Abstract
Automatic visual inspection takes a relevant place in defect detection of industrial production. In this field a fundamental role is played by methods for the detection of superficial anomalies on manufactures. In particular, several systems have been proposed in order to reduce the burden of human operators, avoiding the drawbacks due to the subjectivity of judgement criteria (Kwak, Ventura & Tofang-Sazi 2000, Patil, Biradar & Jadhav 2005). Proposed solutions are required to be able to handle and process a large amount of data. For this reason, neural networks-based methods have been suggested for their ability to deal with a wide spread of data (Kumar 2003, Chang, Lin & Jeng 2005, Garcia 2005, Graham, Maas, Donaldson & Carr 2004, Acciani, Brunetti & Fornarelli 2006). Moreover, in many cases these methods must satisfy time constrains of industrial processes, because the inclusion of the diagnosis inside the production process is needed. To this purpose, architectures, based on Cellular Neural Networks (CNNs), revealed successful in the field of real time defect detection, due to the fact that these networks guarantee a hardware implementation and massive parallelism (Bertucco, Fargione, Nunnari & Risitano 2000), (Occhipinti, Spoto, Branciforte & Doddo 2001), (Perfetti & Terzoli 2000). On the basis of these considerations, a method to identify superficial damages and anomalies in manufactures has been given in (Fornarelli & Giaquinto 2007). This method is aimed at the implementation by means of an architecture entirely formed by Cellular Neural Networks, whose synthesis is illustrated in the present work. The suggested solution reveals effective for the detection of defects, as shown by two test cases carried out on an injection pump and a sample textile.
TopArchitecture
The detailed implementation of each module will be illustrated in the following. Successively the results obtained by testing the suggested architecture on two real cases are shown and a discussion of numerical outcomes is reported.
Key Terms in this Chapter
Histogram Stretching: A point process that involves the application of an appropriate transformation function to every pixel of a digital image in order to redistribute the information of the histogram toward the extremes of a grey level range. The target of this operation consists of enhancing the contrast of digital images.
Real Time System: A system that must satisfy explicit bounded response time constraints to avoid failure. Equivalently, a real-time system is one whose logical correctness is based both on the correctness of the outputs and its timeliness. The timeliness constraints or deadlines are generally a reflection of the underlying physical process being controlled.
Fuzzy Associative Memory: A kind of content-addressable memory in which the recall occurs correctly if input data fall within a specified window consisting of an upper bound and a lower bound of the stored patterns. A Fuzzy Associative Memory is identified by a matrix of fuzzy values. It allows to map an input fuzzy set into an output fuzzy one.
Major Voting: An operation aiming at deciding whether the neighbourhood of a pixel in a digital image contains more black or white pixels, or their number is equal. This effect is realized in two steps. The first one gives rise to an image, where the sign of the rightmost pixel corresponds to the dominant colour. During the second step the grey levels of the rightmost pixels are driven into black or white values, depending on the dominant colour, or they are left unchanged otherwise.
Cellular Neural Networks: A particular circuit architecture which possesses some key features of Artificial Neural Networks. Its processing units are arranged in an M×N grid. The basic unit of Cellular Neural Networks is called cell and contains linear and non linear circuit elements. Each cell is connected only to its neighbour cells. The adjacent cells can interact directly with each other, whereas cells not directly connected together may affect each other indirectly because of the propagation effects of the continuous time dynamics.
Region of Interest: A selected subset of samples within a dataset identified for a particular purpose. In image processing, the Region of Interest is identified by the boundaries of an object. The encoding of a Region of Interest can be achieved by basing its choice on: (a) a value that may or may not be outside the normal range of occurring values; (b) purely separated graphic information, like drawing elements; (c) separated semantic information, such as a set of spatial and/or temporal coordinates.
Image Matching: Establishment of the correspondence between each pair of visible homologous image points on a given pair of images, aiming at the evaluation of novelties.
Automated Visual Inspection: An automatic form of quality control normally achieved using one or more cameras connected to a processing unit. Automated Visual Inspection has been applied to a wide range of products. Its target consists of minimizing the effects of visual fatigue of human operators who perform the defect detection in a production line environment.