In the field of robotics perspective is employed to provide a kind of artificial visual perspective (2nd type) for the autonomous system; whereby the robot is to be imbued with powers of visual or spatial perception.
Computer Vision
COMPUTER VISION is artificial intelligence (AI) that uses computers to analyse visual data (e.g., views images of spatial reality) to identify and understand observed objects and processes correctly.
In some incarnations, computer vision seeks to replicate how humans see and understand the world. In line with our thesis, most human seeing involves the formation, measurement, or representation of perspective views of spatial reality, in which the key factor is extracting data and subsequently acquiring useful knowledge about the visual target(s). Note that robotic vision is a specific subtype of computer vision that deals with live image analysis and associated real-time decision-making.
Earlier, we defined the types of information a perspective image/view provides as perspective products, including object/scene location, size/shape, orientation, focus, arrangement, texture, and colour. Whereby certain perspective functions are enabled, including viewing (capturing, representing), matching (measuring, calculating, classifying), and representing (modelling) aspects of spatial reality.
Patently, such goals also apply to robotic or computer vision; however, these same top-level perspective goals are only a beginning and represent a brief glimpse into a highly complex topic. You only have to do a quick Google search to see that thousands of documents have been written on the labyrinthine subject of computer vision. One way to cut through this complexity is to say that a perspective image seeks to identify image properties containing type/kind and genus/species data for observed objects/processes.
Earlier, we defined technical perspective as any optical process that forms a detailed visual image, measurement, representation, model, or view of a 3-D object or scene. These outcomes relate to the products of perspective: attaining subsumptive data (whole/part), ordinal data (figure/size/scale/position), and/or determinative data (state/activity) from a perspective image and about a spatial reality.
So much for the general goals of computer vision, but specific systems require specific solutions and capabilities. Overall, we can identify two main types of computer vision, being live analysis of visual scenes to aid real-time machine and human decision making, and non-live or database image analysis, which is used as a feedback system for the same goal, but also for many other statistical analysis processes; whereby both have many of the same goals as identified above, yet are different in several distinct ways.
Computer Vision Methods / Systems
Today, many computer vision systems achieve remarkable results.
But how, exactly, does a computer vision system manage to succeed at interpreting the near infinite potential complexity that a perspective image represents?
The answer is that computer vision uses machine learning and artificial intelligence (AI/ML) to process data from devices like security systems, traffic cameras, cars, aeroplane landing and traffic control systems, and smartphones (and even data from 3-D simulations and computer games).
Computer vision is used for a variety of tasks:
- Object/process detection in images and videos
- Robotics: humanoid and manufacturing forms
- Real time spatial awareness: vehicle navigation
- Facial recognition in images and videos
- Image classification on camera databases
- Product search
- Content moderation
- Smart city and Internet of Things
Computer vision is applied to problems across many industries, including manufacturing, automotive, energy, and utilities. It can also be used in healthcare to improve patient outcomes, detect disease, and more. For example, computer vision can be used to analyse medical imaging data from CT scans, X-rays, and endoscopy cameras. It can also be used to remotely monitor patients by analysing their movement data.
Patently, computer vision is a subset of artificial intelligence that allows computers to analyse visual data and interpret objects and events in real-world images and videos, often using machine learning models. The application of perspective techniques to AI’s understanding of vision remains an open question.
You must be logged in to post a comment.