МОДУЛЬ КОРРЕКТИРОВКИ ПАРАМЕТРОВ АЛГОРИТМОВ АВТОМАТИЧЕСКОГО ОБНАРУЖЕНИЯ И СОПРОВОЖДЕНИЯ ОБЪЕКТОВ ДЛЯ ОПТИКО-ЭЛЕКТРОННЫХ СИСТЕМ

Abstract

In order to create an innovative module for automatic correction of algorithms for automaticdetection and tracking of objects with real-time training, a study of world experience in the fieldof general-purpose automatic tracking with the ability to recognize the tracking object for use inembedded computing devices of optoelectronic systems of promising robotic complexes was carriedout. Based on the conducted research, methods and approaches have been selected and testedthat allow with the greatest accuracy, while maintaining high computational efficiency, to provideon-the-fly training of classifiers (online learning) without a priori knowledge of the type of tracking object and to ensure subsequent correction during tracking and detection of the original objectin case of its short-term loss. Such methods include a histogram of directional gradients – a descriptorof key features based on the analysis of the distribution of brightness gradients of an objectimage. Its use allows you to reduce the amount of information used without losing key dataabout the object and increase the speed of image processing. The article substantiates the choiceof one of the classification algorithms in real time, which allows solving the problem of binaryclassification - the method of support vectors. Due to the high speed of data processing and theneed for a small amount of initial training data to build a separating hyperplane, on the basis ofwhich the classification of objects takes place, this method is chosen as the most suitable for solvingthe task. To implement online training, a modification of the support vector machine was chosen,implementing stochastic gradient descent at each step of the algorithm – Pegasos. Anotherauxiliary method is the clustering method of key points – this ensures an accelerated selection ofobjects for classification and training. The authors of the study carried out the development andsemi-natural modeling of the proposed module, evaluated the effectiveness of its work in the tasksof correcting and detecting the object of interest in real time with preliminary online training inthe process of tracking the object. The developed algorithm has shown high efficiency in solvingthe problem. In conclusion, proposals are presented to further improve the accuracy and probabilityof detecting an object of interest by the developed algorithm, as well as to improve its performanceby optimizing calculations.