Contact Us   

Unlock Effective Condition Monitoring with Automated Vision Systems

Feb 1, 2022

Automation continues to press onward thanks to advances in computing technology. Thanks to the proliferation of artificial intelligence, specifically machine learning, visual data processing has never been more powerful. Essentially, computers can see now. And they can use what they see to make decisions faster than any human could. With machine vision systems, you can bring that technology to your business. To better understand these systems, let’s take a closer look at how they work.

While using imaging devices to provide input data for computer systems is nothing new, machine vision has reached new heights in recent years. You’ve probably already seen it in action. It’s the same technology that underpins Tesla’s self-driving system. It drives real-time facial recognition software. However, there are many more pragmatic, mundane applications that can provide significant benefits to your company. Are machine vision systems what your business needs to move forward? Let’s find out.

Automated Machine Vision

Automated Vision: Key to Effective Condition Monitoring

The Purpose of Machine Vision

A fully functional machine vision system can essentially replace a human’s eyes. A machine can see what a person would see and be programmed to make the same decisions, or provide the same inputs, as a person. This has immense potential for all kinds of industries. Quality control could be greatly automated. Machines can be monitored for anomalies without physical inspection by a person. Out-of-spec products or parts could be instantly identified and discarded.

However, it’s not the raw capability that should impress you. The real advantage with machine vision is that it doesn’t shut down. Camera systems don’t need a break. And they can often process images much more quickly and reliably than a person can. So, is there any disadvantage? Machine vision systems are a great long term investment, but do require some time to set up. You can speed up that process by working with experts, however.

Components of a Machine Vision System

How complex is a machine vision installation? That depends on the level of automation you hope to achieve. However, the core components of your vision system will be the same. Of course, no two applications are truly the same, so every vision system needs to be designed to suit your needs. Nevertheless, we can break system components down into a few categories.

Activation Mechanisms

Even though a machine vision system can remain online at all times, for efficiency it’s best to activate them via a sensor or input from another system. Think of this as the shutter on a camera. While you can leave the shutter wide open as long as you like, the resulting image will be overexposed. Likewise, if your system is constantly outputting data, that wastes valuable space and may even reduce accuracy.

Proximity sensors, motion detectors, and temperature sensors can provide reliable inputs for machine vision systems. For instance, if you needed a camera to monitor for defective products, you could place a sensor just ahead of the camera so that whenever a product comes down the line, your camera captures the exact moment it is in the frame. Similarly, a temperature sensor on a machine could trigger an infrared camera’s output.

Imaging System

Your activation mechanisms will trigger an imaging system and instruct it to intake data. Imaging systems use a lens to focus on an object in front of them and capture images at a framerate of your choosing. The sensor you use will play an important role, as it will determine the resolution of the images you collect. The resolution dictates the amount of raw data that the imaging system’s computer must process, and higher resolutions enable more fine-tuned actions.

Many imaging systems will use an onboard processing unit to output still frames saved in an image format. These images will be transferred to the GPU for analysis. Real-time processing in GPUs has become possible with newer hardware, however chip shortages have stifled supply’s ability to meet demand. In either case, this is where software processing allows machine vision to truly shine.

Software Processing

Several kinds of software can be used to extrapolate information from visual data. For instance, Optical Character Recognition, or OCR, is a tool that many office workers are familiar with. Machines are already quite good at recognizing text or symbols. Factories use OCR by placing QR-code-like patterns to serve as markers for mobile machines with visual systems. However, text is only the beginning of what vision systems can do.

Visual systems can easily detect objects and then measure them by comparing known contrast points. This is how facial recognition software works. It looks for known markers such as the corners of your eyes or lips and then takes measurements between other markers. For industrial applications, this same technology can identify an out-of-spec item. A flawed product, damaged box, or unusual item could trigger a disposal action. This is how modern factories have automated QC to a never-before-seen level.

Achieving Full Automation With Machine Vision

With more advanced processing, your visual systems can fuel a stronger level of artificial intelligence in your organization. By integrating your visual systems with advanced processing, you can allow your visual systems to truly become your floor’s eyes and ears. You can automate interventions and reporting to facilitate human activity as well. Machine learning has made it possible for your automation potential to grow as you feed more data into the system.

Machine Learning

Machine learning starts with a modal. By training the modal on the expected, ideal conditions, it can then detect deviations in those conditions and take action. As the modal experiences more deviations, it identifies patterns in that data to attempt to make sense of it. This is how deepfake algorithms work. As they see more and more images of an individual’s face in a variety of positions and expressions, it starts to distinguish different states.

In people, those states manifest as emotions and morphemes. In your company, though, a machine vision system equipped with machine learning could identify a recurring manufacturing error or faulty components. Data collected from drones could turn into insights about crop growth. Images of roads could catalog potholes and become a full maintenance schedule. This level of AI does require substantial processing power, but cloud computing can greatly simplify that step of the process.

Logic Control

When you are confident in your vision system’s ability, you can give it some authority over other systems. Programmable outputs can be used to shut down a factory line due to a machine-identified problem. Engineers can be given direct orders as to what machines to maintain. Predictive maintenance becomes possible since vision systems can also identify erratic behavior, such as vibration or overheating if thermal cameras are used. Let automation make the decisions, and let your team do the work.

Future of Machine Vision

The sudden leap forward in machine vision that we’ve experienced in the last few years is only the start. However, future advances will come from increased processing power and more efficient machine learning algorithms in the future. Optics and imaging systems have little room to grow. Outside of niche applications where extremely tiny cameras or high-speed sensors are needed, imaging technology is capable of fulfilling most companies’ needs at the moment.

We expect to see an increasing amount of processing power dedicated to visual data and more devices that are specifically designed to handle this kind of information. Already, products like Google’s Edge TPU modules are bringing machine vision prices down significantly. TPUs can handle image data quite quickly and while consuming much less power than traditional CPUs or GPUs. They can be installed more conveniently as a result and have made entry-level machine vision much more affordable.

Is Machine Vision Right for Your Company?

Early on, we mentioned that the purpose of machine vision was to act as a set of eyes. So, can your company benefit from a better set of eyes? Think about where workers currently have to see and react. But also consider places where human eyes cannot easily go. Machine vision can provide support in hazardous areas and can see features people cannot, for instance, infrared data. There’s almost always a good application for machine vision.

To learn more about machine vision and how a machine vision system could help your company get a competitive advantage, contact SAAB RDS to schedule an appointment or call us to speak with a representative. We can help you build a system to suit your needs.