Application of Improved Optical Flow Algorithm in Quadrotor UAV

• Published in: 2021 6th International Symposium on Computer and Information Processing Technology (ISCIPT) pp. 269 - 274
• Main Authors: Lin, Zhihao, Lu, Bingliang, Mou, Xiyu, Zhang, Xindong
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.06.2021
• Subjects: Autonomous aerial vehicles; Corner detection; Feature extraction; Four rotor; Indoor positioning; Lighting; Military computing; Power demand; Proportion of gradual; Real-time systems; The Gaussian Pyramid; Visualization
• DOI: 10.1109/ISCIPT53667.2021.00061

With the rapid development of computer science, automation and communication technology, unmanned aerial vehicle (UAV) emerges in various fields, and has become an indispensable platform in both military and civil fields. Due to the integral error of traditional inertial navigation, the errors will accumulate over time. In complex application scenarios, such as dense buildings, indoor corridors, tunnel rescue, it is still a difficult problem to provide accurate positioning for the UAV when the GPS signal is lost. Therefore, it is urgent to study the accurate positioning of UAV indoors. Traditional visual navigation has the disadvantages of low real-time performance and easy to be affected by environmental lighting changes in complex environment. In this paper, an improved optical flow algorithm based on Shi-Tomasi (S-T) corner detection is proposed. The brightness compensation method of front and rear frames is adopted, which makes up for the weak anti-interference ability of optical flow algorithm to illumination change. The proposed algorithm was applied to a quadrotor UAV to verify the results. The results show that the improved algorithm has higher indoor positioning accuracy and better real-time performance compared with the traditional optical flow algorithm Lucas-Kanade (LK).