Skip to main content

Table 11 Overall comparison of OCR stages or systems hardware implementations

From: Optical character recognition on heterogeneous SoC for HD automatic number plate recognition system

OCR algorithm

Hardware platform

Type of characters

No. of testing characters

Character size (cols×rows)

Recognition rate (%)

Processing time per character (ms)

BRAM

DSP

FF

LUT

[Proposed] template matching

Zedboard (Zynq-7000)

10 Arabic digits

2790

22 × 34

99.50

0.63

10

18-Kbit

20 (DSP48E)

4247

5616

[5]

ANN

Mentor Graphics RC240 FPGA

25 English letters and 9 Arabic digits

3570

22 × 34

97.30

0.7

4

18-Kbit

8 (DSP48s)

2711

6868

[12] Feature Extraction

Altera Stratix FPGA

10 Persian digits

100

10 × 7

100

47 ns

N/A

N/A

N/A

N/A

[13]

SVM

N/A

10 Arabic handwritten digits

120

16 × 16

83.30

N/A

N/A

N/A

N/A

N/A

[14]

SVM

TI C64 Fixed-Point DSP Platform

25 English letters and 9 Arabic digits

N/A

6 × 9

94.00

2.88

N/A

N/A

N/A

N/A

[17]

SOM-NN

Xilinx Virtex IV FPGA

22 English letters and 10 Arabic digits

8531

N/A

90.93

N/A

N/A

N/A

43,551

50,310

[30]

ANN

N/A

25 English letters and 9 Arabic digits

665

9 × 21

98.20

4.36 μs

351

9-Kbit

N/A

N/A

13,909