Realization Of Metal Workpiece Defeat Detection Platform Based On ZYNQ

During the production process,metal sheets are prone to various defects that can affect the usability of the products due to their material characteristics.With the continuous advancement of computer and image processing technologies,machine vision inspection has experienced rapid development.Compared to traditional manual visual inspection and nonintelligent detection methods,machine vision inspection offers advantages such as contactless and non-destructive inspection.It has become an efficient,intelligent,and precise means of quality control,reducing production costs while improving detection accuracy and efficiency,thus having practical applications.This dissertation focuses on the detection of copper sheets,which have slightly rough surfaces,are prone to reflection,and have small dimensions.The machine vision inspection platform needs to meet the high accuracy,low false detection rate,and real-time defect detection requirements of these copper sheets.Relying solely on a general-purpose processor or FPGA is challenging to fulfill the design requirements of the entire system.Therefore,this dissertation proposes a metal defect detection platform based on Zynq,an heterogeneous platform that combines dual-core Cortex-A9 processors and FPGAs,leveraging the advantages of both for software and hardware co-design.The dissertation’s main research includes the following:Investigation of the framework of a metal defect detection platform based on Zynq and the design of suitable lighting schemes for defect detection in copper sheets.The platform employs a turntable for transferring the copper sheets,and industrial cameras triggered by photoelectric switches capture the images.The image processing and defect detection functions are mainly implemented on the FPGA side of the Zynq.Analysis of the material characteristics of small-sized copper sheets and the design of detection algorithms targeting dirt and scratch defects.The algorithm consists of three main parts: image preprocessing,ROI extraction,and defect detection and filtering.The image defect detection algorithm IP core is implemented using the HLS(High-Level Synthesis)tool.By analyzing the system’s functional requirements,Zynq chip resources are allocated appropriately.A Vivado hardware project is designed,and the setup and porting of an embedded Linux system on the Zynq platform are completed.A GUI based on Qt is designed,and finally,the detection platform is tested.Performance testing of the defect detection platform demonstrates that the system designed in this dissertation supports the detection of dirt and scratch defects in circular copper sheets with a diameter of 6mm.The minimum detectable defect radius is 0.5mm,and the detection accuracy reaches 93.75%.In continuous triggering mode,the platform achieves a detection speed of up to 18 fps for 480 ×480resolution images,meeting practical production requirements.