基于面向对象的热带林分类方法研究

【目的】为了加强热带林资源的保护,采用遥感技术对热带林植被进行分类研究。【方法】基于SPOT6高分辨率遥感影像,采用ESP多尺度分割评价模型与专家知识结合的方法确定最优分割尺度参数,在分割的基础上充分挖掘目标地物的光谱、形状及纹理信息,合理选择分类特征组合,建立分类规则,构建了一套基于面向对象的热带林多尺度分类方法。【结果】与单一尺度的分类方法相比,该方法分类精度有明显提高,分类总体精度达到84.46%,并且缩短了传统目视确定最优分割参数的时间,提高了分割效率和精度。【结论】基于面向对象的多尺度分类方法能够实现高精度的热带林植被信息提取,可为遥感分类技术在热带林的应用提供参考。

Object-oriented classification of tropical forest

Objective] This study was conducted to improve tropical forests protection by assessing remote sensing-based classification technology based on remote sensing.Methods] Using an object-oriented classification method,tropical forests was extracted based on SPOT-6 high-resolution remote sensing images.Specifically,the estimation of scale parameter (ESP) multi-scale segmentation model,coupled with experts' knowledge,was used to determine the optimal segmentation scale parameters.Through the analysis of the spectral,shape and texture features of the image objects,a reasonable set of these features was established.As a consequence,an object-oriented multi-scale classification method was created to map the distribution of tropical forests using classification rules.Results] The results showed that the proposed object-oriented multi-scale classification method had the ability to extract information about the distribution of tropical forests.This method had an overall classification accuracy of 84.46％,which was an improvement over that of singlescale classification.Moreover,less time was needed and accuracy was improved with this method,compared with the traditional segmentation method.Conclusion] Our object-oriented multi-scale classification method provides a solid technical reference for mapping tropical forests mapping,which is fundamental for monitoring and protection of tropical forests resources.