AN INTEGRATED UAV-BASED OBSERVER PLATFORM HYBRIDISING ONLINE FUZZY CLASSIFIER

Intelligent system-assisted UAV-based observer platforms could achieve various complex observing tasks over traditional methods. However, due to the complexity of their algorithms, UAV’s first-flight route is still challenging to deploy quickly and minimise energy consumption in an emergency. Another challenge is that the UAV-based observer platform severely requires an efficient classifier with high processing speed for higher observing efficiency. As the first research objective, this paper artificially evaluated seven UAV first-flight routes by simulation and real-world flighting environments to identify one proper first-flight route that could be deployed quickly. Secondly, a new integrated UAV-based observer platform, including a new three-colour channel-based online fuzzy classifier, is proposed for quickly detecting abnormal objectives in practical observing tasks. Simulation and real-world flighting experiments identified that the square helix with smooth turn consumes the most miniature battery and can cover the observing area among seven different first-flight routes. The results also proved the proposed integrated observer platform’s feasibility in detecting abnormal objectives while UAVs fly in a real-time, real-world environment. Most importantly, the proposed observer platform has good interpretability because it employs an actual image stream to train its classifier during flighting.