Detecting and Measuring Individual Trees Using an Airborne Laser Scanner

High-resolution airborne laser scanner data offer the possibility to detect and measure individual trees. In this study, an algorithm which estimated position, height, and crown diameter of individual trees was validated with field measurements. Because all the trees in this study were measured on the ground with high accuracy, their positions could be linked with laser measurements, making validation on an individual tree basis possible. In total, 71 percent of the trees were correctly detected using laser scanner data. Because a large portion of the undetected trees had a small stem diameter, 91 percent of the total stem volume was detected. Height and crown diameter of detected trees could be estimated with a root-mean-square error (RMSE) of 0.63 m and 0.61 m, respectively. Stem diameter was estimated, using laser measured tree height and crown diameter, with an RMSE of 3.8 cm. Different laser beam diameters (0.26 to 3.68 m) were also tested, the smallest beam size showing a better detection rate in dense forest. However, estimates of tree height and crown diameter were not affected much by different beam size. Introduction Airborne laser scanning systems now offer the possibility to retrieve three-dimensional information about individual trees. Airborne lidar systems were first used in profiling mode for estimating tree height and crown closure (e.g., Aldred and Bonner, 1985). Lidar-generated canopy profiles were used to estimate stem volume (Maclean and Krabill, 1986) with models similar to those earlier developed using aerial photographs mounted on a stereo plotter. Development of the Global Positioning System (GPS) and inertial navigation systems (INS) made it possible to determine the position of each laser reflection point with high accuracy. Scanning systems were developed, and Nilsson (1996) used the distance between the first and last peak of the returning pulse and the area of the waveform from a scanning system to estimate the stem volume on field plots. The vertical distributions for squares (typically 15 by 15 m) were used to estimate mean tree height and stem volume for forest management in previously mapped stands (Naesset, 1997a; Naesset, 1997b; Magnussen et al., 1999). As scanner technology has developed, the pulsing frequency of systems has increased rapidly. According to Ackermann (1999), it is possible to obtain up to 20 points per mZ from an airplane at an altitude of 1000 m. Magnussen et al. (1999) states that six to ten laser hits per tree crown would be needed to detect individual trees. Samberg and HyyppB (1999) have shown that individual trees can be identified using airborne laser data. Brandtberg (1999) detected individual trees using laser data and validated the results by comparing with aerial photographs. The ability to detect individual trees makes it possible to estimate the height and crown diameter of these trees. Using these estimates, the stem diameter and stem volume can be derived. The correlation between stem diameter and crown diameter was studied early in order to estimate stem diameter by measuring tree crown sizes in aerial photographs (e.g., Jakobsons, 1970). Hyyppa et al. (2001) used laser-measured tree height and crown diameter to estimate the stem diameter of individual trees. Estimated stem diameter together with laser measured tree height was used as input to existing stem volume functions for individual trees, making it possible to estimate stem volumes for all detected trees in a stand. They evaluated the estimates using forest stands with a mean size of 1.2 ha and corrected for sampling error. Mean tree height and stem volume were estimated with a standard error of 9.9 percent and 10.5 percent of the mean values, respectively. These estimates were better than a traditional field inventory of the stands. Detection of trees has usually been evaluated by comparison with aerial photographs, and estimates have been evaluated by summing values for all trees within delineated areas. In this study, evaluation of tree detection was possible because, at a test site in southern Sweden, the positions of all trees within delineated areas had been measured on the ground with high precision. An algorithm was first developed. The algorithm was evaluated over this area, and the ability to link field-measured trees with laser-measured trees made it possible to study which trees were detected. No parameter settings were changed during the evaluation. The measurement of tree height and crown diameter of the detected trees could also be evaluated. The effect of the beam size on these estimates was also possible to investigate because a laser scanning system with a programmable scanner was used. Furthermore, image processing methods for removal of penetration into the crowns based on theories of active contours (e.g., Cohen, 1991; Cohen and Cohen, 1993; Kass et al., 1998) were evaluated for the first time. The objectives of this study were to (1) evaluate how well individual trees could be detected by segmenting a canopy model of the tree crowns; (2) evaluate the accuracy of the tree height estimates, the crown diameter estimates, and the stem diameter estimates of the detected trees; and, finally (31, investigate the influence of the beam size on the ability to detect single trees and on the tree height and crown diameter measurements of individual trees. A. Persson and U. Soderman are with the Department of Laser Systems at the Swedish Defense Research Agency, P.O. Box 1165, SE-58111 Linkoping, Sweden ([email protected]; [email protected]). J. Holmgren is with the Remote Sensing Section, Department of Forest Resource Management and Geomatics, Swedish University of Agricultural Sciences, SE-90183 Umei, Sweden. PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING Photogrammetric Engineering & Remote Sensing Vol. 68, No. 9, September 2002, pp. 925-932. 0099-lll2/02/6809-925$3.00/0 O 2002 American Society for Photogrammetry and Remote Sensing S e p t e m b e r 2002 925