Digital Imaging Optics
|
Choosing a CMOS Image Sensor is the easy bit. Once you have chosen your sensor a suitable
lens and lens assembly needs to be found. A PCB mount lens holder will screw to the PCB,
covering up your sensor from light. The top of the holder will have a standard thread which
will allow you to screw in a lens with the same thread type. The lens holder may also
incorporate a Infra-Red Cut Off filter and is typically made of Black ABS Plastic or
Black Anodised Diecast Aluminium.
The typical thread sizes are detailed below,
|

A S-mount glass lens
encased in aluminium
|
| Mount Type | Thread | Distance from Back Flange to Image Sensor |
| C | 1 – 32 (1 Inch/32 TPI) | 17.526 mm |
| CS | 1 – 32 (1 Inch/32 TPI) | 12.5 mm |
| S | M12x0.5 | Not Specified |
| X | M10x0.5 | Not Specified |
C & CS Mount Lens are typically used by the CCTV Market. Being an inch in diameter they
take up quite a bit of real estate but give a better result. S and X mount lens are
more typically used on the eyeball PC Cameras being only 12mm in Diameter. S Mount
(M12x0.5) Lens are the more dominant standard in eyeball / small cameras sacrificing
image clarity for size.
The Lens will screw into your PCB mount thus ensure it has the same thread. You will
be able to purchase lenses of different focal lengths, made of either glass or plastic.
Plastic are cheaper and typically of lower quality than multi element glass lenses. Lens
can also come housed in plastic or aluminium. Beware of inferior quality.
There are a number of parameters associated with the lens. The Focal Length expressed
in millimetres determines the field of view. A typical human perception is about 40
degrees, thus this is targeted as the normal view. They can range from telescopic
at approximately 20 degrees field of view which has a high image magnification to
fish-eye at approximately 110 degrees or more. The field of view can be calculated
from the focal length and diameter of the sensor, but most vendors will normally
specify both the field of view and focus length.
|
The aperture number, f Number or f / # specifies the amount of light which passes
through the lens. The lower the number, the more light that will pass through the lens,
thus the better performance in low light conditions. Every time the f number doubles,
the light is reduced by a factor of 4. A lower f-number requires more precise adjustment
of focus, where as a lens with a high f-number will be easier to focus. Eyeball cameras
for PC imaging have a typical f-number of 1.8 to 2.0 where 2.0 is standard for
most S-Mount Lenses. Pinhole lens, while having the ability to be concealed behind
a hole have a high f-number thus are not as effective in low light environments.
|

The inside view of a lens holder showing the IR Filter. B/W sensors do not require the filter.
|

Another example of a Lens Holder with IR Filter.
|
The pixel element of a CMOS sensor is normally susceptible to a greater span of the
electromagnetic spectrum than the human eye. The sensors ability will range from the
deep blues to almost infra-red. To get a more accurate measure of what the human eye
is seeing, an IR Cut Off filter is included with most lens mounts or lenses. Colour
aliasing and blurring can be a consequence of omitting the IR Cut Filter on a colour
imaging system.
As the IR cut off filter can either be part of the lens or lens assembly this will
need to check on purchasing. The filter present on a lens mount looks like a bit of
thick glass mounted to the bottom of the mount. This is normally made up of several
layers of optical crystal with a IR Cut Crystal layer. Beware of some lens having a
film on the back of the lens as these are easy to scratch.
|
Typical lens and lens mount vendors are
Sources of additional information
Digital Creation Labs uCFG
Digital Creation Labs has created this nifty analog video frame grabber based on the popular ATMega128 AVR Microcontroller. Using a Philips SAA7113 Video Decoder IC, the uCFG will capture video from up to 4 analog sources which can be downloaded over a serial link. This can allow you to use 4 small analog cameras while still having digital access to the data. Full source code is provided with the evaluation kit.
CMUcam
|
The CMUcam from the Robotics Institute at Carnegie Mellon University shows just what can be achieved with minimal effort. The CMUcam uses a Scenix (ubicom) SX28AX clocked at 75MHz to retrieve pixel data from an Onmivision OV6620 sensor contained on a C3088 module. With a maximum speed of 16.7fps the camera can track the position and size of a colourful or bright object, measure the RGB or YUV statistics of an image region or automatically acquire and track the first object it sees among an array of other applications.
|
|
|
CMUcam
- Uses OV6620 Onmivision Sensor (352 x 488 Colour)
- Scenix SX28AC MCU
- 16.7 frames per second, 80x143 Resolution
- RS-232 or a TTL serial port communication
- 200mA Power Consumption @ 7VDC
|
The CMUcam makes an excellent choice for your next robot vision project. The module is available for purchase from Acroname and Seattle Robotics.
VGA00AIT1 - VGA CMOS Sensor with JPEG Encoder
|
This module combines a VGA CMOS Image Sensor with a Microcontroller and 120Kbytes of RAM. The Microcontroller is pre-programmed to perform JPEG Encoding. A two byte command can be sent to the module to retrieve the image. It includes such commands as GET RGB Pixel Data (80 x 80 Pixels), GET RGB Pixel Data (640 x 480), Get JPEG Image (80 x 80) and Get JPEG Image (640 x 480). This module is a wonderful way to get started if you don't want to get your feet wet converting raw pixel data into common image formats such as JPEG.
|
|
Spectronix Modules
For the faint hearted or strapped for time (but not for cash) you can get pre-build modules from
Spectronix which terminates to a Ribbon
Cable and can provide a RAM based memory mapped interface. Spectronix’s
RoboCam I (164x124) and RoboCam II (356x292) come in both B&W and Colour
Versions based on ST/Visions VV5404/VV6404/VV5300/VV6300 CMOS Images Sensors.
RoboCam RC-1
- 160 x 120 Pixels
- Monochrome and Bayer Colour
- 8 or 4 bit data output, frame and pixel clock and I2C control interface
- Single 2x2 Inch PCB
|
RoboCam RC-2
- 356 x 292 Pixels
- Monochrome and Bayer Colour
- 8 or 4 bit data output, frame and pixel clock and I2C control interface
- Single 2x2 Inch PCB
RamCams RCM-1/RCM-2
- 160 x 120 and 356 x 292 pixel versions
- Monochrome or Bayer Colour
- Flexible 8-bit RAM interface, 3 modes + int
- Double 2x2 Inch PCB - Ribbon Cable Interface
|

Photo of the Spectronix RAMCAM2
|
Neuricam plCAM Intelligent Camera
The Neuricam pICAM Intelligent Camera is a very compact unit which contains an Analog Devices
ADSP-2181 DSP for image processing. It's size makes it an excellent choice for many applications
including robot control.
|
pICAM Intelligent Camera
- Twin-board module measuring just 53 x 53 x40 mm
- 256x256 pixel, grey-scale digital CMOS optical sensor
- Analog Devices ADSP-2181 DSP
- 512 KByte SRAM and 512 KByte FLASH
- RS-232 serial interface and Fast LVDS synchronous serial interface
- Buffered I/O Port
|
Neuricam also manufactures cameras with a microprocessor interface and control logic such as the NC1002 256 x 256 Digital Camera.
Bayer Colour Filters / Interpolation Methods
Other CMOS Sensor Manufacturers
|