A Contrario multi-scale anomaly detection method for industrial quality inspection

Abstract

Anomalies can be defined as any non-random structure which deviates from normality. Anomaly detection methods reported in the literature are numerous and diverse, as what is considered anomalous usually varies depending on particular scenarios and applications. In this work we propose an a contrario framework to detect anomalies in images applying statistical analysis to feature maps obtained via convolutions. We evaluate filters learned from the image under analysis via patch PCA, Gabor filters and the feature maps obtained from a pre-trained deep neural network (Resnet). The proposed method is multiscale and fully unsupervised and is able to detect anomalies in a wide variety of scenarios. While the end goal of this work is the detection of subtle defects in leather samples for the automotive industry, we show that the same algorithm achieves state-of-theart results in public anomalies datasets.

Image Forgery Detection Based on Noise Inspection: Analysis and Refinement of the Noisesniffer Method

“…Instead, we propose a heuristic approach to reduce the number of regions that will be evaluated using the a contrario approach described in the previous section. The construction of such candidate regions is based on the greedy algorithm proposed by Grompone et al [17], and the modifications introduced in [35].…”

“…Starting from a seed cell, the region-growing algorithm iteratively adds neighbor cells that satisfy the region-growing criterion. To define this criterion, we follow the approach introduced by Tailanian et al [35]. Namely, in order to decide if a cell is to be added or not to the region, we evaluate the NFA value of the region with and without this cell.…”

U-Flow: A U-shaped Normalizing Flow for Anomaly Detection with Unsupervised Threshold

“…The a contrario framework [8] is a multiplehypothesis testing methodology that has been successfully applied to a wide variety of detection problems, such as alignments for line segment detection [42], image forgery detection [43], and even anomaly detection [44], to name a few. It is based on the non-accidentalness principle [45].…”