Software And Hardware Co-design Of X-ray Weld Image Stitching Based On FPGA

With the continuous development of science and technology,X-ray inspection and image processing technology is increasingly used in the military,scientific research,medical and industrial and agricultural production and other fields.Due to the characteristics of X-ray inspection technology itself and the length of the negative,the same casting of the weld inspection information often requires multiple negatives lap each other in order to completely display the information of the original casting,which not only reduces the intuitiveness of the test results,there may also be a risk that the negative is falsified and swapped,to achieve the purpose of the second best.At present,the identification of forged radiographic negatives is mainly manual,but due to the excessive number of negatives and other unstable factors,it is likely to cause manual identification efficiency and accuracy reduction.In response to the above problems,this article uses the Zynq-7000 series fully programmable system-on-chip launched by Xilinx,which integrates Xilinx 7 series FPGA and dual-core ARM Cortex-A9 processor,and has a high-speed interconnection communication structure to ensure ARM and FPGA For data exchange,this paper adopts the method of software and hardware co-processing,realizes the image feature point detection in FPGA,and implements the subsequent image registration and fusion algorithm with the feature point detection results in Linux.This makes full use of the FPGA's powerful parallel processing capabilities and abundant logic resources,making the design flexible,reducing power consumption and improving the overall operating speed of the system.This thesis analyzes the commonly used image registration methods in detail,and compares them,and transplants the FAST algorithm and the SURF algorithm to FPGA for hardware acceleration.This article uses high-level synthesis tools to transplant the image registration algorithm to the hardware,through code writing,simulation testing,packaging and output RTL-level packaged IP core,the hardware of the algorithm is realized,and the data type,throughput and other optimization methods are adopted in the design.Optimize code writing.The software part mainly realizes the registration,stitching and fusion of the images processed by the hardware.Experimentally verified,this system can theoretically realize the splicing of radiographic images,with less resource consumption,and finally realize a software-hardware cooperative processing system for X-ray images.