Fabric Detect Detection Method Based On Texture Gradient

In the textile industry, quality control is of vital importance for fabric products. Currently, the inspection task is primarily performed by human inspectors and hence heavily relies on their experience, judgement and attention. Automated inspection of the fabric defects is becoming an attractive alternative to the human visual inspection in modern textile industry. The main goal of fabric automated inspection is able to precisely locate the defect regions in the fabric. In recent years, texture gradientis a popular tool for image segmentation that considerally improves contour detection performance, and it is further used in the inspection of fabric defects. The main work is as follows:1. To develop a texture image segmentation method based on texture gradient. Based on texture gradient and marker-based watershed, simple texture watershed is designed for texture image segmentation. In order to reduce noise (which in a general sense includes fine texture and non-stationary behaviors), nonlocal means (NL-means) filter is used for enhancing the differences between the defect texture and background texture. Combined with the texture watershed transform, it is called texture watershed based on texture enhancemence. The defect regions in fabric images can be defected accurately in this method.2. To develop a method for texture defect detection based on texture gradient and MRF. By analyzing the model of MRF, the automated texture defect detection based texture gradient and MRF is proposed to overcome the drawback of simple texture watershed. In this method, Markov random field model is used in the final segmentation, followed by the texture enhancemence and texture gradient.3. To develop a method for fabric defect detection based on texture gradient. The three methods based on texture gradient are used in the inspection of fabric defects. Experiment results have demonstrated that simple texture watershed can get fast and accurate result of fabric image containing obviously different texture. The texture watershed based on texture enhancement can get accurate result of more difficultly defected fabric image, but it costs more time. The detection based on texture gradient and MRF model can get accurate result automatically.