Safe avoidance path detection using multi sensor integration for small unmanned aerial vehicle

• Main Authors: Ramli, Muhammad Faiz, Shamsudin, Syariful Syafiq, Legowo, Ari
• Format: Conference or Workshop Item
• Language: English; English; English; English
• Published: Institute of Electrical and Electronics Engineers Inc. 2018
• Subjects: TE210 Construction details. Including foundation, maintenance, equipment; TJ Mechanical engineering and machinery; TJ210.2 Mechanical devices and figures. Automata. Ingenious mechanism. Robots (General)
• Online Access: http://irep.iium.edu.my/69114/3/69114_Safe%20avoidance%20path%20detection%20using%20multi%20sensor_cp.pdf; http://irep.iium.edu.my/69114/4/69114_Safe%20avoidance%20path%20detection%20using%20multi%20sensor_scedule.pdf; http://irep.iium.edu.my/69114/1/69114_Safe%20avoidance%20path%20detection%20using%20multi%20sensor_article.pdf; http://irep.iium.edu.my/69114/2/69114_Safe%20avoidance%20path%20detection%20using%20multi%20sensor_scopus.pdf; http://irep.iium.edu.my/69114/; https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8453521
• Institution: Universiti Islam Antarabangsa Malaysia

Achieving a robust obstacle detection system that can provide a safe avoidance path system for small UAV is very challenging. Due to size and weight constraints, very limited detection sensors can be equipped in the system. Prior works focused on a single sensing device which is either camera or range sensors based. However, these sensors have their own advantages and disadvantages in detecting the appearance of the obstacles. In this paper, a combination of both sensors based is proposed for a small UAV obstacle detection system. A small Lidar sensor is used as the initial detector of obstacle and queue for image capturing by the camera. Next, SURF algorithm is applied to find the obstacle regions and free space regions. This is done through the principal of object size changes and distance relationship in an image perspective. The proposed method was evaluated by conducting experiments in a real complex environment which consist of a textured and textureless obstacle. In the experiment conducted, we successfully detect and create a safe avoidance path for both situations. The textured situation gives a high success rate while textureless situation produces acceptable success rate until 60cm distance.