Rick_green_200

Computational imaging comes to industrial machine-vision applications

April 16, 2018

Boston, MA. Computational imaging is finding use in industrial machine-vision applications, according to Steve Kinney, general manager of business development at CCS America Inc. Speaking April 11 at The Vision Show, he cited as a common example of computational imaging the HDR (high dynamic range) function on a cellphone camera. The camera takes two or three pictures at different exposure settings; software combines them into one superimage pleasing to the user.

HDR can be useful in machine-vision applications that include a combination of bright and dark objects. In an article in Vision & Sensors from last May, he presents an example of a PCB-mounted LED. A dark exposure clearly represents the LED and package but leaves the rest of the board dark. A bright exposure shows connectors and components but leaves the area near the LED washed out. Computationally combining those two images with a third medium-exposure image yields a superimage that adequately shows the LED, PCB silk screen, and components.

Photometric stereo is another computational imaging technique. The idea is grab a series of four images each lit at a different 90-degree angle, Kinney said. Software then makes a determination based on shadows to accurately detect surface characteristics like bumps and edges, making visually noisy surfaces easier to inspect.

A key challenge facing machine-vision applications involves the tradeoffs between color and image resolution. Kinney said that a high-speed monochrome camera coupled with full color lighting that supports independent strobing of red, green, and blue can yield images that computational-imaging software can combine into one color superimage at the full color resolution of the image sensor. In contrast, he said, Bayer color imagers lose spectral resolution across several pixels.

Kinney cited several other examples of computational imaging, including combinations of bright-field and dark-field images that, for example, can create superimages of dust on the surface of a sheet of glass as well as bubbles. Still others include extended depth of field, multispectral imaging, and 360-degree object capture.

CCS’s offerings include computational-illumination products that provide illumination in a structured format to support multishot image capture. Other enabling technologies, Kinney said, include high-speed CMOS image sensors, fast processing, high-speed interfaces, and liquid lenses.

Comments

To join the conversation, and become an exclusive member of Electronic Design, create an account today!