Research On Defect Evaluation Technology Based On Ultrasonic Image Processing

The research of ultrasonic testing technology is gradually developing towards automation and intelligence,and combining ultrasonic testing with automatic control and image processing can effectively assist the staff to locate and evaluate the defects.However,due to the noises in ultrasonic echoes such as diffraction and scattering,and the residual signals in reflective scanning,defect signals with relatively small energy are not easily distinguishable from noise signals.The defects of ultrasonic images cannot be well segmented.At the same time,the accuracy of ultrasonic image classification is low with the problem that the ultrasound images defect have less classification samples,more categories and some samples are at the edge of the classification boundary,and the samples may have artificial classification errors.In order to solve the problems in the above-mentioned ultrasonic flaw evaluation,this paper constructed a set of ultrasonic testing system and carried out relevant theoretical and experimental research.To reduce the influence of residual vibration on the effectiveness of ultrasonic image defect segmentation,an ultrasound image defect segmentation algorithm based on automatic seeded region growing was proposed.First,the pre-segmentation of the ultrasound image was performed by the OTSU method.Next,the seed starting points were set automatically by seeking the absolute background area.Then,the defects were segmented from the background area by regional growth algorithms.Finally,the defect recognition was further improved by digital morphological noise reduction method.And then an ultrasonic image defect classification method based on support vector machine optimized by genetic algorithm was presented.Firstly,the feature data of ultrasonic image defect is extracted by image processing.Secondly,the support vector machines were trained as ultrasonic image defect classifiers.Finally,the parameters of the classifiers were optimized by genetic algorithm to obtain the optimal classifiers.Experimental results have shown that the ultrasonic image defect segmentation algorithm not only can accurately segment the defects,but also has better defect boundary information,improves the processing efficiency of ultrasonic images,effectively inhibits most of the image noise,and the maximum relative error between the calculation distance from the top of the workpiece and the actual value is 5.988%.At the same time,the ultrasonic image defect classifier proposed in this paper is better than the other classifier in recognition rate,and the overall recognition rate of the training sample and the test sample is 90%,which can effectively assist the staff to classify the ultrasonic defects.