Mohammadivojdan et al.: Preprocessing and Analysis Strategies .. 
discrete point cloud is a complex task both from computational as well as math- 
ematical aspects. 
Overall preprocessing and analysis of these measurements are usually per- 
formed manually by experts. This process is expensive and time-consuming. The 
aim of this project is to define a pipeline approach that not only improves the 
current state of affairs, but also significantly reduces the dependence on manual 
processing. 
2 Methodologies 
The proposed pipeline consists of different stages: preprocessing, modelling and 
evaluation. A brief overview of the current status of the mentioned stages are 
given in the following. 
2.1 Preprocessing and Data Cleaning 
As described, the data includes measurement errors and outliers. Therefore, in 
modelling the data, it is essential to count for any possible errors in them. Oth- 
erwise, the final model is distorted. The effect of measurement noise could be 
considered in the surface modelling technique. The most critical aspect is deal- 
ing with the outliers before modelling in a pre-processing step. The main part 
of this step is usually handled manually or has at least significant manual steps 
after an initial automatic approach (Lorenz et al. 2021). To reduce the manual 
effort, a data-adaptive algorithm, with a density-based foundation is proposed 
to process the data and identify its outliers. The anomalies are detected based 
on the deviation of the data points to a fitted model in a hierarchical approach 
(Mohammadivojdan et al. 2021). The results of implementing this method show 
more than 70 % agreement to the manually detected outliers. 
2.2 Surface Modelling 
Considering the characteristics of a data set and its challenges, a method should 
be chosen that is able to not only handle large data sets and data gaps but also 
is computationally efficient. The method of Multilevel B-splines Approximation 
(MBA) is adapted to model such a data set (Lee et al. 1997). 
The results of implementing MBA to approximate a sample data set related to 
a section of Kiel Canal are shown in Fig. 1. This data set is captured by multibeam 
echo sounder surveying system. The approximated model is color-coded based 
on the deviations to the real measurements. The mean of the deviations of the 
approximated model to the observed point cloud is close to zero (0.4 mm). This 
DVW-SCHRIFTENREIHE # Band 102/202, 
39 Mißner-Verla: