Sistema de visão computacional embarcado para previsão de trajetória plana e interceptação de projétil em tempo real

Abstract

Computer vision has become an essential technology in the development of autonomous systems, with significant applications in fields such as medicine, sports, robotics and agriculture. Due to advancements in computational power and energy efficiency in electronic components, data processing in embedded systems has proven useful in increasing the autonomy and decentralization of computer vision-based solutions. Among the tasks that this technology can perform, object identification and real-time trajectory prediction stand out. In this context, this work proposes the development of a physical system composed of an intelligent interceptor capable of blocking the trajectory of a projectile in a planar environment. Through the integration of computer vision and control algorithms, the system will be able to identify and predict the trajectory of a sphere, moving a linear actuator to the calculated collision point to prevent the projectile from advancing. The processing of visual data will be performed by a BeagleBone Black board, which will operate under the Linux operating system. The acquisition of sensory information and the sending of control signals to a direct current motor will be the responsibility of an ESP32. Thus, the project addresses multidisciplinary aspects, involving electronics, programming, electric drives, control, and artificial intelligence, providing a compact and efficient solution for high-precision, real-time response challenges. Initial tests were conducted on a personal computer and validated the proposed vision and trajectory prediction algorithms. When processing was transferred to the BeagleBone Black, a decrease in performance was observed due to its lower computational capacity; however, the system was still able to intercept the projectile in a satisfactory manner. The adoption of simple and robust techniques, such as geometric calculations, was essential to achieving accurate predictions. The results obtained were consistent with the proposed objectives.