Published: Jul 25, 2022

computer vision: a human-like vision


Creating things that can think and act like humans have been many people’s long-held ambition for decades. Giving computers the power to “view,” “observe,” and “understand” the world around them was an intriguing concept. What was once simply just a pipe dream has now come true.

The world is abruptly changing into a more diverse, high-technology environment throughout the years. Computer vision works like how human perceives things; object identification through image processing is now in the palm of your hands.

Artificial intelligence, deep learning models, and various computer vision applications are a few things you experience virtually through computer science. We now enable computers to do the job we cannot simply do, such as object tracking, pattern recognition, video motion analysis, image classification, image restoration, and computer vision algorithms.

Applying computer vision technology needs a deep understanding to grasp the idea of what it has to offer the physical world. This article will define computer vision, its history, how it works, and its real-world applications.

Stick around and dive deep into the computer vision field!

What is computer vision?

It is possible to teach computers to recognise and understand the environment around them through the study of Computer vision. The goal of computer vision, a branch of computer science, is to replicate some aspects of the human visual structure and allow computers to detect and interpret images and videos in the same manner as people do. For a long time, computer vision was only capable of a limited number of tasks.

Computer vision is much like human eyesight, except humans have an advantage because they've been around longer. People’s eyes have been trained for decades to distinguish between different objects and whether they are moving or stationary.

In contrast to the human visual cortex, computer vision uses cameras, data, and algorithms to train robots to do these functions in a fraction of the time. Using a machine trained to examine products or monitor a manufacturing asset can swiftly outperform human capabilities, spotting even the tiniest faults or concerns.

Computer vision is a branch of artificial intelligence (AI) that has made enormous strides in recent years. It has now surpassed humans in some tasks relating to object detection and classification because of improvements in deep learning and neural networks.

Data generated nowadays is a major driving force behind the development of computer vision.

How does computer vision work?

The field of computer vision requires a massive amount of information. It performs data analyses until it identifies patterns and eventually recognises images. For example, to teach a computer to recognise automotive tires, it must send large amounts of tire photos and tire-related materials to the computer to understand the differences and recognise a tire, especially one without faults.

Two essential technologies are utilised during this procedure: Deep learning and Convolutional neural network  (CNN).

Algorithmic models in machine learning allow a computer to learn about the context of visual input on its own. Eventually, the computer can tell one image from another if it has fed enough data. Learning algorithms allow the machine to recognise an image without being programmed.

As an image breaks up into pixels, a CNN assists a machine learning or deep learning model to “search” for specific information. It creates predictions about what it is “seeing” using convolutions, a mathematical procedure on two functions to produce a third function.  The neural network conducts convolutions and tests their accuracy in a series of repetitions until the predictions start to come true. A human-like capacity for picture recognition and perception is then demonstrated.

A CNN first looks for sharp corners and basic shapes like the human eye. Then, as it goes through different versions of its predictions, it fills in the details. A CNN works with single photos. In the same way, a recurrent neural network (RNN) is used in video applications to help computers figure out how the images in a series of frames are related.

Here are some of the most common ways that computer vision systems are used:

  • Object classification. The system looks at the visual information and puts a photo or video’s subject into the chosen category. For example, the algorithm can find a cat in a picture full of other things.

  • Object identification. The system analyses the visual content of a photo or video and locates a specific object within that content. For instance, the algorithm can zero in on a particular cat amid the other cats in the picture.

  • Object tracking. The system analyses the video that identifies the object or objects that fit the search criteria and then follows the object as it moves.

The history of computer vision

Most engineers and scientists have been working on making machines more capable of processing visual input for about 60 years. In 1959, neurophysiologists began testing a cat’s brain response to various visuals by exposing it to a series of images. They discovered after experimenting that image processing begins with simple shapes like straight lines. 

Around the same time, computer image scanning technology made it possible for computers to digitise and collect images for the first time. In 1963, computers could turn two-dimensional images into three-dimensional digital images. During the 1960s, artificial intelligence (AI) became a topic of study in academia, and people started trying to solve the problem of how humans see.

In 1974, optical character recognition (OCR) technology was introduced, allowing it to read text printed in any font or typeface. It could also use neural networks for intelligent character recognition (ICR) to decode handwritten text. As a result, OCR and ICR have made their way into document and invoice processing and vehicle plate identification.

Dr. David Marr, a neuroscientist, discovered in 1982 that vision works in a hierarchical fashion and devised algorithms allowing robots to recognise basic geometric patterns such as edges, corners, curves, and the like. At the same time, computer scientist Kunihiko Fukushima created a network of cells capable of recognising patterns for use in his research. The Neocognitron neural network had convolutional layers in it.

As of 2000, object identification was the primary focus of research, and by 2001, the first face recognition applications appeared. Visual data sets began to be tagged and annotated consistently in the early 2000s. The ImageNet dataset became available in 2010. With more than a thousand object classes, it is an excellent starting point for CNNs and deep learning models that are utilised today. CNN was entered in an image recognition competition in 2012 by a team from the University of Toronto. For image recognition, a model called AlexNet significantly decreased errors. The error rates have dropped to just a few per cent since this breakthrough.

There is no longer a scarcity of computing power in today’s world. It’s not just new hardware and advanced algorithms propelling computer vision technology ahead; the enormous amount of publicly available visual data we produce every day is also a factor. When used in conjunction with robust algorithms, cloud computing has the potential to assist in the resolution of even the most complex problems.

The following are some of the most significant developments in the field of computer vision:

  • The first digital image scanner was developed in 1959 by converting images into a grid of numbers.

  • When Larry Roberts, the father of CV, explained how to extract 3D information about solid objects from 2D images in 1963, it was the beginning of the process.

  • Marvin Minksy ordered a Ph. D. student in 1966 to hook up a camera to a computer and have the computer report what it saw.

  • Neocognitron, the forerunner of today's Convolutional Neural Networks (CNN), was invented by Kunihiko Fukushima in 1980.

  • Multiplex recording devices and ATM cover video monitoring were implemented in 1991-1993.

  • The first real-time face detection framework (Viola-Jones) was developed by two MIT researchers in 2001.

  • Self-driving cars were tested on public roads by Google in 2009.

  • Goggles, Google’s image-recognition program for mobile devices, was released in 2010.

  • Facebook started employing facial recognition in 2010 to make it easier to tag photos.

  • 2011 – Osama bin Laden’s identity was confirmed using facial recognition after being killed in a US operation.

  • In 2012, Google Brain’s neural network used a deep learning algorithm to distinguish images of cats.

  • Google released the open-source TensorFlow machine learning technology in 2015.

  • A computer program called AlphaGo, created by Google DeepMind in 2016, defeated the world’s best Go player.

  • In 2017, Waymo made allegations that Uber had stolen trade secrets from the company.

  • In 2017, the year of the iPhone X’s release. The company touted face recognition as a critical new feature.

  • In a Stanford University reading and comprehension test, Alibaba’s AI model outperformed humans in 2018.

  • Rekognition, Amazon’s real-time face-recognition solution, was offered to law enforcement agencies in 2018.

  • Police in India will be able to search photographs using a smartphone app and facial recognition technology in 2019.

  • The United States added four of China’s most prominent AI start-ups to a trade sanctions list in 2019.

  • The UK High Court has found that automatic face recognition technology to search for persons in crowds is legal in 2019.

  • In 2020, Intel planned to enter the GPU industry with the Intel Xe graphics card.

  • From the midst of the pandemic last 2021 until today, computer vision has helped track down COVID-19 patients and those they have been in contact with. The data analysis and artificial intelligence fields have also improved over time.

Computer Vision Applications

Some people believe that computer vision is the wave of the future in terms of design. Yet, the use of computer vision is ubiquitous. Some of the current applications of this technology are as follows:

Agriculture

Many agricultural organisations use computer vision to monitor harvests and handle typical agricultural issues, weeds and nutrient deficiencies. Through computer vision systems, it does help farmers scan photos from spacecraft, drones, or planes and seek to identify problems in the earliest stages and avoid unnecessary financial losses.

Augmented Reality

Augmented reality apps rely heavily on computer vision. For example, real-time detection of physical surfaces and objects in the actual world can be used by augmented reality (AR) apps to position virtual things in the real world.

Autonomous Cars

With computer vision, automobiles can better understand their immediate surroundings. The computer vision software in an intelligent car uses video feeds from various cameras to gather data. When the video feed is received, the system analyses it in real-time, looking for road markings, nearby objects (such as pedestrians or other vehicles), and traffic signals, among other things. Autopilot on Tesla automobiles is one of the best-known examples of this technology.

Facial Recognition

This technology enables a computer to match photos of people’s faces to their identities. Significant products we use every day have this technology built-in. For instance, Facebook uses computer vision to recognise individuals in photographs.

Biometric authentication relies heavily on facial recognition. Face unlocking is an increasingly common feature on many of today’s mobile devices. Through a front-facing camera, mobile devices may identify if a user is permitted to use a device by looking at their face and analysing their facial expressions. This technology has a lot of going for it in terms of speed.

Healthcare

In healthcare, medical diagnostics rely heavily on picture information and image categorisation because it accounts for 90% of all medical data. X-rays, MRIs, and mammograms, to mention a few, are all diagnostic tools that rely heavily on image processing. As for medical scan analysis, picture segmentation has proven its worth. For example, computer vision can analyse images of the back of the eye to determine whether or not a disease is detected. Diabetes retinopathy, the leading cause of blindness, can be detected using computer vision algorithms.

Another notable example is cancer detection. Accuracy in cancer diagnosis is essential. Using computer vision, Google claims it’s possible to detect cancer metastases more accurately than human doctors.

Most popular computer vision applications in the industry

We've compiled a list of industries' most common computer vision applications.

Pedestrian Detection

Pedestrian identification and tracking is a crucial computer vision research topic for designing pedestrian protection systems and smart cities.

It employs cameras to automatically identify and locate pedestrians in images or videos or object recognition while considering body attire and position, occlusion, illuminance, and background clutter.

Parking Occupancy Detection

Parking Guidance and Information (PGI) systems use computer vision to identify parking lot occupancy visually. It’s an alternative to expensive, maintenance-intensive, sensor-based solutions.

Camera-based parking occupancy detection systems have reached great accuracy thanks to CNN, largely unaffected by lighting and weather conditions.

Digital Pathology

When whole-slide imaging (WSI) digital scanners become more widely available, computer vision can interpret medical image data to detect and identify the type of pathology shown in medical image vision models.

What can it do?

  • Analysing and interpreting images

  • Examining tissue samples in great detail.

  • Matching pathology categories to previous cases

  • Accuracy and early detection are the keys to a proper diagnosis.

In digital pathology, with the aid of computer vision, doctors will save time and make better-informed judgments because of enhanced diagnostic accuracy.

Self-Checkout in Retail

Computer vision-based systems have made autonomous check-out possible by enabling computers to understand consumer interactions and monitor goods’ movement via visual data placed for specific object detection.

Intelligent Video Analytics

In the event of questionable behaviour, AI-powered systems can quickly identify it and notify the right individuals, who may investigate and take proper action.

Conclusion

As clear as the daylight, computer vision has the potential to be successfully used across a wide variety of business sectors, particularly those that are reliant on image and video data.

We can automate monotonous operations, acquire higher diagnostic accuracy, boost agricultural output, and ensure safety with the use of this technology.

We can anticipate that computer vision will continue to be the driving force that alters industries of all kinds due to the increasing number of businesses adopting the AI-first attitude.


explore more

Chart your career

Discover how you can chart your career and elevate your skills

Learn more

Jobs at NCS

Find available positions and join our multi-talented team.

Learn More

Talent Programme

Find out how you can advance your career in NCS.

Learn more

what are you looking for?

Contact Us

You can drop us a call or email

6556 8000
We endeavour to respond to your email as soon as possible. When sending in an enquiry, please fill your contact details and indicate the request purpose for our follow-up.

Thank you for your enquiry! We'll get back to you as soon we can.

Thank you for your interest.