A Fabric Defect Classification Based on Two-dimensional Sparse Representations and a Norm Optimization

Abstract

Sampling loss of the structural information of the image for the one-dimensional compression and bring about the loss of recognition accuracy, we propose the concept of two-dimensional compression samples. Using a set of sparse-based perception to get the sparse data on the raw data of the defect, fabric defect two-dimensional sparse. Finally, use of norm optimization method accurately decrypt the compressed data, the eigenvalues of different fabric defect classification. This approach solves the proliferation of data collection and the sensor waste greatly reduces the computational complexity, fabric defect classification, and thus to lay a theoretical foundation for machine vision to identify fabric defects.