MICROSERVICES ARCHITECTURE FOR MULTI-DIMENSIONAL FEATURE RECOGNITION USING A KNOWLEDGE REPRESENTATION MODEL, BOTH BASED ON HUMAN PERCEPTION

Abstract

Pattern recognition, semantic evaluation and classification in artificial vision are complex problems that are being tackled from a wide range of specific approaches. Most of these perspectives are based in the analysis of the information from a specific dimensional perspective (e.g. bi-dimensional images or video) considering a narrow set of indicators, and in the application of particular algorithmic techniques, with less or more success. This work presents a model intended to combine existing and future algorithms in order to evaluate visual information from a multi-dimensional perspective, inferring advanced properties and features by the distributed analysis of multiple source imagery, enabling the identification of environment elements in a similar way human perception works. After implementing a simplified version of the proposed model and executing it under a MPI cluster, low level features of test images are extracted and aggregated, and successful preliminary results are presented.