Design And Implementation Of Accelerator For AdaBoost Detection Algorithm Based On Heterogeneous System

In recent years,artificial intelligence is rapidly developed and amounts of new image algorithm with high accuracy are applied in different field such as safe monitor,advance drive assistant system and automatic producing.Face detection based on AdaBoost is one of these algorithm.Because of its complexity and huge amount of calculation,the algorithm could not be applied in the real-time scene.However,Face detection needs to be applied in real-time scene in realty situation.So the algorithm should be calculated more effectively.The main contribution of this paper are:1.The computation tasks are analyzed.The whole processing tasks are separately allocated to different processing unit.2.Based on analysis result,this paper propose a configurable accelerator aim on Ada Boost algorithm.3.This paper design and realize an ARM&FPGA heterogeneous system based on ZYNQ platform.4.Control flow is designed to make all the system work in high effect situation.This paper present an accelerator aim on Ada Boost algorithm and achieve high performance while using small amount of hardware resource.A complete face detection system is realized.