Today in Tech Talk: Laser triangulation! This technique is used to measure 3D geometry, we use it in our val-IT Smart. One or multiple laser lines follow the surface of an object to gain information about the positions of its shape. For this to work either the 3D scanner or the object must move in respect to each other, and often an encoder is used to detect small jitter in motion, as well as acceleration or deacceleration.
3D scanner configurations are often mounted on a multiple axis the robot, but we don’t recommend this. As servo’s jitter around a position they are not really suitable for linear motion.
Each laser profile captured by the camera represents a 3D slice of the scanned object. All slices combined form a complete 3D model of the object. This means that the more images captured, the higher the resolution of the object in transport direction is. So that’s why we often use a camera with a highspeed sensor that takes high resolution images with thousands of frames per second. As we need to capture several laser profiles in one single capture, the sensor often supports multiple readouts to increase the framerate and reduce the number of pixels that need to be processed.
There are different ways of laser triangulation possible:
Depending on the (reflective) object properties, required accuracy and acceptable occlusion, one of the above configurations must be picked. Table 1 shows the different pros and cons in respect to the triangulation geometry.
The high intensity laser can be problematic for the sensor. That’s why we use sensors with a High Dynamic Range, low Signal To Noise Ratio and special dynamic functions to avoid overexposure. Although CMOS sensor cells are big with a relative small fill factor, they have the advantage to be pixel controlled. So an important feature is that you can control the slope of the video signal with help of one or multiple knee points.
The used laser-projector (and optics) plays an important role in both the X,Y resolution and Z accuracy. If the line has some distortions (which is very common with cheaper projectors) it is pretty difficult to make corrections afterwards. The line is often curved or S-shaped, and is often caused when the incident laser beam enters the beam shaping optics at an abnormal angle of incidence. Most lasers behave Gaussian, meaning that the center of the laser line has a much higher intensity compared to the edges. If you want to detect defects, you want the laser line to be thinner than the measured defect. Fine lasers are pretty expensive and have limited depth range and field of view.
It is important that the lasers can handle vibrations and temperature differences as every difference of the laser line during and after calibration has a great impact on your precision.
Working with lasers in an application that demands high accuracy measurements, always remains a big challenge. Indirect reflections often have a different phase, so polaroid filters can be used to filter those out. As lasers are monochromatic, you could use narrow bandpass filters to filter out all background illumination, which is also helpful when working with multiple lasers in your application. A laser has random speckles that has influence on the accuracy of your application. This can be reduced by setting the camera lens a little bit out of focus, so that it works as a low pass filter.
Texture and color variations of the scanned object have a great impact on the precision of your measurement. The choice of the wavelength of your laser is highly dependent on these properties. Reflecting material is still quite a challenge. Specular reflections can be detected by using lasers from different angles. If one of the lines diverge you at least know that the measured depth position is less reliable.