The first computer vision use cases in the 1950s analysed typed versus handwritten text. Early commercial applications focused on single images, including optical character recognition, image segmentation, and object detection. Pioneering work on facial recognition began in the 1960s, and big tech companies began launching capabilities around 2010.
The computer vision market size was estimated at $14 billion in 2022 and is expected to grow at a compound annual growth rate of 19.6% from 2023 to 2030. While there are many new computer vision breakthroughs and startups, its market size is small compared to other AI technologies. Generative AI, for example, is estimated to become a $1.3 trillion market by 2032.
Emerging use cases for computer vision
Wherever you go today, cameras are likely scanning you, and computer vision algorithms are performing real-time analytics. Computer vision’s top use cases include document scanning, video surveillance, medical imaging, and traffic flow detection. Breakthroughs in real-time computer vision have advanced self-driving cars and driven retail use cases such as cashierless stores and inventory management.
You've likely experienced or read about these and other consumer-facing use cases, especially the top computer vision applications in the automotive and consumer markets. You may know less about how manufacturing, construction, and other industrial businesses use computer vision technologies.
Businesses in these industries are generally slow to invest in technology, but initiatives like Industry 4.0 in manufacturing, digital construction, and smart farming are helping industrial leaders better understand the opportunities with emerging technologies.
Reducing waste in manufacturing
Computer vision presents a significant opportunity in manufacturing, with computer vision algorithms reaching 99% accuracy. That is especially impressive considering that only 10% of companies use the technology. Is a digital revolution brewing in the industrial sector, or will these businesses continue to lag in adopting computer vision technologies?
Arjun Chandar, CEO at IndustrialML, says identifying product quality on materials in motion is a primary use case in manufacturing. “With the help of a camera with a high frame rate and applying a machine learning model frame by frame, it is possible to identify defects at production lines with high speed.”
Global manufacturers waste as much as $8 trillion annually, and computer vision can help monitor equipment, manufactured components, and environmental factors to help manufacturers reduce these losses.
The underlying technologies for many manufacturing use cases are mainstream, says Chandar. “These mostly use 2D cameras, albeit with a high resolution and frame rate of 20 frames per second or higher, and a convolutional neural network (CNN).”
To increase accuracy, manufacturers will need a strategy to augment that data. “To add training capacity as in typical manufacturing environments, the number of images with good product quality vastly outweighs defects,” adds Chandar.
Jens Beck, partner of data management and innovation at Syntax, says manufacturers can start with basic visual inspection steps and then lead to greater automation opportunities. “We see computer vision and AI combined for visual inspection, such as in automotive to check glue tracks,” he says. “The business value for the customer is not only the option to increase its overall equipment effectiveness (OEE) by automating manual steps but to document the check, and then integrate computer vision into their manufacturing execution system (MES) and then finally, enterprise resource planning (ERP).”
Improving safety on the factory floor
Beyond quality and efficiency, computer vision can help improve worker safety and reduce accidents on the factory floor and other job sites. According to the US Bureau of Labor Statistics, there were nearly 400,000 injuries and illnesses in the manufacturing sector in 2021.
“Computer vision enhances worker safety and security in connected facilities by continuously identifying potential risks and threats to employees faster and more efficiently than via human oversight,” says Yashar Behzadi, CEO and founder of Synthesis AI. “For computer vision to achieve this accurately and reliably, the machine learning models are trained on massive amounts of data, and in these particular use cases, the unstructured data often comes to the ML engineer raw and unlabeled.”
Using synthetic data is also important for safety-related use cases, as manufacturers are less likely to have images highlighting the underlying safety factors. “Technologies like synthetic data alleviate the strain on ML engineers by providing accurately labeled, high-quality data that can account for edge cases that save time, money, and the headache inaccurate data causes,” adds Behzadi.
Sunil Kardam, SBU head of logistics and supply chain at Gramener, says, “Computer vision’s benefits include real-time analysis, improved efficiency, and enhanced decision-making.” Kardam shares several other example use cases:
- Track material movement, identify defects in products and packaging, and reduce waste
- Enforce protocols by monitoring unauthorised personnel behaviours
- Simplify document processing, optimise inventory, aid insurance claims, and enable efficient logistics management
Kardam shares that computer vision relies on cameras and advanced algorithms like YOLO, Faster R-CNN, and OpenCV. He says machine learning models for computer vision “can be processed on edge devices or in the cloud, with smart cameras and cloud-based APIs providing powerful capabilities.”
Monitoring the power grid
Most manufacturing is indoors, where engineers have some control over the environment, including where to place cameras and when to add lighting. Computer vision use cases are more complex when they involve analysing outdoor areas and landscapes using mounted cameras, drones, airplanes, and satellites.
Vik Chaudry, CTO, COO, and co-founder of Buzz Solutions, shares an example of using drones. “Computer vision is used to monitor for and identify faults in the power grid and utilities as well as substations to ensure a reliable and connected grid across the US,” he says. “Relying on thousands of images collected from various utilities across the US, computer vision can accurately identify threats, faults, and anomalies.”
Powerline fires are a significant area of concern. From 1992 to 2020, there were more than 32,00 powerline-ignited wildfires across the US, according to the National Interagency Fire Center, and California’s second-largest wildfire was caused by a powerline too close to a tree. Utility companies are now evaluating AI opportunities to optimise repairs and minimise risks.
“Because this software uses real data and images from existing utilities, it is extremely accurate and can identify a range of threats from weather, poorly maintained infrastructure, and rising temperatures,” says Chaudry. “The technology enables fast and efficient maintenance while preventing widespread outages and grid-sparked disasters.”
Looking toward the future, Naeem Komeilipoor, CTO of AAVAA, shares a new frontier for computer vision and emerging technology devices. “Brain-computer interface (BCI) technology used within an industrial setting can be a complementary approach to certain industrial computer vision use cases, especially in environments with low visibility, extreme temperatures, or hazardous conditions where the use of cameras is restricted,” he says.
Completing repairs in hazardous conditions is one use case, but a more common one is when construction, work on refineries, or other fieldwork requires using both hands to inspect or operate machinery.
“Take a repair on a job site. BCI technology used within smart glasses provides an alternative method for tracking the eye gaze without needing an external camera so they can still perform in challenging conditions,” says Komeilipoor. “Instead of relying on cameras, BCI monitors eye movements by interpreting brain and biosignals like electrooculogram (EOG). The technology involves advanced signal processing and machine learning algorithms to analyse eye movements captured through specialised sensors.”