A Doorway Detection and Direction (3Ds) System for Social Robots via a Monocular Camera

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Robotic navigation in indoor environments requires accurate perception of surroundings.
• Social robots often rely on monocular cameras due to cost and size constraints.
• Detecting doors and their orientation is crucial for effective indoor navigation.

Purpose of the Study:

• To propose a novel algorithm for detecting doors and their orientation using a monocular camera.
• To address the challenge of 2D image-based door detection and orientation estimation for social robots.
• To enhance the capabilities of social robots in indoor navigation tasks.

Main Methods:

• A convolutional neural network (CNN) model was trained on the Social Robot Indoor Navigation (SRIN) dataset for door detection.
• A system integrating Depth, Pixel-Selection, and Pixel2Angle modules was developed for orientation estimation.
• Simulation and real-time experiments were conducted to validate the algorithm's performance.

Main Results:

• The proposed algorithm successfully detects doors in indoor settings from monocular camera views.
• The system accurately estimates the orientation of detected doors relative to the robot.
• Experimental results demonstrate the algorithm's effectiveness in both simulated and real-time scenarios.

Conclusions:

• The developed algorithm provides a viable solution for door detection and orientation estimation using monocular vision.
• This research contributes to improving the navigation capabilities of social robots in indoor environments.
• The findings have potential applications in various robotic navigation systems.