Development Of Image Target Tracker Based On FPGA+DSP Architecture

With the development of airborne photopods on the ground and sea surface detection technology and its extensive application in the military and civilian fields.The need for automatic positioning and tracking of sensitive targets using photoelectric pod is also increasing.Target tracking is carried out on airborne photopods,which can automatically track interested targets and assist in flight targeting and navigation of aircraft.Therefore,the development of airborne image target tracker device is of great significance.Because in the airborne system,the computing resources are limited,but as the precision and speed of the acquisition of images are improved,the amount of image data collected is also increasing.The airborne system,which requires real-time processing of images,brings the pressure of processing performance.Therefore,on the basis of the research of image tracking algorithm,we choose a fast and efficient tracking algorithm,and develop an image target tracker based on DSP+FPGA architecture for the interface requirements of airborne optoelectronic pod.Firstly,the related requirements of the image target tracker in the project are analyzed,and the technical indicators that the system needs to meet is given.On the basis of analyzing the comparison of system technical specifications and hardware selection,a general scheme of image target tracker based on DSP + FPGA architecture is designed.Which FPGA is mainly responsible for the realization of image acquisition,transmission and display interface.For the hardware structure of DSP,a multi-core cooperation framework is designed.DSP handles internal data based on a multi core collaboration framework.Based on the data flow model of data exchange between processor cores,the framework uses high-speed transmission interface SRIO to communicate with FPGA.This not only ensures the high efficiency of the internal processing of DSP processors,but also satisfies the fast and real-time transmission of image data.Consider the performance of tracking scenarios and algorithms that the image tracker may encounter.The KCF target tracking algorithm is chosen as the core processing algorithm of the image target tracker.Compared with other algorithms,the algorithm can process images in real time,and objects that may be encountered in airborne optical boxes are blocked by short circuits and generate motion.KCF target tracking algorithm has good processing robustness when it is difficult to track such as blur and rotation.Secondly,according to the design of the system,the image target tracker is implemented on the hardware board.It mainly includes the standard interface Camera Link protocol module for FPGA reading and collecting images,the data exchange transmission module between FPGA and DSP,the standard interface HDMI module for displaying images,the multi core cooperation framework for DSP operation and the DSP for image tracking.The algorithm realizes and optimizes the algorithm module.The image target tracker developed in this paper,its input and output interface modules are standard protocol interfaces,and can be used as part of the airborne optoelectronic pod system,which is directly related to the matching and installation.The multi-core DSP algorithm module can process the HD video captured in the photoelectric box in real time.The power consumption of the target tracker is less than 10 W,and has good adaptability.Finally,the image target tracker is tested on the basis of function and system processing performance.The tracking problem faced by the image tracker in airborne optoelectronic pod system has been thoroughly tested.The test results show that the image target tracker developed in this project can process the image captured by the forward looking camera in real time.The image target tracker developed in this paper has strong robustness to the problem of temporarily blocking airborne optoelectronic box targets,background clutter,target image motion blur and possible problems when the target is rotated.