Abstract:
This paper proposes a learning-based algorithm for real-time license
plate detection. Two kinds of features, statistical gradient features and Haar-like
features, are used in the algorithm. Firstly, two statistical features are extracted
from vertical gradient images. Classifiers based on these two features are
constructed through simple learning procedures respectively. Using these
classifiers, more than 80% of background area can be excluded from further
training or testing. Then the AdaBoost learning procedure is used to build up
the other classifiers based on selected Haar-like features. Combining the
classifiers using the statistical features and the Haar-like features, we obtain a
cascade classifier which can real-timely detect license plates from various
complex backgrounds. In the experiments, high detection rate and low positive
false rate are achieved when the algorithm is used to detect license plates from
images taken in various complex environments.
