Once the realm of science fiction, machines that are able to see and take actionare now very much a reality. Computer vision is a field of artificial intelligence that trains computers to interpret and understand the visual world. Using digital images from cameras and videos and deep learning models, machines can accurately identify and classify objects. And machine vision refers to the industrial usage of computer vision in automatic inspection, process control and robotic guidance.
A machine vision system uses cameras (its “eyes”) or optic sensors to acquire images so that computer hardware and software (its “brain”) can process, analyze, and measure various characteristics of this information for decision-making. Very often, machine learning principles are applied to this so that the system can be taught to execute tasks without human intervention.
The machine vision system’s software helps identify the pre-programmed features – a variety of set actions- according to its findings and analysis. In factories and laboratories, for example, we turn to machine vision when we need to execute a certain function based on the image analysis performed by the vision system. Here the machine vision system is likethe human inspector who performs visual quality control; more on this, in a moment.
Limitless potential in manufacturing, insurance, more
Machine vision systems have long proven to be invaluable in what are called narrow artificial intelligence (AI) types of applications, where machines can be taught to perform narrowly defined tasks and to do these so well, that they outperform humans.For example, we used a low-cost, single-board computer (a Raspberry Pi) to build a machine vision system that can detect potholes on the road and report them for repair.
Another common use of machine vision narrow AI is in manufacturing, where the technology is used to perform quality checks. Scanning through thousands of similar-looking objects in a day, cameras on machine vision systems efficiently single out items and products that are different or defective, and sort them out as rejected.
There are other, more elaborate ways we can use machine vision. For instance, we recently worked on a project in the DXC Labs that involves putting machine vision and machine learning into a humanoid robotics solution. This robot is designed to serve asan “agent” that can help financial companies or insurance companies recommend insurance products that are personalized for each walk-in customer.
Machine vision allows the robot to use facial recognition technology to identify existing customers and interact intelligently with them. The robot is also able to recommend insurance products to a new customer based on vital image statistics. It can also refer a customer to a human insurance agent to complete the insurance solutions sale process.
The application possibilities for machine vision arewide ranging. Some common areas for its use include:
• Calibration and testing
• Counting or sorting
• Data collection
• Machine monitoring
• Quality assurance
• Real-time process control
• Robot/machine guidance
The benefits that a machine vision system brings arealready clear. In a factory environment, machine vision keeps production lines going. Indirectly, the system can help manufacturers enhance productivity, ensure quality control and lower costs.
Nonetheless, we are just beginning to see the full potential of machine vision. Smartphone designers are now embedding machine vision into future-generation phones to make them even smarter. For example, a real-time language translation app is delivered using smartphone cameras. A user simply points their smartphone camera toward the words requiring translation – such as on a signboard – and the app can immediately translate the words and display them on screen in a language chosen/set earlier by the user.
Machine vision is also fast replacing the cumbersome and expensive radar systems on autonomous cars. Cheaper to produce, easier to deploy and taking up less space, machine vision is now helping autonomous vehicles effectively detect nearby obstacles and navigate safely through traffic.
Technologists are finding ways to leverage machine vision technology in projects across areas like aeronautics and aerospace, agriculture and veterinary care, banking and finance, building and construction, education and research, medical science and healthcare, and more. As noted, the emerging technology is now getting embedded into devices we use every day. Soon, machine vision is going to be so common and so much a part of our personal and professional lives that wemay not even be aware of its existence while using the capability.Application of machine vision is limited only by our creativity and imagination.