Machine vision’s vital role in automation

What can direct a robot to grasp a part, extract data from a label, and verify that a bottle cap is properly placed? Technically a human worker could perform these tasks, but that would open the door to human error, safety concerns, and excessive labor costs. As machine vision becomes more and more capable of locating and inspecting objects of interest, there’s little to no reason to have people standing by a conveyor, mindlessly completing the same repetitive tasks day in and day out.

Even flexibility is increasingly becoming a benefit of machine vision rather than something it’s deficient at, which is certainly a driving factor behind the projection that the worldwide machine vision industry will reach USD 17.2 billion by 2027 (Research and Markets). When combined with collaborative robotics, machine vision helps provide a highly flexible, autonomous solution for bin picking, machine tending, and a variety of other applications.

First things first: What exactly is machine vision?

Machine vision is the automatic extraction of information from digital images. A typical environment for this technology would be a manufacturing production line where hundreds of products are flowing down the line in front of a smart camera. The camera captures the digital image and analyzes it against a pre-defined set of criteria. If the criteria are met, the object can proceed. If not, the object will be re-routed off the production line for further inspection.

Manufacturers use vision systems instead of human inspectors because they’re faster, more consistent, and inexhaustible. For example, a beverage manufacturer would typically have human inspectors watching thousands of bottles move down a production line. The workers would need to ensure that every bottle cap was secured correctly, every bottle was filled to the appropriate level, label placement was appropriate, and all information contained on each label was 100% correct and readable.

Vision solutions are used heavily in conjunction with robots to increase their effectiveness and overall value for the business. These types of robots resemble a human arm with a camera mounted at the “hand” position. The camera acts as the robot’s “eyes,” guiding it to complete the assigned task.

The process for setting up a vision system

Machine vision is not a single product guiding automation. Rather, it’s a compilation of technologies that require the cooperation and intermingling of multiple technologies working together to solve problems. One of the key technologies that machine vision solution is heavily dependent on is camera products, which can range from simple industrial cameras to smart cameras and then to full vision systems. These cameras contain highly advanced lenses, which are a mission-critical element in any machine vision solution.

Arguably the most difficult portion of machine vision solution setup is lighting, which is frequently the cause of issues within a vision system. Even within the same production setting, the image results can be very different depending on the lighting. Fortunately, today’s smart cameras and vision systems often come with a variety of lighting options to ensure that your application’s lighting is consistent and repeatable.

Once you have good-quality cameras, lenses, and lighting set up, the next step is to capture and process the data. This can be done by implementing controllers, smart cameras, industrial PCs, or touchscreen HMIs. Finally, you need the right software to analyze the data and make it useful for problem solving.

The hottest trends in machine vision today

Given the incredible growth that the machine vision industry is currently experiencing, it’s no surprise that multiple groundbreaking innovations are occurring within this space as well. The trendy new technology that is happening right now in machine vision is known as hyperspectral imaging. Traditional vision is done on the normal visual wavelength and can be delineated between monochrome or color image processing, whereas hyperspectral imaging is pushing machine vision to a different visual spectrum. The key methods included in hyperspectral imaging are Ultraviolet (UV), Infrared (IR), and Short-Wave Infrared (SWIR).

Manufacturers are demanding high-speed cameras with high resolutions to improve overall equipment efficiency, and the development of rapid image sensing technology keeps pushing boundaries in terms of resolution. On the other hand, high-speed cameras require high-speed processing. Driven by data acquisition, IPCs and smart cameras need to be capable of handling larger amounts of bandwidth.

All these advanced technologies help to collect and analyze large amounts of data fast. This data acquisition is driving the use of artificial intelligence (AI) alongside machine vision. In order to achieve any level of AI, data must be processed over a long period in order to teach the software how to function. Leaps in AI for machine vision are driven by the amount of data that can be acquired and processed.

The four pillars of vision-enabled automation applications

Most traditional vision applications will fit into four pillars: inspect, locate, measure, and identify. Flexible manufacturing demands for mobile intelligence, along with developments in autonomous intelligence, will dramatically change the manufacturing floor. While logic controls manufacturing processes, motion and robotics bring manufacturing processes to life.

Inspect

Quality inspection requires the ability to detect defects and incorrectly made products. Vision systems are used heavily for inspection purposes in the food and commodity and life sciences industries to ensure product quality, helping protect consumers from defective and potentially dangerous products. Vision allows manufacturers to automate defect detection in places where typically manual labor would be needed.

Locate

Robots are used heavily in automation today, primarily for replacing monotonous and repetitive tasks traditionally performed by humans. One of the most common applications combining robotics and vision is pick-and-place. Vision provides the data that robots need to properly accomplish their tasks. Without vision, it would be difficult to properly guide robotics. 3D vision makes the ability to properly locate objects in 3D space even easier and with higher accuracy.

Measure

One unique vision application that performs measuring tasks is liquid level inspection. Manufacturers need to be able to provide a consistent amount of product, whether it be bottled soda, household cleaners, or something else. Utilizing machine vision, manufacturers can measure the exact level of the liquid in containers to ensure the highest quality. SWIR is a great example of a new image technology driving innovation for this field.

Identify

The best way to perform identification tasks is using 1D and 2D barcode reading. Since manufacturers need to be able to confirm that a given product is what it is supposed to be, they use barcodes for product and supply chain tracking, regulatory requirements, and quality control. Today’s vision systems make it possible to scan multiple products at a time, and even accomplish more tasks than just barcode reading for an application. Barcodes will also be verified for accurate information, which can only be accomplished with vision.

Conclusion

Machine vision and automation are in a symbiotic relationship — essentially, vision is the “eyes” of automation. When innovations in machine vision technology happen, automation becomes more efficient. While certain applications and tasks in machine vision can daunting, the reality is that there are many applications that can be solved simply and with incredible ease.