Diseño de un prototipo de entorno de video vigilancia integrado Open Source utilizando cámaras integradas en vehículos.

Abstract

The objective is the development of a video surveillance system using artificial vision for the automatic detection and real-time monitoring of pedestrian flow in the parking lot of the Faculty of Informatics and Electronics, using the experimental scientific method since there is manipulation of the independent variable directly and indirectly. A database containing four characteristic images of the system was created. Subsequently, an evaluation of content extraction methods using CBIR such as SURF and SIFT was performed, which allows detecting points of interest of an image invariant to scale, rotation and illumination change. Finally, based on the detection and description of characteristic points, the images that have a portion in common were combined, for which the keypoints of both were located and a panoramic image was obtained. It was determined that the implementation of the system has an efficiency of 89% using the Surf algorithm with respect to 56% of the Sift algorithm. It is concluded that the open source video surveillance system based on cameras integrated in vehicles did improve the location of the location and greater visual coverage for the detection of incidents in the parking lot of the Faculty of Informatics and Electronics, using the Surf algorithm. It is recommended to perform a combination of feature extraction systems, as this will allow building more robust image systems.