Design Of Weld Defect Intelligent Recogition System Based On GPU

The oil and gas pipeline transportation industry is an curcial part of China’s energy deployment,with the characteristics of uninterrupted work,large capacity and small footprint.Therefore,it is crucial to ensure the quality and safety of weld joints at pipeline joints.However,the existing inspection methods mainly depend on manual completion of the quality assessment work,resulting in the emergence of problems such as low inspection accuracy,strong subjectivity of examination results,and low inspection efficiency.In order to solve the above problems,this thesis develops and designs an intelligent,efficient and digital weld defect identification system from the actual engineering needs to provide technical guarantee for the safe transportation of oil and gas pipelines.The major research contents of this paper are as follows.Firstly,In image processing for high depth welding seam film images,the detail loss and feature extraction are difficult due to the mismatch between bit depth and the general display system.A pseudo-color enhancement algorithm based on the reconstruction of chromatographic mapping table was proposed to smoothly map the negative image of high depth welding seam into color space.The difference between background and weld characteristics was enlarged.At the same time,the details of image information were retained,which laid a foundation for the subsequent construction of the weld defect data set and weld defect detection.Secondly,Aiming at the problems of low accuracy and intense subjectivity of manual evaluation,this paper introduces the idea of deep learning into the field of weld defect detection,completes the construction of a weld defect detection network model based on YOLOv4,and completes it according to the characteristics of the high-level deep weld negative image optimization of network model training samples and pre-selected anchor boxes.The experimental results show that the optimized network model has improved detection accuracy compared with the original model,meets the actual engineering technical index requirements,and can complete weld defect detection with high quality.Thirdly,To address the problems of manual evaluation and the low efficiency of existing digital inspection systems,this paper aims to improve the operational efficiency of the system by building the system using CUDA-based GPU parallel acceleration technology,analyzing the system workflow,designing software function modules,and implementing GPU-side programming of image processing algorithms.The experimental results show that compared with CPU serial computing,GPU parallel computing achieves a quantum breakthrough and reduces the execution time of image processing algorithms from seconds to milliseconds,contributing to the realization of high-efficiency weld inspection.Finally,By completing the deployment of the work environment on the QT platform,the login interface,overall interface,and database of the system software are designed to achieve file processing,image processing,defect detection,user interaction,defect file management,and other functions.The experimental test results show that the software structure of the system is reasonable,and the operation effect is good.It can effectively supervise and monitor project construction units and oil and gas pipeline quality management departments.It has good application value and market prospects in the oil and gas pipeline transportation industry.