Sistema de detección y reconocimiento de intrusos de bajo costo basado en visión artificial, aplicado a sistemas de videovigilancia

This degree project deals with the construction of an intrusion detection system based on artificial vision, applied to video surveillance systems, aimed at solving the limitations of motion and presence sensors, in addition to establishing a versatile, efficient and low-cost solution. For its development, a bibliographic review was used to obtain an adequate detection model, in addition to acquiring the necessary knowledge to establish the correct parameters and generate a software capable of detecting human presence given a specific environment. Consequently, the necessary algorithms were built to obtain a set of data that gave way to the training of the neural network by applying criteria that allow increasing learning information, generating an efficient solution capable of differentiating the human form from a scenario provided by a CCTV system. In addition, a mobile application was designed in conjunction with a web API to provide the ability to deploy the solution to a realistic environment allowing anyone who requires its access to monitor the situation of the monitored environment. WebSockets were used as the data transmission protocol, which allowed real-time updating of the information displayed in the application using single execution threads, avoiding adding process overload to the system. Thus, once the different algorithms were generated for the correct operation of the system, a performance test was established in different monitoring scenarios to measure its effectiveness, which showed that the final product is suitable for production deployment. Keywords: neural networks, digital images, digital processing, web services, mobile applications

• Hovedforfatter: Ortega Jaramillo., Jorge Enrique (author)
• Format: bachelorThesis
• Sprog: spa
• Udgivet: 2022
• Fag: <REDES NEURONALES>< IMÁGENES DIGITALES> <PROCESAMIENTO DIGITAL><SERVICIOS WEB> < APLICACIONES MÓVILES>
• Online adgang: https://dspace.unl.edu.ec/jspui/handle/123456789/25093