Feasibility Study of OpenCV Implement Real-time Object Tracking by Image Recognition of Aerial Photos in Android Systems for UAV

• Main Authors: JYUN-HONG SHANG, 尚俊宏
• Other Authors: HUAN-JUNG LIN
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/6886fv

碩士 === 國立虎尾科技大學 === 航空與電子科技研究所 === 102 === Now, Unmanned Aerial Vehicle (UAV) is developed rapidly, where the multi-rotor UAV is the primary research and usage for surveillance and image capture. The main reasons are that multi-rotor UAV can hover, and its flight control is ease. On the other hand, the image capture is the main payload of multi-rotor UAV. Those camera payloads cannot process or deliver image promptly. Recently, the smartphone is very popular and suitable for UAV payload with its handy, power saving and fast computing. Especially, multi-rotor UAV can use the high-performance camera, Wi-Fi communication, data storage and also have sensor like gyroscope sensor, magnetometer sensor and GPS of smartphone for aerial photography, flight record and attitude control. The main research of this thesis is that uses an Android smartphone for image capture, image delivery, image process, image recognition and target tracking of multi-rotor UAV. Where the thesis adopts OpenCV(Open Source Computer Vision Library) to achieve target recognizing, as well as delivery image via Wi-Fi to ground station really time, then ground station Android tablet can assign target from ground image and switch image recognition function. Finally, we use USB interface between Smartphone and Arduino Mega 2560 flight controller of multi-rotor UAV. The Flight control controls rotor motors and keep tracking target in sight to implement real-time target tracking for multi-rotor UAV. In the feature,the system can integrate with well-develop multi-rotor UAV for flight test, and take full advantage of smartphone to improve the UAV research.