Projeto e implementação de um robô de precisão com visão computacional baseada em algoritmos de inteligência artificial

Abstract

This work aims to design, build and program a Sniper Robot with two degrees of freedom for the vision of two cameras. A sniper robot has similar characteristics to the sniper observer team, used widely by all armed forces in the world. The first is dedicated to information from the field, such as the location of a target and its distance. The second tries to triangulate the variables of the field with the gun and shoot the target. These combinations of elements are created by the proposed robot, with each camera being a part of the observer-shooter team. The logic of this study is based on capturing video from cameras that interact with artificial intelligence (AI) algorithms, suitable for detecting and chasing targets. They return the position of the target in Cartesian coordinates as an extra layer on the screen. Coordinates are pixels based on camera resolution. These pixels are converted into decimal values corresponding to the necessary movement that the motors must make to triangulate the weapon with the target. The weapon is positioned in the center of the shooter’s camera, so when the target coincides with the center of the camera, the robot will be able to shoot. The detection and tracking algorithms chosen in this project are Google MediaPipe (MP) and Tracking Learning Detection (TLD). The first will be on the sniper camera, a cutting-edge technology developed by Google and provided for free use. Capable of identifying a person and marking parts of a human body, it was configured to center the middle of the eyes with the center of the screen. Each eye has its own coordinate, and when averaging has a coordinate used for aiming. The second open code, developed by the python community, seeks to identify any object on the screen. Used by the observer’s camera, this algorithm seeks to differentiate similar targets and determine which one should be pursued. The final result is expected that through an interface, the user must indicate a person and the Sniper Robot identifies and pursues the target leaving it centered on the screen. It will be up to the operator to pull the trigger of the gun and shoot. This prototype is developed thinking about occupying the space of technologies of the fourth industrial revolution in systems for the defense of Brazil.