Research On The Key Generic Technology Of An Automatic Identification System For Textile Fiber Based On Image Processing

Digital image processing technology has been widely used in industrial measurement, biomedical and other fields. In textile industry, automotive detecting equipment has been developed to replace the time-consuming, subjective greater impact artificial recognition based on digital image processing. This way has a promotion of both the precision and efficiency. In this paper, the main research of the common key technologies based on digital image processing technology to achieve the automatic recognition of cotton, ramie.In the hardware part, as the microscope table’s automatic temperature control system has its deficiencies, an automation temperature control system for microscope table is invented, which is based on C8051F021. The system uses the high accuracy platinum resistance, PT100as the temperature sensor and Fuzzy-PID algorithm and control temperature range from-10℃to40℃. Temperature control precision is less than±0.5℃.In the digital image processing part, firstly, because there are uneven illumination, focus blurring, dust and such impurities which are not conductive to identification in the image collected from microscope, this paper proposes a gray stretch algorithm in order to enhance image. Then the Otsu binarization procedure is taken on to deal with the image. After image binarization, the two-dimension size is taken account to discriminate the target area and non-target area, meanwhile seed-filling method is used to eliminate noise.Secondly, because of fiber crossing and overlapping, the idea of extracting the public fiber skeleton firstly, and then pairing fibers is proposed. To reduce the difficulty of pairing, deburring of fiber skeleton is carried out. Then break the intersection point to get separated fiber skeleton. For the separated fiber section, extract public fiber skeleton segment, pair separated fibers according to inclination slope of fitted fiber segmentation, length of central distance projected to the inclination slope, distance between fiber segments projected to the slope. Find the longest segment that belongs to the same paired fiber and record coordinates of the center. According to the coordinate point value, collecting fiber images to be discriminated in10x objective lens.Again, extracting the outline to be identified fiber then calculating diameter of each micro segmentation based on area method and fiber skeleton iteration method, statistical average diameter.Calculating hue value of micro fiber region where gets the maximum diameter. Selecting the average diameter and hue values as input feature vectors to support vector machine (SVM) to determine fiber type.Lastly, according to the proposed fiber detection and recognition algorithms, a set of software with good human-machine interface is done to detect the fineness and component of fiber with Matlab and Vc++mixed programming.