Indoor/Outdoor Transition Recognition Based on Door Detection

Seohyun Jeon, Minjung Kim, Seunghwan Park, Jae-Yeong Lee

Abstract: An autonomous navigating mobile robot uses various sensors for localization, such as GPS, laser scanner, IMU, odometer, and camera. These sensors use different weights and parameters according to the location, whether the robot is in indoors or outdoors. For both indoor and outdoor (IO) seamless navigation, the robot needs to know where the robot is, either indoors or outdoors, and when the IO transition has been made. This way the robot can switch the appropriate weights and parameters at the right time.This paper describes the method of identifying the moment when the indoor and outdoor transition is made. In order to recognize the time of IO transition, door detection with YOLOv5 is used. As the robot approaches the door, the detected bounding box of the door shows changes in size. This feature is used for IO transition recognition. The algorithm is tested with two scenarios and it shows that this method is simple and applicable to real world and the result is reliable to the detection failure.