Model-based conifer-crown surface reconstruction from highresolution aerial images. Photogr

Abstract Tree crown parameters such as height, shape and crown closure are desirable in forest and ecological studies, but difficult to measure on the ground. The stereoscopic capability of high-resolution aerial images provides a way to crown surface reco

Model-Based Conifer-crown Surface Reconstruction from High-resolution Aerial Images

Y. Sheng, P. Gong, G.S. Biging

Abstract Tree crown parameters such as height, shape and crown closure are desirable in forest and ecological studies, butdifficult to measure on the ground. The stereoscopic capability of high-resolution aerial images provides a way to crown surfacereconstruction. However, existing digital photogrammetry packages designed to map terrain surfaces cannot accurately extract treecrown surfaces, particularly for conifer crowns with steep profiles in the vertical direction.

In this paper, we integrate crown features derived from images to stereo matching, and develop a model-based approach toreconstruct crown surfaces for conifers. The idea is based on the fact that most conifer crowns are in a form of solid geometry. We modela conifer crown as a generalized ellipsoid; establish the optimal tree model using a geometric equation; and then apply the optimal treemodel to guide a conventional pyramid image matching in crown surface reconstruction. The effectiveness of the proposed - approach isillustrated using an example of a redwood tree on 1:2,400 aerial photographs.

1. Introduction

The description of three-dimensional crown shape is useful in estimating the amount of foliage, and further the photosyntheticability of trees. In forestry inventory, many parameters are collected from crowns such as crown diameter, crown height, crown closure,etc. Crown width and crown height are important inputs to forest models (Deutschman et al., 1997), and are of critical importance inmodeling forest fires (Keane et al., 1999).

It is a time-consuming and labor-intensive process to measure crown diameter and crown height in the field, let alone thethree-dimensional crown surface. This led us to develop photo-ecometrics techniques for forest inventory (Gong et al., 1999). Aerialphotography provides a practical means to tree measurement. Large-scale aerial photos have long been used for measuringing suchparameters. Andrews derived tree height from aerial images in a stereo pair as early as in the 1930s (Andrews, 1936). Tree heightreadings from 1:1,000 scale photos using a stereoplotter was found even more accurate than field measurements using tapes andclinometers (Kovats, 1997). Moessner (1949) developed a crown-density scale as a reference to interpret crown closure. Sayn-

Wittgenstein (1961) applied crown characteristics (crown density, size, and marginal and apex shape) in tree species recognition. Crowninformation is estimated most easily from aerial photos, however, the approaches are mainly based on visual interpretation. As aconsequence, the process is less efficient, subjective and error-prone (Biging et al., 1991).

Precise measurement of crowns from aerial photographs needs relatively accurate crown surface data. Due to the perspectiveview of aerial photographs crown closure is overestimated when subject location moves away from the principle points of the

photographs.. Theoretically, the parameters derived from orthophotos are free of the displacement influences. However, crown surfacedata are needed for generating orthophotos from perspective photos.

Another requirement for crown surface data comes from the automated photo interpretation. As a result of improvement inboth the computing power and spatial resolution of remotely sensed data, more attention is being paid to individual tree-based photointerpretation (Gougeon, 1992; Gougeon, 1995; Larsen, 1998; Pollock, 1996). On one hand, it is good to have photo resolution highenough to capture information on individual trees; on the other hand, the displacement of crown surface is by no means negligible. Thecurrent work in this field has focused on tree delineation using spectral information from monocular images. When results of photointerpretation of forests are inputted to a geographic information system (GIS) , the coordinate data contain errors due to the

perspective view of aerial photography. To eliminate the geometric errors, we need to use the 3D crown coordinates to orthorectify theairphotos.

In summary, the marriage of crown surface and spectral information will substantially benefit tree delineation; and crownsurface can be used to produce orthographic tree map for a GIS.The available literature body on crown surface reconstruction is ratherlimited. Laser range detection (Lidar), radar interferometry, and photogrammetry are three major techniques for surface reconstruction .Airborne laser scanning systems detect range using laser signals in the visible or near infrared wavelengths. To measure crown surfaceheights, a lidar records multiple echoes: the first one is supposed to be reflected from a crown top while the last from the ground afterpenetrating through the canopy. Crown surface heights can be derived by subtracting the first range reading from the last one. Thepenetration ability of a laser signal is critical to the quality of surface reconstruction. It was found that with near vertical incident anglesof laser systems, 20-40% penetration rates can be expected through European type of coniferous and deciduous forests. In particular,the penetration capability of laser signals through dense forests remains questionable (Ackermann, 1999). The requirements of highsampling rate, high signal-to-noise rate, and multiple-echo recording capability for crown surface measurement would make the systemquite expensive.

Radar interferometry detects elevation by phase correlation of radar echoes received by two radar antennas or by the sameantenna at two fixed locations. The height can be acquired at an accuracy of centimeters. Space-borne interferometric radar (around 30meter resolution) has been used in earthquake monitoring, glacier movement monitoring, and digital elevation model (DEM) generation.The only experiment found on tree crown characterization was the use of 1.5 m resolution airborne interferometric SAR data at X- andC-band over tropical rain forests in Indonesia (Hoekman and Varekamp, 1998; Varekamp and Hoekman, 1998). The results show that theinterferometrically derived tree height can dramatically underestimate tree height.

Stereoscopic surface reconstruction from digital images is an important field of computer vision and photogrammetry. Bothcomputer vision and photogrammetry use image matching method to reconstruct surfaces. Photogrammetry is commonly practiced asthe standard approach in terrain surface generation. However, photogrammetric efforts to generate crown surfaces are rare. Instead,trees are usually treated as undesired “disturbance” to be eliminated in photogrammetric operations. Gong et al. (2000) tested the useof digital photogrammetry for oak wood land monitoring. They suggest that a digital surface models that contain the elevation oflandscape features such as buildings and tree canopies instead of the commonly used digital elevation models that only describe

Model based conifer crown surface reconstruction from highresolution aerial images. Photogr相关文档