Computer vision
It extracts operational information from images and video, such as an element's presence, position, class, size or change. The answer has to fit its setting: a factory camera, a phone photo and a medical scan have different lighting, resolution and risks.
Text reviewed on
← Back to the 7 pillars
When it emerged
In 1966, MIT's Summer Vision Project proposed splitting an image into regions, describing shapes and identifying objects. The plan documents early computer vision research; it does not mean the field was solved that summer. Geometric and learning methods expanded what could be analyzed in later decades.
What it is
Analyzes images and video to locate, classify, measure or track elements. It can combine geometry, pattern recognition and machine learning.
How it works
A camera or file supplies pixels. A system may correct the image, locate edges and regions, detect objects or estimate motion; trained models can also learn features from examples. Evaluation must distinguish tasks: saying that a car is present, marking its position and tracking its movement are different problems.
What it is for
It extracts operational information from images and video, such as an element's presence, position, class, size or change. The answer has to fit its setting: a factory camera, a phone photo and a medical scan have different lighting, resolution and risks.
Where it is used
Visual inspection of parts, document reading, traffic analysis and support for reviewing medical images are examples. On a production line, a camera captures a part, the system flags a possible defect and a validation step decides whether to separate it.
Limits and safeguards
Finding a visual pattern is not the same as understanding the whole scene or producing a reliable diagnosis without validation. A detection does not prove complete understanding of a scene. Reflections, occlusion, images outside the training distribution and annotation errors can change the result. In health or safety, the system should support a validated professional decision.