Color provides critical information that can improve the reliability of many machine vision inspection applications and make it possible to inspect many products that cannot be inspected in grayscale. Typical examples include ensuring that a tan rather than a taupe automotive interior component is installed, and checking the position of a black label on a black single-serving coffee cup.

Color machine vision identifies position of black carpet against black background.

Selecting the vision sensor type is important, with the trend currently running away from charge coupled device (CCD) sensors and towards complementary metal-oxide semiconductor (CMOS) sensors because of the latter’s ability to operate at higher speeds. The type of color processing is also important with the most advanced type of processing representing color images in 3D red-green-blue (RGB) space.

The first step in any machine vision application is defining the specifications, such as obtaining agreement on the differences between a good and a bad part, and obtaining samples of each. The application is normally developed in the lab where different approaches can quickly be evaluated. Software-based inspection items that process the image to determine whether to pass or fail the product are critical to robust inspections.

Color Vision Applications

Automotive interiors represent a very important color vision application. Automotive interior colors are often so similar to each other that it is often difficult to distinguish between them, even when they are placed in close proximity to each other. Each automobile moving down the assembly line has dozens of different components ranging from the dashboard and seats, down to tiny fastener covers that must be exactly the right color.

Another challenging automotive interior application is inspecting buttons that have light emitting diodes behind them that must be the right color and often are designed to cast different colors in day and night conditions. Inspection is typically done in a closed box with a light turned on to simulate daylight conditions.

Black Velcro pad is easily identified against black carpet.

With the recent requirement for floor mats to be locked into place, some automobile manufacturers are putting Velcro pads on the underside of the mats. Color machine vision systems are used to detect the presence of the pads by identifying the subtle difference in the tone of black between the underside of the mat and the pad. Color marks are also used for identification purposes on many automotive components that would otherwise be difficult to distinguish from similar parts.

There are also many color machine vision inspection applications outside the automotive industry. A very simple application is the inspection of expiration codes that are printed on labels in the food industry. The codes are printed in different colors, which poses a challenge for conventional grayscale vision systems in that different lighting is often required to read the different colors. A color machine vision system will be able to read any color of text with white light.

Single-serve cups are typically made of black plastic and are fitted with a label that also has a black background. The position of the label needs to be tightly controlled with respect to vertical and horizontal position as well as rotation. This is accomplished with a color machine vision system that distinguishes between the label and the cup based on subtle differences in the tone of the two blacks. Color vision is also used for inspecting the filter paper used in single- serve machines, which is cut in a circular shape, folded like an accordion, and heat sealed to the cup. Color machine vision can tell the difference between defects such as holes in the paper and acceptable variation such as slight discoloration that could not be distinguished by grayscale vision.

Color Vision Sensors

There is an ongoing evolution in the type of color machine vision sensor that works best for different types of applications. Charge coupled device (CCD) sensors transfer each pixel’s charge through a limited number of output nodes and the charge is then converted to a voltage, buffered, and sent off-chip as an analog signal. An advantage of CCD sensors is that the entire pixel can be dedicated to light capture which, combined with high output uniformity, tends to provide a high-quality image. Another advantage is that the CCD’s built-in electrical shutter eliminates the need for a mechanical shutter. On the other hand, CCDs tend to consume relatively large amounts of power, have lower frame rates, and are relatively expensive to manufacture and integrate.

CMOS sensors, on the other hand, convert the charge from the photosensitive pixel to a voltage at the pixel site, and multiplex the signal by row and column to a collection of digital-to-analog converters (DACs) located on the chip. Each site contains a photodiode and three transistors that perform the functions of resetting, amplification, charge conversion, and multiplexing. Performing these functions on-site provides higher speed and reduces power requirements.

CMOS sensors have delivered significant improvements in resolution and noise levels that have come close to closing the quality gap with CCDs. With quality levels now essentially the same, the higher speed of CMOS sensors has become the deciding factor in many applications.

The method of processing the color image is equally important as the sensor performance. In the earliest and simplest approach to color processing, called color segmentation processing, the color images acquired by the camera were first converted into black and white pixels. A limited number of colors of interest, typically between 1 and 8, are represented by different grayscale values and all other colors as black. This approach enables very high-speed processing; however, it is highly sensitive to lighting variations. If the lighting conditions are not as expected, some or all of the selected colors may not be recognized, resulting in false negatives or positives.

In the next level of color processing, called color image processing, the color images are converted into 256 levels of grayscale. This approach provides more precise and stable measurements than color segmentation, but has difficulty capturing subtle variations in color.

The most advanced type of processing, called real color processing, represents color images as different positions in 3D RGB space. Subtle variations in color can be recognized by representing them as distances between different color pixels comprising this space. Thus, scratches and dirt can be detected accurately even in images with low contrast.

Setting Up the Vision Application

The first step in developing a color machine vision application is for the manufacturer to define the specifications of the project, starting with the specific differences between a good and bad part. Are we looking for the presence or absence of a part? Does the part need be in a certain position? Does it need to be a certain color? Do the dimensions need to be measured? Other requirements should be defined at this point such as the time available for inspection, restrictions on positioning the vision sensor, ambient lighting, etc.

Special vision tool calculates glue width and path.

The vision process is then defined in the lab setting where it is easy to evaluate different types of equipment and software approaches. Inspection items are selected to perform the inspection process such as judging objects based on their shape, judging the position of objects, judging according to dimensions, judging based on color, etc. Specialized tools can be applied at this point, such as one that accurately calculates glue width and path as well as identifying any gaps regardless of glare, changes in lighting conditions, or variations in metal workpieces.

Setting up an inspection item requires configuring the inspection item, teaching the inspection by acquiring images from good and bad parts, and setting judgment parameters that determine the measurement tolerances. The next step is to test the process. If the measurements are stable, the process is complete. If measurements are unstable, then detailed items need to be set up and re-teaching should be performed. Manufacturer-provided samples of good and bad parts are then run in the lab in order to ensure that the proposed inspection process is robust.

The last step is to take the process to the manufacturing site and evaluate its performance in the real-world environment. Here it may be necessary to make adjustments in the lighting, judgment parameters, or other aspects of the process in order to ensure robust performance.

Conclusion

Color imaging provides valuable information that can improve inspection reliability on many manufacturing operations. Today’s color vision sensors and software have made tremendous advances that have improved their precision, robustness, and ease of use.

This article was written by Alex Nowak, Commercial Marketing Engineer – Vision, Measurement, RFID and Auto ID at Omron Automation and Safety (Hoffman Estates, IL). For more information, contact Mr. Nowak at This email address is being protected from spambots. You need JavaScript enabled to view it. or visit http://info.hotims.com/55593-201 .