Parameter estimation for digital images, using K-NN and Tesseract classifiers

Abstract

Introduction: For this study, there is an estimation of parameters from a comparison between two algorithms for the recognition of numerical characters in images: K-NN and Tesseract, in order to determine the one with the highest degree of similarity. Methodology: The inductive and experimental method was used to acquire information and data such as: precision, recognition time, percentage of consumption of RAM and CPU memory. This research is of a quasi-experimental type due to the techniques chosen for the recognition of the digits applied to images and later to evaluate in K-NN and Tesseract electric energy meters captured in photography to obtain an automatic consumption reading. The research is of an applicative type since it was based on existing knowledge from previous research aimed at technological development to improve new processes. It can also be taken as experimental by the acquisition of data through laboratory tests where important elements can be appreciated and a simple view a capture of the phenomena of the case. Conclusion: Through tests to determine character recognition using the K-NN and Tesseract algorithms, the precision estimation results of 439.3% were obtained with the K-NN algorithm and 29.34% with Tesseract using a time average of 1.2 and 0.06 seconds in each algorithm.