Terrestrial mobile LiDAR systems have demonstrated significant application value in vegetation structural parameter extraction and forest resource inventories due to their close-range scanning and automatic point cloud registration capabilities. However, the operational mode commonly adopted in urban forest surveys—rapid mobile scanning along roads to capture peripheral data of green patches—lacks systematic validation regarding its vegetation parameter observation capacity and measurement accuracy. This study employed a backpack-mounted LiDAR system to conduct field scanning in two representative urban forest types: sparse and closed-canopy stands. Comparative analysis between point cloud-derived parameters and field measurements revealed measurement accuracies (R2) of 0.96, 0.98, and 0.97 for tree height (TH), diameter at breast height (DBH), and crown width (CW) in sparse stands, while corresponding values in closed-canopy stands were 0.95, 0.99, and 0.65. These results confirm the reliability of rapid scanning for TH and DBH measurements, but highlight significant errors in CW estimation for dense stands. By incorporating measured DBH and TH into species-specific biomass regression equations, we established a regionally applicable generalized biomass estimation model. Building upon individual tree segmentation, this approach achieved large-scale aboveground biomass quantification at individual tree level across urban forest areas. The developed methodological framework provides a novel technical solution for digital urban forest resource inventories, demonstrating significant practical value in enhancing precision management of urban green spaces and supporting ecological conservation decision-making processes.