Research On Targeting System Based On Zynq

With the development of industrial intelligence,machine vision is used to identify and locate products on many industrial lines.Adaptability and real-time are important indicators for judging the target positioning system.However,in some occasions with high requirements on volume,power consumption,cost and stability,the traditional PC vision-based machine vision system can not meet these requirements.For this reason,the research on Zynq-based target positioning system is carried out.This article has completed the following research content:(1)In terms of target localization,this paper uses a shape matching algorithm based on edge direction.This algorithm first needs to extract the edge of the template image,obtain the gradient value and gradient direction of each edge point,and then calculate the edge of the template.The inner product sum of the corresponding points in the image to be matched,and normalized to determine the matching situation.Experiments show that this method can also be very good when dealing with the target object being subjected to illumination changes,occlusion,rotation,etc.Strong stability,and can also reduce the time-consuming matching by speeding up the strategy.(2)The first thing to do in the shape matching based on the edge direction is the edge extraction.In this paper,the Canny operator is used to remove the noise of the template image,non-maximum value suppression,and hysteresis threshold to obtain the edge of the target object in the template image.At the same time,template making is an indispensable part of target positioning.In the case of using the image pyramid acceleration strategy and the target object has a deflection angle,it is also necessary to make templates of different angles in different levels to extract the gradient vectors of the corresponding edge points.The design of target location algorithm is completed by studying edge extraction,template creation,shape matching and acceleration strategy.Experiments show that the algorithm can be free from illumination and occlusion,and pyramid hierarchical search can greatly speed up the matching.(3)Zynq chip integrates ARM processor and programmable logic FPGA,and takes advantage of ARM in program control and FPGA parallel computing.Through the analysis of the function of the target positioning system,the most time-consuming template creation and template matching algorithms are divided into FPGAs for hardware acceleration processing.In the target positioning system,the template and the image data to be matched need to be transmitted from the ARM end to the FPGA end.Through the analysis of the characteristics of the AXI4 bus in Zynq,the AXI4-Stream protocol is selected as the transmission method of the image data.The protocol takes the data stream as the core,and does not The address bus is required,which greatly improves the transmission efficiency.(4)In order to realize the target positioning system in Zynq chip,firstly transplant the target positioning algorithm into Vivdo HLS,set the parameters of the top function of the algorithm to the port type for external use,and then optimize and loop the data type of the algorithm.The optimization is carried out,and finally packaged into an IP core to complete the hardware integration of the target positioning system.The software side realizes the hardware acceleration of the target positioning algorithm by calling the IP core,and finally tests the system.Through the analysis of the test results of the Zynq-based target positioning system,the hardware-accelerated shape matching algorithm is 10 times faster than the ARMonly system,which greatly improves the execution efficiency of the algorithm.