•Maize and soybean heights were successfully estimated using UAV-LiDAR data.•Method based on LiDAR variables yielded higher accuracy than CHM-based method.•Maize height prediction model outperformed soybean height prediction model.•Prediction results of crop height were poor when point density was below 1 point/m2. Crop height is a key structure parameter for the modelling of crop growth, healthy status, yield forecasting and biomass estimation. Unmanned aerial vehicle (UAV) LiDAR systems can quickly and precisely acquire vegetation structure information at a low cost. UAV LiDAR data are increasingly used in vegetation parameters estimation. In this study, we estimated maize and soybean heights using two methods, i.e., based on LiDAR-derived CHM and based on LiDAR variables. The results show that UAV LiDAR data can successfully estimate maize and soybean heights. We found that the method based on LiDAR variables can produce more accurate estimates than CHM-based method. The estimation model of combined maize and soybean had a better prediction performance than those of the specific maize and soybean. Moreover, the soybean height estimation models derived from both methods yielded the lowest prediction precision. We studied the influence of LiDAR point density on crop height estimates through reduced point density (0.25–420 points/m2). When LiDAR point density was less than 1 point/m2, the estimation precision for the specific maize and soybean dropped rapidly with the decrease of point density. However, the point density had no significant influence on crop height estimation precision while LiDAR point density was greater than or equal to 1 point/m2. Moreover, the original point density did not generate the highest estimation precision in our study. Therefore, high LiDAR point density may be not required for estimating vegetation parameters, and a good balance between the point density and data acquisition cost should be found.