Pinus Massoniana Identification Based on Object-oriented Convolutional Neural Network and Random Forest
This study aims to address the challenge of identifying Masson pine in the complex tree species composition and large spatial scale of southern hilly areas. An identification model combining convolutional neural networks (CNN) and random forest (RF) was developed using object-oriented and deep learning methods and applied to high-resolution imagery from the Gaofen-2 satellite in Jianou,Fujian province. The experiment showed that the model combining CNN and RF outperformed both the model using only CNN and the model using only RF classification algorithm,with overall good classification accuracy. The extraction of Masson pine forest under the guidance of the forest inventory of the third national land survey also achieved good results. Based on this,the spatial distri-bution of Masson pine in Jianou city was analyzed,which can effectively predict the spatial distribution of Masson pine forests and has practical value.
convolutional neural networkrandom forestmulti-scale segmentationdepth learningtree species recognition
万方数据
维普
