| The research work of this dissertation is under the background of the important national "Eighth Five-year Plan" science and technology project(85-302-02-06). The automatic flaw detection of welded tubes is one of the most important steps to ensure the quality of the tubes. Nondestructive inspection on welding seam of tube is required in the tube production,and real time X-Ray radiography inspection is an effective means. Along with continuous improvement of the productive ratio,the demand for the automatic inspection to the welding seam becomes more and more pressing,so implementation of the automatic inspection possesses important significance on both theory and reality. Wavelet transform is a powerful tool in the signal and image processing,and its fundamental theory has been formed. From the view of engineering applications,however,the wavelet transform is still in the elementary stage,the further researches are required for the practical uses. In this thesis,we concentrate mainly on using wavelet analysis for welding seam image processing and recognition,and some related techniques are developed.For constructing welding seam positioning and detection control system,the multiple computers configuration for weld seam image recognition is proposed. The system adopts the architecture in which multiple CPUs process parallelly under the control of the master IPC computer. The system can perform storing the weld seam images,positioning,flaws recognizing and quality rejudging. The Watch-Dog interface card is successfully developed,it can improve the system reliability by redundance technique of saving breakpoint data and restoring them.The hardware supporting the system makes use of the high speed digital signal processor TMS320C30 from Taxax Instruments Company. The frame grabber can capture 25 frames of welding seam image per second continuously and make it possible to fulfil the real time welding seam image processing and recognition. The one kind of improved FWT(Fast Wavelet Transform) algorithm for a finite sequence is proposed after studying theory of multiresolution analysis and analyzing technical characteristics of DSP. The implementation of the periodic extension of the FWT on DSP is described in detail and the corresponding FWT assembly code is described for the DSP TMS320C3X series.This dissertation suggests scheme of image denoising based on two-dimensional discrete wavelet transform. The denoising algorithm is describedwith some operators. By thresholding the wavelet transform coefficients-of noisy images,the original image can be reconstructed correctly. Different threshold selections and thresholding methods are discussed. The new robust local threshold scheme is proposed. Quantifying the performance of image denoising schemes by using the mean square error,the performance of the robust local threshold scheme is demonstrated and is compared with the universal threshold scheme. The experiment shows that image denoising using the robust local threshold performs better than that using the universal threshold.In order to improve the accuracy and the real time performance of edge detection,a method need to be found to match the detection of low contrast blurred welding seam image. This dissertation analyzed the main sources of noise as well as the different characteristics of noise and signal under wavelet transform,and proposed a multiresolution edge detection method based on wavelet transform. The experimental results show the effect of this algorithm is advantageous over that of traditional edge detection algorithm.The geometrical relation of elliptic imaging is studied for welding seam image of the butt welds in straight tubes. The region model of welding seam image is proposed,It furnishes a evidence theory to further process to welding seam image. Combining with the region model,a model-based adaptive target segmentation algorithm is proposed. One basis of the algorithm is Otsu's discriminant criterion. The adaptive target segmentation of welding seam image is realized. The...
|
PDF Full Text Request