Machine Vision (MV) automated systems are greatly enhanced when fused with Artificial Intelligence (AI) and Machine Learning (ML), notably at the edge. These ML-based computer vision systems, a staple of Industry 4.0, are highly proficient at complex image analysis, detecting irregularities, and making sense of large datasets that a human never could. While smart manufacturing may be the most obvious Industry 4.0 use case for ML-enabled MV processes, other areas where the technology will be valuable include retail, logistics, transportation, automotive, healthcare, public safety, and urban design—making all these sectors ripe for computer vision innovation. Below is a list of several market forecast statistics to be mindful of as it relates to camera system shipments, installed base, and revenue.
Registered users can unlock up to five pieces of premium content each month.
Market Overview
Camera systems are integral to the use of computer vision as it’s the mechanism for which object detection/inspection is made possible. Whether it’s an Industry 4.0 application in a manufacturing plant or a smart city use case like dynamic traffic light signaling, a camera is essential to edge-ML-based computer vision. To accurately evaluate the totality of the computer vision market, it’s important to highlight the market opportunities for camera systems, notably those enabled with AI enablement.
- In 2027, 198 million camera systems will ship, which is up from 126 million shipments in 2022. The Compound Annual Growth Rate (CAGR) between 2022 and 2027 is 9.5%.
- The smart city vertical currently accounts for 70% (88.3 million) of camera systems, but that figure will decrease to 58% (115.5 million) by 2027.
- Smart buildings come in a distant second, with 36.4 million camera system shipments expected by 2027, closely followed by the automotive sector at 31.1 million shipments.
- Of the more than 976 million AI-enabled camera systems that will be deployed worldwide by 2027, more than half of them (50.9%) will have AI implemented on the camera itself. The other primary methods of AI implementation for camera systems used in MV are the cloud, gateway, and on-premises services.
- Between hardware and software/services, total revenue is growing at a CAGR of 11%, increasing from US$21.4 billion in 2022 to US$36.1 billion in 2027.
- Hardware will maintain at least 84% of total revenue throughout the forecast window. However, software/services have been and will continue on a sluggish, yet steady, upward spiral. Compared to accounting for 11% of total revenue in 2020, that number will climb to 16% by 2027.
“At present, Machine Vision (MV) is becoming one the most central technologies of automation, and now this technology is merging with Artificial Intelligence (AI) and Machine Learning (ML) to lead the transition to Industry 4.0.” – David Lobina, Analyst at ABI Research
Key Decision Items
Vendors Must Balance Solutions between Edge ML-Based and Traditional Computer Vision
While there is a lot to be excited about in computer vision at the edge, it shouldn’t be seen as the end all be all. Vendors should be careful not to overstate the boons of ML-based computer vision, as classic systems are still very much in play.
Some applications are better suited with classic MV or a combination with ML. For example, traditional machine vision software is still preferable for jobs that require exactitude, as opposed to the probability-based process that edge ML entails. Moreover, Deep Learning (DL) would be impractical for decoding barcodes because the process involves specific algorithms.
Edge ML-based computer vision is undoubtedly a game-changer for complex situations and detecting object discrepancies, but it also stipulates extensive training to achieve at least the 99% process accuracy that Industry 4.0 requires. For these reasons, it’s paramount that vendors ensure that ML-based computer vision software integrates seamlessly with existing MV systems in manufacturing and logistics. This could even mean that the initial ML-based machine vision deployment is initially done in the cloud instead of directly at the edge.
Identify the Best Industry 4.0 Use Cases and Key Sectors for Edge ML-Based Computer Vision
As previously alluded to, edge ML-based computer vision is not a fit for every Industry 4.0 application. Besides, classic MV is more than good enough in many scenarios.
Adding to this, not every computer vision problem calls for edge ML. Vendors and service providers must identify where edge ML-based computer vision stands to gain the most from its deployment, often in complicated and changeable situations that could use ML modeling.
As for the sectors that machine vision vendors need to identify, manufacturing, smart cities, retail, transportation, and logistics use ML-based computer vision at the edge most often.
However, the automotive, healthcare, and smart building sectors are untapped and poised for huge payoffs. Identifying the use cases and industries where edge ML-based computer vision shines the most will enable vendors to commercialize offerings optimally.
Take a Wide View of the Market
Customers need computer vision products that work cohesively within the larger MV ecosystem, which encompasses hardware (e.g., cameras, chips, etc.), software, and a means to dissect the data. In fact, some computer vision vendors offer the complete package to capture the market and ensure no compatibility issues for the end user.
If a vendor does not offer the whole package, the company must provide hardware-agnostic computer vision software and software-agnostic data analysis for edge ML-based MV. Take the time to understand what products are leveraged by enterprises and other organizations—and factor this information in during product design. Technology agnosticism is especially important in smart city development, healthcare, and transportation.
Consider Integrating with Orbiting Sectors
The advantages of edge ML in computer vision, as great as they are, won’t necessarily be comprehensive in MV systems, nor broad in scope. Indeed, it’s not going to be a matter of “flipping a switch” and instantly acquiring the benefits of computer vision use cases. Instead, the best results will often come about through small, incremental advancements, such as defect detection, edge extraction, and object classification.
Moreover, in the opinion of ABI Research, overlapping sectors, such as TinyML processes, microcontrollers, and neuromorphic chips, will have a crucial role in advancing the edge ML-based computer vision market. While each individual procession may not seem like much, a collective improvement in various computer vision applications will result in cumulative results with a profound impact.
Key Decision Items
Dig Deeper for the Full Picture
For a more comprehensive overview of how edge ML fits into computer vision/machine vision applications, download ABI Research’s Edge ML-Based Machine Vision Software and Services research report.
Not ready for the report yet? Check out our TinyML Is A Perfect Match for Environmental Data Research Highlight. This content is part of the company’s AI & Machine Learning Research Service.
Edge AI
