Research On Dynamic Target Detection And Tracking System Based On FPGA

At present,dynamic target detection and tracking technology is one of the most important research directions in the field of computer vision.This technology has been widely used in many aspects of military and civilian use,especially in the fields of intelligent monitoring,machine navigation,industrial manufacturing,military reconnaissance and so on.Currently,a large number of target detection and tracking algorithms have been deeply explored and analyzed.However,the implementation of these algorithms using hardware is still the current research direction.FPGA is implemented in hardware programming language with the characteristics of repeatable programming;it has parallel operation and pipeline structure characteristics,this feature makes its running speed faster,It can significantly improve the real-time performance of image processing,and effectively reduce the product development cycle and cost,and is obtained in the field of video image processing.This thesis first established a basic video image acquisition and display system,and designed the image acquisition module,data cache module,image display module,etc.and simulated and verified it.Secondly,the dynamic target detection and tracking algorithm is compared and studied.The current common algorithms are analyzed in depth.The advantages and disadvantages of inter-frame difference method,background subtraction method and optical flow method are compared.Then according to the difference of target representation method in dynamic target tracking process,the basic principle and advantages and disadvantages of each target tracking algorithm are analyzed in detail.Finally,the inter-frame difference algorithm was used to detect the moving target,and then the template matching algorithm was used to track the target.At the same time,the morphological changes occurred during the movement were taken into account.Therefore,the Harris corner algorithm was added to update the template.Afterwards,the simulation of the studied algorithm was verified in matlab software.Compared with the traditional template matching method,the tracking of the target after the addition of the Harris corner algorithm becomes more flexible and stable.At last,the implementation process of the detection tracking algorithm in FPGA is described in detail.The algorithm implementation part is mainly divided into five modules: data format conversion module,median filter module,motion detection module,motion tracking module,and template update module.The system uses Quartus II software,according to the overall flow chart of the system,using Verilog HDL language combined with Altra's DE2-115 development board to design a dynamic target detection and tracking system,and through the modelsim software for each function module for simulation testing.Finally,the designed video image acquisition,image display,data buffering,and detection and tracking of moving objects are combined.Throughexperimental simulation and image display,we can see that this design can realize real-time detection and tracking of moving objects in the scene,and the system has good stability.