Research On Detecting And Tracking Method Of Infrared Target In Anti-UAV System

With the development of modern science and technology,UAV(unmanned aerial vehicle)technology are increasingly mature,it applied in many civilian industries areas,with high efficiency,low cost,large output and other advantages.Further,UAV also causes a serious threat to airspace security,so many countries begin to develop their own anti-man-machine system.The application of infrared imaging technology to anti-man-machine system,not only has anti-interference,good concealment and adaptability to weather and other advantages,but also has relatively low price and maintenance costs.How to improve the detection range of UAV targets,and to discover and track the UAV earlier is a key and difficult problem in this system.The UAV is represented as a point in infrared imaging,which is determined by the image distance and area.For the point target is easily disturbed by noise and background clutter,and the UAV group show a strong maneuverability during flight,which greatly limits the anti-man-machine system to detect and track the UAV,especially in the extreme cases where multiple-UAV or UAV group appear in the monitoring area.In this paper,we discuss the related technology of target detection and tracking in infrared anti-man-machine system.The main research work of this paper is introduced next.In the aspect of image preprocessing,we discuss the structure and infrared imaging characteristics of UAV in this paper.Analyzing the infrared UAV images,we obtain a more practical pretreatment method.For infrared UAV target detection,we discuss the DBT and TBD algorithms,and propose a CFAR multi-frame association detection method based on dynamic threshold,which can effectively improve the detection accuracy of single frame image and reduce the false alarm rate,and it is more easier to realize in project.For infrared UAV target tracking,we apply the Bernoulli filter algorithm based on stochastic finite set in the infrared UAV target tracking.And it is proved by experiment that the label multi-Bernoulli algorithm can track a plurality of specified targets stably and can distinguish each target well by the tag.For UAV target tracking,we propose a the LMB group target tracking algorithm based on clustering in this paper,which avoids the track cross-chaos and intra-group target number estimation error,and realize a stable tracking of single group target center.Finally,for the instability of clustering center which is caused by multi-group intersection,we use a secondary clustering method to provide a more reliable small center for the LMB.After merging the small center trajectory,we can get the target center trajectory.Experiments show that this method can solve the loss of target caused by clustering center is unstable when the target intersects.