New Line Scan Camera Technology Digs Deeper
| By: Winn Hardin, Contributing Editor
Line scan cameras are the sensor of choice for high-speed applications in sorting, printing, textiles, agriculture, and related industries. By providing single-row images at a rate of up to 70,000 per second, line scan cameras are capable of creating “infinitely” long images up to 16,000 pixels wide. Meanwhile, customers spend a fraction of what it costs to do it the old-fashioned way, which required employing approximately 16 standard area array-based cameras with sensors composed of 1000 or more columns and half-again as many rows per image.
In addition to high-speed and high-resolution image acquisition, line scan cameras can also provide high-resolution color images using a variety of different camera architectures, each of which excels in color fidelity, ruggedness, speed, or a combination thereof. The latest line scan cameras combine the latest high-speed CMOS sensors with new sensor architectures to provide even more color and polarity information to solve the world’s fastest industrial applications.
Polarity looks below the surface
Teledyne DALSA (Waterloo, ON, Canada) has been a market leader in the supply of tri-linear and quad-linear color and multispectral line scan cameras for high-speed color images in printing and agricultural applications. Teledyne DALSA applies bandpass filters to each sensor in the camera to allow only certain wavelengths of light to reach the sensor. This way, each sensor provides an image based on the intensity of red, green, blue, or infrared light striking the camera sensor.
This year, the company will also offer a new polarization line scan camera as part of its Piranha4 color line scan cameras. According to Xing-Fei He, Senior Product Manager at Teledyne DALSA, the new quad-linear camera applies polarization filters to three of its four sensors: at zero degrees, 90 degrees, and 135 degrees. The fourth channel doesl not have a filter for comparison with the other channels.
“There are three fundamental properties of light: intensity, wavelength, and polarity,” explains He. “Using conventional monochrome or color cameras, for example, you cannot see the stress built up in glass panels because the glass is transparent. However, stress causes the polarization to change the material, which affects the material’s birefringence, which in turn affects the material’s polarization state. The same is true of transparent coatings applied to silicon [semiconductors.] You cannot directly image the coating to see if it’s been applied correctly, but by looking at polarization information, you can identify holes or gaps in the coating.”
Polarization images also reveal the hidden health of nontransparent materials, such as carbon fiber, where hidden material stresses can lead to delamination of the extremely expensive material.
Polarized images can also enhance contrast for common line scan agricultural sorting applications, says He. Sorting white corn kernels from contaminants is particularly challenging because the rocks and pieces of plastic that may mix with the food stuff look very much the same. In many cases, agricultural customers will use multispectral cameras that add an infrared channel to the standard RGB channels to detect differences in water content that separate organic food stuffs from inorganic contaminants. By adding polarization to this sorting process, He says customers can have much more confidence in the accuracy of their sorting machines.
Today, monochrome line scan cameras account for approximately $800 million in North American annual sales versus $300 million for color line scan cameras. “Before our new Piranha4 polarization camera, these kinds of cameras were very large and cost upward of $50,000. They were too big, too slow, and too expensive,” notes He. “We now make a very small, cost-effective solution, and that’s key for our industry.”
No one does color like line scan
While the latest CMOS line scan sensors are helping camera makers to double the resolution of their highest speed cameras, they also bring challenges related to special response and filter-coating applications.
According to Paritosh Prayagi, Line Scan Product Manager at JAI, the latest CMOS sensors have enabled JAI to offer speeds of up to 200 kHz in monochrome and up to 66 kHz in trilinear for its 4K resolution line scan cameras. This state-of-the-art sensor technology is also being incorporated into a 4K, 66 kHz 3-CMOS prism line scan camera, making it the fastest prism camera on the market. “We are using a unique sensor technology for these new models,” says Prayagi. “While the competition is promoting the TDI approach for high-speed readouts, we offer an innovative multi-line sensor design where the user can switch between two pixel sizes or combine lines for vertical binning using shared floating diffusion nodes between lines to provide increased sensitivity at maximum speed and sensor resolution. When also combined with horizontal binning, responsivity is increased by up to four times, providing high quality results even under low-light conditions.”
Traditionally, JAI uses a prism inside its cameras to separate the different color components from the incoming light into RGB (and now NIR or SWIR) channels. According to Prayagi, this approach yields several benefits. Dichroic coatings grown on the prism block offer steep cutoffs between color channels, which minimizes spectral crosstalk, facilitates better spectral differentiation, and results in sharper contrast. Because there is no coating to absorb light, prisms reduce optical losses compared to traditional polymer dye-based color filters, resulting in better sensitivity per channel.
The prism approach also means that users don’t need to worry about the physical spacing between the red, green, and blue line sensors like they do in a trilinear camera. “Prism technology is ideal for applications that have variable and unpredictable speeds because propagation of light through the camera is along a single optical path, so each of the three sensors is aligned to the exact same place on the target, as compared to trilinear cameras where the spacing of the individual lines creates some parallax issues which must be accounted for,” Prayagi explains.
In addition to offering higher throughput for traditional uses in agriculture and sorting applications, the new prism camera’s higher speed means it can compete for print inspection applications that might benefit from higher color fidelity.
All that said, the new trilinear camera from JAI takes advantage of a simpler architecture than a prism camera to offer an extremely attractive price-performance ratio for price sensitive color sorting applications. “Our prism technology is well established in the market. With this new trilinear camera, we wanted to introduce an attractive alternative to existing line scan solutions favored for use in more price-sensitive systems across a range of food and sorting applications,” Prayagi says.
Using time to differentiate colors
Teledyne DALSA is also turning the cost justification of multispectral cameras on its head with the introduction in Q1 2018 of its newest Linea four-channel camera. Unlike its other quad-linear, multispectral cameras that capture images within RGB and IR spectral bands, the new Linea will capture each band in sequence and allow the choice of LED narrow-band illumination to do the color filtering.
“Time division multispectral imaging capability enables definition of the color wavelengths using pulsed LEDs without spectral crosstalk,” explains Teledyne DALSA’s He. “This allows the customer to define the spectral bands instead of Teledyne DALSA defining the spectral bands based on the filter coated sensor.” This approach wasn’t viable before because CMOS sensors lacked independent start and end of exposure controls for each channel. “You could a single line sensor rather than multiline sensors on a single die, but it would slow down the application by a factor of three or four,” He says.
Line scan cameras continue to target some of the most challenging machine vision applications. According to JAI’s Dickerson, his company is exploring 10GigE and fiber channel networks to allow even faster camera speeds. With line scan cameras pushing the boundaries of what’s possible to achieve with machine vision, it is no surprise that designing and installing these systems takes advanced engineering, optics, single processing, and other knowledge to be successful. But don’t worry: Machine vision has a long history of designing intelligence into their products to make the complex appear simple. And that’s the one thing that is not going to change.