基于随机森林算法的冬小麦提取研究北大核心CSCD

全球气候变化对粮食安全和农业可持续发展造成威胁,冬小麦作为全球重要粮食作物之一,其快速和准确的信息提取对保障区域粮食稳定具有重要意义。采用在农作物识别和提取领域具有明显优势的随机森林算法,结合典型冬小麦种植区光谱特征、纹理特征和主成分特征实现了30m空间分辨率遥感影像下的冬小麦地块的特征选择和快速提取,并分析了不同特征空间组合方式下的提取效果。研究表明:在"光谱特征"、"光谱特征+纹理特征"、"光谱特征+纹理特征+主成分特征"3种特征空间组合下,第3种组合方式下的冬小麦提取效果最佳,总体精度可达到84.85%,分别高于前两种方式8.08%和6.88%。因此,利用随机森林算法结合多源特征信息,可以有效实现特定农作物如冬小麦的快速提取,并为区域作物进一步应用研究提供有效数据支撑。

Research on Extraction of Winter Wheat based on Random Forest

Global climate change poses a threat to food development and sustainable development of agriculture.As one of the world's most significant food crops，rapid and accurate information extraction of it is of great importance in ensuring food stability and security.In this paper，the random forest algorithm with obvious advantages in crop identification and extraction was selected.The feature selection and rapid extraction of winter wheat plots based on 30 m spatial resolution imageswere achieved by combining the spectral features，texture features and the principal components of the typical winter wheat growing areas.The extraction results under different feature spaces combination were compared and analyzed.The results showed that under the combination of three spectral spaces：“spectral feature”，“spectral feature+texture feature” and “spectral feature+texture feature+principal component feature”，the third combination method had the best extraction efficiency and the highest overall accuracy up to 84.85%，respectively higher than the previous two 8.08% and 6.88%.Therefore，by using random forest algorithm combined with multi\|source feature information，the rapid extraction of specific crops，such as winter wheat，can be effectively achieved and provide effective data，thus supporting for the further application of regional crops.