A Multi-Style License Plate Verification Method Based on Character Classification under Illumination Variances

License plate verification is essential for locating the license plate region precisely in a vehicle image. Existing verification methods implemented after location are effective for the single-style license plates, but will not do well when distinguishing multi-style ones from complex backgrounds. We propose a method based on segmented character classification, combining Principal Component Analysis and Support Vector Machines, to deal with the verification of license plates with multi-style. Compared with previous methods, the segmented character regions can be analyzed more effectively, and the proposed method gets the high confidence level by calculating the classification result of the segmented characters but not the whole license plates. The experiments were done under complex illumination conditions, and the results show that the proposed method is effective for multi-style license plates for both fixed and moving camera systems.