Novel method for sphere target detection and center estimation from mobile terrestrial laser scanner data

• Published in: Measurement : journal of the International Measurement Confederation 2019-04, Vol.137, p.617-623
• Main Author: Liu, W.I.
• Format: Article
• Language: English
• Subjects: Adaptive dynamic random sample consensus; Adaptive sampling; Algorithms; Center estimation; Coordinates; Data processing; Lasers; Mapping; Mobile terrestrial laser scanner; Optimization; Scanners; Sphere target detection; Surveying; Target detection
• ISSN: 0263-2241; 1873-412X
• DOI: 10.1016/j.measurement.2019.02.025

•Adaptive dynamic random sample consensus (AD-RANSAC) proposed for detect target.•Nonlinear least squares (NLS) proposed for estimate the center of sphere.•Fusion method of AD-RANSAC and NLS proposed for detect and estimate sphere target. Sphere targets have been widely used in mobile terrestrial laser scanner (MTLS) registration, using their center coordinates the multiple viewpoints clouds data can be transformed into a unified coordinate system. However, in practice, the task of sphere target detection and center estimation is still a labor intensive and time consuming for large surveying and mapping projects. For this reason, a novel method of fusion adaptive dynamic random sample consensus (AD-RANSAC) and nonlinear least squares (NLS) is proposed to automatically detect and optimally estimate center of sphere target. The basic procedure of the proposed method is that, firstly, a newly AD-RANSAC algorithm is developed to automatically and rapidly detect the sphere targets from MTLS point clouds data, which extends the traditional RANSAC algorithm to detect the sphere targets in dynamic states and recursively update the state estimations using sequential measurements, also improves the accuracy and speed of detection of sphere target. Secondly, the sphere target center coordinate is determined through the NLS optimal estimation algorithm, which has regions mathematical calculation and minimizes the errors in estimation of sphere targets center. Finally, twenty pre-surveyed sphere targets distributed in indoor field were conducted to check the validity of proposed method, experimental results are presented that validate this method and demonstrate our proposed method can automatically detect the sphere targets in more than 1 million points data within 1 min, the accuracy of the sphere center coordinate estimation is less than 2 cm.