A Military Vehicle License Plate Automatic Recognition System

With the construction of military mechanization and continuous in-depth information on a variety of new vehicles, troops and equipment included in the sequence of pairs of military standardization modern professional management and put forward higher requirements, the need to use high-tech means to conduct effective supervision. Military vehicle license is an effective authentication of military vehicles, we can by military license plates and identification, monitoring the operation of military vehicles to achieve the purpose of the management of military vehicles.First introduced the current vehicle license identification RFID system concept, significance and composition, and pointed out the possible security issues. To solve these problems is proposed based on digital image processing and pattern recognition of the military vehicle license plate recognition system, the system consists of hardware and software system components. From the high-definition industrial cameras, Gigabit Ethernet network cable and other parts of the image acquisition, transmission and processing hardware systems, ensuring access to high-definition license plate image.Secondly, in the software system structure, the system is divided into vehicle license plate location, character segmentation and character recognition in three parts. In the orientation section, describes the characteristics of military license plates; use of the image gray level, image enhancement technologies such as binary images of the license plate pretreatment. Presents a compact based on the ideas of regional and rectangle fitting method of vehicle license plate location. Set the ROI area, with an adaptive template width method, the license plate image convolution, and subtracted with the Original, to be characteristic map, horizontal scanning method, statistical frequency hopping black and white pixels to determine the upper and lower plate boundary . With rectangular fitting methods, the license plate of the candidate region.Once again, presents a character-based segmentation of the car brand of image pre-processing algorithm to enhance the image, and then carried out on the license plate tilt correction and removal of the upper and lower left border processing, the final binarization for licenses and to binarization car a unified brand image for the black white text, the character segmentation for follow-up has laid a good foundation. License plates is presented based on the proportion of features and a combination of the vertical projection of character segmentation method, using plates to determine the proportion of general characteristics of the scope of the partition, and then split the region from the combined value of the vertical projection to find the best split position. The creation of a cluster generalization of the characters in libraries, to split with the character after the character in the template library seeking Euclidean distance, and make judgments on the characters that do not meet the conditions for re-segmentation.Finally, in the license plate characters based on the image pre-processing, using the LIBSVM ways to bring the image data into text data as input samples, to avoid over-reliance on the extracted image features deficiencies. The results show that the nuclear function of different parameters and different parameters on the identification error penalty factor had a big impact the results, to discuss the optimal parameter selection method and verified with the experimental data SVM algorithm is suitable for small sample cases, the identification of unique advantage.Experiments show that the imaging system, a clear image capture, transmission speed, real-time strong, positioning, segmentation more accurate and robust stronger is better in light conditions, the recognition rate can reach 90%.