Remote Sensing Interpretation On The Scenic Forest Landscape Based On Object-Oriented Classification

The quality of scenic forest is mainly reflected in the demand for people aesthetic satisfaction of scenic forest, is a comprehensive reflection of human about scenic forest structure rationality and in harmony with its surrounding degree. This study is based on high-resolution remote sensing image classification, with ground survey information and data, inverse modeling of the test area scenic forest tree species composition, stand density, forest spatial structure, tree spatial structure, canopy structure and some others landscape elements, and research its effect on the remote sensing interpretation of scenic forest landscape. The main conclusions are as follows:(1) In this paper, we use split---rules-based classification---based on classifier of object-oriented classification method. By using four classifiers of eCognition software including Bayes classifier, support vector machine classifier (SVM), K nearest neighbor node algorithm (KNN), decision tree classifier (Decision Tree), we compared the accuracy of forest classification, and the results show that classification accuracy of decision tree classifier> K nearest neighbor node algorithm> Bayes classifier> SVM classifier. Overall classification accuracy of decision tree classifier is90.20%, Kappa coefficient is89.61%, precision of production has a good stability, and obtain relatively satisfied classification results.(2)Based on the results of remote sensing classification, we conduct ground survey in Banshihe forest farm, and analyze the test area194plots of various types of scenic forest structural conditions. Studies show that the test area is composed mainly dominated by mixed forest, among birch, linden, pine, poplar, elm as the dominant species, and planted larch, spruce forest; Most forest stands are in middle-aged, mature-aged stage, followed by young forest, nearly mature forest, almost none overmature forest. Forest stand density is between600/hm2and1500/hm2, middle-aged forest stands have a great variety of species composition; The vertical diversity of broadleaf forest mixed stand is higher than coniferous forest, the height of middle-aged forest stands is greater than the young forest, among larch forest, mixed forest is significantly higher than other forest (height>10m). With the growth of forest age, the distribution of various types of forest stands change from gather strong gradually to random. Broadleaf forest leaf area index is higher than coniferous forest, and the leaf area index and the average leaf angle index value is gradually increasing from young to middle-age stage; while the canopy gaps parameter value is decreasing; with the growth of forest age, the leaf area index and the average leaf angle index values are decreasing, while the canopy gaps parameter value is increasing.(3) In this paper, we use SPSS statistical analysis software to build a linear regression model on the stand spatial distribution of landscape elements. The results show that the average leaf area index and the leaf angle index have a high coefficient by using relatively few variables. The estimation model of stand density, DBH dispersion, K value of the negative binomial are poor, the coefficient of determination were0.500,0.535and0.516respectively, indicating there is a nonlinear relationship among stands sparse density, stands species differentiation degree, distribution of trees on spectral texture features.(4) By ANOVA analysis we conclude the average density, the average diameter, the mixed degree, the vertical diversity, the average leaf area, the canopy gaps P≤0.05, indicating that these structural parameters have a significant effect on the accuracy of remote sensing interpretation. The results of cluster analysis show that the average density of801-1100plants/hm2and1101-1500plants/hm2, the average diameter of9.0-17.0cm and17.1-26.0cm, the mixed degree of0.41-0.60and0.61-0.80, the vertical diversity of0.20-0.50and0.81-1.20, the average leaf area of2.81-3.80and3.81-5.31, the canopy gaps of0.10-0.20and0.21-0.30have a significantly relationship with various types landscape of remote sensing interpretation separately.