It is of vital significance to efficiently collect the information of crop canopy structure for new cultivar breeding and field management optimization. At present, methods such as three-dimensional digitizing have been used to obtain canopy structure information of field-grown crops, but most of them require manual intervention, which is time-consuming and laborious. Therefore, it is urgent to develop novel methods with high-efficiency. A micro UAV was used in the field to acquire image sequences of maize canopy at the seedling stage, and individual plants as well as several neighboring plants at the later mature stage. Considering the heavy shading among plants at the late stage, surrounding plants of the target plants were removed before images were taken. Based on the point clouds reconstructed using the UAV images, the canopy structure model was efficiently built by creating pseudo poles. Then, the model was evaluated according to the field measurements of plant height, leaf length, max width and leaf area. There was a good agreement between the measured and calculated plant height, leaf length and max width with R2 no less than 0.91 and RMSE, rRMSE and ME were small for both growth stages. The R2 of leaf area at both growth stages were 0.96 and 0.76, respectively. RMSE, rRMSE and ME were small at the seedling stage while marginally larger at the mature stage. The proposed method provided a novel way for high-throughput plant structure modeling and phenotyping of field-grown crops.