Information Technology Reference
Fig. 8. ROC curves of ALPR and
HelpMe ( k =2,3,4)
Fig. 9. Time consuming of HelpMe
( k =2, 3, 4)
than ALPR. And among ROC curves of HelpMe, HelpMe with k value equals 4
has the highest quality of veracity, and HelpMe with k value equals 3 is just a
little inferior than that with k value equals 4.
All in all, HelpMe provides a better performance in veracity than ALPR. Thus
our method would provide a higher veracity to satisfy the requirements of high
level trac applications in ITS.
Time Consuming Evaluation. Time consuming factor mainly indicates the
eciency of HelpMe.
ThetimeconsumingofHelpMewith k value equals 2, 3, 4 is shown in Fig.
9. Experiments with k equals 2, 3, 4 have been conducted 10 times respectively.
As far as the scale of current data set, the higher k value of HelpMe, the time
consuming becomes larger. It is easy to see that all cases with different k values
are capable of meeting the requirements of upper applications of ITS. With
consideration of all conditions, HelpMe with k equals 3 consumes less time than
that with k equals 4 and has higher veracity than that with k equals 2. Therefore,
HelpMe with k equals 3 is more comparable than those with other k values in
Generally speaking, comparing with different k values, HelpMe has a stable
time consuming. Moreover, HelpMe on HANA cluster has satisfactory perfor-
mance on eciency.
5 Related Works and Comparison Analysis
Most methods of license plate correction focus on ALPR algorithm optimiza-
tion. And there have been many automatic license plate recognition (ALPR)
methods developed in both academia and industry. In , the author shows that
license plate algorithms in images or videos are generally composed of following
three processing steps: 1) extraction of a license plate region; 2) segmentation
of the plate characters; and 3) recognition of each character. In , a license