License Plate Recognition Based On Spline Template Filtering And Double Function Method

License Plate Recognition (LPR) confirms vehicles'identity by recognising license plate automatically. It is a computer visual system that aims at given object. This system can abstract license plate image from a picture automatically, segment plates automatically, and then recognise characters. It is a key technology of Intelligent Transportation Systems (ITS). And it has been widely used in Electrical Toll Collection, City Traffic Monitoring and Controlling System and Special Vehicle Identification System. Scholars all over the world have researched into this area for dozens of years, and applied image engineering, pattern recognition, artificial intelligence, computer vision, computer graphics and relevant mathematical tools to vehicle plate location, character recognition and some other related aspects, and make great improvement both in theory and practice.First step of License Plate Recognition is to obtain clear vehicle images, and remove the disruptions caused by light, shadow and weather; second is to find out the plate position from the vehicle image; third is image preprocessing; fourth step is plate segmentation and character recognition. Article details the hardware component, software structure and treatment scheme of License Plate Recognition from the 4 aspects mentioned above. For some key algorithm, in-depth and detailed research is conducted to improve processing effect and Recognition rate.Plate image preprocessing method is demonstrated in detail. The comparison result of Multi-Resolution Analysis (MRA) and spline function convolution filtering shows that although MRA with Daubechies wavelet (n=4) has better filtering effect than Haar wavelet, they have same defect that the low frequence part present squareness, and this will make difficulties during image binarization segmentation. The low frequence part obtained by templet convolution has no squareness and brightness changes smoothly. The quality of Image Binarization Segmentation is much better than wavelet filtering based on MRA. So spline function collection is researched and a new spline function collection is brought forward which can approach gaussian function infinitely. Consequently, the relationship between spline function and gaussian function is illuminated in theory. At the same time the simplified algorithm is derived and the computation is cut down greatly.Double function method is put forward. The main idea is that a moving pixel slide along the border of the 2-D object in the image, then the projection on the x-axis and y-axis become the functions of time. So all the curves can be transformed to functions with the time as independent variable. With double function theory, processing method of object borders and curves are greatly extened. Fourier transform, wavelet analysis and other methods can be applied to object border or curve processing. The character abstraction of letters and numbers is based on double function theory and provides good results.Masking technology is employed to open a specific window for automobile detection. It substitutes for conventional sensors. So the hardware is simplified, and this technology describes detection window much more accurate then conventional virtual coil which composed of rectangles, trapezias and lines. The influence of complex background such as trees, grasses is smoothed away, and false detection is reduced.Some conventional algorithms are improved. For example the substitution of Link list structure for recursion algorithm reduces system resource expenditure. An improved rotation and shear algorithm is applied to remove serration texture. 8-neighborhood judgement is put forward to take place of morphological operation. It keeps the basic strokes'of characters while remove noise with one pixel width effectively. This avoids the Excessive influence to thin strokes by morphological operation.License Plate Recognition implementation is introduced, including image data format, License Plate Recognition System survey, operating platform, development environment and relevant software tools. And then the Recognition result of License Plate Recognition System is given.Finally, the progress and creative research are summarized. And also look foreward to the future of the License Plate Recognition technology.