How to select a prism camera?

Research - Multispectral Articles

So you are aware of what multispectral technology has to offer, and specifically, prism cameras. But how do you select the camera that best fits your needs from the multitude of different cameras that are available? In this article the properties of prism cameras are explored in order to help you make a checklist for selecting the best camera for your application.

CMOS or CCD?

The two different technologies used for camera sensors are CMOS and CCD. Historically CCD is known for its high quality image with low signal to noise and high sensitivity. CMOS was often used as a cheaper solution that was less sensitive, noisier and used a rolling shutter but could achieve very high frame rates. That distinction has disappeared fast over the past years. Nowadays you can encounter high quality scientific CMOS sensors or cheap CCD sensors in webcams or mobile phones. Making a choice between CMOS and CCD is often no longer a matter of technology but of other criteria, which we will examine now.

Bigger is better?

One of the first set of criteria to examine is about size and resolution. The size of a sensor has many implications on the quality of the images. Larger sensors can sport larger physical pixels sizes and/or more pixels in each dimension. Larger pixel size tend to lower the typical noise level of a sensor and increase the sensitivity.

One of the consequences of increasing the physical sensors size is that a larger prism is needed, which is harder to produce. The lens mount and size also changes with larger prisms dimensions. Due to the back-focal distance requirements, a prism for 2/3” is not compatible with e.g. C-Mount lenses. Typically such prisms are used in combination with F-Mount or similar mounts.

Lenses

A prism camera is often selected for its high image quality and color accuracy. To benefit from the capabilities of the camera a good lens is needed. While any high quality lens can be used, it is recommended to select a lens that is specifically designed for minimal chromatic aberration.

Optimal image quality can be obtained when the sensor alignment is done in the very same setup that will be used during the application. This allows the manufacturer to perform an optimization whole optical system including lens, mount, prism and sensors.

Color selection

The visible spectrum can be captured several ways. The most common method use either grayscale sensors (no color information), Bayer pattern sensors or a RGB prism component.

Which method is best depends on the application. Is there a need to distinguish colors in the visible spectrum? If there is no such need then a grayscale sensor will suffice. If color imaging is needed the nest question is what the color quality must be. A prism generally results in high quality color imaging due to its very precise chromatic splits. But when total system cost is an issue a Bayer sensor may be a better solution.

There are two distinct methods to separate colors (wavelengths) from each other: filters and beam splitters. Filters let pass through only a part of the beam, while absorbing the rest. This lead to an inherent loss of energy by absorption. A beam splitter does not absorb light by separates the beam in two directions.

Prism splits

Prisms are glass objects that split a light beam in different directions. Each separated beam can be directed to an image sensor to result in exactly matched images from several sensors. But there are several methods to make a split. And each method has different properties.

Chromatic splits

The light beam can be split based on whether the wavelength of the light is above or below a certain value. For example a split can be placed at 700nm, effectively separating the visible spectrum from the infra red. Chromatic splits can be combined to create a system that can capture several smaller bands.

 The principle of a chromatic beam splitter; no light is lost.

Panchromatic splits

A panchromatic split does not discriminate between colors but instead acts as a half-coated mirror. Its main parameter is how the intensity of the output beams are related. This is often described by giving the percentage of the light that follows each output path. Common configurations are 50/50 and 67/33. It is even possible to combine 67/33 with 50/50 to obtain 33/33/33.

Note however that combining a panchromatic split with sensor-side color filters will reduce the sensitivity of the camera. At the split only a percentage of the energy is directed to the sensor and at the sensor the filter will block some of the remaining light. For this reason panchromatic splits are not the optimal choice for color cameras. Chromatic splits outperform such systems because no light is absorbed.

A 50/50 beam splitter combined with filters results in suboptimal performance; only half of the light is directed to either sensor, where a filter absorbs a part of the remaining light.

Sometimes a panchromatic split with an additional filter is useful to obtain a specific wavelength characteristic using cost-effective components. Also for polarimetric cameras that capture the polarization of light, this is exactly what is needed. Here there are several pixel-to-pixel aligned sensors that capture exactly the same image (using panchromatic splits). The only difference between the images is a differently oriented polarization filter in front of each sensor.

Custom filters

As indicated in the previous section, each sensor can be equipped with filter. In principle any type of filter can be used. In multispectral cameras this often is a band-pass filter to make certain channels more selective. In the aforementioned polarimetric cameras it obviously is a polarization filter.

Alignment accuracy

Alignment accuracy may or may not be a mayor selection factor: it depends on the application. When the alternative is to use two cameras side-by-side, a prism camera is a more accurate solution even without pixel-perfect alignment. And it has the added bonus that there is no need of (re)alignment of a multiple-camera system in the field. On the other hand, a camera for applications such as film industry or print inspection requires a much more accurate alignment. In these cases alignment errors of less than a quarter of a pixel are required.

Frame grabber interface

Every camera needs a frame grabber interface to transfer the images to an external device. In this area there are many different options available. A prism camera contains several sensors that provide images simultaneously. Each sensor increases the amount of data that needs to be transferred. As a result the slower frame grabber interfaces such as USB2.0, FireWire and to some extend also GigEVision can only be used at reduces image sizes or frame rates. For full speed and resolution it is best to use a frame grabber that is capable of transferring several GB/s.

Just selecting a very fast frame grabber does not solve the problem completely. It may only move the performance bottleneck to the PC that needs to process all that data.

< Prev		Next >