Improvement of the Geospatial Accuracy of Mobile Terrestrial LiDAR Data
Leslar, Michael David
MetadataShow full item record
Many applications, such as topographic surveying for transportation engineering, have specific high accuracy requirements which MTL may be able to achieve under specific circumstances. Since high rate, immersive (360 FOV), MTL is a relatively new device for the collection and extraction of survey data; the understanding and correction of errors within such systems is under researched. Therefore, the goal of the work presented here is to quantify the geospatial accuracy of MTL data and improve the quality of MTL data products. Quantification of the geospatial accuracy of MTL systems was accomplished through the use of residual analysis, error propagation and conditional variance analysis. Real data from two MTL systems was analyzed using these methods and it was found that the actual errors exceeded the manufacturers estimates of system accuracy by over 10mm. Conditional variance analysis on these systems has shown that the contribution by the interactions among the measured parameters to the variances of the points in MTL point clouds is insignificant. The sizes of the variances for the measurements used to produce a point are the primary sources of error in the output point cloud. Improvement of the geospatial accuracy of MTL data products was accomplished by developing methods for the simultaneous multi-sensor calibration of the systems boresight angles and lever arm offsets, zero error calibration, temperature correction, and both spatial and temporal outlier detection. Evaluation of the effectiveness of these techniques was accomplished through the use of two test cases, employing real MTL data. Test case 1 showed that the residuals between a control field and the MTL point cloud were reduced by 4.4cm for points located on both horizontal and vertical target surfaces. Similarly, test case 2 showed a reduction in the residuals between control points and MTL data of 2~3cm on horizontal surfaces and 1~2cm on vertical surfaces. The most accurate point cloud produced through the use of these calibration and filtering techniques occurred in test case 1 (27mm 26mm). This result is still not accurate enough for certain high accuracy applications such as topographic surveying for transportation engineering (20mm 10mm).