Edited by Leland Teschler
Consider Escherichia coli (E. coli). It is one of a large group of bacterial germs that inhabit the intestinal tract. There are more than 700 different types of E. coli, but the type that contaminates foods and beverages is one that produce Shiga toxin (Stx). The most notorious Stx-producing E. coli is E. coli O157:H7, first recognized in 1982 as a food-borne pathogen during an investigation of bloody diarrhea among people who consumed bad hamburgers.
Some 30 E. coli O157:H7 outbreaks were recorded in the U.S. in the next 10 years, and estimates are the actual number was really much higher. The Centers for Disease Control has estimated that 85% of E. coli O157:H7 infections are from contaminated food. In fact, consumption of any food or beverage that becomes contaminated by animal manure invites the illness, which also brings on stomach cramps, nausea, and vomiting. Foods that have been sources of contamination include ground beef, venison, sausages, dried (noncooked) salami, unpasteurized milk and cheese, and unpasteurized apple juice and cider.
No wonder, then, that food and beverage processors put their equipment through intense caustic wash downs every day or even more frequently. Nobody wants to leave a residue that could potentially introduce E. coli O157:H7, salmonella, or other pathogens into finished products.
Unfortunately, many kinds of electronics do not perform well in the harsh environments characterized by caustic wash downs and other kinds of extremes. Vision sensors, often used in manufacturing areas to carry out high-accuracy, complex inspections, are in this camp.
Sensors take a variety of forms. The simplest commonly used in harsh environments are photoelectric sensors. They can be designed with sealed housings that withstand wash downs and other extremes typical of food-processing lines. Their sensing capabilities are generally limited to presence/absence detection. This type of sensor — also called a photoeye — can detect whether a bottle cap is present, for instance, but not whether it has been tightened half a turn.
Vision sensors provide the “big picture”—with image-based results that can capture a target image and inspect it for multiple features, such as cap seal and fill level, in a single step. Along with this quality assurance, vision sensors also identify products, confirming that a label is present on each object and that it contains the right information for consumers.
But in harsh manufacturing environments, vision equipment has been difficult to apply. The dust, dirt, and grime common in many industrial areas can create fog that clouds the vision-sensor lens so it can’t see its target object clearly. And the housings typically used in off-the-shelf vision sensors typically corrode easily in the excessive moisture and caustic chemicals used for wash downs. That’s because ordinary sensor housings are usually plastic or nickel-plated aluminum and are constructed to only guard against dust.
Machine builders often worked around this challenge by sealing the vision sensor in a protective enclosure. But sealed enclosures can be costly and may have to undergo testing to ensure they adhere to food-industry regulations.
In recent years, though, vision sensors have increasingly been offered with stainless-steel housings that resist corrosion and which employ curved surfaces for a more-hygienic design — allowing wash downs to more easily remove contaminants. The sensors also come sealed to meet IP68 and NEMA wash-down requirements.
Other packaging has been designed for industrial manufacturing environments wherein vision sensors may see some moisture but primarily stand up to dirt and dust. Here, a nickel-plated aluminum housing that has been enhanced to guard against these elements may be enough to fit the bill.
The introduction of ready-made housings for harsh environments has sometimes made it easier to apply vision sensors rather than their photoelectric counterparts. A photoelectric sensor can confirm that a single yogurt cup contains yogurt. But to properly inspect a carton containing 12 cups, one would need to employ multiple photoelectric sensors — each of which would need to be configured, installed, and maintained. On the other hand, a single vision sensor could do the same job. It would simply acquire one image encompassing the full carton. If any cups remained unfilled, the carton would fail the inspection.
Vision sensors would have similar benefits if the application parameters changed. Suppose in the yogurt cup example the 12 cups/container jumped to 16. In the case of photoelectric sensors, more would have to be added, and the existing sensors might have to move to accommodate different cup locations within the carton. But it would be likely that the same vision sensor could accommodate these changes with basically no change in the hardware. Changes would be confined to reprogramming the sensor so it identifies a “pass” image as one that contains sixteen filled cups rather than 12. There are obvious advantages in setup time and costs.
Specifically, vision sensors are most often applied in harsh environments to verify labels, including bar-code and date/lot code reading; and quality control, which frequently involves color verification and matching.
Verify the label
Each bottle, box or other container coming down a packaging line requires a label that contains the right information. A chocolate bar containing peanut butter that is labeled incorrectly could cause a severe reaction in a consumer with a peanut allergy, for instance. People who take pharmaceuticals rely heavily on the validity of these labels. Taking the wrong medicine could have catastrophic consequences. A vision sensor confirms each product contains a label and that it is easily readable. Sensors with optical character recognition and verification (OCR/OCV) capability additionally confirm that the label reads properly, and that the label includes the correct product information.
One of the most common and pivotally important ways vision sensors verify labels is through 1D and 2D bar-code reading (BCR). Designers must evaluate both the size of the bar code and the speed of the production line — the number of parts/min — before selecting a sensor for a BCR application. Each object triggers the camera to take a picture, so the line speed corresponds with the necessary camera speed. Vision sensors are commonly able to read up to 600 codes/min.
The proper camera lens and resolution for a BCR application depends upon the bar-code orientation and size, as well as its placement on the product. The bar-code size — in addition to the distance between the sensor and the target object — determines the required field of view (FOV), which directly affects lens selection.
Resolution depends upon label placement: If the bar code sits in a repeatable location, standard camera resolution will suffice. If it varies due to hand placement, the situation demands a larger field of view to accommodate this variance — and within this extended FOV the bar code appears smaller. Therefore, many such applications use a high-resolution 1.3-megapixel camera to sufficiently distinguish between bars and spaces within a bar code.
Beyond the vision sensor itself, aspects of the bar code such as background color determine what color lighting the application demands. Over 90% of machine-vision installations use dedicated lighting. This guarantees constant, consistent light conditions, regardless of the ambientlighting. In BCR applications, the right kind of dedicated lighting will create ultimate contrast between the bar code and its background, making it easier for the vision sensor to read the code.
Is the date/lot code right?
Another job for vision sensing is to verify the date/lot code on each product. The pharmaceutical industry has used date/lot codes in conjunction with bar codes for many years. Date/lot coding plays a role in product track-and-trace, ensuring each medicine is tracked throughout development, production and shipping. It thus allows end users to verify the product’s authenticity and keeps counterfeit products from seeping into the process. The food and beverage industry is now using this tactic more often to assist in recalls, so a recalled product can be simply traced back to its initial batch.
Vision sensors in these applications ensure each container has a legible date/lot code, so product lacking the necessary code doesn’t ship. Vision can also prevent product loss on the line by determining immediately if the date/lot code printer is working improperly or has stopped working all together — stopping the line and thus keeping subsequent containers from skipping this important step in production. The key to properly identifying this
date/lot code is to create the optimal contrast between the code and its background color.
A vision sensor’s ability to evaluate several different product qualities simultaneously is particularly useful in food, beverage, and pharmaceutical quality control. Few inspection processes are as critical as those that confirm containers are properly sealed before shipping. Vision sensors can examine each product to confirm not only that there is a secure cap or lid seal, but also that the container has been filled accurately.
Another reason vision sensors are used in these industries is to assure consistency. When a pizza crust is baked, for instance, a vision sensor may confirm the crust is of the correct size, shape, and color — not burnt — before toppings are added later in production. The sensor’s advanced capabilities let it examine a product for all these qualities at once, expediting the inspection process.
Consistency may often involve color verification and matching. A grayscale vision sensor may work for these applications if the user can create proper contrast between the target object and its background — usually accomplished using white lighting. If the system must analyze two colors, colored lighting can be used to create the needed contrast between them. For instance, if the system must identify both red and blue tablets within a blister pack, red lighting can be used to make the blue object appear black and the red appear white. The vision sensor can easily analyze the resulting high-contrast image. Conversely, a color vision sensor is used primarily when three or more colors must be identified or when the color of the target object varies, and therefore must be observed and reported. A typical example is that of a baked roll inspected for “doneness” based on its shade of brown.
Contrast is particularly crucial in harsh environments because the camera lens is prone to scratches and grime in washdown areas. It is harder to discern the feature of interest if the contrast is merely marginal. Optimizing contrast and keeping the vision sensor properly maintained are both key in such cases.