Research On Wavelet Detection Algorithm And DSP Implementation For Fabric Defects

In the process of textile production, textile fabric defect detection isone of the most important parts. For a long time, textile fabric defectdetection and level assessment depend on human eyes. This traditionalway creates the bias of the visual impact of subjective factors, so there arefalse detection rate and the high rate of undetected defects. With thedevelopment of the computer vision, the digital image processing andneural network technology, the automatic detection of fabric defects hasbecome an inevitable trend in image processing and computer platform.Firstly, the common method of the detection of fabric defects basedon spatial domain and frequency domain was described, and theadvantages and disadvantages of the method were analyzed. Secondly, themain principles and applications of wavelet analysis were introduced,including the properties of continuous wavelet, discrete wavelet transformand orthogonal wavelet. At the same time, the construction of the adaptiveorthogonal wavelet was discussed in detail. The experiment results showthat the fabric defects in the image appear apparently phase distortion byusing the orthogonal wavelet of two-layer decomposition. The reason isthat low-pass filter coefficient of the orthogonal wavelet does not satisfythe symmetry, that is the linear phase non-existent. Therefore thebiorthogonal wavelet detection for overcoming the distortion wasproposed, and time consuming was compared with the lifting orthogonalwavelet. It is shown that the operation speed of the biorthogonal waveletwhich is structured by the lifting scheme is improved by an average of47%comparing with the biorthogonal wavelet. Meanwhile the experimentresults show that there is no phase distortion by means of the liftingbiorthogonal wavelet detection, which is more suitable for on-linedetection of fabric defects.After the simulation of biorthogonal wavelet lifting scheme algorithmof fabric defect detection on MATLAB, the software design of the fabric defect detection algorithm was completed on DM642image processingsystem, and the results analysis were given. The detection system includeshardware and software. The hardware is composed of the DM642structure, linear array CCD camera and host computer etc. The softwaredevelopment is based on the CCS integrated environment which issupporting DSP and C language for programming. The results indicatethat the detection of fabric defects on the DM642image processingsystem may be realized.