What Is Machine Vision?
Machine vision is the ability for machines to see their environment and make decisions based on visual input, such as image data or video data in an industrial setting. It’s carried out through a combination of specialized hardware and software that provide image-based sensing and processing in industrial settings. machine vision is typically deployed via PC-based systems or smart cameras. It can also be used to maintain a safe and secure industrial environment.
Machine vision has been one of the key drivers of industrial automation in recent years. It helps automate factory and plant processes by collecting and intelligently responding to patterns in visual data. The smart factory of the future will leverage machine vision and analytics to achieve a comprehensive view of factory operations from product to people to processes, which ultimately will enable better business decisions. For instance, applications such as remote monitoring and analytics at the edge will provide the insights for predictive maintenance, which in turn will reduce maintenance costs while improving machine uptime.
Advances in image sensors, vision processors, and a wider range of hardware and software applications offer more options in compute intensity and power consumption to enable solutions that fit specific industrial end user needs.
When augmented with deep learning, a form of artificial intelligence, machine vision systems are now capable of performing visual tasks more accurately than human eyes can. The latest machine vision technologies solve problems that were never before possible for machines to address—for example, recognizing characters under low light and on curved surfaces. They are also capable of processing higher-pixel-count images and handling higher-speed line rates for defect detection.
Machine Vision vs. Computer Vision
Machine vision usually refers to using visual processing technologies in industrial applications. Computer vision is a broader term as the fundamental technology that enables vision across retail, transportation, and digital surveillance.
Machine vision systems are powered by specialized vision algorithms that interpret data at high speed or in harsh industrial environments, which may involve low light, heavy vibration, fast-moving products, or high temperatures. Industrial machine vision may also call for incredibly high accuracy and real-time responses.
Machine Vision Systems
Machine vision systems come in two main form factors:
- Smart cameras, which have vision algorithms built in, capable of simple inferencing tasks
- PC-based systems, which could be a vision control box, industrial PC, or vision server with options for accelerators, and vision analytics software
The latest smart cameras provide compact, basic compute nodes for machine vision applications. These have processing speeds that rival those of PC-based systems while offering flexible and easy configuration and setup.
PC-based vision systems consist of a PC with an operating system and one or more cameras. A frame grabber or Ethernet card in the PC interfaces to the camera to import the images. A software application on the PC analyzes the images, can make judgments and decisions based on the data, and sends this decision data (such as quality pass or fail information) to other systems.
The smart factory of the future will leverage machine vision and analytics to achieve a comprehensive view of factory operations.
Machine Vision Applications
Machine vision has a number of compelling applications that help reduce costs and increase efficiency in industrial settings, including:
| Applications | |
|---|---|
| Quality assurance and inspection | Systems can inspect for conformance to design and quality parameters to not only detect whether goods pass or fail in a production line, but also identify the defect and its location. |
| Positioning and guidance | Machine vision systems can automate the positioning of goods in production, so they are in place for the next step. |
| Measurement | Machine vision uses an object’s edges to locate, identify, and gauge the object. |
| Identification | Machine vision leverages traditional and deep learning algorithms to identify barcodes and characters found on items on the production line. With the addition of deep learning, machine vision can also identify nonstandard characters and numbers, even in challenging conditions. |
Machine vision is useful in factory operations for applications that include:
| Operations | |
|---|---|
| Safety | Helping protect your most valuable asset—your employees—by reducing the potential for unsafe situations. |
| Security | Helping protect people and assets against threats by monitoring equipment and workers’ movements. |
| Worker behavior | Ensuring smooth operations as workers interact with robotics, other workers, devices, and products moving through the factory or plant. |
| Asset management | Tracking and managing assets, including raw materials, work in progress, finished goods and inventory, and their flow from the shipping dock of a factory floor to the warehouse. |
Machine Vision in Manufacturing
Machine vision is being used for a variety of use cases that drive business value for manufacturers. For example, a vision application for quality assurance can bring near-real-time image processing to the factory floor to proactively detect mechanical anomalies along a manufacturer’s production line. This can streamline quality assurance processes, minimize downtime and defects, and increase production line speed.
Machine Vision Software and Hardware
To support high-speed production, camera-based sensors must be accurate, fast, and reliable. A smart camera with specialized algorithms enables applications such as defect detection, vision localization, pattern matching, measurement, and optical character recognition (OCR).
Machine vision systems can be powered by Intel® Core™ and Intel® Xeon® processors for PC and server-grade computing, as well as by Intel® Movidius™ Vision Processing Units (VPUs) for deep learning acceleration at the edge. Accelerators such as Intel® FPGAs can help deliver additional performance to these systems.
The Intel® Distribution of OpenVINO™ toolkit helps accelerate the development of deep learning–enabled machine vision applications. Based on convolutional neural networks (CNNs), the toolkit streamlines the deployment of code across Intel® Hardware and accelerators while optimizing performance on the targeted hardware.
Intel’s Edge Insights for Industrial is a modular, validated software that helps companies combine their edge-based devices, data, and assets for faster insights. With built-in video ingestion functionality, it assists machine vision systems in their learning process. With the Intel® Distribution of OpenVINO™ toolkit, machine vision systems can expand upon their capability using deep learning and neural networks.