Hanks on Machine Vision: Choosing a camera for your vision application

By John Hanks and Gail Folkins, National Instruments Inc.

Table of Contents
Online or offline?
Increasing speed
Monochrome or color?
Line scan cameras
Software assistant

The camera is the most fundamental part of a vision system, but the myriad of choices faced by the new user can be daunting. Digital or analog? Area scan or line scan? Color or monochrome? Before you can make an informed decision, you need to understand your requirements.

Online or offline?
The first question to ask when choosing a camera for your vision application is whether your project involves an inline manufacturing process or an offline inspection system.

You can use analog image acquisition boards for many in-line applications. The frame rate for analog monochrome cameras normally sets the minimum acquisition rate and processing time of the acquisition board. The standard frame rate of an 8-bit analog camera is 1/30 of a second. PCI image acquisition boards typically can keep up with acquiring 30 frames/s, which is approximately 11 Mbytes/s throughput.

The limitation on throughput is either the camera's frame output rate or the amount of time need for processing between images. If using a single camera is the limiting factor in a vision system, expanding to two cameras could double the acquisition and processing bandwidth along different positions on the production line.

Increasing speed
Several methods exist to improve image acquisition performance. For example, some image acquisition boards feature a programmable region of interest (ROI). By using a programmable ROI, you can minimize the data that is transferred across the PCI bus and processed. For example, instead of acquiring a full-sized frame (640 x 480), you can limit the image size to (200 x 200) and effectively reduce the number of pixels from 307,200 to 40,000. Less datain this case one-seventhyields faster results.

If you need faster frame rates, more pixel depth (greater than 8 bits), or more spatial resolution (larger than 640 x 480 pixels), a digital camera and image acquisition board would serve as a better option than a standard analog camera and image acquisition board.

Depending on your application, you should consider using a progressive scan camera for applications in which the object is moving. Progressive scan cameras minimize the blurring of a moving object. These cameras come in analog and digital formats.

Monochrome or color?
To a new vision user, the question that comes to mind almost immediately when purchasing a camera for a vision system is whether to use a monochrome or color model. Although visually, you may prefer a color image, in actuality the eye has an improved perception of spatial differences in black and white.

Often, a monochrome system is more efficient. Industrial applications requiring a computer interface typically operate with a black and white camera because a color image requires more processing time. A monochrome image is typically only 8 bits while a color image is 24 bits, or even 32 bitsthese extra bits are costly when it comes to processing. Plus, when processing a color image in a machine vision application, the first step is often to separate out one of the color planes and process on only that color plane. In essence, you are back to processing a monochrome image.

Overall, color often does not yield significantly more information about the object inspected, especially when the analysis focuses on the spatial relationships of objects in the image, such as the number of edges and the distances between edges. Color can prove an important factor, however, when it is the only differentiation between the objects being inspected. In applications such as fuse, capacitor, resistor, and medical pill inspection, color is more often than not the only difference in the objects under inspection.

If high-color resolution in an application is important, you can use a three-chip or RGB camera. Many camera vendors, such as Sony and Cohu, have inexpensive industrial monochrome cameras. Sony and JVC provide high-resolution three CCD or RGB cameras. If you need faster than 30 frames/s, digital cameras are available from Roper, Dalsa, PULNiX, Hamamatsu, and others.

Line scan cameras
Another decision to make is between line scan cameras and area scan cameras. An area scan camera incorporates an m x n detector array that captures an entire image at once. A line scan camera uses a 1 x n detector arraybasically a single line of CCD sensors instead of a rectangular array. A line scan camera acquires an image only one or a few pixels wide.

The line scan camera is generally focused on a narrow line. As a part moves past, the camera produces 10,0000 or more lines of video per second. You can individually process the linesor, as is more common, you can stitch them together into a two-dimensional image.

Line scan cameras are especially useful for inspecting rounded or cylindrical parts. As the part is spun and imaged, you can unwrap its surface into one long image, which is easier to inspect. In addition, lighting techniques are not as complex because you only need to light the on-scan line of the part and not the whole part.

Line scan systems are usually for applications where the imaged item moves, such as production line systems. Although it is possible to use an area scan camera in such applications, camera framing timing control or strobe lighting is usually is necessary to ensure a complete image without blurring into one frame. Dalsa and Basler both supply line scan cameras.

Software assistant
To help determine their camera needs, engineers and scientists can turn to the National Instruments Camera Advisor (www.ni.com/camera). Users can compare camera features and specifications for more than 100 cameras; the package also explains how various cameras work with National Instruments software and hardware.

The software allows users to compare different models and makes of cameras, such as line scan, area scan, progressive scan, and digital and analog cameras. They can also use Camera Advisor to compare technical details of the cameras.

About the authors…
John Hanks and Gail Folkins are with National Instruments Inc. Gail is the Vision/Motion MarCom Manager and John is Vision/Motion Product Manager. John has worked for National Instruments in application engineering, marketing, and product management. He has also worked for Siemens Medical Systems Magnetic Resonance Imaging Division as an imaging support engineer. Degrees include engineering B.S. from Texas A&M and an engineering M.S. degree from the University of Texas. As a researcher at the University of Texas he studied image-processing algorithms and develop a system for biomedical cell counting. John has more than fifty published articles on signal processing, machine vision, and measurement technologies. He can be reached at National Instruments, 11500 North Mopac Expressway, Bldg. B, Austin, TX 78759. Phone: (512) 794-0100; fax: (512) 683-5569; e-mail: John.hanks@ni.com.