Linescan Cameras Expand Image Resolution

To use linescan cameras, you need to ensure that the object you want to inspect moves in synchronization with the camera

Yair Kipman, KDY, Nashua, NH, Scott Cole, RVSI Acuity CiMatrix, Nashua, NH -- Test & Measurement World, 10/1/1998

Linescan cameras have been used in inspection systems for years. The quintessential line-scan application is web inspection, in which an inspection system continuously inspects an almost endless image of a moving sheet of material. In electronics inspection, linescan cameras are used in non-web applications, typically ones that involve capturing a large amount of image data in a short time. The one overriding requirement is for relative motion between the camera and the object.

In an area camera, a lens projects a complete image onto a rectangular array of photodetectors (Fig. 1a). Each detector collects a charge proportional to the number of photons that strike it. After “exposing” the detectors for a set time, the camera sends the video information to a frame-grabber board, which assembles the pixel information into a complete image. The number of detectors in a typical area camera roughly equals the number of pixels in a 640x480 pixel display, but you can purchase cameras with higher resolutions—at a much higher cost.

Figure 1. Photographs of the sensors in (a) an area camera with a 768x494 detector and (b) a linescan camera with 2048 detectors show their relative geometry and size.

In contrast, a linescan camera employs a sensor that provides only a single row of photodetectors, from 128 to more than 8000 in number (Fig. 1b). The detectors acquire a long narrow image one pixel wide. To create a useful image, the object you want to inspect must move in synchronization with the camera’s acquisition of line images. The movement allows the object to pass through the camera’s narrow field of view one line at a time in the same way a document passes through the scanner in a fax machine. In many applications, a position encoder triggers the camera or the frame grabber to acquire each line image (Fig. 2).

Figure 2. A linescan camera acquires long narrow images across an object. A sensor triggers the camera when the object moves into position to have another image taken. Later, a computer assembles all the lines into a complete image you can view or process.

When the speed of the parts you want to inspect is constant, you can operate the camera without an encoder. The camera simply acquires sequential line images over an entire object. If you use a position encoder, the line rate will vary with the speed of the moving object. Unfortunately, as the object speeds up, the encoder produces pulses at a faster rate, and the lines forming the image tend to get darker because the photodetectors have less time to gather light. You can partially overcome this problem by programming a constant, reduced exposure time.

In an area camera, the vertical and horizontal resolutions are generally equal or very nearly equal. With linescan cameras, however, things are slightly more complicated: The number of photodetectors and the lens determine the horizontal resolution, but the motion of the object you want to inspect determines the vertical resolution.

For a constant scan rate in lines per second, the vertical resolution is linked to the motion. Thus, the vertical resolution is finer when the motion is slower—the camera acquires more lines across an area. If you use an encoder to sense the object’s position, each line of the image corrresponds to a “tick” from the sensor, which corresponds to a fixed distance traveled.

Cameras Take Images in Slices
A linescan camera, like an area camera, connects to a computer through a frame grabber, which assembles the individual line images into a two-dimensional image. For linear motion, linescan images look very much like images acquired using area cameras. If the object is rotating, its linescan image will look totally different. The linescan image of a spinning spark plug (Fig. 3) shows how a linescan camera can unwrap the curved surface to obtain lettering around its circumference. An area camera would distort the curved surface.

Although a rotating spark plug provides an interesting example, you’re more likely to inspect a BGA or components in trays. To inspect the components, you need a linescan camera set with the proper lighting, lens, and resolution to resolve the details you want to examine. Then, you move the tray through the line-scan camera’s field of view so the camera can image individual components or the entire tray in one pass, depending on the amount of detail you need to inspect. As a mechanical stage or platform moves the tray, the camera acquires line images, which a frame grabber reconstructs into a useful image.

Xerox (Canandaigua, NY) uses a linescan camera system developed by KDY to inspect test printouts from ink-jet print heads (Fig. 4). After a test print head prints a pattern on a paper sample, a carrier moves the paper into the camera’s view. As the carrier moves the paper past the camera, a position sensor triggers the camera to begin acquiring new line images across the paper strip. After the system acquires all the image data, its image-analysis software determines whether the test head passed or failed.

Movement Causes No Blurring
You may wonder why the camera doesn’t blur the image of the moving paper. Exposure times, which are programmable, can be as short as 50 ms. Thus, the only blurring that occurs corresponds to the distance the object moved in 50 ms. A non-strobed area-scan camera that requires a 33-ms exposure time would significantly blur a moving object.

Although the short exposure times of linescan cameras reduce blurring by several orders of magnitude (660 times in the example above), you must increase the amount of light used to illuminate the object by approximately the same factor. You can meet this requirement by adding more light or by placing an image intensifier in front of the line of photodetectors. Unlike area cameras that require uniform illumination in two dimensions, linescan cameras require uniform light only in a one-pixel-wide line.

Integrate an Image
Instead of using an intense light or an image intensifier, you could buy a camera that incorporates a time-delay integration (TDI) sensor. A TDI camera works much like a linescan camera, except its sensor provides between 4 and 96 rows of photodetectors, depending on the model. As the object moves past the camera, its image shifts from one row of detectors to the next. Simultaneously, the camera’s electronics move the stored electrons so they match the movement of the image. In this way, the sensor integrates the image of each line over several rows of sensors, thus gathering more light per exposure.

To properly use a TDI camera, you must use an encoder. Furthermore, you must match the encoder with the optical setup—the lens, the distance to the object, and the field of view—or you’ll get images blurred along the direction of motion.

Lenses play an important role in inspection systems, so you must pay attention to the type of lens you use with a linescan camera. The C-mount lenses used with most area cameras may work well for a 1024-detector linescan camera, but longer sensors may require larger-diameter F-mount lenses.

There is more to setting up a linescan camera system than we can describe in a short article. As you specify the hardware, you must investigate the clock frequency you need for the camera, the required line rate, how to control exposure, how to synchronize the camera with a position, and your choices of trigger signals (see “For Further Reading,” below).

With all these variables to consider, just taking a simple picture may seem like a big project. In many cases, though, a linescan camera provides an economical way to acquire images that would otherwise prove difficult to obtain.  T&MW
 

FOR FURTHER READING
Boroero, Pierantonio, and Robert Rochon, “Match Camera Triggering to Your Application,” Test & Measurement World, Newton, MA, August 1998, pp. 53–58.

Catalogs from camera manufacturers often contain useful application notes.
 

Yair Kipman is founder and president of KDY. Kipman has an M.S. in mechanical engineering from the University of Connecticut and a bachelor’s degree from Technion Institute of Technology.
Scott Cole is a senior software engineer at RVSI Acuity CiMatrix and he has worked for 14 years in machine- vision applications. He has B.S. and M.S. degrees in physics from the University of Florida.