The emergence of high-speed 3D image processing technologies is a major milestone for the vision industry. 3D measurement methods have been around for a while, but were not initially used in industrial vision applications due to the computational loads involved and the difficulties associated with calibration and set-up. However, not only has processing power vastly improved in recent years, but laser and camera systems are available at lower cost and provide higher accuracy, meaning that 3D imaging is playing an increasingly important role in processing and quality control applications for defect detection, measuring volumes, or measuring elevation dimensions in a host of industries.
Traditional 3D measurement systems comprise a structured laser source, camera and optics to accommodate the laser triangulation measurement technique. By moving the product through the laser line, the camera can record profile images of the object, with image data being transferred to a host PC for 3D analysis by sophisticated 3D processing software.
The emergence of a new generation of 3D ‘smart cameras’, analogous to the smart cameras well-established in traditional vision applications, is set to revolutionise industrial 3D measurements. 3D smart cameras such as the Gocator series from LMI Technologies not only feature an integrated laser source and optics (Figure 1) but provide on-board processing of 3D data without the need for an external PC, allowing the direct output of the measured results. A major benefit of these cameras is the reduction in the complexity of user calibration. They are factory pre-calibrated to simplify setup and measurement, meaning that only a simple calibration adjustment is required on installation. In addition, it is possible to link multiple cameras together to make extended measurements with straightforward calibration.
Accurate calibration, or the conversion of pixel values into real-world measurement units, is an essential requirement for any industrial vision system used for making measurements. Because we are dealing with an optical system, any calibration process must allow for perspective, lens distortion and temperature effects.
Calibration routines for 3D measurement systems involve moving a known object accurately through the field of view using a positioning stage. From this, the system can build a lookup table for converting xyz pixel values to real-world coordinates. This type of equipment is expensive and must be used on the actual production line for traditional 3D systems utilising independent cameras and laser sources. In the smart camera configuration, however, the integral laser, camera and lens are housed in the same assembly. This means that the calibration using the precision stage can be carried out at the factory before delivery to the user to provide consistent, reliable measurements in real-world coordinates, even in applications where temperature variation normally would introduce measurement errors.
In the smart camera, 3D calibration must allow for factors such as triangulation non-linearity, distortion due to imaging lens errors (Figure 2), focus errors (Figure 3), triangulation angle errors and temperature effects. These can all be addressed through a comprehensive factory pre-calibration process. For 3D sensors, thermally induced movement leading to measurement errors is caused by several factors.
The projected laser spot or line position can shift with temperature due to changes in mechanical movement of the optical path or a shift in laser frequency. By using a proprietary laser and optics package, excellent pointing stability of the laser beam can be achieved. Mounting all optical components on a solid mounting frame or spine CNC machined from a solid aluminum block, keeps components in the required relative positions. In addition, as part of the manufacturing process, sensors are placed in an environmental chamber, where temperature can be varied through a programmed cycle, and thermal compensation factors are generated for each individual sensor. This compensation provides a dramatic improvement in measurement performance.
The final stage of factory pre-calibration is to calibrate each unit against precision targets moving on precision slides (Figure 4). This process sets the camera up for measurements within a measurement range indicated by the red area in Figure 5. The size of field of view (and therefore the active measurement area) is a function of the camera model selected, but can range from 14 mm to 1260 mm with measuring ranges from 25 mm to 800 mm with a linearity of 0.02% of the range.
Although Gocator sensors are pre-calibrated and ready to deliver profiles out of the box, a user calibration is required to compensate for sensor mounting inaccuracies and to determine the resolution and speed of the transport system by conducting an ‘encoder ticks to mm’ transformation. User calibration is also required when multiple sensors are linked, to align these sensors into a common coordinate system. The user can perform either an alignment calibration or a travel calibration. Travel calibration performs essentially the same role as alignment calibration, but calibrates encoder resolution and y-axis offsets in addition to the corrections provided by alignment calibration. Alignment calibration can be used to compensate for mounting inaccuracies by aligning sensor data to a common reference surface (often a conveyor belt). A calibration bar (featuring holes of a known size and position) can also be used.
When using the conveyor belt surface, any irregular objects that might interfere with alignment calibration are removed from the sensor's field of view. A single calibration button automatically performs the calibration. Travel calibration can be used to achieve alignment calibration and motion calibration in a single procedure that uses a fixed size calibration disk (40 mm or 100 mm diameter) that travels through the field of view at a constant speed on a conveyor. When the calibration target has passed completely through the laser plane, the calibration process will complete automatically. To properly calibrate the travel speed, the transport system must be running at the production operating speed before the target passes through the laser plane. Once calibration has been completed, the values derived will be saved automatically and reloaded each time the sensor is reset or powered up.
Multiple sensor calibration
As mentioned above, Gocator can link multiple sensors to increase the field of view. One of the more common arrangements features two sensors and Gocator automatically recognizes a second sensor called a “Buddy”. The dual sensor mode seamlessly combines profile data from both Main and Buddy sensors as if they were one. Figure 6 shows the most popular configurations for dual sensors: ‘wide orientation’ where a Main and Buddy are mounted side by side to measure objects that are wider than a single sensor's field of view and ‘opposite direction’ where the Main and Buddy perform top and bottom differential measurements to calculate true thickness when the object cannot be referenced to a known surface such as a conveyor. Alignment calibration or travel calibration can be used to establish a common coordinate system for the Main and Buddy sensors. Calibration determines the adjustments to X, Z, and Tilt (rotation in the X–Z plane) required to align the data from each sensor.
System coordinates are aligned such that the system x-axis is parallel to the calibration target surface and the system Z-origin is set to the base of the calibration target object. For Wide and Opposite layouts, profiles and measurements from the Main and Buddy sensors are expressed in a unified coordinate system. For wide, multi-sensor systems, a calibration bar of the design described above must be used which matches the length of the system.l