Processing 3D Data from Laser Sensor into Visual Content Using Pattern Recognition
Jaromír Klarák, Ivan Kuric, Miroslav Císar, Ján Stanček, Adrián Hajdučík, Karol Tucki · 2021
This paper is focused on processing data from a laser 3D scanner working on the triangulation principle. Generated point cloud processed by an algorithm in a programming language python using mainly OpenCV library to generate 2D data - images. This type of generated image achieves better pixel stability compared to images captured by camera vision, which are highly dependent on lighting conditions. At present, mainly inspection systems using camera vision, which are used to capture images being analyzed. This paper describes processing scanning surface of an object performed at an experimental inspection stand. The object of interest is an automobile's tire-like object. The main goal is to develop an inspection system for tire manufacturing. The first Part-Task is developing a matching system for combination more source data and merges them. Generation template data and compare them with inspected scanned data by the same methods. This process is performed by pattern recognition, where the possible methods applicable to these specific data types are described. The result of this work is the preparation of the 3D and 2D data and their analysis for further experiments in the inspection system. Especially, for the analysis of specific principles to identify abnormalities in the geometry of the scanned surface are used. The final system that will be used in mass production will be a compromise between conditions such as the price of hardware, the time to evaluate the possibility of finding abnormalities and analyze them.