Research On Color Measurement Of Dyed Yarn Using Image Processing

Color is an important indicator in the textile field.In the past,colors were judged manually.In recent years,the use of spectrophotometer to measure color,but the measured object needs to have a certain size in order to meet the measurement requirements of the spectrophotometer aperture,so the color measurement of the dyed single yarn is very difficult to sample.Therefore,based on the RAW raw digital image obtained by the digital camera,this paper established the polynomial regression model of the RGB original response information to the CIEL*a*b* color space to realized the color prediction of the pure color sample cloth and the dyed single yarn.The main work of this article includes the following aspects:(1)Set the camera mode manually and calculated the white balance gain parameters according to the Bayer filter rules of the camera.The calculation criteria of the gain parameters and the selection of the basic parameters set by the camera manually are based on the fact that the resulting RGB value of the whiteboard is 220 to 240 approximately,which prevented distortion due to color saturation of the photosensitive element.Verification results: lens focal length was 35 mm,aperture was f6.3,shutter was 1/4 second,light was ISO100,R channel gain was 2.3039,B channel gain was 1.2672.(2)Obtained the RAW format image of the sample and created a polynomial regression model of the RGB raw response information and the CIEL*a*b* color space.The digital camera is used to obtain the packaging file of the sample,and then Dcraw is used to analyze the image format required by the experiment,and the RGB original response information was extracted to establish a polynomial regression model with the CIEL*a*b* color space.Experimental results show that RGB response information is directly converted to CIEL*a*b* color space,which is more accurate than RGB response information through XYZ and CIEL*a*b* color space.Based on the optimized model,the color of various textile samples in the experiment was predicted better.(3)An artificial brightness segmentation method for knitted fabric images is proposed.Because the knitted fabric is knitted by the coils of the coils,recessed voids will appear in the fabric image.These voids appear as black areas on the image.The pixels in these areas will greatly hinder the accurate extraction of the entire fabric image color.Therefore,this paper proposed an image segmentation method based on lightness.The experiment showed that after artificial brightness segmentation,the predicted color of knitted fabric converted by the model approaches the DigiEye color measurement result.(4)An artificial threshold yarn image processing method was proposed.First,the original RAW image of the yarn was obtained,and after the white balance,the color bitmap of yarn was restored.Then,the color bitmap is subjected to artificial threshold segmentation processing,closed arithmetic processing,and connection domain processing.Then,the remaining pixel corresponds to the original RAW image of the yarn,and the RGB response information corresponding to the yarn is extracted to perform color prediction.The experiment shows that after processing of this process,the image of the dyed single yarn is completely segmented,and the hairiness that is not required for the experiment is removed.(5)Five methods for extracting RGB original response information are proposed,and yarn color prediction was well achieved.The five methods for extracting RGB original response information proposed in this paper include central region method,edge region method,transition region method,68.269% normal distribution method and kernel density estimation method.The experimental results show that after processing by the five methods,the prediction of the yarn color for the preferred model conversion is better than the DigiEye measurement result except for the transition area method.