Algorithm for the Camera Scene 3D Model Calculation

Modern 3D reconstruction algorithms offer a wide range of methods and tools for building a model of the surrounding space. Most of the approaches require the use of additional equipment, such as: lidars, laser scanners, stereo cameras. At best, a video camera with high-definition resolution is required. The cost of auxiliary equipment limits the availability of these methods. In addition, using 3D reconstruction approaches for the task of obtaining a video surveillance scene leads to the problem of determining the location of the camera in the resulting model. To avoid this problem, the spatial information obtaining by the surveillance cameras itself is suggested. Since the cameras using in surveillance are usually not a stereo pairs and in a worst scenario placed without overlapping, the suggested approach is to extract the spatial information during the calibration. This article presents an algorithm for obtaining a 3D model of the camera scene using the reference points acquired during calibration.