Study On The Forest Classification And Crown Density Extraction Method Based On ZY-3 Image Data

Taking Jiangle state-owned forest farm, Fujian province as the research site, an object-oriented classification method combine with texture information is applied in this paper to classicify forest and extraction Canopy closure information by means of ZY-3 remote sensing image. Resoults of the research promote the application of ZY-3 image in forest resources inventory. Through analysis and research, conclusions of this research paper are as follows:(1) Popular fusion methods of high resolution images are adopted, including the high-pass filter (HPF), multiplicative fusion, brovey transform, panchromatic enhancement fusion (Pansharping), transformation of principal components (PC), Gramm-Schmidt fusion, etc. Combining subjective evaluation and objective analysis, it’s concluded that the high-pass filter is the best image fusion method when studying images of the ZY-3.(2) Through the study of image variables such as variance, skewness, coefficient of variation, average and correlation, then selected 3(ZY-3) and 9(QB) variables from 58 characteristic variables to image classify.(3) Using the method of supervised classification and threshold classification, the overall classification accuracy reached 0.840 and 0.883 respectively, and the overall Kappa coefficient was 0.794 and 0.850 respectively, shows that the classification results of images from both satellites are good.(4) With 58 data by means of stepwise regression, an estimating model is established to predict the crown density. The R2 in this model is 0.714, and the parameters variation coefficients is within 20%, which shows the model is relatively stable.19 data for test,accuracy test shows the prediction accuracy of this model reaches 88.640%, a relatively high level.