Increasing the operating efficiency of sorting robotic complexes based on multi-projection processing

Abstract

Introduction: this paper explores computer vision techniques for automated sorting of objects based on their geometric shape, color, and brightness. The research addresses two primary scenarios: objects moving along a conveyor belt and objects placed unordered in a common container. Method: the sorting system utilizes computer vision algorithms that incorporate edge pixel extraction, cellular automata, and the Radon transform. Edge detection is achieved using the Prewitt operator to extract object contours. Cellular automata are employed to generate object backgrounds and define polygonal regions, improving shape recognition. The Radon transform is applied with a hexagonal image grid to produce six projections, aiding in noise reduction and accurate shape and orientation detection. Results: the combined use of six Radon projections and cellular automata enables the system to distinguish individual objects even when they are placed together in a single container. The approach effectively detects and sorts distorted or variably shaped objects with high precision, regardless of their arrangement—either random or orderly. Conclusions: the proposed computer vision-based sorting method is robust and versatile, capable of handling complex object configurations. It offers a reliable solution for sorting objects by shape, color, and brightness in diverse industrial or logistical settings.