Biomedical Engineering Reference
In-Depth Information
achieved by matching 3D model representations obtained from the visual object with 3D
model representations stored in memory. The 3D model representations obtained from
the object are formed by first identifying the concavities of the object, which separate the
stimulus into individual parts, and then determining the axis of each individual part of
the object. Identifying the principal axis of the object assists in the normalization process,
via the mental rotation that is required, because only the canonical description of the
object is stored in memory. Recognition is acquired when the observed object viewpoint
is mentally rotated to match the stored canonical description [64].
An extension of Marr and Nishihara's model, the recognition by components theory
suggested by Biederman [65], proposes that the visual information gained from an object
is divided into simple geometric components, such as blocks and cylinders, also known
as 'geons' (geometric ions), which are then matched with the most similar object repre-
sentation that is stored in memory to provide the object's identification (see Figure 10.7)
[65]. A small number of geons, appropriately arranged in space and relatively sized, is
proposed to be sufficient to compose any familiar object. The process of pattern recogni-
tion consists, in part, of extracting edges, determining geons from their spatial layout and
combining them into an object, then comparing that object to representations in memory
corresponding to object categories.
The basic ideas about edge extraction and combining the elements of object images
from edges are widely accepted in theories of visual pattern identification. Applying such
a model to haptic pattern recognition is problematical, however, because the haptic system
is simply not very good at extracting information about the spatial layout of edges. In a
haptic display, three types of edges might be presented: (i) 2D patterns in the range of the
fingertip scale, (ii) 2D patterns extending beyond the fingertip scale, and (iii) contours of
a fully 3D object. Examples of 2D representations are embossed numerals or alphabetic
characters and raised line drawings used as illustrations for the blind. The ability to encode
edges from patterns lying under the fingertip was systematically investigated by Loomis
[66], who evaluated the ability of both sighted and blindfolded people to identify letters
and digits that protruded 6 mm from the surface of the larger dimension.
Besides the kinesthetic and cutaneous groupings, haptics can be categorized as being
both active and passive, as shown by Gibson [67, 68]. Although other dividing lines exist,
it is these that are the two most important in which passive sensing concerns the analysis
Figure 10.7
How objects can be broken into geons
