The key to the detection of mesh fabri c defects is to completely segme nt the mesh from the complex background.The adhesion and defects of the mesh se gmentation will lead to a lower accuracy of mesh segmentation.Aiming at the pro blems of low contrast of mesh fabric image and the existence of shadows at the j unction of fabric warp,resulting in mesh adhesion and incompleteness in the seg mentation results,this paper proposes an improved FCM (fuzzy C-means) clusteri n g segmentation algorithm.First,in order to solve the problems of reduced infor mation entropy and insufficient contrast after the image is processed by the his togram equalization algorithm,a gray-scale transformation curve is established to improve the image quality.Finally,in order to improve the accuracy of mesh segmentation again,the membership matrix extracted by the FCM algorithm and the double threshold classification method are used to reclassify the pixels with h igh segmentation uncertainty.Experimental results show that the algorithm in th is paper has a good segmentation effect,effectively enhances the gray value of pixels at the mesh junction and solves the problems of mesh adhesion and incompl eteness in the segmentation results.The accuracy of mesh segmentation is about 96%.