AT logo - The Machine Vision Specialists  
  Frame Grabbers  
  Vision Systems  
  Contact AT  
  Site Map  
Request product information


(08) 9242 5411

(02) 9905 5551

(03) 9384 1775

Defence Recognised Supplier Scheme Logo


Using prism based cameras for multispectral applications

Back to Newsletter
Go to Jai Cameras

Selecting the right multispectral method for your application

Multispectral imaging is a term that can be used to describe RGB images that we are all accustomed to, however it's normally used as a term in machine vision when selective spectral bands are combined to create an image that highlights and contrasts the specific features of interest. When deciding what method of multispectral imaging will suit your application, it's not only the principal spectral wavelength bands that are important. Other things to consider include whether bands are to be captured at the same time, if you need pixel correlation between bands and other technical aspects along with cost, ease-of-use and overall functionality. When needing to capture pixel correlated small details at the same time and at the maximum resolution possible, prism based multispectral technology is an excellent choice.

What is the difference between hyperspectral imaging and multipectral imaging?

Hyperspectral Imaging
- Used to capture full spectral signatures for spectral analysis
- Typically hundreds of wavebands captured
- Wavebands are equally sized with no gaps between them
- The principal wavebands of interest that contribute to the anylsis are selected

Multispectral Imaging
- Used for targeted application requirements when principal wavebands are known
- Typically 2-15 wavebands
- Wavebands are discretely positioned and sized
- Each waveband has a specific purpose for a given application

Using Multi-sensor dichroic prism camera to achieve multispectral imaging

Using a multi-sensor dichroic prism camera for multipspectral imaging empowers users with a fast and compact solution. Prism based cameras for Multispectral imaging use multiple sensors aligned to a common optical path. This ensures simultaneous waveband capture with pixel correlation between bands and so fully detailed images. Prism based cameras are typically fine tuned to look at specific features and allow users to identify what is required.

In the past we have typically seen multispectral imaging solutions built with multiple cameras and so require image manipulation to overlay wavebands causing a loss of sharpness in the resultant multispectral image. They are larger and also typically cost more than a prism based solution. Additionally, multi-sensor prism based cameras with dichroic filters deliver strong spectral separation with little crosstalk at full spectral intensity as well as high frame rates.

Applications best suited to multi-sensor dichroic prism cameras for multispectral imaging:
Fruit and vegetable inspection/sorting systems: A multi sensor dichroic prism camera with RGB plus NIR data is perfectly suited to this due to its high spatial detail and precise spatial fusion. Useful in identifying ripeness and deformities. Since nearly all fruit and vegetable inspection systems involve some form of conveyance or vehicle movement, a prism camera with instantaneous multi-waveband capture and global shutter sensors remove all motion related artefacts.

Intelligent farming: Intelligent farming systems typically operate when mounted to a continuously moving vehicle. Dichroic prism cameras with global shutter, high sensitivity and short exposure-times removes problems with motion and vibration associated with agricultural vehicles in motion.

Today there is a growing demand for multispectral systems. Multispectral systems can be tailored to work with a range of different applications. In the past, multispectral prism cameras were manufactured with standard wavebands and so suited a small number of applications. More recent advancements in the manufacture of prism cameras now allows for customised waveband centre wavelengths and band widths and so can be tailored to a large variety of specific applications. It is often the case that hyperspectral imaging is used to determine the principal wavebands for an application but when deployed in the field and in volume customised multispectral cameras save time and cost.

Other applications include:

- Crop health, intelligent farming, environmental applications
- Surgical guidance, life sciences, forensics
- Electronics, recycling, scientific research
- Meat and poultry inspection
- Inspection of fruits, vegetables, nuts, grains, tea leaves
- Pharmaceuticals, cosmetics, packaging
- Currency inspection, passports, biometrics
- Fish farming

A comparison of the different multispectral techniques is described below. Each has its own advantages and disadvantages.
Two separate camera method
The two separate camera method to achieve multispectral imaging is perhaps one of the earliest methods used. For example, an RGB camera with an IR cut filter and a monochrome camera with a IR pass filter, create a colour and NIR combination.

- No specialised technology required
- Easy to source with lots of options
- Customisable waveband selection

- Difficult to achieve accurate pixel alignment/image fusion due to optical parallax
- Users are forced to deal with two separate imaging setups to build and maintain

Suitable applications:
Where the application can tolerate separate analysis of each waveband and where separate pass/fail steps are acceptable
Multispectral filter array method
The Multispectral filter array method features a predetermined mosaic pattern of spectral band pass filters on the sensor. The repeating pattern might capture a few spectral bands in a single shot. Often called "snapshot" multispectral.

- Single sensor/optical path, larger number of bands possible
- All bands are captured in a single snapshot

- Users will have to use demosaicing software methods to interpolate waveband values other than those of the filter on a pixel. This restricts spatial precision and blurs data.
- Has significant spectral crosstalk between wavebands and pixels
- Batch-to-batch consistency effects spectral accuracy
- Wavebands not customisable

Suitable applications
Where lower resolution is acceptable and in instances where some spectral uncertainty is tolerable.
Filter-wheel camera method
The filter-wheel method addresses both the parallax issue with the two camera method and the demosaicing issue with the multispectral filter array method. This method puts a filter wheel either in front or behind the lens to achieve multiple images of specific wavebands.

- Single sensor/optical path
- Full spatial resolution per spectral band
- Ability to customise filter set

- Sequential image capture results in a reduction of frame rates and possibility of position shift between wavebands.

Suitable applications:
- Where targets are static
- Where different filter sets are needed for different targets
- Where lower Mean Time Between Failure (MTBF) is acceptable
Two cameras + a beam splitter method
Similar to the two camera method, this technique overcomes the optical parallax issue by using a beam splitter. The beam splitter gives each camera the same optical path to the target.

- Single optical path
- Full spatial resolution per spectral band
Customisable waveband selection

- Large and expensive
- Loss of light intensity from split

Suitable applications:
- Where the target or the system is moving
- Where the system can be large/heavy
- Where illumination is plentiful
Multi-Sensor, Multi-path imaging
The multi-sensor method eliminates the need for a beam splitter by simply using multiple sensors each with its own filter. This gives full resolution and strong spectral separation to each band. Because we give each sensor its own optical path, this method reintroduces the issue of optical parallax and off-angle viewing. This requires complex alignment algorithms.

- Simultaneous capture of wavebands
- Full resolution of each sensor
- Strong spectral separation and full intensity
- Compact all-in-one systems
- Customisable waveband selection

- Optical parallax and off-angle viewing. This requires complex alignment algorithms that never produce perfect pixel correlation.
- Typically a higher cost
- Bulk and weight

Suitable applications:
- Where the target is flat and perpendicular to the camera
Multi-sensor dichroic prism camera for multispectral imaging
Unlike the multi-sensor, multi-path method we have eliminated the need bulk and weight seen with multiple camera methods simply by using multiple sensors in a single camera. Sensors are mounted on prism faces and are precisely aligned to a common optical path.

- Common optical path
- Simultaneous capture of wavebands
- Full resolution of each sensor
- Strong spectral separation and spectral intensity
- High frame rates
- Lightweight, ultra-compact solution
- Customisable waveband selection

- Resolution and number of channels limited by current prism designs

Suitable applications:
- Where both spectral and spatial precision are required including precise fusion of multiple wavebands
- Where there is a large amount of target or camera motion
- Where there is angled viewing, uneven surfaces or varying object sizes.
- Where light weight is important (e.g. drones)


Need a price or more application information on any of these multispectral methods Adept Turnkey can help. Please email Adept Turnkey or call our offices
Adept Turnkey Pty Ltd is "The Machine Vision and Imaging Specialists" and distributor of multispectral products in Australia and New Zealand. To find out more about any multispectral product, please call Adept Turnkey at Perth (08) 9242 5411 / Sydney (02) 9905 5551 / Melbourne (03) 9384 1775 or contact us online.



If you like this page, please recommend and share it.

Google+ Facebook Twitter More