UAV Abnormal State Detection Model Based on Timestamp Slice and Multi-Separable CNN

With the rapid development of UAVs (Unmanned Aerial Vehicles), abnormal state detection has become a critical technology to ensure flight safety UAVs. The position and orientation system (POS) data, etc., used evaluate UAV status are from different sensors. traditional model ignores difference POS data in frequency domain during feature learning, which leads loss key information limits further improvement performance. To deal with this improve safety, paper presents method for detecting based on timestamp slice multi-separable convolutional neural network (TS-MSCNN). Firstly, TS-MSCNN divides reasonably time by setting set specific timestamps then extracts fuses features avoid information. Secondly, converts these into grayscale images reconstruction. Lastly, utilizes convolution (MSCNN) learn more effectively. binary multi-classification experiments conducted real Air Lab Fault Anomaly (ALFA), demonstrate that outperforms machine learning (ML) latest deep methods terms accuracy.