Sistema de monitoramento de segurança para canteiros de obras

Abstract

The use of monitoring and surveillance systems in public, residential and commercial environments has already been validated as a tool for reducing crimes and assisting in their investigations. However, an environment that has not yet been reached by these systems is civil construction site, which presents challenging characteristics for the implementation of current systems. Therefore, in this work, we present a study of the main technologies used in the area, and we design and develop a system, as proof of concept of the chosen technologies, for remote monitoring of small and medium-sized civil construction sites. The monitoring system consists of three main parts: computer vision, data transmission, hardware and mobile application. The computer vision and data transmission parts were studied, tested individually, and integrated into the hardware and application. In computer vision, two approaches were studied as a possible solution: Histogram of Oriented Objects (HOG) and Artificial Neural Networks, with the performance of the first surpassing that of the second in all aspects specified in the project. The final precision of the model trained to detect people during the day reached 68.02%, during night reached 71,56%, whereas the precision for the detection of trucks during the day was 79,84%. In the data transmission part, four approaches were studied: Wi-Fi, 3G, 4G and LoRa (Long Range), and, after carrying out the comparative study, it showed that despite all having some limitation, the one that best suits the proposed system is 4G. Furthermore, the hardware project was created with the aim of lower battery consumption and better system performance in mind. Ultimately, once all parts are integrated, the system is capable of identifying elements of interest through a camera and sending images and other data to the application on the user’s cell phone.