Automated control of the size and shape of products can reduce rejects, increase labor productivity, reduce raw material consumption and logistics services costs.
Control of geometrical parameters is probably the most demanded process at industrial enterprises. Vision systems allow to measure and control various objects in automatic mode and without contact at almost any speed.
Unfortunately, there is no universal vision system, equipment and algorithms for control of different types of products. Vision systems based on video analytics and neural networks work well with images with good contrast and the presence of characteristic features. On the contrary, objects with holes, edges, stepped corners, zones of sharp changes of surface reflection coefficient are poorly controlled by 3D laser scanners, because the reflected probing beam of the device is highly distorted, which leads to large errors.
Therefore, the object to be measured must be carefully analyzed before selecting a measurement and control method.
The accuracy of the vision system is influenced by systematic and random measurement errors. Systematic errors are removed by the software developers for a particular type of product with mathematical processing. Accuracy can thus be increased many times over. Special attention should be paid to this.
- control of object contours and edges – search for deviations of object contours, borders and edges in the image.
- measurement of geometric parameters – control of geometric dimensions of products, linear dimensions (length, height, thickness), parallelism, coaxiality, roundness, angles, comparison with the standard and checking tolerances.
- inspection of complex shape – comparison of 3D scan with a reference CAD model, identification of holes, steps, angles, etc.
Object contour and edge inspection
Contour and edge inspection is an effective method for determining object geometry. As a rule, two methods of video analytics are applied – one based on light reflection and one based on backlighting (“shadow” method). The shadow method is almost always more accurate, but in many applications it is not applicable due to the inability to position the illumination behind the object being measured. The contours and edges of the object can be imaged with very high accuracy – better than 1 µm due to the use of the telecentric optics, because of which diffraction effects at the edges are avoided, there is no “perspective” effect and the sensor is not affected by oblique rays (only those running parallel to the optical axis of the telecentric objective).
Measurement of geometrical parameters
Control of geometric dimensions of products, linear dimensions (length, height, thickness), parallelism, coaxiality, roundness, angles, comparison with the benchmark and tolerance check using video analytics and 3D scanners. As a rule, multi-sensor technology is used for complex objects; for edges, steps, holes, slots – video analytics with back and front illumination; for linear dimensions and smooth shapes – laser scanners with illumination of “point” and “line” types.
Inspection of complex shapes
It is advisable to digitize products with a complex shape as a three-dimensional point cloud, obtained with high accuracy and resolution. If scans from different angles are required, the algorithm stitches everything into a single model. Special computational geometry algorithms are used to overlay the resulting point cloud with the reference CAD model. The algorithm can recognize different parts of the part for more accurate processing. The algorithm then calculates tolerance deviations in real time. The part that does not pass the check is removed from the conveyor.