Sensors emerge as machine vision tools

A new category of machine vision has emerged: vision sensors. They are increasingly seen as the low cost general-purpose form of today's machine vision technology.

Some of the more obvious uses for a vision sensor would be for traditional vision applications where machine vision has already been applied. But due to ease of use, lower cost, and functionality, vision sensors can also be applied to applications that traditionally use mechanical gauging or sensor technology. Measurement sensors, proximity and photoelectric sensor arrays, or mechanical gauges that perform inspection and measurement can, in many applications, be replaced by a single vision sensor. Even though the part cost of these inspection and measurement techniques may be lower than vision sensors, their engineering, installation, constant calibration, and maintenance costs (including change over time for different or new products) make a vision sensor a potentially competitive and cost-effective solution. And in many cases, due to their functionality, vision sensors can add new inspection or measurement capability with fast, easy on-line modification.

Depending on the choices made, several different kinds of detection methods can be employed, including binary, gray scale, or single CCD color. The type of detection method usually depends on the inspection or measurement required. For example, high-contrast images using back or simple front lighting can use binary (black or white only) imaging to count pixels for area or basic dimensioning or presence/absence applications. Applications that require searching for features, pattern matching, or simple defect detection benefit from the use of gray scale detection (usually 256 levels of gray). Color detection can be used for isolating features by their color, then inspecting presence/absence based on color, or for advanced defect inspection by subtle color definition or shift.

Among different hardware configurations are: the all-in-one "smart camera" approach, where light source, lens, camera, and processor/controller are in a single package; and a (typically) smaller remote camera, lens and light source with a separate processor/controller. The camera's imaging element sizes (usually a CCD array) can vary to provide pixel resolutions from 200H x 190V to 640H x 480V. The single all-in-one approach has advantages of acting as a stand-alone unit with I/O or communication outputs to other devices like PLCs, SCADA, or an information-collection network. The transfer of the image to the processor is not exposed to a cable transmission. This can also offer a slight speed increase in total processing time.

A remote camera with a separate processor/controller allows for a more compact camera assembly, typically 2- x 2-inch square or smaller (not including the lens or light assemblies). The use of a separate processor/controller also allows for I/O wiring termination points to be located on the controller housing. Room for additional connections like external monitors and communication ports is also available. A separate processor/controller allows the controller to be located in a control cabinet or enclosure, close to the PLC equipment for example, to minimize wiring.

Some types of vision sensors combine several software methods to maximize flexibility. A pushbutton interface can be used for pattern matching, presence/absence and feature comparison by simply putting good and/or bad parts in the field of view and teaching the sensor the parts (usually with a single push of a button on the device). Any deviation or change can be analyzed or compared to a good model, and a pass or fail judgment or data can be given with setup or changes being easily made.

Drop-down menus or interfaces with a Windows^(R) appearance or based GUI offer the user a customized configuration capability that allow individual settings or changes to be made to many aspects of each measurement or inspection. The measurement tools, their execution order, and the data they produce can all be configured or changed without the use of programming commands or languages.

Today's vision sensor de-signs can accomplish what was once only possible with high-end, high-cost vision systems or lower cost, but complexly engineered, high-maintenance sensor arrays. One of the fundamental advantages of vision sensors is that their low cost and ease of use allow the process of inspection, measurement, and part rejection to be distributed out along the production line at each point of production, instead of being concentrated at the end of the line after value has been added to a part or parts have already assembled. This is similar to the advantage seen when the PLC became distributed along a production line also. Besides the reduction in wiring runs, problems and troubleshooting also become a much simpler task, with no one sensor shutting down the entire vision system. Downtime and changeover costs, as well as tooling for different or new parts to a production line, are also reduced. Feedback of location or measurements can also be provided to motion control systems or robotics at any point required.

Another advantage of vision sensors is that design considerations don't only apply to a new production line. Vision sensors can be added to an existing line or replace existing sensor arrays without the expense of a capital appropriation or extensive training and engineering man-hours. What designers should consider is where they want to make these inspections, or where feedback is required, and identify where critical inspections and rejection of parts must take place to improve quality and efficiency. Existing mechanical gauging, measurement, proximity, or photoelectric sensor arrays can also be considered as good locations for vision sensors, especially if mechanical changes must be made for different part runs and/or constant calibration or if maintenance is required by the sensors or arrays.

A designer or user must also consider what the demands are for the application-for example, if it only requires pattern matching or simple feature comparison with pass/fail output only and no customized lighting or lenses, then a vision sensor with simple functions, built-in light and lens may be all that is required. If lighting is critical for seeing features and/or a variable focus lens is required, then controllable lighting and an industry standard C-mount lens capability should be considered. In applications where dimensioning, feature inspection, or defect detection is required, a vision sensor with menu or Windows^(R) based GUI and configurable inspection and measurement tools may be best to solve the application.

Vision sensors can solve many non-contact inspection and measurement problems, replace complex sensor arrays, provide feedback to robotics, and solve problems where machine vision was once thought too expensive and complex.

As vision sensor technology progresses, look for higher speed processing, more advanced measurement and inspection tools, better resolution, and greater networking capabilities for information exchange and multiple camera capability. As each new version of vision sensors is released and as new players enter the market, look for continued low cost and improved functionality to benefit all users.