In forestry supervision, timely and accurate estimation of the felled tree volume is a very important task. Obtaining the felled tree volume by remote sensing using an unmanned aerial vehicle (UAV) is an effective means to reduce cost. Quantitative inversion of the felled tree volume of individual trees by visible light remote sensing using a UAV has major advantages but its accuracy is still a common concern in practice. The objective of this paper is to verify the feasibility and reliability of quantitative inversion of the felled tree volume by UAV visible light remote sensing. The proposed workflow is as follows. Firstly, raw images are obtained by UAV oblique photography and processed into a digital orthophoto map (DOM) and digital surface model (DSM) based on 3D reconstruction technology implemented in DJI Terra software. Secondly, based on before- and after-cutting DSM data, treetops are extracted using the conventional local maximum method, then are matched to the corresponding stump points. The tree height of the felled tree is estimated from the elevation difference between the treetop and the stump. Using the after-cutting DOM data, the root diameter and the geometric center of the felled tree are determined by the proposed circumcircle method. Thirdly, three regression models [estimated tree height and measured tree height, measured tree height and measured diameter at breast height (DBH), and measured root diameter and measured DBH] are tested to verify the correlation between the main parameters of the volume model. Lastly, three volume models [unary tree height–volume model (M1), unary root diameter–volume model (M2), and binary tree height and root diameter–volume model (M3)] are built and compared and analyzed through the calculation of the felled tree volume, taking the measured volume as the reference data. The experimental results showed that the correlation coefficient and root mean square error of the three volume models are 0.9093 and 0.1233, 0.9589 and 0.0831, and 0.9796 and 0.0584, respectively. This study demonstrates that by combining UAV visible light remote sensing with artificial intelligence, highly intelligent forest harvesting supervision can be achieved at a much lower cost than traditional field investigation.