Computer vision: Extracting business value from images

Computer Vision (CV), an innovative field of artificial intelligence (AI) that enables computers to identify and classify objects and persons using image/video recognition and deep machine learning, is rapidly gaining momentum. Industries ranging from retail to manufacturing and automotive are experimenting with the use of CV to engage in-store shoppers, boost quality control in factories and enhance the safety of connected cars.

Computer vision can understand, analyze and extract meaningful data from images or a series of images faster and more accurately than humans. These images include traditional photos and videos, as well as images taken from thermal, infrared sensors and medical MRI (machine resonance imaging) scans. In fact, CV technologies can interpret 1D, 2D or 3D images of any object.

New algorithms, such as convolutional neural networks (ConvNet/CNN), which are particularly powerful at analyzing images, are driving the market alongside growing demands for new Industry 4.0, health diagnostics, autonomous vehicle and drone applications. A ConvNet is a deep or machine learning algorithm that can take in an image, assign importance to various aspects of the image (using self-learnable weightings) and differentiate between images.

There is no end of innovative applications for CV, which is already beginning to transform traditional industries such as automotive, defense, insurance, healthcare and retail.

“By adding computer vision to components into organizational applications, enterprises can add significant value by increasing efficiency, automating business processes and/or reducing costs,” explains Dave Schubmehl, Research Director, AI Software Platforms at IDC.

Quality control in manufacturing

In a bid to speed time-to-market and at the same time ensure quality, manufacturers are looking to automated quality inspection with little or no human intervention. This is where CV innovations come into their own.

Business & Decision, an Orange Business company, has run a pilot with a global pharmaceutical company to use AI and CV to detect and analyze bacterial growth in petri dishes containing samples of vaccines in production for quality assurance purposes.

Pharmaceutical firms test a sample of each batch of vaccine they produce. The process of visual human manual inspection is costly and labor intensive. CV has enabled the pharmaceutical company’s staff to focus their time on assessing abnormal levels of bacterial growth, significantly shortening the inspection process and improving accuracy in detecting production problems.

Effective data management is paramount

Richard van Wageningen, CEO of Orange Business in Russia and CIS, recently participated in a panel at the Machines Can See conference in Moscow on the topic of computer vision.

According to van Wageningen, “The ability to manage gargantuan amounts of data is central to an enterprise’s success with CV. Yet still, many enterprises find data management a struggle. Many simply don’t possess a coherent strategy for managing big data, despite data management tools being readily available.”

A report by the Economist Intelligence Unit (EIU) found that only 42% actually have what they consider to be a well-defined data management strategy. Manufacturers surveyed said they find it difficult to integrate data from diverse sources and hire skilled personnel to analyze it.

“Pulling together an effective data management strategy is something Orange Business excels at, whether it’s putting in place data lake infrastructure or applying the right data analytics techniques,” added van Wageningen.

Manufacturers that can efficiently manage their data have gained very valuable insights. Around 75% of companies surveyed by the EIU said they are seeing production efficiencies and annual cost savings of 10% or more in terms of improved product quality control through less wastage and rework.

Van Wageningen adds, “Quality control is carried out throughout the production process. So many variables can affect the quality of the end product – from the raw materials to the production line equipment, manufacturing environment (humidity, heat or cold can all be issues) and the packaging.”

“CV is far more systematic and reliable than humans for quality control and assurance across all these tasks and frees up staff for complex trouble-shooting tasks, continued van Wageningen. “It can be used, for example, to spot color, welding and other irregularities in car assembly lines.”

Guiding autonomous vehicles

The automotive industry is going through a huge shift towards autonomous cars. According to ABI Research, eight million vehicles will ship with some level of autonomous technology by 2025. CV will play a central role in autonomous cars.

Orange, for example, is currently running trials of autonomous guided vehicles (AGVs) and 5G connectivity in a co-innovation program with Groupe PSA, which owns a number of automotive brands including Peugeot, Citroën, DS, Opel and Vauxhall and UTRAC-CERAM, a private automotive testing company.

Wireless AGVs are likely to be adopted fastest in manufacturing. The factory floor is a far more controlled and predictable environment than the open road. Today, mobile robots typically follow what are known as “flow lines” marked on the factory floor and are wired. Wired mobile robots have various disadvantages, however. They need to follow a fixed route, which means making changes to the production line layout is complicated. There have been attempts at using Wi-Fi to provide greater routing flexibility, but the handovers between Wi-Fi cells is not robust enough. 5G faces no such problems.

In the automotive sector, many advanced driving features that use CV are being introduced into cars being launched onto the market, and features like collision avoidance and platooning are on the horizon.

Drones with “sight”

“CV is an important enabling technology for emerging drone-based applications,” suggests van Wageningen. “Drones, for example, are increasingly being used for asset inventory and management in agriculture and utilities, for example, as well as surveying, mapping, public safety programs and security surveillance.”

For data analysis applications, images and data captured by drones can be drilled down into for more granular information, such as potential crop yields or the state of infrastructure like train tracks. Data can be analyzed quickly and efficiently without the expensive use of low-flying helicopters.

Orange is working on 5G technology that will support drones. The high-bandwidth connectivity power of 5G, together with lower latency, will give drones far greater maneuvering precision and a quicker data transfer rate.

Improving event efficiency

The innovative applications for CV are endless. Orange Business has a particular focus on the use of CV for “object identification,” “facial recognition/biometrics” and “indoor analytics,” in addition to “quality control/assurance in manufacturing” and “automotive applications,” as mentioned previously.

For example, CV has the ability to better ensure the safety of the public at large and crowded events. It can detect potential issues with over-crowding and trigger messages via digital displays or audio announcements, for example.

At the Roland Garros tennis tournament in Paris this year, Orange piloted the innovative use of real-time analysis of CCTV images using CV to enable staff to guide visitors to their seats, reducing the risks of overcrowding and improving the sports fans’ visitor experience. In this type of scenario, edge computing is vital to protect the privacy of citizens.

The CCTV cameras were connected to an advanced private network built inside the stadium. Images were processed locally at the edge of the network for real-time image analysis. Staff could see a simple diagram of the court on their smartphones to check seating availability.

Computers will see more of our world

As the use of CV expands, computers will get a broader view of our lives in images. For the moment, CV has to be trained by feeding it a continuous diet of images. But, as AI-powered CV becomes more adept at training itself, we’ll see a host of visual applications appear that we’ve yet to imagine.

Want to know more about AI and how it works? Read our blog series beginning with Everything you always wanted to know about AI (but were afraid to ask) and read our ebook: Smarter data management for AI-enabled business success.

Jan Howells

Jan has been writing about technology for over 22 years for magazines and web sites, including ComputerActive, IQ magazine and Signum. She has been a business correspondent on ComputerWorld in Sydney and covered the channel for Ziff-Davis in New York.