ABSTRACT |
||
|
||
The theory of object recognition I am investigating proposes that recognition occurs in two ways. These two ways work in parallel but also interact at various stages. The first way is like focusing a telescope: First an object is just a fuzzy blob, and then as more information enters the eye, as the telescope is focused, more details are observed. The initial details will be things like the edges of the object (luminance and colour contrasts), and the distance of it (size and context). This pathway leads to the development of a theoretical 3D structural description of the object. This 3D structure remains theoretical as we can never see in 3D we just reconstruct it based on what we already know about reality. For example, we know there is a side to the object we cannot see, and we construct what that side looks like based on experience. Combining what we see with what we know provides the subjective experience of reality (thus no one can ever see the same reality… hymmm). The other way object recognition proceeds is through a comparison of what we know with what we see. Just like the theoretical 3D construction, we take the concepts (what we know) and overlay this on top of what we see to compare the two (“meeting”). But the concept also includes a structure, and so now we can compare the known 3D structure the object has with the theoretical 3D structure of the object. Our ability to find a match will primarily depend on our experience and on our memory. But other issues also relate to this match. First, the context in which an object is seen can either help or hinder its recognition. For example, seeing a car in a field or a tractor on a road will slow their recognition. Then, presenting an object in an unusual orientation will also slow or prevent recognition. For example, if objects within a category have features that allow each object to be distinguished, then presenting them in unusual orientations will increase the time required to find these features. And more generally, categorization of objects will be like a fractal: The category “vehicle” is made up of many different subcategories (certain features will need to be identified to categorise the object as a vehicle). The subcategory “car”, is made up of many subcategories. The subcategory “racing car”, is made up of many subcategories. The subcategory “F1 car” is made up of many subcategories, and so on until you get to a single car with features that separate it from all other vehicles. Presumably the brain could organise information like this and this will depend on experience and memory. And presumably reality is organised like this (i.e., a fractal). |