Hardware Design Of Defect Detection System For Digital Textile Printing Based On Embedded GPU

In recent years,digital textile printing technology has developing rapidly.It has been widely accepted by the industry with advantages of low cost,low pollution,and short lead time for sample order,etc.However,problems exist in the high-speed printing process like nozzle clogging and step deviation,which lead to the printing errors.The traditional artificial detection has a lot of shortcomings such as heavy workload,high missing rate and poor timeliness.With the development of machine vision technology,the automatic defect detection technology based on machine vision has attracted great attention.Compared to manual method,automatic defect detection technology has made obvious progress in efficiency and timeliness,which has high value of practical application.Therefore,a defect detection hardware system for digital textile printing based on embedded GPU is designed in this thesis.It achieves automatic defect detection.The system receives images from industrial line array camera through the Camera Link interface.After receiving the data,FPGA completes the data processing and transfers it to embedded GPU Tegra K1 with PCI-E.The Tegra K1 performs a series of operations on the images with GPU cores,such as preprocessing,extraction of characteristic parameters,analysis and judgement.The defect detection system also extends a Solid State Drive to store image data,Gigabit Ethernet to achieve network transmission,DVI-D and USB interfaces to achieve interaction.The test shows that the defect detection system for digital textile printing based on embedded GPU is stable and reliable.At present,the system has got through the testing process of prototype machine,which achieves automatic defect detection for digital textile printing and meets the design requirements.