Laser triangulation method is not applicable in all possible areas, due to the presence of numerous disturbances.
The method of predictive segmentation of laser line resistant to interference, which allows use in industrial
quality control of the glossy elements, has been proposed. It requires precise synchronization with the CAD
model of the scanned item. Very high accuracy of positioning is required to enable predictive segmentation of
laser prole. -The worse positioning, the widest condential intervals, and worse algorithm's ability to reject
The most universal way to synchronize the model of a part under laser scanning, with its virtual model,
is to use an extra camera and some 3D matching software. This solution, however, is rather of poor accuracy
-even if a high quality optics is used. Kalman ltering is used to reduce the deviation of results. To avoid
problems in description of such complicated system, a coherent and homogeneous mathematical description has
been proposed. It uses so-called Denavit-Hartenberg notation, widely used in robotics.
Several tests has been carried out to verify the 3D matching algorithm for object-model synchronization during
laser scanning. The results show clearly that the accuracy of matching can be improved by using Kalman ltering
-even up to 10 times. Thanks to it, its accuracy increases to face the requirements of predictive segmentation
Accuracy and robustness of 3D inspection methods, which are based on structural lightening, strongly depend on surface
reflectance and its geometry. If the intensity of light profile undergoes disturbances or the image is interfered with
reflection artefacts, then the algorithms for height determination fail.
The paper presents a method for predictive light profile line segmentation which employs information from 3D model of
inspected object. Besides calculation of profile position, based on confidence intervals, the advantage of surface normal
pitch and surface scattering are utilized. The necessary information is calculated based on STL object model and surface
description extension. The developed algorithm includes error model of camera and laser optics, what corrects distortion
The developed solution increases accuracy and robustness of machine vision quality inspection systems especially based
on laser triangulation.