Research On Fabric Defect Detection Algorithm And Its Implementation On DM3730Platform

Fabric defect is an important factor affecting the quality of textile. The automatic detectionfor fabric defects have been a challenging problem and research focus in the field of machinevision. In china, the fabric defect detection is mainly done by manual with these disadvantagessuch as low efficiency and higher false detection rate due to human visual fatigue and so on.With the rapid development of industrial automation and machine vision, it is an inevitable trendthat the automatic detection system will take the place of human to pinpoint fabric defects inorder to improve the quality of textile and reduce labor costs.In order to detect fabric defects that are various in sort, shape and size, two new detectionalgorithms are proposed in this paper, one of which is based on Butterworth filters and the otherbased on NonSubsampled Contourlet Transform (NSCT). Experimental results show that twoalgorithms can effectively detect defects of different texture images. In the second algorithm,four kinds of NSCT composed of different filters are used. On the basis of the defect detectionresults of both two algorithms, the conclusion can be drawn that in terms of fabric defectdetection, the detection algorithm based on the NSCT consisted of9/7wavelet with rationalcoefficient has better performance.In addition, the DM3730embedded development platform’s two dual-core communicationmechanisms are introduced in this paper, which can be implemented by using the VISA moduleand the C6Accel module respectively. Then, the program’s collaborative development methodbased on the DVSDK Software development kit is given under two mechanisms. Finally, byusing the VISA module of the DM3730platform, software design of the fabric defect detectionalgorithm based on NSCT is accomplished. The software completes the defect images storage, processing and the displaying of detection results.