Orthogonal Msplit Estimation for Consequence Disaster Analysis

Nowadays, the data processing used for analyzing multifaceted disasters is based on technologies of mass observation acquisition. Terrestrial laser scanning is one of those technologies and enables the quick, non-invasive acquisition of information about an object after a disaster. This manuscript presents an improvement in the approach to the reconstruction and modeling of objects, based on data obtained by terrestrial laser scanning presented by the authors in previous work, as a method for the detection and dimensioning of the displacement of adjacent planes. The original Msplit estimation implemented in previous research papers has a specific limitation: the functional model must be selected very carefully in terms of the mathematical description of the estimated model and its data structure. As a result, using Msplit estimation on data from laser scanners is not a universal approach. The solution to this problem is the orthogonal Msplit estimation method proposed by the authors. The authors propose a new solution: the orthogonal Msplit estimation (OMsplit). The authors propose a modification of the existing method using orthogonal regression and the Nelder-Mead function as the minimization function. The implementation of orthogonal regression facilitates the avoidance of misfitting in cases of unfavorable data acquisition because the corrections are calculated perpendicularly to the estimated plane. The Nelder-Mead method was introduced to the orthogonal Msplit estimation due to it being more robust to the local minimum of the objective function than the LS method. To present the results, the authors simulated the data measurement of a retaining wall that was damaged after a disaster (violent storm) using a terrestrial laser scanner and their own software. The conducted research confirmed that the OMsplit estimation can be successfully used in the two-plane detection of terrestrial laser scanning data. It allows one to conduct the correct separation of the data set into two sets and to match the planes to the appropriate data set.