Tree crown volume measurement method based on oblique aerial images of UAV

Abstract: Tree crown volume is an important basis for monitoring tree growth and estimating tree biomass. Accurate measurement of tree crown volume has always been a difficult issue in forest measurement research. Traditional method of estimating tree crown volume is to bring crown breadth and tree height measurements to inherent empirical models, but it's faced with problems such as rough operation mode and no precision guarantee of measurement results. In recent years, the emergence of the modern equipments and technologies lay the foundation to achieve high precision tree crown volume measurements. Unmanned aerial vehicle (UAV) oblique aerial photography technology with high-resolution images changes traditional measurement ways, which can use oblique aerial images to generate point cloud data and extract different types of tree geometry parameters by point cloud information. In this paper, a consumer-level multi-rotor UAV named DJI Inspire-1 was used as data acquisition platform, which was equipped with RGB band of ordinary digital camera named Zenmuse-X3. In the Beijing Jiufeng Forest Farm, comprehensively considering the flight operating conditions of the UAV and tree size specifications, we selected eight target trees with different types and sizes. Using the UAV in the manner of spiral flying, we obtained multi-angle oblique aerial images of these target trees. During operation, the main remote controllor controlled the flight status of the UAV and the auxiliary remote controllor controlled the status of the camera haeundae. These two were operated at the same time to collect the oblique aerial images. The following points were the schemes for collecting UAV data: a) In the case of ensuring a safe distance, taking the trunk of the target tree as the center for low-speed flight photography. b) Adjusting the camera pose in real-time during the hovering process so that making sure obtain images of the target tree at different positions and angles. c) Ensuring the overlapping rate of adjacent images collected at the same height exceeded 90%, and the overlapping rate of images collected at different height exceeded 60%. The acquired images were processed through the principle of aerial triangulation for generating three-dimensional point cloud models of target trees. Based on three-dimensional point cloud models, the research segmented the tree crown point cloud by contour lines method and determined the optimal segment number of tree crown point cloud. To extract tree measurement factors, projection method was used to reduce the dimension of the point cloud data. And the measured values of tree height and the arbitrary cross-sectional area of tree crown were calculated by using the key points. According to the established algorithm, the measured volume of the entire tree crown was calculated by accumulating the volume of each rule body after segmentation. Taking actual values by total station as reference, the accuracy of the tree height and tree crown volume measurement results was examined. The results showed that it was a feasible and effective method that the oblique aerial images of UAV were used to establish the three-dimensional point cloud models of single trees and to calculate the tree crown volume. In addition, the average relative error of tree height and tree crown volume of eight target trees was 2.88% and 9.42%, respectively. The accuracy met the standard for tree height and tree crown volume measurement resulted in forestry surveys. In conclusion, three-dimensional point cloud models generated by oblique aerial images of multi-rotor UAV can realize the extraction of measurement factors of single trees, which could be applied to the investigation and protection of ancient and famous trees. This method provides a new approach for the extraction of single trees geometry parameters.