HMM-Based Defect Localization in Wire Ropes – A New Approach to Unusual Subsequence Recognition

Abstract

Abstract. Automatic visual inspection has become an important application of pattern recognition, as it supports the human in this demanding and often dangerous work. Nevertheless, often missing abnormal or defective samples prohibit a supervised learning of defect models. For this reason, techniques known as one-class classification and novelty-or unusual event detection have arisen in the past years. This paper presents a new strategy to employ Hidden Markov models for defect localization in wire ropes. It is shown, that the Viterbi scores can be used as indicator for unusual subsequences. This prevents a partition of the signal into sufficient small signal windows at cost of the temporal context. Our results outperform recent time-invariant one-class classification approaches and depict a great advance for an automatic visual inspection of wire ropes.

Combining Structure and Appearance for Anomaly Detection in Wire Ropes

“…We compare our results to the one obtained with the Hidden-Markov model (HMM) approach of Platzer et al [6] which leads to the best published results so far with regard to an individual analysis of each camera view. Fig.…”

“…Their results underline the importance of context information for the problem of surface defect detection, especially with respect to the complex structure of wire ropes. In [6] Platzer et al focused on contextual anomaly detection by modeling the intact class with help of Hidden Markov Models. Haase et al [2] diagnosed contextual anomalies in the rope surface with help of an autoregressive model which predicts the intact surface appearance given its neighborhood.…”

Enhanced anomaly detection in wire ropes by combining structure and appearance

“…other Approaches We compare our results to those obtained with the Hidden-Markov model (HMM) approach of Platzer et al (2009) which leads to the best published results so far with regard to an individual analysis of each camera view. Figure 9(b) shows the averaged ROC curves (averaged over all four views) obtained on the same dataset with the HMM approach and with the model-based approach.…”

“…Platzer et al (2010) compared the performance of different textural features for the problem of defect detection in wire rope surfaces. In Platzer et al (2009), they focused on contextual anomaly detection by modeling the intact class with help of Hidden Markov Models. Haase et al (2010) analyzed contextual anomalies in the rope surface with help of an autoregressive model which predicts the intact surface appearance given its neighborhood.…”

An Analysis-by-Synthesis Approach to Rope Condition Monitoring

“…As the most important problem in this context is a lack of missing defective examples for supervised learning strategies, Platzer et al present a oneclass classification approach to surface defect detection in wire ropes based on different features [2]. In [3] the same authors make use of Hidden Markov models to solve the problem of defect localization in wire ropes. However, none of these approaches focuses on an automatic estimation of meaningful rope parameters, which would close the gap between all the different inspection techniques.…”