Study On Color Difference Detection Of Printing And Dyeing Fabrics Based On Machine Vision

Accurate monitoring and control of color difference is an important guarantee to determine the quality of fabrics in textile dyeing and printing industry.The manual detection has high cost and poor stability,and the colorimeter has the small detection range and the poor overall detection uniformity,so a lot of researchers have carried out the study on color difference detection of dyeing and printing fabric based on machine vision.From the related literature,The existing method of evaluating color difference introduced background noise,which made the accuracy of evaluation results poor,the speed and efficiency needed to be improved as well as.Therefore,it was of great significance to improve the accuracy,speed and efficiency of color difference detection of dyeing and printing fabric based on machine vision.In order to improve the accuracy of color difference detection of dyeing and printing fabric,considering the influence of the background color noise,MATLAB were taken as algorithm development platform,improved fabric color difference evaluation method based on saliency algorithm and CIELAB color space were proposed.Firstly,characteristic law of fabric texture clearance region based on the area and roundness distribution of saliency region was developed by the improved CMC(l:c)saliency algorithm,which extracted feature of untreated fabric texture clearance region.Secondly,two steps color difference evaluation flow were designed to complete the untreated fabric color difference grating.Lastly,experimental results showed that improved fabric color difference evaluation method based on saliency algorithm and CIELAB color space had the less deviation to reflect the true color difference of fabric and provided a new method of fabric color difference grading.Based on MATLAB platform,improved fabric color difference evaluation method based on saliency algorithm and CIELAB color space had a significant increase in accuracy,but the speed still needed to be improved.Therefore,on the basis of the theoretical study mentioned above,the research of rapid color difference detection was carried out by using Halcon as the algorithm development platform.Through studying neural network classification learning algorithm and optimization method of mean square error minimization,research on fabric color difference detection based on MLP multilayer perceptron in Halcon was proposed,which provided a specific implementation scheme suitable for industrial production.Experimental results showed that color difference area of different class could be distinguished,color difference detection had high sensitivity and the lowest average accuracy could be reached at 89.7%,which could meet the requirements of accuracy and sensitivity of industrial color difference detection.The existing color difference detecting methods are all processing images collected by cameras,which had limited visual field and could not make the entire width of the fabric image collected.The usual solution was using two camera to collect fabric image,and then taking color difference detecting,the problems was that the images collected by the two cameras overlapping in some areas made the efficiency lower,so the fabric images collected by two camera should firstly be mosaicked and then taking the color difference detection.Taken Halcon as algorithm development platform,the method of image mosaicking based on the geometric space position relationship between the applied camera and the world coordinate systems was studied and analyzed,finally image mosaicking based on camera calibration was proposed,and the results of experiments were better.On the basis of the above theory and application research,3D model of fabric color difference detection platform was designed to meet actual demand for industrial chromatic aberration detection and provide technical support for the next practical development and application.