UAV Tracking by Identification Using Deep Convolutional Neural Network
Yongguang Mo, Jianjun Huang, Gongbin Qian · 2022
In recent years, the frequent occurrence of UAV black flight and intrusion incidents has made their effective detection and tracking a research hotspot. This paper focuses on the problem of poor detection and tracking when traditional detection and IMM tracking algorithms are used to detect and track UAVs, we improved the tracking accuracy by introducing the flight pattern recognition probability into the IMM to assisted adjusting the model update probability. Firstly, a multi-layer deep learning UAV detection and identification algorithm is proposed to detect and identify the presence and flight patterns of UAVs. Secondly, an IMM algorithm based on flight pattern assisted adjustment (FAMP-IMM) is proposed by introducing the flight pattern recognition probability from neural networks as auxiliary knowledge into the IMM algorithm and continuously correcting the model update probability. The experimental results demonstrate that the detection and recognition accuracy reaches as higher to 99% with the multi-stage DL approach, the F AMP-IMM algorithm greatly decrease the model switching lag and tracking performance compared with the IMM algorithm.