Introduction: Convert a 1980s Video Camera Into a Real-Time Polarimetric Imager
This is an entry in the
Trash to Treasure
Polarimetric imaging offers a path to develop game-changing applications across a wide range of fields – spanning all the way from environmental monitoring and medical diagnostics to security and antiterrorism applications. However, the very high cost of commercial polarimetric cameras has hampered research and development on polarimetric imaging. This paper presents detailed instructions for converting a surplus 1980s-era, 3-tube color camera into a real-time polarimetric imager. The camera used as the basis for this conversion is widely available in the surplus market for around $50. This trash-to-treasure Instructable will show you how to convert a camera that is suitable only as a prop into a useful scientific instrument, commercial versions of which would be worth many tens of thousands of dollars.
You’ll need the following items to perform this conversion:
- Working surplus JVC KY-1900 camera (models KY-2000 and KY-2700 seem similar to the KY-1900 and may also be suitable)
- Ø25.4mm wideband 70T/30R beamsplitter (e.g. Thorlabs BSS10)
- Ø25.4mm wideband 50/50 beamsplitter (e.g. Thorlabs BSW10)
- 3D-printed beamsplitter adapter rings
- Sheet of polarizing plastic (e.g. Edmund Optics 86-188)
Step 1: Understanding Polarimetric Imaging
A light wave is characterized by its wavelength, which we perceive as a district color; its amplitude, which we perceive as an intensity level; and the angle at which it oscillates with respect to a reference axis. This last parameter is called the wave’s “Angle of Polarization”, and is a characteristic of light that unaided human eyes cannot distinguish. However, the polarization of light carries interesting information about our visual environment, and some animals are able to perceive it and rely critically on this sense for navigation and survival.
A detailed, and easy-to-understand description of polarimetric imaging and its applications is available in my whitepaper on the DOLPi polarimetric cameras available at:
http://www.diyphysics.com/wp-content/uploads/2015/10/DOLPi_Polarimetric_Camera_D_Prutchi_2015_v5.pdf and its presentation on YouTube at: https://www.youtube.com/watch?v=A7bXkp8SWCA
Step 2: Buying and Aligning the Camera
The KY-1900 was introduced as a professional-grade color camera in the late 70’s. It was one of the few models to be produced with a plastic orange body, making it very distinctive, and a mark of high-end professionalism for camera crews. Back in 1982, this camera retailed for around $9,000.
Today, you should be able to find one in the surplus market for around $50. The KY-1900 was built like a tank, so chances are very good that it will be fully functional if it looks good cosmetically. Just connect it to a NTSC color monitor and supply it with 12VDC (the camera draws around 1.7A).
Before proceeding with the modification, make sure that the camera is in working order and well aligned. Use the instructions shown in Appendix II of the project’s whitepaper to align your camera and check that it works correctly.
Step 3: Accessing the Optical Assembly
The first step in the conversion is to access the camera’s optical assembly, which involves the following steps:
- Take apart the camera’s left cover
- Remove the DF printed circuit board
- Peel-off the plastic isolation sheet that is attached with double-sided tape to the optical assembly’s outer cover plate
Step 4: Opening the Optical Assembly
Pry off the inner optical assembly cover plate. This plate is glued to the assembly. The plate won’t be used again, so don’t worry about distorting it. However, be careful not to damage the optical elements within the assembly.
The bottom pane of the figure shows the optical assembly of the unmodified JVC KY-1900 camera. Incident light through the First Relay Lens is split into three colored images by the dichroic beamsplitters before they are sent to their respective Saticon tubes via Second Relay Lenses. The modification into a real-time polarimetric imager involves exchanging the original dichroic beamsplitters of the Dichroic Beamsplitter Assembly by wideband beamsplitters, eliminating the color trimming filters inside the Second Relay Lenses, and adding polarization analyzers.
Step 5: Removing Dichroic Beamsplitter Assembly
The Beamsplitter Assembly is held with three screws, one from the front and two from the back. As such, the camera’s right-side cover, PCB, and plastic film must be removed to make these accessible.
Step 6: 3D-Printing Beamsplitter Adapter Rings
The dichroic beamsplitters originally used in the KY-1900 camera have a non-standard diameter, so I decided to use 1”-diameter wideband plate beamsplitters for the modification. My friend and colleague Jason Meyers designed and 3D-printed a retainer ring to hold the 1” beamsplitters in place. CAD and 3D-printing files are available at this DropBox.
Step 7: Replacing the Dichroic Beamsplitters by Wideband Beamsplitters
The next step in the conversion process is to replace the dichroic beamsplitters by wideband beamsplitters. The image needs to be more-or-less equally split into three images, so the first beamsplitter needs to reflect around 33.33% of the incident light, while allowing 66.66% of the light to go to a second beamsplitter that should then split this portion evenly. I used the following beamsplitters:
- Ø25.4mm wideband 70T/30R beamsplitter (Thorlabs BSS10)
- Ø25.4mm wideband 50/50 beamsplitter (Thorlabs BSW10)
The wideband beamsplitters within the retainer rings should be installed in the assembly, and the modified Beamsplitter Assembly can then be installed back in place. Temporarily reconnect the circuit boards. Making sure that nothing shorts against the exposed parts of the optical assembly, power-up the camera. Only minor adjustment of the horizontal/vertical potentiometers should be needed to reach alignment if you correctly placed the beamsplitters. You will notice that the image is still in color, albeit a bit washed-out in comparison to the original image. The image still shows up in color because there are very strong filters within the Secondary Relay Lenses that need to be removed.
Step 8: Accessing the Second Relay Lenses
Removing the Second Relay Lenses (that's JVC's name for them) from the optical assembly takes some additional disassembly of the camera. This is because the image pickup tubes must be removed before the Secondary Relay Lenses can be taken out.
Start by taking out and disconnecting the printed boards from the cable assemblies. Then remove the back of the camera. The tube assemblies can then be pulled off the tube housings of the optical assembly, giving access to the Second Relay Lenses.
Step 9: Removing and Disassembling Second Relay Lenses (One at a Time!)
The Second Relay Lenses are held in place by well-hidden,small setscrews accessible from the right side of the optical assembly. Once the setscrew is open, pull out the Second Relay Lens on which you are going to work. Wrap a few layers of thick electrical tape over the two sides of the optical tube and open it using pliers.
Step 10: Removing the Color Filters and Second Relay Lens Reassembly
The color filter should be removed by unscrewing the retainer ring using a spanner wrench or very pointy tweezers. After removing the filter, simply reassemble the lens and finger-tighten.
Eliminating the color filter shifts the focal point of the Secondary Relay Lens, so it shouldn’t be reinserted all the way into the optical assembly. Instead, the modified Secondary Relay Lenses should protrude only about 2.5mm.
The camera can be reassembled after installing and securing with setscrews all of the modified Secondary Relay Lenses. Leave the optical assembly accessible, and only reconnect the DF board temporarily, making sure that it doesn’t short-circuit with the optical assembly.
Step 11: Realigning the Camera
Now is time to align the camera very carefully so that it produces a perfectly black-and-white picture. Some level of color fringing will always be seen because the Secondary Relay Lenses were designed for a narrow band of wavelengths, and are now being used over the full bandwidth of visible light. The fringing is especially noticeable at the edges of the image when the zoom is pulled all the way back, but decent registration can be achieved by patiently following the procedure outlined in Appendix II of the project’s whitepaper.
Step 12: Making Polarization Analyzer Filters
Cut three 1.42”×1.42” squares out of a polarization sheet. I used an Edmund Optics 86-188 150 x 150mm, 0.75mm Thickness, Polarizing Laminated Film. I chose this film instead of cheaper offerings because it features a very high extinction ratio, as well as high transmission, which make for better polarimetric images. Notice in the figure that one of the squares is cut at 45° with respect to the other two.
Step 13: Adding the Polarization Analyzers
Attach the polarization analyzers with clear tape within the optical assembly such that they are placed within the optical paths to the tubes as shown in the figure.
That’s it! The conversion is complete. You can test the camera at this stage before reassembling the optical assembly’s cover (I discarded the inner cover), reattaching the plastic sheet, reconnecting the DF board, and closing the camera’s enclosure.
Step 14: Using the Camera
The figure shows results with a sample target made with pieces of polarizing plastic at angles between 0° and 180° along with a colorbar. The target as captured from the modified JVC KY-1900 camera shows the colorbar and other non-polarized elements of the picture in gray-scale, while the pieces of polarizer film are brightly colored, encoding their angle of polarization in NTSC’s RGB space.
We have a be nice policy.
Please be positive and constructive.
How can you turn individual tubes off and back on at will, to see what different tubes see, and how the final image changes with them?
If you go the digital camera route, in applications where slow capture or with distant objects so parallax isn't much of a problem, how do you embed the different separate pics together?
If you use 3 different polarization positions, i.e. 0, 45 and 90 degrees on each of 3 filters, say R-G-B you get 9 combinations.
How to digitally process all that to get useable info?
Can polarimetric imaging be used to discover ancient, buried buildings, like a cheaper LIDAR alternative?
Agree with you on suitability of 3 ioptically-independent cameras for imaging far-away objects, or if slow capture (or off-line) to give you time for image registration. In fact, I have been working on a setup like that using 5MP cameras, but am still trying to do optical alignment because good digital registration is very computationally-heavy. In my book on UV photography (https://www.amazon.com/gp/product/1682031241/ref=a...) I discussed the Matlab-based “SIFT Flow” registration method that I use for hyperspectral imaging. It warms up the Xeon processor when running it...
Regarding processing to yield usable images, please take a look at the DOLPi whitepaper at:
There I give a very detailed explanation and Python/Matlab code on how to process the three images together to yield useful information. For a quick view of the DOLPi cameras and their capabilities and applications, please watch the 10-minute video that I submitted to the Hackaday Prize in 2015 (this project was one of the HAD Prize winners that year):
Regarding your last question, there are at least two universities that have built DOLPi cameras based on the ones described in my whitepaper and are flying them on drones for archaeological site discovery. At the Hackaday Superconference last year I met both Professors who have been doing it and discussed with them ways of increasing the real-time sensitivity (through stretching the dynamic range on the rendering code) to enable visualization of subtle polarization contrast when flying over vegetation. They were just starting their field experiments, so I haven't heard yet of results.
Would you mind explaining why an old camera is required and if modern digital cameras can be used?
Can you arrange several digital cameras each one with a different filter and polarizar take a pic at the same time and digitally compare their pics?
I did try the same conversion on a JVC 3-CCD camera. However, I wasn't able to take apart the beamsplitter and dichroic filters. These newer cameras use a beamsplitter prism with integrated filters, and no amount of attacking with dedicated optical adhesive solvents managed to disassemble it. In addition, the CCD sensors are assembled directly onto the prism, so I doubt that I would have been able to put it back together and achieve good registration.
Regarding your second question, in my original DOLPi project I used one camera and three filter positions in sequence (either mechanically or electro-optically switched), which gives excellent results, but the sequential image capture is slow with the Raspberry Pi because of the way in which individual frames are captured by the GPU (which has closed firmware, and thus is extremely difficult to bypass).
Placing three modern cameras to take independent pictures indeed speeds up the process, but that requires either geometrical transformation in software (to counteract parallax), or a beamsplitter arrangement to have a single optical input that is copied three times over so that the images can be optically registered. Both are possible, but require more complex hardware and/or optics. Please take a look at the comprehensive discussion in the DOLPi paper mentioned in the article for a detailed discussion about these.
As such, the old, tube-based camera hack is a much easier path for whomever wants to start playing with polarimetric imaging.
I'm not sure, but if I understand well, colors represent the polarization angle of the beam light ?
The attached picture (taken using DOLPi-Mech, as described in the DOLPi whitepaper) is a nice example. Note that the slight tinge of some of the coloring pencils, as well as of the plaque behind are because dielectric surfaces polarize reflected light. This makes polarimetric cameras useful for finding surface antipersonnel mines in humanitarian demining (the original goal of the DOLPi project)
Yes. Randomly-polarized light shows as grayscale (because it reaches all three tubes equally), while polarized light reaches the tubes differently, and hence its angle of polarization is represented as a color.