Research And Application Of Multi Domain Perceptual Detection Algorithm For UAV Small Target

Drone technology is developing at an astonishing speed,with vast application prospects and huge market demand.Drone detection is a challenging problem in drone technology and serves as fundamental research for other issues such as multi-drone cooperation and avoidance of drone attacks.This thesis focuses on exploring the problem of detecting drones.Firstly,a semi-supervised learning-based drone image detection model is proposed for small drone targets.Then,a drone video detection model is proposed by integrating spatio-temporal information.Based on these models,an excellent drone detection system is designed and implemented.The main contents of this thesis are as follows:1.To address the problem of sparse drone detection datasets,a semisupervised learning-based drone object detection model is proposed for small drone targets.The model utilizes unlabeled data to improve its performance and reduce its dependence on labeled data.At the same time,the model alleviates the impact of the small object foreground-background imbalance problem under the semi-supervised object detection paradigm.2.To address the problem of inaccurate drone detection,a video drone detection model based on the attention mechanism is proposed.The model enhances the spatio-temporal consistency of drone video object detection by utilizing both spatial and temporal information,which improves detection accuracy.3.A drone detection system is designed and implemented.Firstly,a new drone dataset is built.Then,semi-supervised learning is used to improve the training process of real-time drone detection algorithms.Based on the considerations of computing power limitations and real-time system requirements,a drone detection system is designed and implemented.