The theoretical foundations of building decision-making systems based on the results of image recognition accompanied by texts are considered. The approximate structure of the image recognition system is given. The basis of image recognition systems is the selection of text inscriptions on existing photos, their pre-processing, selection of isolated areas on the image, performance of mathematical operations on individual groups of pixels to bring them to known forms and comparison with them. The description of various methods of image preprocessing is performed. An analysis of the feasibility of using such methods of image binarization as adaptive Bradley-Roth binarization, median filtering, Gaussian filtering, methods of balanced histograms and class variances, discriminant analysis, logistic, probit regression, etc. was carried out. Different algorithms for dividing the image into separate areas for the purpose of their further recognition are considered. among them the moving average algorithm, the algorithm for estimating the probability of finding an object in a selected area based on boundary analysis, Category-independent object proposals, Constrained Parametric Min-Cuts, Mu
📖 افتح في inklap 🔗 DOI 📮 اطلب بحثاً