ability of Digital images aerial identify tree species the of Caspian mixed forests, the aim of this study research is investigation of capability of UltraCam-D aerial digital photographies in identify tree species the Caspian Mixed Hardwood forest regional of area of district 1 of Shast Kalate forest in GORGAN and compare Operation two pixel-based algorithm the maximum likelihood and support vector machine. There are several ways to extract information from this type of image.
provided the ground truth map position of 128 dominant trees by accurate registration with DGPS. Identify and classification of tree species using pixel–based method and Collection of original bands and artificial bands was derived processing bands using the maximum likelihood algorithm and support vector machine was used. Accuracy assessment of Maps derived from classification done with the use of 25% of the ground truth
After filtering accuracy evaluation results showed the map of the classification with maximum likelihood algorithm to the overall accuracy and kappa coefficient of 63.63% and 0.51, and support vector machine algorithm to the 42.37% and 0.2 respectively.
By comparing the results show that the pixel-based classification method in identifying tree species because of the salt and pepper effect or not using of the auxiliary data (slope, elevation, etc.) in the classification process has been relatively effective. The use of other methods such as object classification method based on identify of tree species is recommended. In addition, need to assess the capability of the images data in different tested habitat conditions.