Surface Defect Classification for Hot-Rolled Steel Strips by Selectively Dominant Local Binary Patterns

Abstract

Developments in defect descriptors and computer vision-based algorithms for automatic optical inspection (AOI) allows for further development in image-based measurements. Defect classification is a vital part of an optical-imaging-based surface quality measuring instrument. The high-speed production rhythm of hot continuous rolling requires an ultra-rapid response to every component as well as algorithms in AOI instrument. In this paper, a simple, fast, yet robust texture descriptor, namely selectively dominant local binary patterns (SDLBPs), is proposed for defect classification. First, an intelligent searching algorithm with a quantitative thresholding mechanism is built to excavate the dominant non-uniform patterns (DNUPs). Second, two convertible schemes of pattern code mapping are developed for binary encoding of all uniform patterns and DNUPs. Third, feature extraction is carried out under SDLBP framework. Finally, an adaptive region weighting method is built for further strengthening the original nearest neighbor classifier in the feature matching stage. The extensive experiments carried out on an open texture database (Outex) and an actual surface defect database (Dragon) indicates that our proposed SDLBP yields promising performance on both classification accuracy and time efficiency. INDEX TERMS Automatic optical inspection (AOI) image classification local binary patterns (LBP) steel industry, surface texture.

Research Progress of Automated Visual Surface Defect Detection for Industrial Metal Planar Materials

“…It cannot be ignored that the useful description information in the non-uniform patterns in all these LBP variants has been elided. Luo et al creatively used the reverse thinking to explore the non-uniform pattern to supplement the description information hidden in the uniform pattern in [ 7 ] and [ 61 ]. As lightweight feature descriptors, LBP and its variants can be applied to defect detection and classification at the same time, and researchers should follow and develop LBP variants or LBP-like descriptors with better noise robustness and scale invariance, which is also in line with the future development trend of AVI.…”

Automated Visual Defect Detection for Flat Steel Surface: A Survey

“…Using reverse thinking, Luo et al [3] proposed a generalized completed local binary patterns (GCLBP) by first exploring the non-uniform patterns to supplement the descriptive information in uniform patterns. Further the work of GCLBP, Luo et al developed a more effective LBP-descriptor (namely SDLBP) in [46], which has remarkable advantages in anti-interference and simplicity of calculation. As a lightweight feature descriptor, LBP variants can be applied on both defect detection and classification, developing more noise-robust and scale-invariant LBP variants or LBP-like descriptors is highly encouraged and coincides with the AVI future trends.…”

Construction of an Energy Saving Model Based on FA and Its Applications for Detection of Battery Power Consumption

“…A homomorphic mapping is given by the lattice of FDFA in the homomorphic mapping layer. At the same time, a fuzzy inference algorithm by FA is given in the fuzzy inference layer, as well as the inference operator is also novel, because it adopts the fuzzy operator method instead of the usual detection operator method [12]- [17]. Compare it with the old method in the simulation.…”

“…At present, there are some disadvantage in the existing literatures [12]- [17], [25]- [33] on the detection methods for battery power, because these methods only detected the state of charge, but couldn't achieve automatic charging or add electrolyte to the battery automatically.…”

“…In this paper, the detection method by the detection model constructed is applied in consumption tracking of battery The above experimental process was repeatedly carried out 30 times, and then their average is taken. Compare the accuracy of detection of the detection model constructed and existing detection algorithms [12]- [17] in the detection of SOC. The experimental results show that the correct average detection rate of the detection model constructed is 92.79%, that of Everest method is 59.…”