Comparison of outlier detection at the edges of point clouds using statistical approach and fuzzy methodology: ground-based laser scanner field experiment and randomly simulated point cloud
The random error is following the features of normal distribution function (NDF) which those random errors deviated from the NDF's characteristics can be considered as outliers. In fact, the outliers exist inevitably in any observed parameter that is an undesirable part of the measurement's procedure due to its negative influence on the sensitivity analysis. It is therefore necessary to investigate more efficient methodologies especially for current remote sensing data processing and assimilations. In this paper, the comparisons of Baarda method as the conventional statistical methodology with the Fuzzy approach are presented to detect the outliers at the edges of two data groups: 1. The point cloud of ground-based laser scanner field experiment from one side of a wall, and 2. A group of randomly simulated distributed 3D point cloud. The results show that the Baarda method eliminates the outliers as soon as they are being found while the Fuzzy approach works critically based on the outputs of the statistical tests. Thus, the Fuzzy approach deals mostly with the residuals and those observed errors in the adjustment computational procedures. The obtained results about the successfulness rate of outlier detection for each method are separately presented in both graphical and statistical overview. Also, the capabilities of Fuzzy approach to detect the outliers in different point cloud's size and numbers of existing outliers at the edges of point cloud are investigated and discussed in details.
This work is licensed under a Creative Commons Attribution 4.0 International License.