Paraconsistent artificial neural networks applied in the optical monitoring of goods

Abstract

This work presents the use of the Paraconsistent Logic (PL) combined with techniques of the photogrammetry aimed at processing images to obtain data of interest in a production line quality control scenario. Photogrammetry has been successfully applied in monitoring structures while PL, in turn, has been applied in process control circumstances. Since PL considers uncertainty when performing calculations, it is useful in scenarios where measurement plays a key role. Here, the steps for assembling and executing a logical structure denoted the Paraconsistent Convolution Unit (PCU) are described. PCU performs transformations in images and extracts properties, used by Paraconsistent Artificial Neural Cells for classifying the objects present in the images and identifying deviations in their dimensions. Results were
encouraging, presenting a decrease of the variance in the frequency distribution of grey levels in the image 13 times higher than that achieved by the use of a standard programming library, the realization of measurements on objects with an observed maximum error of 1.77% in relation to the expected theoretical value, a 100% accuracy rate in the object classification process with the use of low computational-intensive algorithms and an ability to identify deviations in dimensions of the object as low as 1 mm.