A Study On The Extraction Of The Stand Description Factors From High Resolution Remote Sensing Images

In this paper, with QuickBird images, the method of how to extract some important stand description factors was discussed. According to the rich texture and structural information bearing in the high-resolution remote sensing images, a method of how to extract woodland area, which based on knowledge discovery, was proposed. Integrating the information of the images bands, spectral, NDVI (Normalized Difference Vegetation Index) and texture characteristics of images, the classification precision was improved evidently. Studying spectral reflectance characteristics of the crown, the method of calculating the stand density was proposed, after the points whose reflectivity being the greatest in crowns were extracted through Local maximum filtering method from QuickBird images. The crown diameter of stem can be used to estimate DBH(the diameter at breast height) indirectly, then the single-volume(i.e. the volume of individual tree) can be evaluated. For the forest part with high canopy density, which can not be identified clearly in the images, after studying the spatial structure characteristics of crowns, which presented in the images, a method of estimating the average crown diameter with the range of the semi-variance function of images slices, that is subset images cut from the images, was put forword. Also, A simulated stands images method was used to prove the effectiveness of the method. As to the forest part with low canopy density, which is easy to identify clearly in the images, a method of extracting the crown, with the combination of NDVI contour lines and the points whose reflectivity being the greatest in crowns, was proposed. The shadow in the images contains the three-dimentional information. After study the effection of sun high angle and azimuth angle, and that of satellite high degree and azimuth angle, towards the stem, with QuickBird images, taking stem shadow length, slope, sun high angle and azimuth angle as parameters, the model, which is used to estimate the tree height, was explored. In the paper, the research result was tested with the investigation data of Baisha forest farm, of Minhou county in Fuzhou, and the effectiveness of the methods mentioned in the paper are proven.