Download as doc, pdf, or txt
Download as doc, pdf, or txt
You are on page 1of 3

Q. What is Recognition- by- Components Model? How is it different from Feature - Matching Model?

Each and every century has its own characteristics in scientific and technological development. The first
phase of twentieth century may be regarded as the era of Electronics and Electrical
development,whereas the second phase showed the marvels in development of Electronics and
Telecommunication. The birth of Computer Science was the most remarkable happening of this phase.
In the last part of 20th century, we observe an unprecedented growth of computer industry,
unbelievable advancement in research, that made a dramatic change in the technological scenario.
Around 1950, a renovation occurred in psychology, anthropology, linguistics; computer science and
neuroscience that got unveiled as a new theory resulting the emergence of cognitive science, a
combination of psychology, linguistics, computer science, neuroscience etc.
Human being has the lust to know the unknown, see the unseen. But it is not enough, rather it is
impossible to experience the whole empirical world only through visual perception. Human being
perceives the world, collects and stores experiences in memory and recognises a similar object as same
as the stored object through cognition. 'Recognition' is the process of matching representations of
organised sensory input to stored representations in memory.
So, in psychology and cognitive neuroscience, recognition describes cognitive process that matches
information from a stimulus with information retrieved from memory. Recognition occurs when
information from the environment is received and entered into short-term memory, causing automatic
activation of a specific content of long-term memory.
But sometimes human cognition fails in recognising a object. Though it seems simple, even with an
intact brain, recognition is not a trivial act. It remains difficult for even the most sophisticated computer
programs. Making the connection between memories and information perceived is a step of recognition
called identification. Recognition requires repetition of experience.
There are four main theories of recognition: template matching, feature matching, recognition-by-
components theory and configural model.
Now, we will discuss on Recognition-by-componentsmodel or RBCmodel.
The recognition-by-components theory, or
RBC theory, is a process
proposed by Irving Biederman in 1987 to
explain object recognition. According to this theory, we are able to recognize
objects by separating them into geons (the
objects main component parts).
Biederman suggested this method for recognising three dimensional objects accross the variation in
viewpoint. According to this model, a set of 24 basic geometrical three dimensional shapes like
cylinders, cones etc can be assembled in various arrangements to form a virtually unlimited number of
objects. This model proposes that any three-dimensional object can be described by its components and
their spatial relations. So, observing a part or geon we can recognise a object.
RBC model depicts that a object can be specified by its structural description. In observing a object, we
should focus on two components:
Edgeswhich enables us to maintain the
same perception of the object
regardless of viewing orientation.
Concavities,the area where two edges
meet.
Suppose, a mug is a combination of two components : a cylinder and a curved handle on the side. In
case of bucket, we can break it down into two parts: a cylinder and a curved rod on top.
So basically, these two components i.e. edges and concavities enable us to observe the separation
between two or more geons.
In 1992, Hummel and Biederman included that the geons and their soatial relation try to match their
assembled parts to a stored three dimensional representation of a known object. Perceived geons are
compared with objects in our stored memory to identify the object which is being observed.
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. Geons, the most useful units, play a vital role to describe objects as their
properties are viewpoint independent. The invariant edge properties are as follows: Straight lines,
Corners, Curvature, Parallel lines, Symmetry and asymmetry, Co-linearity etc.
It is proposed that the reason for this effect is the invariant edge properties of geons. Each geon is
associated with a set of viewpoint - invariant properties which uniquely specify it from other geons. So
the structural description of an object is viewpoint invariant even if the viewing condition is totally
changed.
In experiencing this model participants can easily recognise geon renditions of man made objects.
Infact,experimental evidence comes from the use of visual priming which produces faster recognition
when an object is observed for second time. Using this technique, in 1995, Biederman created
complementary pairs of images of a given object with some contours deleted. Each image in a pair had
half configuration of the whole and the two images had no contours in common. A second pair of
contour-deleted images presented the same object, though it was different in shape and contained
different geons.Here, participants were shown on member of a pair then one of another pair(same
object with different geons). Recognition was faster when the second image presented had the same
geons as the first.
The main strength of this model is its simplicity. As geons are based on object components that are
stable across viewpoint (viewpoint invariant), and all geons can be differentiated from one another, a
single geon description is sufficient to describe an object from all possible viewpoints.
But this theory is not free from weakness.
This model can be applied only on man made objects. It is less clear about its application on natural
objects like animals or plants. RBC model fails to detect individual differences between different faces.
As faces generally includes two eyes, nose and mouth in the same arrangement, RBC model would
construct the same arrangements of geons for every faces. Though this model is good in finding the
most commonly used categories of an object, it gets trouble in identifying a specific or a special
exemplar. RBC theory is not in itself capable of starting with a photograph of a real object and producing
a geons-and-relations description of the object; the theory does not attempt to provide a mechanism to
reduce the complexities of real scenes to simple geon shapes. RBC theory is also incomplete in that
geons and the relations between them will fail to distinguish many real objects. For example, differences
between two people or the difference between a pear and an apple, which are easily distinguished by
humans, RBC theory lacks to recognize their differences.
Now we would focus on the differences between RBC model and feature matching model.
As the former works on three dimensional objects, the latter focuses on two dimensional objects.
We can recognise objects through features in feature matching model and in RBC model, we recognise
objects by its geometrical components named geons.

RBC model takes less space in machine than the feature making models.
RBC model is also a feature analysing model but it is exclusive in its idea of geons where as the feature
matching model works with line segments. For example, in feature matching model, letter "A"can be
recognised through three features: left slanted line, right slanted line and horizontal line. On the
contrary, in RBC model, a mug can be recognised by two geons: a cylinder and an curb on the side.
RBC model works on different paradigm of man made objects through geon. Feature matching model
can not be applicable in different exemplar.
Feature matching model fails to distinguish objects with the same component features but arranged in
different spatial relationship whereas RBC model is more flexible and fairly successful at recognising
objects in a specific category as it focuses on the components of the object and their spatial relations.
However, Irving Biederman has argued that RBC theory is the "preferred"mode of
human object recognition, with a
secondary process handling objects that
are not distinguishable by their geons. Shortcomings of this theory allows the configural model as a
theory of recognition.

You might also like