The Research On License Plate Detection And Recognition For Low Illumination Image

License plate detection and recognition technology is an important tache inintelligent traffic system, it plays an indispensable role in traffic supervision, trafficcontrol and vehicle management. The License plate and other important information arenot so distinct in the low illumination level images, so it is hard to detect and recognizedirectly. According to the problem, the dissertation proposes a method which could helpto solve the problems in License plate detection and recognition for the low illuminationlevel images; it enhances the image’s luminance and details, and improves the accuracyand applicability of License plate detection and recognition. The main works of thedissertation can be organized as follows:1. The features of the domestic License plate and the fundamental theory of itsdetection and recognition algorithm are introducted. It summarizes the general methodsof License plate detection, character segmentation and character recognition, andanalyzes the merit and demerit of some typical methods.2. A new JPEG image enhancement method based on Retinex theory is proposed toimprove the existing enhancement methods. The existing ones separate enhancementfrom compression for JPEG images, and emerge some block artifacts, the proposed oneembeds the JPEG image in the compression frame and enhances images beforequantization procedure to make full use of the original images’ information. DCTcoefficients are divided into the incident component (DC coefficients) and the reflectioncomponent (AC coefficients) based on Retinex theory, the DC coefficients are adjustedin the image dynamic range, and the AC coefficients are adjusted to enhance the detailswith enhancement factors of DC coefficients in the new algorithm. The block artifactsare suppressed by smoothing the DCT coefficients. Compared with other traditionalenhancement algorithms, the experimental results show the proposed approach is betterat details enhancement and color preserving, and sharing the same compression ratio withthe normal JPEG compression.3. The approach of JPEG compressed domain enhancement is applied to the Licenseplate detection in the low illumination image and a new approach for License platedetection is successfully designed and implemented. The approach can enhance theLicense plate information efficiently and restore its color feature. A rough location of theLicense plate in the image is made after enhancement by using Adaboost algorithm, and then the second precise location are made combined with the color feature. This approachnot only gets a high detection rate but also provides optimum conditions for characterrecognition. The experiments suggest that the proposed approach has well results forLicense plate detection and recognition in low illumination conditions.