Computer vision - Cameras

Cameras are built trying to imitate how eyes work. Light comes through an opening, a pinhole or a lens, and hit an image plane where some photosensitive elements detect it and mesure it.

• Lens provide more light at the expense of having objects out of focus – depth of field.
• The image plane is divided in pixels, which form the final image that has to be processed.

The reduction in hardware and improvements on software have made cameras a common sensors on the robotics field 20210602191112.

Color representation on cameras

The most common way of representing color is RGB, representing color on 3 planes (red, green, blue). However, it has two disadvantages: it is sensitive to light changes caused by movement and position on the environment, and also the devices used to capture color are more insensitive to the red plane than the other two.

HSI measures the Hue (dominant wavelength), Saturation (lack of whiteness) and Intensity (quantity of light received). This deals with some of the problems of RGB, but are usually more expensive and require more complex algorithms.

Notes References

20210709195051 INDEX - Computer Vision

20210602191112 Cameras and Computer vision in Robotics - Basics

References

(Murphy 2000)

Murphy, Robin. 2000. Introduction to AI Robotics. Intelligent Robotics and Autonomous Agents. Cambridge, Mass: MIT Press.