Visual information
Vision for robotics

The aim of this section is to show how to exploit the information contained in the image. This can be achieved by using either sub-sections or the entire image. When sub-sections are used and characterized by their geometry (point, segment, circle, etc.), we speak of geometric primitives. When sub-parts are characterized by their intensity or color, they are referred to as appearance-based primitives. A primitive can thus be defined as a characteristic element found in an image. This element must be detectable, repeatable, precise and robust to noise.

Detecting and matching visual information is a major challenge in computer vision. They are essential for image processing and pattern recognition (faces, plants, animals, etc.). They also represent the most difficult stage in the process, although significant progress has been made, notably with neural networks. In...