Recognition-by-components theory

The recognition-by-components theory, or RBC theory, is a bottom-up process proposed by Irving Biederman in 1987 to explain object recognition. According to RBC theory, we are able to recognize objects by separating them into geons (the object’s main component parts). Biederman suggested that geons are based on basic 3-dimensional shapes (cylinders, cones, etc.) that can be assembled in various arrangements to form a virtually unlimited number of objects.

The recognition-by-components theory suggests that there are fewer than 36 geons which are combined to create the objects we see in day-to-day life. For example, when looking at a mug we break it down into two components – “cylinder” and “handle”. This too works for more complex objects, which in turn are made up of a larger number of geons. Perceived geons are then compared with objects in our stored memory to identify what it is we are looking at. The theory proposes that when we view objects we look for two important components.

In his proposal of RBC, Biederman makes an analogy to the composition of speech and objects that helps support his theory. The idea here is that about 44 individual phonemes or "units of sound" are needed to make up every word in the English Language, and only about 55 are needed to make up every word in all the languages spoken on the earth. Though small differences may exist between these phonemes, there is still a discrete number that makes up languages.

A similar system may be used to describe how objects are perceived. Biederman suggests that in the same way speech is made up by phonemes, objects are made up by geons, and as there are a great variance of phonemes, there is too a great variance of geons. It is more easily understood how 36 geons can compose the sum of all objects, when the sum of all language and human speech is made up by only 55 phonemes.

One of the most defining factors of the Recognition-by-Components theory is that it enables us to recognize objects regardless of viewing angle, this is known as viewpoint invariance. It is proposed that the reason for this effect is the invariant edge properties of geons. The invariant edge properties are as follows:-

Our knowledge of these properties means that when viewing an object or geon, we can perceive it from almost any angle. For example, when viewing a brick we will be able to see horizontal sets of parallel lines and vertical ones, when considering where these points meet (co-termination) we are able to perceive the object.

Using geons as structural primitives results in two key advantages. Because geons are based on object properties that are stable across viewpoint ('viewpoint invariant'), and all geons are discriminable from one another, a single geon description is sufficient to describe an object from all possible viewpoints. The second advantage is that considerable economy of representation is achieved: a relatively small set of geons form a simple "alphabet" that can combine to form complex objects. For example, with only 24 geons, there are 306 billion possible combinations of 3 geons, allowing for all possible objects to be recognized.

...
Wikipedia