Research Article

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

Table 3

The recall rate (%) of different methods on the fastener database.

Number of the training samples of each subjectIntact fastenerPartially worn fastenerMissing fastener
501001502505010015025050100150250

SVM84.387.589.093.577.581.383.785.392.594.895.398.0
Improvement to SVM88.891.295.896.389.591.695.798.594.596.897.398.5
NNA86.388.592.894.581.683.387.892.391.594.395.897.3
Improvement to NNA87.589.894.596.589.891.396.899.092.595.896.397.8
SRC87.389.393.395.585.588.389.891.394.394.896.397.5
Improvement to SRC89.890.395.896.389.393.797.596.894.895.597.298.3