Search and selection of license plate images in TV systems with video analytics

Algorithms for searching and extracting images of state vehicle license plates from video sequences are discussed. The most common is the search algorithm, in which the image of licence plate is considered as a connected closed foreground area with the specified sizes. This algorithm uses the edge detector Kanny to select the boundaries of objects. The use of this algorithm leads to errors with poor image quality. Errors are manifested in the presence of significant boundary breaks around the license plate. The use of morphological processing methods to eliminate errors leads to the complexity of the algorithm. The algorithm proposed by the authors does not search for the number plate, but directly for the symbols of the state license plate, which are then used to localize the location of the number plate in the image. In this case, the object of interest is considered as a chain of symbols with known characteristic features. At the first stage, the applicant areas for symbols of the licence plate are found. At the second stage, a mask is formed to select the image of the licence plate. The results of modeling algorithms using MATLAB are given.