Department of Civil Engineering
IIT Kanpur, India
This paper describes the development of a LiDAR simulator aimed at replicating the process of LiDAR data collection using mathematical models. This simulator will be useful in better understanding of LiDAR data collection process and errors. The data produced by simulator can be used to evaluate and develop information extraction algorithms
Laser altimetry is a direct method to efficiently map the earth’s surface in three spatial dimensions. Topographic mapping with LiDAR seems on the verge of supplanting traditional photogrammetry in the near future for detailed terrain mapping. This technique has a large and growing number of actual and anticipated applications in disciplines ranging from urban analysis to natural resources.
Initially the focus of research in this field was on how to obtain terrain geography and objects such as buildings, trees etc. Several other researchers were interested in identifying the errors involved in LiDAR data. In all these studies the raw LiDAR data were employed and attempts were made to compare results with ground truth. For a detailed review of these efforts please refer Lohani (1999) and special LiDAR issue of ISPRS Journal (ISPRS, 1999). However, the literature clearly indicates that a limited success has been achieved in these efforts and a lot needs to be done to make full utilisation of this technology.
The basic drawback of the methods relying only on field data is that, it is not always possible to collect true ground truth for validation of these approaches, besides the difficulty in ground truth collection.
In view of the above it is of utmost importance that research be carried out to develop theoretical equivalents of LiDAR data collection process. Realising this, recently the research related to simulation of LiDAR data has started taking place. The purpose of these researches should be to study the effect of various flight parameters, LiDAR instrument parameters, wrong instrument calibration and environment on LiDAR data generated for different types of objects and terrains being scanned. The researches conducted in this area hitherto, though only a few, are limited in their scope. These have studied the effect of only single parameter on one kind of object, assuming the effect of other parameters either negligible or independent of other parameters as in case of Holmgren et al (2003). However, in real world all parameters affect together and have varying degree of effect on different terrain conditions. Furthermore, considering the data points to be randomly distributed, as in the case of Beinat and Crosilla (2002), does not reflect the way LiDAR collects data.
In view of the above, the aim of present work is to develop a simulator that can replicate LiDAR data collection process with variation of several associated parameters on different types of surfaces.
Methodology for Simulation
The simulator has basically two components: 1) Ground component, and 2) Sensor component. In its simplest form mathematical equations are written to generate a laser vector from a moving platform towards the ground. The equation of this vector is known and is intersected with the equation of the ground surface (also known, as specified). The point of intersection gives X, Y, and Z coordinates which are equivalent to LiDAR data generated (Figure 1).
This component is formed using multiple mathematical surfaces. The bald terrain is represented by a single or multiple surfaces. The surfaces chosen resemble the simple ground surfaces (Figure 2: A, B, and C) and complex ground surfaces (Figure 2: D, E, and F) that vary spatially in terms of their curvature, slope, altitude etc. The equations shown in Figure 2 are used to