CCDs Capture Dynamic Motion

Toshi Hori, JAI Pulnix

Whether the application involves high-speed machine vision, live-cell imaging or surveillance, dynamic motion capture is a key task of the CCD image sensor. The need to capture an image fast and freeze it, while immediately exposing the photodiode to capture subsequent images, typically requires an interline transfer or even an intensified interline transfer CCD in extreme cases. A full-frame CCD architecture won’t work in this application because the photodiode basically is the CCD and will read out while being exposed. The device must read the stored image pixel by pixel.

An interline transfer CCD differs in that it has photosensitive diodes on one part of the pixel that couple to a CCD storage region called a vertical shift register. Once the signal accumulates on the photodiode, it can transfer into the shift register in just microseconds. Because of this, the detector architecture tends to have fewer problems with smearing and thus works well in short-exposure imaging of fast-moving objects. Required CCD specifications, however, will vary with the imaging applications.

With traffic cameras mounted on a gantry to look downward as vehicles pass underneath, the vertical resolution requirement of the camera is more critical than the horizontal. For a typical optical character recognition reading, it is necessary to have at least 10 pixels per inch, which translates to a 2 × 1.5-m field of view with 12 × 6-in. plates for a 768 × 480-pixel imager format.

Consider, for example, high-speed imaging of license plates on moving vehicles, which has become a fairly routine surveillance task for CCD cameras. The imager must accommodate diverse day and night conditions, which translates into a need for a very wide dynamic range of lighting capabilities. One can think of dynamic range as the device’s ability to detect both dim and bright parts of an image. Its mathematical definition is the linear full well of the device divided by read noise. Full-well capacity, the amount of photons each pixel can hold before saturating, depends in part on the pixel size; i.e., larger-pixel sensors will have larger full wells.

In this type of surveillance application, the sensor resolution needed will depend on the field of view, and a typical imaging application will require at least one- to two-megapixel resolution. This conflicts somewhat with the requirements for well capacity. In general, besides the limitations in how many photocharges can accumulate in the photodiode of an interline transfer CCD sensor, there are limitations related to the size of the vertical shift registers into which electrons must transfer.

Although high dynamic range requires a higher well capacity (and thus larger pixels), an increase in sensor resolution typically calls for the use of smaller-size pixels. The end user must compromise and determine the acceptable minimum resolution for the dynamic range required in the application.

The camera’s sensitivity also comes into play in imaging license plates because the sensor will use an electronic shutter and have an exposure of very short duration. In its most basic sense, sensitivity is a measure of how many photons enter the CCD vs. how many electrons come out, and it is affected by both the pixel quantum efficiency (the parameter measuring the effectiveness of the CCD to produce electronic charge from incident photons), as well as the charge-to-voltage conversion factor of the CCD’s output amplifier and the camera gain.

Another basic issue to be examined at this stage involves analog vs. digital output, with the former costing less but involving some loss of pixel information. In the license plate example, a progressive-scan CCD with digital output will be more advantageous because it can process high-resolution output at the pixel level. This includes the ability to interpolate color data to focus on areas of the image in more detail.

All of the required specifications — sensitivity, dynamic range, resolution and output format — will have an effect on the cost of the camera. Depending on the application, end users will be able to trade off requirements for certain parameters to maximize system efficiency and hold down overall inspection costs — or at least transfer the cost component to the more critical specifications. With machine vision inspection, for example, lighting is usually well-controlled, so dynamic range is less of an issue than it would be for surveillance. Speed of image capture is perhaps a more critical parameter to focus on.

Meet the author

Toshi Hori is president of JAI Pulnix Inc. in Sunnyvale, Calif.; e-mail: thori@pulnix.com.