Multi-View Images Fusion Model

Abstract

The tasks of recognition actions and classification objects are fundamental in computer vision systems. Even subtasks, such as recognition of atomic motion and single objects form the basis for understanding the situation in the work area and the scene in general. This is especially important in video surveillance systems designed to ensure security. Thus, the effectiveness of recognition and classification methods is one of the primary tasks of computer vision. But the visual methods implemented in similar video surveillance systems, encounter some difficulties, such as inhomogeneous background, uncontrolled operating environments, irregular illumination, etc. To address these drawbacks, the paper presents a model for combining visible range images and depth images. This model allows to improve the quality of recognized images, provides the construction of a more informative descriptor, which also positively affects the recognition efficiency. Our results show that it has good performance in fusion visible image and depth map.