An Algorithm for Unmanned Aerial Vehicle Landing Scene Recognition based on Deep Learning and Computational Verbs
Chao Tian, Chaobing Huang · 2019 IEEE 1st International Conference on Civil Aviation Safety and Information Technology (ICCASIT) · 2019
Safe landing is a very important part of making the Unmanned Aerial Vehicle (UAV) truly capable of autonomous flight in the field. In order to determine whether the landing site is suitable for UAV landing, it is necessary to study the scene recognition problem. Aiming at the problem, this paper proposes a learning rate attenuation method based on computational verb theory, which can effectively improve the recognition accuracy. The parameters of the four exponential decay learning rate curves were selected to calculate the verb similarity (that is, four computational verb rules are constructed), then the verb similarity was used to derive the learning rate used for each round of optimization, finally, we trained the network with the obtained learning rate. Compared with the usual exponential decay method, the proposed algorithm can effectively improve the recognition accuracy of landing scenes.