Research On Image Stabilization Algorithm Of Video Acquisition For Nondestructive Testing Of Heating Pipe Inner Wall

The construction quality of the heating pipeline and the subsequent maintenance of the pipeline directly affect its performance and service life.Once a leak or explosion occurs,it will cause unpredictable economic losses and social harm to the entire society and the lives of residents.The non-destructive flaw detection of the inner wall of the pipeline based on video images is currently the clearest,intuitive and very convincing detection method.The arrangement of heat pipes is complicated and the caliber is extremely small.Usually,technicians manually control the inlet and outlet pipes of industrial endoscopes to perform video capture.In this process,unavoidable shaking and the inability to control the uniform speed of the lens collection will occur.It will have a huge impact on the subsequent inspection area statistics and precise positioning of the inner wall.Therefore,it is an indispensable link to perform image stabilization in time for the video of the inner wall of the pipeline.Through analyzing the reasons for the extremely poor quality of pipeline video from multiple angles,it is found that there are mainly the following problems:blur,non-uniform speed and jitter.This paper will deal with these three aspects separately to eliminate the problems in the video.First,on the basis of Retinex theory,combined with multi-scale decomposition and self-adaptive contrast enhancement algorithm,the improved method has been made to make the details of the inner wall of the pipeline more clear,and the contrast between the foreground and the background is more obvious.And accurately distinguish the pipeline center area of interest and determine the pipeline center position.Next,the similarity between video sequences is detected by the non-negative matrix factorization algorithm,and redundant frames with too high similarity are eliminated,and the frames with too low similarity are predicted by the frequency multiplication frame interpolation method based on motion estimation.The prediction result is inserted as a transition frame between the two frames whose similarity is too low,so as to complete the processing of the video uniform speed problem.Finally,for the jitter in the video,the deviations of the lens in the three directions of angle,level and rotation are corrected respectively,so that the lens axis coincides with the central axis of the pipeline,and combined with the special imaging characteristics of the pipeline image,the circular projection method is performed The optimization improves the calculation accuracy of the rotation component,so as to realize the image stabilization of the video of the inner wall of the pipeline.Aiming at the improved optimization algorithm proposed in this paper,110 consecutive frames are selected from the video sequence for comparison experiments.The experimental results verify the superiority of the improved enhancement algorithm in various evaluation indicators.The positioning accuracy of the pipeline center point can be stabilized within 10 pixels,the calculation error of the jitter offset component is reduced by about 50%on average,and a higher quality pipeline video is obtained.