Pipeline Architecture for High Speed License Plate Character Recognition - Advances in Image and Graphics Technologies - page 77

q

1



d = E

T

(I − μ ) − E

T

( I − μ )

ik test, k ijk

q j = 1

(9)

q

1



= E

T

I − E

T

I = E

T

I − γ

test, k ijk test, k ik

q j = 1

In Equation (9), µ is removed, and γ ik is a constant which can be obtained in training
phase because of all needed variables are available off-line. In the proposed
architecture, all operations in training phase are run on a PC. And in Equation (7),
since s is a constant, it can be ignored in hardware Implementation.

3.2 Structure of Processing Elements

The core of proposed hardware architecture is the processing elements. The projection
element is designed to project a sub-block of input character into eigen space, which is
essentially a matrix multiplier. The function of distance element is computing the
Euclidean distances between input character and each class in training database. And
nearest neighbor classification is carried out in the classification element. The structure
of processing elements is shown in Fig. 1. The k th sub-block is described in detail, and
others with same structure are omitted.

Fig. 1. Structure of processing elements