The Recognization Of License Plate Based On EMD And Its Application

Intelligent Transportation System (ITS) is an inevitable development trend of the highway traffic management system while License Plate Recognize technology is one of the key technologies of the ITS. Empirical Mode Decomposition (EMD) is a data-driven adaptive decomposition method, which doesn't depend on the primary function. It decomposes a complex signal into a sum total of a series of Intrinsic Mode Function (IMF) and residual terms completely based on scale features of the signal itself. The paper applies EMD method to the license plate identification, which not only pours new vigor into the development of mature license plate recognition technology but also further expands the application field of EMD method. Therefore, it has some theoretical significance and practical value. The main task of this paper is as follows:Firstly, Based on the brief introduction of traditional non-stationary signal analysis, this paper discusses in detail the EMD theory and algorithm.Secondly, Since it is difficult to locate license plate quickly and precisely under complicated background and different illumination conditions, this paper presents an improved method based on texture analysis and projection method.Thirdly, Influenced by factors like mud, frame, rivet and plate slant, plate character segmentation tends to be inaccurate and even results in mistakes. In order to tackle these problems, this paper puts forward an improved plate character segmentation method based on projection feature and prior knowledge.Fourthly, Owing to noise jamming, all the segmented characters are not standardized ones. According to the characteristics of the noise with high-frequency information and self-feature of IMF, the integral projection wave of the characters will be decomposed and high-frequency information will be regarded as the denoising feature vector. At last, the characters will be resonstructed by denosiong Gabor filter of the vector.Fifthly, the layer and outline of the characters can better reflect their complex space structure. The writer of this paper disintegrates the layer and outline of the characters by utilizing good time-frequency localization analysis ability of EMD and regard IMF as the feature vector of the characters. In the end, the writer will adopt minimum distance classifier to realize character recognition.Finally, the writer gives out the license plate recognize prototype system and applies license plate recognize technology to the highway overloading detection that is helpful for improving overload management intelligence.