Development of sequential optimizational algorithms for object detection in images

Development of high quality object detection system is a challenging task, not fully solved nowadays. The relevance of this study is stipulated by the necessity of designing techniques, algorithms, and programs improving the efficiency of automatic objects detection on images with complex backgrounds. Purpose: The aim of this work is to improve the efficiency of automatic number plate detection on images with complex backgrounds using methods, algorithms, and programs invariant to affine and projective transformations. Findings: the problem of detection or detection of the number plate of the vehicle images can be effectively solved using CNN algorithms based on the adaptive boosting. Two convolutional neural networks (CNNs) with different configurations are designed. The first convolutional neural network (CNN) provides the preliminary plate detection while the second provides its final detection so as to compensate classification errors received by the first CNN. As a result, the optimally efficient training algorithm has been selected. The software system based on these algorithms is suggested to provide the high-efficiency automatic plate detection.