Research On Image Detection Method For Defects Of Composite Prepreg Tapes

In recent years,with the development of science and technology,materials have been widely used in aerospace,medical and other fields.However,with the continuous improvement of industrial technology,the use of the single material hasn't been meeting the industrial requirements,which resulted in the appearance of composite materials.Compared to single materials,it has better performances.The glass fiber reinforced thermoplastic in composite materials has been widely used because of the low weight,low cost and excellent mechanical properties.Due to the influence of the processing technology,continuous glass fiber reinforced polyolefin prepreg tapes were prone to cracking and appearing the phenomenon of uneven fiber dispersion,which affected the performances of the product.In this paper,image processing methods were used for testing continuous glass fiber reinforced polyolefin prepreg tapes.This paper introduced the manufacturing process of prepreg tapes.when using melt-impregnated methods to prepare the prepreg tapes,owing to the influence of factors such as technology and raw materials in the production process,it would cause cracks in prepreg tapes and fiber aggregation defect.At the same time,there were some scratches on the surface and some debrises during prepararation.In order to realise the detection of prepreg tapes defect,the ways of median filtering,adaptive median filtering,wavelet transform,and the Lo G operator were firstly used to preproccess prepreg tapes images,then according to the analysis and comparition,the algorithm combined with fractional differential and Gaussian filters was used to reserve defects and get rid of interference of scratches and debrises.After preprocessing the prepreg tapes images,the defective area was extruded by the image segmentation method.The specific process was: Firstly,the improved Bats optimization Otsu algorithm was used to get best threshold.Then,it was used to segment image for obtaining the binary images.Last,Canny operator was used to detect edge of the binary images and extract defective edge information.The low threshold of Canny edge detection was the lowest gray level of the preprocessed images.The high threshold was the optimum threshold rounding in threshold segmentation.This way could get edge information of prepreg tapes detect clearly and lay the foundation for the next feature extraction and identification.The texture feature of gray level co-occurrence matrix was used to extract feature and classify from the preprocessed prepreg tapes images.In the existing data set,prepreg tapeswith good impregnation and defective prepreg tapes were separated by extracting their contrast characteristic value.Then,the way of grayscale histogram and wavelet energy were used to extract the features from crack images and uneven fiber dispersion images.The smoothness,third-order moment and consistency of the preprocessed image grayscale histogram and the 2nd,3rd,4th and 5th layers of coif3 wavelet energy characteristic value after the segmented image were formed into input vector,using support vector machine to classify.The 40 cracking images and 40 uneven fiber dispersion images were selected respectively.The 20 images were used as a training set in every defects.The surplus 20 images were used as a testing set.The result of the test set was that 1 image was misjudged to be cracked for uneven fiber dispersion defects,and 3 images were misjudged as uneven fiber dispersion for the crack defects.