Estimating Mean Tree Crown Diameter using UAV Imagery Based on Multi Resolution and Watershed Segmentation Methods (Case study: Kheyrud Forest)

Orthphoto mosaic, and digital elevation models (DEMs) that created from UAV imagery can be used to delineate tree crowns. The goal of this research is to compare two segmentation techniques in the estimation of crown diameter. In this regard, two successful flights were carried out in two different seasons (leaf-off and leaf-on). Then we generated accurate CHM through a photogrammetric workflow using DTM and DSM. We used invert watershed (IWS) and multiresolution segmentation (MRS) to detect tree crowns on CHM and orthophoto mosaic, respectively. To compare the estimates of mean crown diameter from UAV images with actual values, 95 trees were measured. The results of comparing the estimated and field measured values showed that all two algorithms effectively delineate tree crowns. The results of linear regression showed there is a high agreement between estimated and measured values, which were (R2=0.88) for MRS and (R2=0.92) for IWS. The best result was obtained using IWS techniques (RMSE= 7.02 % and MAE = 5.97 %). T-test results showed that there are no significant differences between the field measurement and IWS estimated values. Although the t-test result showed there are significant differences between the means of MRS estimated and measured crown diameters, but based on RMSE (8.74 %) and MAE (8.11 %) of the MRS technique, the error of crown diameter estimation was small and therefore acceptable. Finally, the results showed that this methodology, as an accurate and low-cost process, could be used to estimate mean tree crown diameter.