Collaborative Acceleration Design Of Runway Detection Algorithm Based On Saliency Analysis

Aiming at the acceleration problem of runway detection algorithm based on saliency analysis in autonomous landing of fixed-wing UAV,this thesis studies the acceleration method of hardware and software co-design,and finally completes the design of acceleration system.The main work includes the following:1.For the saliency-based runway detection algorithm,several saliency calculation methods are compared and the method suitable for this thesis is selected.How to obtain the region of interest of the runway by saliency is analyzed,and the system image scaling factor is determined by experiment.The mechanism of several image processing algorithms used in saliency calculation is studied.2.The design method of hardware and software co-design is studied.According to the design requirements,a suitable development platform is established;the design process and rules of hardware and software co-design are studied;advanced integrated design methods and optimization strategies are proposed for hardware acceleration of complex algorithms;and a module for high-speed image transmission is designed for the visual algorithm in this thesis.3.The frame of image processing system based on on on-chip system is designed.According to the application requirements,the relevant components of the system are selected;the hardware module of the visual framework is designed,including the hardware logic design of camera capture and image display module;and the appropriate operating system and driver design are selected in the design of the software part.4.The IP core of image processing technology used in saliency calculation is designed.In the design of image scaling module,a scaling algorithm suitable for hardware acceleration in this thesis is proposed,and then the hardware design is carried out.The two-dimensional Fourier transform module and the mean filter module are designed and optimized by advanced synthesis method,and are encapsulated after simulation and verification respectively.5.In order to verify the hardware and software co-design method,the simulation of UAV landing runway video is used to test,and the real-time and accuracy are verified.The experimental results show that the processing speed of the cooperative acceleration system designed in this thesis is up to 30 frames/s,which is much faster than that of the original platform.