Research On FPGA Acceleration Of UAV Vision Algorithm

Unmanned Aerial Vehicle(UAV)are currently widely used in agriculture,transportation,and power troubleshooting.The application of drones is inseparable from the vision system.As the UAV application field becomes more and more extensive and the tasks undertaken are becoming more and more important,the UAV vision system designed has also become larger and more complex.There are two problems with an overly complex vision system: one is that the drone itself cannot meet the required powerful computing power;the other is that a complex vision system will generate a lot of energy consumption.In order to promote the rapid development of UAVs,it is very important to design real-time,low-power UAV vision systems.In order to solve the above problems,this research proposes a universal UAV vision algorithm acceleration architecture,and designs a novel UAV return-to-home system based on this acceleration architecture.The specific research content and main contributions are as follows:1)First,based on the existing FPGA acceleration principles and methods,and the particularity of the UAV field,the optimization method of related algorithms is proposed.In addition,this research adopts a design scheme of software and hardware coordination to achieve the largest speedup under effective resources.Finally,the final general acceleration architecture is determined according to related optimization methods and software and hardware co-design methods.2)The RKLT algorithm is proposed for the application of UAV returning.Most researches apply the KLT algorithm to tracking moving objects.This research mainly applies it to tracking corner points in a static background.RKLT performs corner selection and tracking based on regions,and the offsets of different regions are processed differently according to the level of confidence.In addition,this research combines the proposed general acceleration architecture to accelerate the RKLT algorithm,and the acceleration result is 53 times faster than the PC implementation and greatly reduces energy consumption.3)Propose a novel UAV returning system.So far,most UAV returning technologies have relied on the Global Positioning System(GPS).This has led to more and more people using false GPS signals to abduct drones and make them deviate from the expected flight trajectory.Therefore,a return-to-home system that does not rely on GPS is particularly important.In order to solve this problem,this research proposes a novel UAV return system.The system uses FPGA to accelerate the high-complexity RKLT algorithm so that it can achieve real-time and low power consumption,meets the needs of UAV.