License plate recognition using complex network feature

This paper proposes a novel license plate recognition method based on complex network to improve the accuracy of character recognition under the interference environment. An adaptive multi-threshold method based on the block separation and the feature lines of the image is proposed for the fast image binarization. Then the characters segmentation is performed using region labeling algorithm and the priori knowledge of license plate. The skeleton of the segmented characters is sampled to decrease the data redundancy. The distance between the sampled point of the skeleton and the reference point is calculated. The difference of the distance of two points is used as the weight to construct a weighted dynamical complex network. The degrees of the complex network under the different thresholds can be used as the feature vector for preliminary character recognition. The contours of the characters with the similar skeleton are sampled as additional features to construct complex network, then, the refined recognition is conducted to improve the accuracy of the recognition. This proposed method is simple and faster for the license plate recognition. Meanwhile, it is scale invariant, rotation insensitive, and has strong anti-interference ability and robustness. The experimental results show that the effectiveness of the proposed license plate recognition method.