License plate recognition method and license plate recognition system

Abstract

The invention is applied to the technical field of multimedia (especially to the field of image processing), and proposes a license plate recognition method and a license plate recognition system based on multi-classifier selection. The license plate recognition method comprises the following steps: extracting the features of a data sample and a to-be-tested sample, and training the samples to get multiple single classifiers; selecting neighbors for the to-be-tested sample, and classifying the to-be-tested sample and the selected neighbor training samples; performing multi-classifier selection through weight calculation of the neighbor training samples; and classifying the to-be-tested sample based on multi-classifier weighted fusion. Through the steps, the weights of the multiple classifiers are changing dynamically for each to-be-tested sample. Therefore, multi-classifier classification fusion based on the method is of high adaptability, and the precision of classification is improved.

Description

technical field [0001] The invention relates to a computer vision recognition technology, in particular to a license plate recognition method and system based on multi-classifier selection. Background technique [0002] License plate recognition is an important aspect of the application of computer vision in the engineering field, and plays an important role in vehicle and traffic management. The essence of license plate recognition is the classification of license plate characters. Now there are many classification techniques, such as decision tree method, fuzzy classification method, artificial neural network method and support vector machine classification method. However, the recognition performance of a single classifier is always limited, and the classification performance of the same classifier for different samples is also different. Make full use of the classification performance and applicability of each single classifier for different samples, and adopt the method...

AI Technical Summary

Problems solved by technology

[0005] The weighting method based on the mean value assigns the same weight to each single classifier. Obviously, its disadvantage is that it cannot reflect the performance difference of each single classifier, so that it cannot make full use of the classification advantages of each single classifier.

Images