Optics - Q-engineering
Q-engineering
Q-engineering
Go to content
images/empty-GT_imagea-1-.png
images/RatcliffXS.jpg

We do optics

Every vision project has optics, even if it's only the camera lens and the lighting.
When off the shelf solutions are not available, let us think together with you.

Why bother about optics?

 
Because you need it. Every computer vision project uses a lens and some kind of lighting.
When purchasing the camera, it is often supplied with a lens. Or the supplier gives you a choice between different types. On the internet, you will find excellent sites about lenses. How to calculate the focus, which aperture is needed, which mounting, etc. Or ask your supplier for support.

Just a few words about the Modulation Transfer Function, the MTF of a lens, to prevent a common disappointment. In a nutshell, the MFT determines how detailed your image will be. In practice, light cannot easily be forced through a small opening. Sharp contrast-rich transitions are weakened by a small aperture. The MFT gives a figure to this phenomenon. The higher the number in cycles per mm, the better the lens.
MFT lens

Look at the detail of the felt of the π register on our maths page and compare this with the embedded sculpture. The latter has much fewer details, while the image was taken with many more pixels than the photo of the register. This is only caused by the MFT of the lenses used.
It can be said that the behavior of the MFT is similar to a low pass filter in audio. That is absolutely correct. The same Fourier math can also be applied to the MFT. So high-resolution cameras without associated lenses offer no real improvement.

Some cases.

Tessar lens

In this optical design, the back focus distance is 32.4 mm, so no standard lenses are available. A custom designed lens is the only solution. The well-known Tessar lens design has been adapted to meet all requirements shown in the first slide. The first element is a plano-convex Ø 40.0 mm, f = +80 mm. The second lens is split into two symmetric-concave elements. One is focal is -30 mm and the other focal is -20 mm. This results in one virtual lens with a focal length of -14.7 mm. The last lens is also composed of two separate achromats, both Ø 18.0 mm. The first is f = +50 mm, the second f = +40 mm resulting in f = + 26.6 mm. By reversing the first achromat, aberrations are reduced substantially. Remarkable is also the high MTF value of 680.
  

gallery/System_thumb.png gallery/L_thumb_yku09vt7.png gallery/Field_thumb.png gallery/Chroma-BFL_thumb.png gallery/Longitudinal_thumb.png gallery/MTF_thumb.png gallery/Lens_thumb.png gallery/PCB_thumb_sog99xm2.png

Plastic lens

When capturing high-speed images, a short shutter timing is needed to prevent motion blur. The shorter the exposure time, the more light is needed to illuminate the object. Shutter times as small as 15 μSec can be a serious problem.
High power white LEDs are a cheap and widely used light source. Due to there characteristic Lambertian emission patron a large portion of the light is lost. In this case, only 3.5% of all light falls on the object. The most efficient way to increase this amount was using specially designed optics. A normal lens generates strange, non-uniform patterns. The software has been written to model a PMMA body defined by B-splines. After many iterations, a boot-shaped lens turned out to give the best results. Now 33% of the light is distributed homogeneously over the object. Not bad for just a few cents of plastic. With the lenses and a peak current forward pulse of 4 Amp, eight LEDs now pour 450 Klux on the object at 71mm distance for the duration of 15 µSec. The used Windows app can be downloaded here.
gallery/Motion-blur_thumb.png gallery/Lambertian_thumb_fe0cku5r.png gallery/Drawing_thumb.png gallery/T_thumb.png gallery/L_thumb.png gallery/IMG_4470_thumb.png gallery/PCB2_thumb.png gallery/Profiel_thumb.png

Document scanner

Many documents, such as passports, have special features that are only visible with infrared light. That is why the document scanner has two cameras, one for visual light, the other for IR. The biggest challenge was a homogeneous illumination. A straight row of LEDs only gives a gigantic bubble of light in the middle and leaves the corners untouched. An evolutionary algorithm has been used to determine the correct positions of the LEDs. The original population consists of 24 sets of exposure PCBs. Each board has 16 white and 16 IR LEDs. The algorithm could freely choose the location of the LEDs. Lens vignette was also taken into account. The result is shown in slides 5 and 7 for the white light, slides 6 and 8 give the IR profile.
gallery/Euro_thumb.png gallery/Scanner_thumb.png gallery/Board_thumb.png gallery/Overview_thumb.png gallery/LuxOut_thumb.png gallery/IROut_thumb.png gallery/LuxOut_Lens_thumb.png gallery/IROut_Lens_thumb.png
Raspberry and alt
Info
Raspberry Pi 4
Back to content