The urgency of the discussed issue is caused by the need to develop methods, algorithms and programs to ensure efficiency of car number plate recognition on images with a complex background. The main aim of the study: to increase the efficiency of character recognition on images with a complex background by developing methods, algorithms and programs invariant to affine and projective transformations of the input data. The methods used in the study: To solve the task the author has used the methods of the artificial Intelligence, identification and pattern recognition in images, theory of artificial neural networks, convolutional neural networks, evolutionary algorithms, mathematical modeling, probability theory and mathematical statistics with the help of software Visual Studio and MatLab. The results: The author developed the software allowing the recognition of car number plates on images with a complex background. The seven-layer convolutional neural network for character area selection on images is proposed. The algorithm based on the average pixel intensity histograms for individual characters selection is used. The six-layer convolutional neural network for character recognition on images is implemented. The represented software system can recognize license plates with deviation horizontally, vertically and in a plane angels. The system has high speed work.