Geometric accuracy evaluation of mobile terrestrial LIDAR surveys with supporting algorithms
Mobile Mapping System (MMS) technology is widely used for many applications, hence quantifying its accuracy is a very important and essential task and is a primary focus of this research. In general, to perfrom geometric accuracy evaluation of MMS data, validation points/features are needed. A method is needed to capture a point feature off the roadway in a position where a target on the ground surface would not be visible to the scanner. In this study, eight sphere targets with 14" diameter were placed on the shoulder of the roadway over validation points on the ground. The sphere targets were constructed from injection molded spherical light fixtures. Through a calibration process, they were verified as consistent in size and shape at the 1 mm level. The targets were scanned by four different MMSs (two of design grade and two of asset grade) on two established Test Sites representing different roadway environments (highway and urban settings). Two different selectable data rates (250 KHz and 500 KHz) were also exercised in the data collection as well as two different vehicle driving techniques for data collection (with and without acceleration while the vehicle is turning). Absolute and relative accuracy of the dataset obtained from MMS are of interest. All of these characteristics and factors have been geometrically evaluated through the developed procedures. An automatic sphere target detection/estimation algorithm was developed to detect and extract the scanned sphere target points by eliminating most of the adjacent non-sphere points via a 3D Hough transform process. Following this, the sphere center is robustly located through estimation via L1-norm minimization which allows outliers (ex. tribrach points) to be detected and automatically eliminated. Subsequently the final sphere target center is estimated through least squares. This procedure is robust to several sources of non-random noise. Through error propagation, the precision of the center point estimation is SE90 = 0.20 cm (radius for spherical error, 90%). The case of disturbed targets was able to be detected with the results from this algorithm as well. Although such geometric targets have been widely used in static laser scanning, their use in Mobile Mapping has not been thoroughly studied. Another contribution from this research is that L1-estimation has been applied to all methods of forming condition equations. Those are indirect observations (line fitting), observations only (level network), and mixed model (dependent relative orientation of stereo pair images) problems. Existing published work has exclusively been applied to the indirect observations form of condition equation representation. In this test, outliers which were intentionally added to observations of all the problems were correctly detected. Additionally, L1-estimation was implemented to each of the problems by two different approaches: 1) by using a linear programming approach solved by the simplex method, 2) by a brute force method (exhaustive search for all possible sets of solutions). Results from both approaches are identical. This has verified the idea that the linear programming approach can be used as a convenient tool for implementing L1-estimation for all methods of forming the condition equations.
Bethel, Purdue University.
Off-Campus Purdue Users:
To access this dissertation, please log in to our