NOKTA BULUTU VERİSİ İLE KURAL TABANLI AÇIK KAPI BULMA YÖNTEMİ

Abstract

In indoor environments, the door detection problem is addressed in many different applications such as localization, navigation, semantic classification, building information modeling and autonomous wheeled vehicle. In previous studies, the methods, which handle door detection problem, generally tried to identify closed-doors and open-doors via visual and range information, respectively. In recent years, with the use of sensors that can generate point cloud data and/or depth image in the robotic applications, the methods that attempt to determine the door position with that data have also been developed. The main purpose of this study is to introduce a method that can detect open-door location in real-time through point cloud data in indoor environments. Firstly, the point cloud captured according to the robot's local coordinate system is transferred into a global coordinate system. Then, a set of rules based on the global coordinate system is defined to detect open-door location. “OGUROB KAPI” dataset is constructed to measure the effectiveness of the proposed method. The dataset contains point cloud data gathered in GAZEBO simulation environment for different angles between robot and open-door location. The tests performed on this dataset were examined in terms of the door detection success and the elapsed time.The test results showed that the average door detection time was 10ms with the proposed method. In addition, it has been observed that the true positive rate is 91%.

Keywords

• open-door detection
• point cloud data
• rule-based
• gazebo