Acta Geodaetica et Cartographica Sinica ›› 2020, Vol. 49 ›› Issue (2): 235-244.doi: 10.11947/j.AGCS.2020.20190272

• Photogrammetry and Remote Sensing • Previous Articles     Next Articles

Forest resource classification based on random forest and object oriented method

WANG Meng1, ZHANG Xinchang1,3, WANG Jiayao2,3, SUN Ying4, JIAN Ge1, PAN Cuihong1   

  1. 1. School of Geographical Sciences, Guangzhou University, Guangzhou 510006, China;
    2. Research Institute of Henan Spatio-Temporal Big Data Industrial Technology, Zhengzhou 450018, China;
    3. The College of Environment and Planning, Henan University, Kaifeng 475004, China;
    4. Department of Geography and Planning, Sun Yat-Sen University, Guangzhou 510275, China
  • Received:2019-06-27 Revised:2019-11-25 Published:2020-03-03
  • Supported by:
    The National Key R & D Program of China (No. 2018YFB2100702);The Key Program of the National Natural Science Foundation of China (No. 41431178);The Key Program of the Natural Science Foundation of Guangdong Province (No. 2016A030311016);The Smart Guangzhou Spatiotemporal Infomation Cloud Platform Project (No. GZIT2016-A5-147);Supported by the Research Institute of Henan Spatio-Temporal Big Data Industrial Technology(2017DJA001);The Fundamental Research Funds for the Central Universities (No. 19lgpy44)

Abstract: Given that there are few studies on forest resource classification with the lack of relatively simple and effective methods, a forest resource classification method integrating object-oriented segmentation and random forest is proposed in this paper. Object-oriented segmentation technology could efficiently reduce the "salt and pepper effect", and random forest classification algorithm has the advantages of high accuracy, strong anti-noise ability and satisfying stability. Therefore, we built the optimum random forest classification model by adjusting the object-oriented segmentation parameters, constructing the optimal feature space and estimating the number of decision trees in random forests. Besides, the SVM algorithm is taken into comparison. The results show that the overall accuracy of the classification algorithm in this study is 83.34%with the Kappa coefficient reaching 0.789 2, which are significantly higher than that of SVM algorithm. It proves that object-oriented random forest classification can effectively improve the accuracy of forest resource classification.

Key words: forest resource classification, object-oriented method, random forest

CLC Number: