Triangulation based optical measuring systems are an important element of industrial quality assurance. Due to their robustness and cost-effectiveness Laser Light Section Sensors have become a widespread solution for Geometry measurements. In order to reconstruct the scene, it is necessary to identify the corresponding laser line, which is distorted due to the geometrical properties of the specimen, in the camera image. In Order to achieve the highest precision possible, the line segmentation has to be performed at sub-pixel accuracy. Furthermore, in an industrial environment, interfering light sources may be present. A distinction between these influences and the laser light ensures a robust measurement. The projected Laser Line of a triangulation sensor is usually formed by a Powell lens from a point source, which results in a uniformly distributed intensity. Another option to achieve highly uniform intensity distributions is by means of a lenticular lens. A side effect of these optics is that the fine-structure of the projected line is formed by a chain of equidistant dots. Across the laser line the intensity distribution can be considered as a Gaussian profile. Challenges to the segmentation are from the fine, dotted structure of the line. Although conventional methods, such as centroid based algorithms can be applied, with the drawback of imprecise peak detection. In order to insure both segmentation accuracy und robustness, this paper introduces a novel segmentation method based on wavelet-transformation. In a first step the periodic fine-structure of the line is utilized for a definite identification of the line with distinction from scattered light. In a second step a gaussian wavelet is used to achieve sub-pixel accuracy in peak detection. The developed method is compared to conventional peak detection methods.
|