Recently, a German manufacturer of tyre-mounting systems asked a project-management company that focuses on developing industrial automation systems, to develop a prototype system. This system is required to automatically capture and analyse alphanumeric information (such as the name of the manufacturer, type of tyre and date codes) engraved on a tyre by multiple mould inserts during manufacture.
In the manufacturing process, moulding systems create distinct markings on a designated position of the tyre sidewall so that tyre makers and their customers can trace the individual product years after the date of manufacture. Because this information is so important, it is imperative that each mark can be clearly read during both automotive manufacturing and later in the after-sales market. To do this, an imaging system must clearly illuminate the tyre rim to increase the contrast between the raised or engraved serial numbers and the base of the sidewall. Unfortunately, imaging these characters using standard light sources is very difficult, since the black characters on the tyres' black background absorb most of the light. Achieving good contrast to read the characters is almost impossible.
To overcome this problem, a prototype system was developed that uses a laser line projector, a C3-1280 CMOS 3D imager from Automation Technology, a CameraLink frame grabber, and PC-based software. Individual tyres are mounted on a rotating platform, on which is mounted a laser line generator and the camera.
The C3 camera acquires height profiles and height images by means of a laser triangulation. The camera, which incorporates a high-resolution image sensor and a powerful processor, is capable of acquiring and processing up to 28 000 frames/second. The image processor detects a laser line projected onto the tyre and calculates a single height profile from each frame. By scanning the laser line over the tyre as it rotates, a complete height image of the object is acquired. Rotational position information from a shaft encoder is acquired by the camera, which features an RS422-based shaft encoder input, with tick counter and direction evaluation, so that no external hardware is required.
Because tyres can vary in dimension, the system has to illuminate a laser line that can vary in length from 150 to 250 mm across the width of the tyre. A uniform light intensity is essential, and a laser line projector that projects a line of uniform intensity along its length is used (non-Gaussian intensity distribution).
To compute the required data, raw 3D data is output from the camera over a Camera Link interface and captured using a PC-CameraLink PCI frame grabber. The raw image data consist of x, y, and z positional information and an 8-bit intensity level for each point on the scan. As data is acquired, the information is reformatted as a height profile for each line that has been scanned.
The C3-1280-CL CMOS camera is based on a high speed 1280 x 1024 x 10-bit CMOS imager. It has been especially designed to target the 3D structured light imaging market. With a Camera Link interface and opto-coupled trigger inputs, the camera's on-board processor has been especially optimised to perform up to 28 800 profiles per second with 36 million 3D data points per second.
"What makes the camera unique are the parallel processors that have been incorporated to process image data at this extremely high data rate," explains André Kasper, managing director of Automation Technology. "In any laser-based imaging system, the height values of the object under inspection are determined by computing values determined by the focal length of the lens, the distance of the object from the camera, and the triangulation angle between the laser and camera. Because the cross section of the line of reflected light is of Gaussian shape," he says, "computing the actual position can be performed in several different ways." Perhaps one of the simplest of these methods is to detect the peak (highest intensity) pixel across the laser line, which results in a pixel-accurate value. Alternatively, a threshold value is determined and the threshold for the two points below which the Gaussian falls is computed. Averaging these two points provides an estimate of the centre point of the Gaussian and results in subpixel-accurate measurement. Other methods can be used to provide greater accuracy. "By sampling multiple points along the laser line and determining the centre of gravity of the Gaussian curve," says Kasper, "the most highly accurate centre point is determined."
From the user's perspective, all three methods of computing the position value can be obtained by configuring the camera using Automation Technology's C3Lib API with source code.
Westplex is the South African distributor of Automation Technology's C3 cameras, Cemarligne non-Gaussian line laser projectors, Matrox frame grabbers and image analysis software, and is a value-added re-seller of National Instruments hardware and software. Westplex specialises in the development of turnkey machine vision systems.
© Technews Publishing (Pty) Ltd | All Rights Reserved
printer friendly version