Energy Education Science and Technology Part A: Energy Science and Research
2014 Volume (issues) 32(4): 2189-2196
Affine invariant locally linear embedding
for license plate character recognition
Boyu Gu1,*, Shuli Huo2, Zhenhuan Zhao3, Qiang Zhang3, Yechun Li4
1Changchun University of Science and Technology, School of Electronics and Information Engineering, Changchun 130022, China.
2Seagate Technology International (Wuxi) Co. Ltd. Wuxi, 214000, China.
3Continental Automotive Corporation (Lian Yun Gang) Co. Ltd. Changchun Branch. Changchun 130033, China.
4Jilin Agricultural University. Changchun 130118, China.
Received: 02 Febuary 2014; accepted: 05 April 2014
Locally linear embedding is an efficient manifold learning approach, for appling which to the practical application of license plate character recognition, and adapt to the images with interferences of all kinds of affine transformations, an affine invariant locally linear embedding algorithm is proposed. The distances of image samples are described by tangent distance instead of Euclidean distance, so that the influences of affine transformations can be eliminated. Also, in the sample space, the neighborhood number of each point can be determined automatically based on the similarity of images. Experimental results show that the manifold structure is created more reasonably, and the recognition rate is improved.
Keywords: License plate recognition; Character recognition; Manifold learning; Locally linear embedding;Affine invariant