| Birds hitting aircraft is a traditional threat to the safety of flight takeoff and landing,in recent years,with the rapid development of consumer-grade drones,drones "black flight" disturbance problem is increasingly prominent.Compared to birds,drones are more rigid and can cause more serious consequences in case of collision with aircraft.The above-mentioned birds and drones can be classified as low-altitude slow-speed and small-size flying objects,referred to as "low and slow small" flying objects,whose common characteristics are flight altitude below 1 km,flight speed below 200 km/h,and radar cross section below 2 m2.In order to protect the flight safety of the airport clear zone,it is necessary to develop an early warning system with the function of anti-low and slow small flying objects.The first task of countermeasures is to detect the target,and only by knowing the time and location of the target can we take corresponding countermeasures.After detecting the moving target,it is also necessary to identify the target,and take different countermeasures for different types of low and slow small flying objects.Meanwhile,for the shortcomings of conventional image processing platform with poor real-time performance,this paper designs and implements a real-time detection and recognition system for low and slow small flying objects based on FPGA(Field Programmable Gate Array)platform,and the specific research contents are as follows:(1)To achieve low and slow small flying objects detection over a large sky area,this paper selects a camera array to acquire video data in parallel,which expands the field of view compared to a single camera.Based on the dual-FPGA platform,three algorithms of median filtering,open-operator morphological filtering and dither processing are used to filter the image noise to reduce the incidence of false alarms.To facilitate hardware implementation,this paper proposes the parity shunt algorithm and combines it with the inter-frame differential algorithm to complete the moving target detection.The minimum enclosing box algorithm is used to obtain the top,bottom,left and right four boundaries for the detected moving targets,and then a rectangular box with crosshairs is used to identify the moving targets on the display.(2)The small size of low and slow small flying objects,coupled with the long shooting distance,leads to their small number of pixels on the imaging surface and lack of feature information such as shape,texture and contour,making it difficult to accurately identify them using conventional target recognition algorithms.To address the above problems,this paper uses the trajectory clustering algorithm to further remove the false targets,and then identifies the "moving trajectory" and "moving form" from the real targets.In order to adapt to the FPGA platform,the above algorithm is fixed-point improved in this paper,so as to divide the low and slow small flying objects into two categories:flying birds and drones.(3)In actual use,the early warning system is usually deployed outdoors,while the staff is operating indoors,so the video data needs to be transmitted over a long distance.For the shortcomings of too many conventional data transmission lines and close transmission distance,this paper combines Gigabit Ethernet,switch and an optical fiber,which can support 3 km long-distance transmission.Finally,this paper objectively evaluates the target detection accuracy of the system in terms of center position offset error and area statistical overlap respectively.The overall system has a comprehensive average detection accuracy of 93.47%and recognition accuracy of 84.62%at a frame rate and image resolution of 17 fps@l 024x768 pixel.Compared with the conventional software-based serial processing,the system designed in this paper has the advantages of high real-time,high parallelism,low power consumption,high energy efficiency ratio and small size,and is suitable for application in practical scenarios. |
PDF Full Text Request