QRGrader: desenvolvimento de um aplicativo Android para correção automática de gabaritos utilizando visão computacional e marcadores ArUco

Abstract

QRGrader is a native Android mobile application prototype for the automatic correction of objective exam answer sheets through computer vision and digital image processing. The application operates in three stages: reading and decrypting a QR Code contain- ing the correct answers, protected by hybrid RSA-OAEP and AES-CBC encryption; automatic capture and perspective rectification of the answer sheet using OpenCV, ArUco markers, and the CameraX API; and automatic grading through contour analysis and relative thresholding, supporting multiple-choice, true-or-false, and short-answer questions. The system was validated through six controlled-variable test sets on Sam- sung Galaxy A32, Galaxy Tab (SM-P625), and Galaxy S23 FE devices, achieving 100% accuracy under ideal conditions — ink pen, distance between 20 and 30 cm, and flat surface. Identified limitations include low-contrast markings (graphite pencil: 50%) and distances outside the operational range, for which the system alerts the teacher or interrupts the capture. Results demonstrate that QRGrader is a viable, portable, and secure solution for automated answer sheet grading without specialized equipment.