Detection and matching of object using proposed signature - Emerging Trends in Image Processing, Computer Vision, and Pattern Recognition - page 497

FIGURE 2 Images of famous regular shapes in (a) and from (b) to (f) other types of different
objects.
In Figure 3 , all objects are individually be defined, detected, and matched by its signatures
in the proposed algorithm.
FIGURE 3 Different objects and their signatures.
It is interesting to note that the similarity is cleared in signatures for the stand ellipse (its ma-
jor axis parallel to the y -axis) and horizontal one (its major axis parallel to the x -axis) because it
is the same shape but different position. Evidently, the square shape has four identical peaks
in its signature because the equality of its sides; furthermore, the circle's signature is one-line
parallel to the x -axis and far away its radius length. On the same way, many objects' signa-
tures are nearest to each others, for example, the object (1, 2) (i.e., in the row 1 and column 2 in
Figure 3 ) is closer in signatures with objects in cells (9, 2), (9, 4). In the same context, signatures
of objects (1, 3) and (5, 4) are seemed to correspond to some. These last cases are happened
because the shape's nature of the original objects not because mistakes or big errors in the pro-
posed algorithm.
Actually, the proposed algorithm has applied on about 120 different shapes, positions, ori-
entations, and intensity luminance of objects in RGB images. Furthermore, signature's determ-
ination for all objects has achieved 100% without any errors (in data, or wrong signature con-
struction) for all objects.
Second part of the proposed algorithm is matching of input and saved signatures. Table 1
presents the matching process that is depending on Equations (4) and (5) ; the decision in this
process is based on the least value in Equation (5) . Obviously, Table 1 shows all input objects
images in first row. Regularly, the first column represents all positions (1-11) of objects in their
main image if are scanned from left to right. Sequentially, Table 1 consists of x rows and y
columns, which contain a set of values, represent the smallest value of errors calculated by
Equation (5) . For example, in the cell (1, 1), the proposed algorithm is selected a least value
(0.1097) in a row (1), which indicates to the first object in the image. Clearly, this value in-
dicates to the exact object position selected in the main image of Figure 2(b) . In this case, the