Assemble the right combination of tools to construct an effective vision system
Seeing the advantages machine vision provides in automation environments isn’t difficult. Inspecting products on the line saves the costs and consequences of recalled items, and properly labeling items before shipping ensures customers receive the correct product – a critical concern in the food, beverage and pharmaceutical industries.
Seeing the advantages machine vision provides in automation environments isn’t difficult. Inspecting products on the line saves the costs and consequences of recalled items, and properly labeling items before shipping ensures customers receive the correct product — a critical concern in the food, beverage and pharmaceutical industries.
What can be difficult is selecting the right machine vision system from the plethora of options on the market today with the capabilities needed, as well as the speed and reliability high-speed production lines require.
The right combination of smart cameras, lighting and software with sophisticated vision tools results in a comprehensive inspection solution that can solve many common applications with ease. Here are a few examples of how manufacturers can utilize vision practically, effectively and cost-efficiently.
The wheel deal
In aluminum wheel manufacturing, identifying each wheel model may come down to something as minute as varying hole patterns. One of the best ways to differentiate one wheel from the other as it travels down a conveyor — and confirm each wheel is placed in its correct container before shipping — is to measure hole pattern dimensions.
The challenge is not only to recognize which pattern belongs to which wheel model, but also to find a camera with high enough accuracy within a specified field of view (FOV). For example, the patterns may be so similar that the vision system may need to provide as low as a 4-micron pixel resolution in an inspection area that is a couple hundred millimeters wide.
Solving this application begins with selecting both the correct camera(s) and mounting technique. If two smart cameras, delivering image capture, processing and analysis in a single bundle, are mounted onto a multi-axis motion controlled fixture, manufacturers can use both cameras’ FOVs to their advantage. The cameras are placed below the conveyor so that as an aluminum wheel enters the first camera’s FOV, the camera can capture a picture of the bolt holes, then calculate the holes’ coordinates.
These coordinates are sent to the PC controlling the multi-axis fixture, telling it to move the second camera, featuring a much smaller FOV of perhaps 30 mm by 20 mm, into a position where it can view each bolt hole individually and measure the center point of each.
The next step is where employing robust vision software is of particular importance. While smart camera hardware is certainly critical, especially depending upon mounting and space requirements on the production line, the software is what makes these cameras intelligent.
Using specialized Circle Gauging algorithms, the software finds all the edge points of a bolt hole and calculates a best fit circle. The diameter of this fitted circle is then used to determine the wheel model. This inspection and verification process only requires 10 seconds per wheel, and it provides a cost-effective alternative to many high-end precision measurement systems.
Error-proof packaging labels
For safety, informational and aesthetic purposes, all consumer goods must be properly labeled. When varying labels are being produced on the same line, especially those that appear very much alike, a cookie-cutter machine vision software interface often isn’t enough.
After bottle and packaging labels are printed, they may be stacked for shipping. During this stacking process, labels similar in appearance may be inadvertently mixed, potentially leading an end-user to mislabel a product. Combining the smart camera technology with a standalone conveyor and proper lighting delivers an automated inspection solution.
The conveyor transports stacks of labels past a camera that is using diffused on-axis LED lighting — which provides even illumination and creates high contrast between the label’s features and its background, making it easier for the camera to identify the subtle differences between label varieties.
The camera position is adjustable so that a specific area within this label — one that differs slightly from all other similar labels — is within the camera’s FOV. This maximizes the camera’s ability to identify one specific label over another. The system additionally allows camera height to be adjusted, in order to accommodate stacks of varying label quantities.
Along with the camera and lighting adjustments, the system can incorporate a custom control panel that allows operators to configure and train this system to identify each label using a combination of inspection tools, such as Optical Character Recognition, pattern recognition and image comparison.
During this inspection process, the system validates that all labels in each stack are the same, and that the right labels are selected to meet a specified order before packaging.
Connecting the system
Sometimes inspecting a larger product with machine vision requires more than the average vision tools. It also requires software that can take several images and process them as a whole. For example, a connector must be accurately examined and measured in order to ensure it will properly mate up with other components during final assembly. This is another application where high precision, and high cost, measurement systems would traditionally be used. But machine vision can provide a more cost-efficient solution with the right camera and software combination.
To verify connector assembly using machine vision technology, a smart camera must be able to zoom in on specific sections of the connector in order to identify each component with high accuracy. Once it hones in on these sections, the camera’s FOV becomes too constricted to be able to fit the entire connector length.
To combat this challenge, a high-resolution camera is first adjusted so that only 1/6 of the length of a connector will be within the camera’s FOV at one time. After the connector is assembled, it is placed onto a conveyor for inspection, and as the connector moves to the next point on the conveyor, the camera takes a new picture of each connector section, allowing the camera to examine each section with high accuracy.
When all six images are stored, the full connector image is constructed using a Multiple Image Stitch software tool, which identifies overlapping features in each image and uses them to develop the full image. This tool is similar to old-fashioned panoramic modes on commercial cameras: it takes several individual images in succession, and then stitches them all together.
The entire connector can now be viewed as one image so that the full length and accurate measurements can be taken. This confirms that each assembled connector meets the required criteria, including pin/terminal height, pitch and coplanarity.
One of the keys to making this low cost solution easy to use is its software interface. An advantage of machine vision is that, with the right tools, users can actually see the inspection in process as it takes place.
Many systems are available with a user interface that allows new part inspections to be set up directly from the control panel. This interface may allow users also to view each image taken individually, as well as the full connector after the image is stitched together.
Since all images are saved and loaded, inspection setup and connector testing are performed quickly and simply. This combination of vision hardware, lighting and software ensures products are inspected thoroughly and efficiently on the line — saving manufacturers time, money and hassle in the short and long run.
Combining a smart camera with sufficient lighting and multi-faceted vision software ensures a comprehensive inspection solution.
Circle gauging algorithms are used to identify the edge points of each hole and calculate a best fit circle that corresponds with a wheel model.
To differentiate one aluminum wheel from another, machine vision may be used to examine hole patterns, which vary with each wheel model.
|Bradley Weber is director of application engineering at PPT Vision.