Data-driven business transformation using computer vision and real-time video analytics

Applications of computer vision technology are numerous. It can help businesses to make data-driven decisions and strategies. Let’s learn how.

Imagine every single CCTV camera in your building to be an ‘eye’ looking on, analyzing, identifying, and processing visuals similar to the human eye, and understanding what’s happening out there. In the simplest terms, this is Computer Vision (CV). Using a combination of Artificial Intelligence and Digital Signal Processing, computer vision technology can enable machines to understand what’s happening in the physical world.

Why computer vision technology is relevant to your business?

Because this technology enables your business to see the hidden opportunities to make your business perform better. The opportunities come in a wide spectrum, such as detecting anomalies, identifying patterns that can help increase your store sales, automating security surveillance, and a lot more can be done. Before jumping into the applications and opportunities, let’s understand a bit in detail about computer vision technology and real-time video analytics.

The History of Computer Vision

A popular misconception is that computer vision is a new technology but history tells us differently. As early as the 1960s, researches on computer vision technology began at several universities. A deeper understanding of sensory processing in animals served as an inspiration for the ‘SIFT descriptor’, which is a local feature used in computer vision for tasks such as object detection.

It started with two neurophysiologists, David Hubel and Torsten Wiesel , who experimented with a cat’s visual system to understand how the visual stimulus is linked to the response system.

In the experiment, an anesthetized cat had electrodes placed on the primary visual cortex area of the brain. The aim was to study neuronal activity in the region while displaying various images. These experiments made it more clear to understand how the visual system builds an image from simple stimuli into more complex representations. Many artificial neural networks and the fundamental components of deep learning, may be viewed as cascading models of cell types inspired by Hubel and Wiesel’s observations.

The precursor of the modern computer vision was discovered in 1963 by Lawrence Roberts when he published a Ph.D. Thesis on “Machine perception of three-dimensional solids”, in which he described the visual world in dimensional objects of 2D and 3D. His finding in the process of 2D to 3D construction and the following display of 3D to 2D is the right starting point in the development of advanced computer-aided 3D systems.

In 2001, the Object detection framework theory was introduced by Paul Viola and Michael Jones which used a binary classifier to identify facial features. Five years after the introduction of the Viola/Jones algorithm, Fujitsu the Japanese camera maker released a real-time face detection feature in their new devices which was the first foray into modern computer vision.

The Present

Computer vision has gone from being theory on paper to real-life run time application on our hands. We can identify objects and faces, recognize them, and even track them successfully through a combination of computer vision with other technologies. Modern computer vision is used in tandem with other technologies such as machine learning, deep learning, etc to implement features like pattern recognition, event prediction, event detection, etc.

Business Intelligence

Real-time video analytics software like Emotyx can generate actionable insights from your CCTV video footage. By analyzing the in-store behavior of customers, their movement patterns, demographics, etc, it facilitates data-driven business intelligence for you. The software employs several activities as listed below to generate intelligent insights that can help grow your business.

• Customer counting
• Heatmap tracking
• Theft detection
• Product affinity analysis
• Customer emotion analysis
• Customer demographics based intel
• Insights based on customer traffic analytics

Furthermore, several processes in a business such as security surveillance, anomaly detection, etc can be automated with help of real-time video analytics and robotic process automation technology. Crime detection, accident detection, automated security surveillance, safety & process violation detection, vehicle counting, and classification, object detection are a few examples of capabilities offered by real-time video analytics.

Applications of Computer Vision Technology

One could easily think of numerous applications of computer vision in different industries. Be that agriculture, healthcare, manufacturing, retail, or wherever CCTV are presently used.

In agriculture, precision cropping uses a technology where aerial imagery is combined with computer vision to weed out good crops and bad crops by ultra analyses of the available visual data. In healthcare, during the era of the pandemic, computer vision was successfully employed to weed out and detect anomalies present in chest X rays to confirm the diagnosis of pneumonia. This was based on an experiment conducted by medical students at Cranfield University. In manufacturing, RPA (Robotic Process Automation) and machine vision are utilized for detecting hindrances in manufacturing processes, accidents, etc.

The high-resolution thermal cameras along with deep learning were used to capture infrared images of passengers in Taiwan during the COVID-19 pandemic and these high-resolution images are run over computer vision algorithms to detect possibly infected passengers.

Self-driving cars are being tested by different major international conglomerates such as Google and Tesla. They are an expert use of real-time video analytics with multiple tools like ultrasonic sensors and LiDAR along with computer vision to operate autonomous vehicles. This technology helps in successfully identifying road signs, obstacles, incoming traffic to safely orient the vehicle accordingly. Using video surveillance to determine patterns of human behavior, crowd analysis, object recognition, and human movement could be used for security or in improving performance at workplaces.

Future

The question that remains unanswered is that ‘Have we successfully tapped the full potential of computer vision?’. The answer would be a resounding ‘No’, there are so many untapped and unexplored benefits and real-time applications of computer vision technology are yet to be inspected.