Integrating the Symmetry Image and Improved Sparse Representation for Railway Fastener Classification and Defect Recognition

摘要: The detection of fastener defects is an important task for ensuring the safety railway traffic. earlier automatic inspection systems based on computer vision can detect effectively completely missing fasteners, but they have weaker ability to recognize partially worn ones. In this paper, we propose a method detecting both partly and proposed algorithm exploits first second symmetry sample original testing image integrates them improved representation-based recognition. This scheme simple computationally efficient. underlying rationales are as follows: First, new virtual symmetrical images really reflect some possible appearance fastener; then integration two judgments recognition somewhat overcome misclassification problem. Second, sparse representation discarding training samples that “far” from test uses small number “near” represent sample, so perform classification it able reduce side-effect error identification problem image. experimental results show outperforms state-of-the-art methods.