Research And Application Of The Key Technology On Oscillographic Screen Image Recognition Based On Image Features

Ultrasonic flaw detector,one kind of nondestructive testing equipment,is widely used in modern industrial manufacturing such as aviation,aerospace,navigation and rail transportation.As an important testing equipment,accuracy,reliability and credibility of the ultrasonic flaw detector are directly related to the structural strength,service life,as well as the wholesome quality and safety of a product or equipment.However,the measurement calibration technology for ultrasonic flaw detector in China is still in the development stage of measuring manually and interpreting graphs artificially.With the further development of multi-functional,digital and intelligent ultrasonic flaw detector,the traditional way of interpreting graphs by eyes has many difficulties in adapting to the needs of the measurement and calibration work.And it is an irresistible trend for the development of modern industry is to realize automatic measurement and calibration and to replace traditional interpreting graphs by eyes with reading graphs by machine vision.Starting from the periodical measurement of ultrasonic flaw detector and by using machine vision to complete the process of graphs interpreting,the author has studied several key points including oscillographic screen window segmentation,salient image edge detection,and recognition of characters and shapes based on image moment invariants in this dissertation,and designed a kind of automatic calibration system for the ultrasonic flaw detector based on image characteristics.The major achievements and innovation points of the research are as follows:1.In view of the problem of incomplete segmentation for oscillographic screen window easily caused by varieties of shape,color matching and structural configuration of ultrasonic flaw detector,the author proposes an image segmentation method based on geometric shape features.By using the geometric shape features of the target object as descriptors,the problem of segmentation is described into one related to shape recognition.Screening and discriminating of the oscillographic screen window area can be completed through the way of obtaining the edge contour line of the target object by image edge detection and combining the minimum enclosing rectangle.In order to reduce the time of image recognition process,the method of seeking for the major axes,which is the calculation method for minimum enclosing rectangle,is optimized in this dissertation,so as to accelerate the convergence speed of the minimum enclosing rectangle.Experimental results show that,the geometrical shape features of the oscillographic screen window are better in discretion ability and the segmentation method has good stability and robustness for noise and illumination changes,and can complete perfectly the segmentation of the oscillographic screen window for the calibrated ultrasonic detector.2.Regards to the poor universality and difficulty of feature acquisition in obtaining thewaveform of oscillographic screen window,a kind of salient edge detection method for oscillographic screen window images based on the theory of spectral residual approach is proposed in this dissertation.In this method,the spatial domain image is transferred to the sequency domain by Walsh transform,a residual sequency spectrum corresponding to the salient edge is established by seeking out the sequency spectrum domain corresponding to the horizontal and vertical directions of the spatial domain image,and finally,the salient edge of target image can be obtained via operations such as Walsh inverse transform.It can be seen from experiment results that,the salient edge detection method based on Walsh transform is of high efficiency and can be realized easily,and it is very sensible to the salient edge of target image.Moreover,the detection and extraction of waveform can be completed effectively by using the salient edge detection method without setting specific image feature parameters for each waveform,which shows that the method has better universality.3.The author proposed and derived a kind of new image moment,that is,Walsh image moment,as the characteristic descriptor of the target character and shapes in regarding to the problems such as selection and rapid and identification of characteristic descriptors for shapes and characters in the oscillographic screen window,and completed the recognition work by using simple similarity measurement(weighting Euclidean distance).Experiment results show that,the Walsh image moments based on the system structure of Walsh orthogonal functions with high calculation efficiency and the characteristic of invariant translation,rotation and zoom,has a strong ability to describe the characters and shapes.By using weighting Euclidean distance which is simple and can be realized easily,an average recognition rate of 98.09% can be achieved for recognizing common character and shapes in the oscillographic screen window.4.Aiming at the problems existing in the calibration and calibration device of ultrasonic flaw detector such as low efficiency,excessive human factors,poor repeatability and stability in measurement and calibration,the author has designed an automatic calibration system of ultrasonic flaw detector based on image features and applied the research results mentioned above successfully.During the design process,the author has analyzed the measurement calibration process of major measurement parameters in the measurement calibration in details,provided the software and hardware structure and the design of the system,and derived the transmission relationship of test data between the change of the waveforms of the oscillographic screen window and the conventional parameters of the calibrated ultrasonic flaw detector.Experimental data shows that the automatic measurement and calibration system can achieve high detection accuracy,which can meet the demands of automatic calibration and calibration for ultrasonic flaw detectors.