| With the development of the automatic and intelligent technology,welding manufacturing will be applied in various fields of industrial production.Generally,PC was used as a platform for weld image processing system,which features high cost,big size,high energy consumption,poor real-time performance,and did not fit need of high reliability.With the development of embedded system technology,DSP chips were provided with more and more processing capabilities,and it was advantaged to process vision image by an embedded system.This dissertation was based on the practical engineering projects as the research background.The DSP real-time processing system of welding seam image could automatic identification as a goal,and processed welding seam image,including image pre-treatment and post-treatment.Finally,it established the real-time weld image processing system based on DSP.The DSP processing tasks of welding seam image was composed of image capturing in this paper,welding seam image processing and image displaying.In order to decrease the difficulty of system development,we used DM642 development board as the video processing platform to software development.Image capturing module finished the task of capturing image.It was composed of CCD and decoder.A technology of active vision was adopted.It would send to weld image collected by EDMA to DM642 for processing,then finished the task of real-time weld image processing and identified the position of welding seam.Finally,it would finish the display of the image processing results by the video encoder and monitor.First of all,this dissertation analyzed the system's requirement and selected TMS320DM642 as the core chip of this system.Then it analyzed the hardware composition and principle of the DM642,and established the DSP image processing system hardware platform.At the same time,the dissertation analyzed the structure and functions of the kernel of DSP/BIOS.Based on the class/micro driver model and according to RF5 reference frame model,it would develop FVID model video processing diver program.In the end,it would implement of weld image acquisition,processing and display.In the process of industrial welding process,a lot of noise would always appear.This dissertation analyzed what kind of noises will appear in the welding process,and the corresponding solution was given.The image processing method mainly contain that optimized the traditional median filter algorithm,threshold processing,based on statistical regularity of all maximum of gray value to determine the optimal threshold,eliminated the noise of small area and isolated based on mark binary image connected area,improved each line(column)search method,found the weld seam center through the sliding window scanning per line(column)method,used the longest monotonical rising(falling)subsequence to find the center position of the weld,etc.At last,finished the DSP video capturing and displaying experiment,and seam tracking experiment.The experiment results show that the DSP processing system of welding seam image could real-time complete image processing task.Weld image processing algorithms could get rid of the noise effectively,extract the weld center line,calculate weld center coordinates.Finally it would real-time provide the location information of the weld center to the control mechanism of seam tracking system.The system satisfied the requirement of real-time performance. |
PDF Full Text Request