 This study proposes a hybrid approach combining random forest and texture analysis to improve the accuracy of urban vegetation mapping from unmanned aerial vehicles, UAV. The study found that random forest outperforms traditional maximum likelihood classifier and performs similarly to object-based image analysis. Additionally, the inclusion of texture features improves classification accuracy significantly. Furthermore, the study demonstrates that the accuracy of urban vegetation mapping follows an inverted U-relationship with texture window size. This article was authored by Kwamong Fong, Jantel Lu, and Jantwa Gong.