Search

With the fast growth of intelligent manufacturing industry, developing advanced industrial inspection robots is becoming a research and application hotspot in the fields of both computer vision and robotics. This kind of industrial inspection robots is expected to automatically detect anomalous structures (e.g., defects, damages, rejects, etc.) from the images of the manufactured products. Generally, the existing visual anomaly detection (VAD) methods mainly focus on modeling the complex and high-dimensional distribution of normal data, while neglecting the specific visual properties of abnormal data since their frequency of occurrence is much less than that of the normal data. In this paper, inspired by the human cognition on extracting abstractly visual properties and to distinguish the anomaly patterns from the observed data, we propose a novel cognitive VAD method for industrial inspection robot. Specifically, we introduce a constrained latent space to mimic the cognitive ability of humans, where the abstraction learned from the observed normal and anomaly data are represented. We build our method based on a convolutional generative adversarial network and a denoising auto-encoder, where the adversarial learning mechanism is adopted to establish the boundary between the normal and anomaly data. In the experiment, we evaluate our method on a real-world dataset where the images are captured for the manufactured products. The comprehensive results comparing with several recent VAD methods show that the proposed method is effective to detect the anomaly images of different categories with a high accuracy.

Fast and reliable industrial inspection is a main challenge in manufacturing scenarios. However, the defect detection performance is heavily dependent on manually defined features for defect representation. In this contribution, we investigate a new paradigm from machine learning, namely deep machine learning by examining design configurations of deep Convolutional Neural Networks (CNN) and the impact of different hyper-parameter settings towards the accuracy of defect detection results. In contrast to manually designed image processing solutions, deep CNN automatically generate powerful features by hierarchical learning strategies from massive amounts of training data with a minimum of human interaction or expert process knowledge. An application of the proposed method demonstrates excellent defect detection results with low false alarm rates.

The inspection of moving specular freeform surfaces is an industrial challenge so far largely unsolved, even for the qualitative case, i.e. the mere determination of the presence of surface defects as opposed to the quantitative reconstruction of a surface. Products produced in high quantities therefore still have to be inspected manually which is labour intensive, expensive, monotonous and subjective. We propose a novel hardware setup and methodology to overcome this shortfall. The reflection of a line laser from a moving surface is captured on a translucent screen; surface defects show as gaps or bulges. Two methods to extract the resulting information are proposed and ways for its interpretation are shown. The proposed method is very cost effective and easy to implement. While limitations to surface geometry exist and absolute precision is not achievable, it is shown that the system is able to reliably detect, characterise and localise a range of typical surface defects on moving glazed ceramic tiles, our example application. The method is however applicable to a wide range of hybrid and specular surfaces.

The Industrial Metrology and Industrial Inspection systems make student conversant with the latest trends of measuring efficiently and build industrial inspection and metrology systems. Therefore, this lab manual has been developed based on the common inspection and metrology platform, combining real-time data processing into sophisticated measurement systems for industrial applications.

The traditional job of a quality control inspector, whether a human or a machine- vision device, is to discriminate products into “acceptable” and “rejectable” items. Acceptable items are those which confirm to a set of predetermined standards or quality characteristics. The performance of inspectors can be measured in terms of rejecting conforming items (type I error) or classifying non-confirming items as acceptable (type II error). Such inspection systems tend to be error prone. Reinspection is often used as a final means to remedy this situation and ensure higher outgoing quality. Effectiveness is limited by the fault rate effect, which shows decreasing defect detection performance with successive stages of reinspection. This paper uses a model of the inspector as a conservative signal detector to examine the various ways in which the positive attributes of human and machine-vision inspection systems can be combined to achieve enhanced system performance.

This paper presents an integrated and knowledge-driven boundary extractor, which can extract closed boundaries of touching objects directly from grey-level images in specific, but generic, real-world industrial environments.

This paper focuses on the application of an object-oriented design methodology (OMT) in an industrial research project The aim of the project is the development of a flexible machine vision workbench environment for the rapid-prototyping of real-time industrial inspection systems. This tool provides for the rapid configuration of systems by an engineer and, using novel verification techniques, allow rapid assessment of candidate solutions. The objective is to support the development of advanced machine vision systems which offer a high degree of reliability, together with known logical and temporal performance. The paper discusses the design methods and quality assurance measures taken to ensure compliance with international standards, describes how the design methodology chosen can be used to model knowledge bases, and finally describes the development of the delivered tool.

Neural networks have been successful at pattern recognition and discovery of hidden relationships amongst parameters and as such are likely supplements to the sensory systems employed in industrial applications. This paper examines four resulting issues imposed upon any industrial inspection system using a neural network: the feature set which the sensory system must provide, the accuracy of neural-network-based inspection, the robustness required of the sensory system for accurate inspection, and the computational burden imposed by accuracy requirements. This is accomplished in the context of web-process inspection, which requires rapid examination of vast amounts of data for on-line detection of faults in the sheet material. Each of the four crucial issues is addressed: 1. (i) Feature vectors with nine or 17 dimensions, created by a simulated segmented photodetector using measurement of the angular distribution over a 25° cone angle of the scattering were evaluated for inspection CrO 2 -coated sheet steel samples. The scattered coherent light from the surface of the material being processed could be directly conditioned by a photodetector so as to produce this small set of features which are then examined by a neural network trained to find and categorize unsatisfactory surface conditions. Details are presented to show how a modified feature set was developed and tested after an examination of feature space. This new, smaller set proved to be more accurate than the larger set. 2. (ii) Classification by fault or no fault categorized 133 samples correctly out of 135, while there were seven errors in one attempt at classification into the various common surface faults out of the same number of test samples and nine in another. It is shown that a bit of insight in feature selection can improve the capability of the network to recognize faults. 3. (iii) The robustness issue is important since the inspection system must function in the industrial environment, where maintaining an exact alignment of the optics is not feasible. Tests are described where it is shown that fault classification using the proposed system is reasonably robust to slight variation of the angle between the laser beam and camera. 4. (iv) The computational issue is discussed in the context of the data handling requirements of the inspection system.

Low cost, flexibility and deterministic response times are three major requirements of modern industry. To address these issues, industrially-relevant control systems must increasingly be made capable of adaptive and dynamic behaviour by being endowed with explicit knowledge and the ability to process this knowledge in a deterministic fashion. The question of representation of knowledge has thus become critical, as it is fundamental to knowledge management. In this paper, two essential points - knowledge acquisition and knowledge reasoning - both of which have a direct bearing on the issue of knowledge representation, will be examined. It will be shown how the combination of appropriate representation together with well-chosen acquisition and reasoning procedures, can lead to a flexible and industrially-relevant solution. To illustrate and justify these claims, the paper will conclude with an extensive set of experiments in on-line inspection.

A brief survey is given of the eye movement parameters which contribute to visual search strategies. The literature on eye movements and industrial inspection is reviewed and some recent studies carried out at the University of Birmingham into the eye movements underlying four inspection tasks are described. These tasks included the inspection of tin cans, electrical edge connectors, men's briefs and woven fabric.

A methodology is described for reliably locating pre-defined features in constrained images. The method uses a geometric model incorporating knowledge about the images involved, and a grey-level model capable of identifying specific features in a grey level profile. Results from a real application are presented.

Inspection is a widespread activity in industry with many problems. This article describes one of the methods, for analysing human inspection performance, which has been developed and applied recently, with some success.

One of the major determinants of inspection accuracy is speed of working. From a knowledge of the speed-accuracy trade off in inspection, it is possible to predict the optimum economic speed of inspection and hence to determine manning levels for inspection. This paper describes seven studies and shows that similar results were obtained irrespective of the type of product examined. Theoretical models of inspection are described which partially account for these effects. A method for the determination of optimum speed in economic terms is given.

In this work a new approach/method for detection of hidden defects by lateral thermal flows is proposed. The method is based on propagation of the thermal wave in a lateral direction in the sample volume and observation of the surface thermal distribution. This approach allows to reveal the hidden defects/cracks which are not detectable by conventional thermal inspection methods. The experimental investigations were conducted on steel samples with the defects which do not satisfy the detection criteria empirically established for thermal methods. The defects were successfully revealed. The proposed technique has a high sensitivity to hidden structures and can be applied in industrial inspection procedures.

Vibrothermography, or sonic IR, is a nondestructive evaluation technique used to find surface and near surface defects—such as cracks and delaminations—through observations of vibration-induced heat generation. This method has significant interest as an industrial inspection method, however, a lack of understanding of the fundamental physics governing the heat generation process has limited its application despite extensive theoretical, numerical simulation, and experimental work. Significant theoretical and numerical simulation work has been performed, but has yet to be rigorously verified experimentally. This paper presents experimental verification of the sources of heat generation in vibrothermography; specifically friction, plasticity, and viscoelasticity. Specific experimental evidence is presented that verifies each of these heat-generating mechanisms.

When an operator first detects unusual and/or infrequent or irregular signals in a system, the operator does not need to take any action, but must increase his/her level of attention and be alert for any more serious signals that may confirm a problem with the system. This is referred to as extended vigilance. The purpose of this study was to construct a fuzzy vigilance-measuring model for countering the loss of extended vigilance. The model extends two-valued logic (“Yes” or “No”) to multi-valued logic through fuzzy sets. Then a fuzzy logic alarm was developed by the model for combating the extended vigilance decrement. The first experiment compared the effect of the fuzzy measuring model with signal detection theory (SDT) and discussed the relationship between preliminary and extended vigilance for a simulated feed-water monitoring system. The results indicated that the sensitivity of the fuzzy vigilance-measuring model is better than index d ′ and β , and that the preliminary vigilance significantly influences the extended vigilance. The second experiment verified the effect of the fuzzy logic alarm. The results indicated that the effect of the fuzzy logic alarm for combating the extended vigilance decrement is significant. When the preliminary vigilance is poor, an appropriate alarm will improve the extended vigilance. However, if the preliminary vigilance is very poor, the improvement of the extended vigilance will be limited. Relevance to industry: The extended vigilance is a new and important issue in human performance problems in industry, particularly in such areas as air-traffic control, industrial inspection and power plant monitor. The measuring model of vigilance could be applied to develop an alarm for promoting supervisory performance of human and human–machine detectors.

A diffractive optical element (DOE) based sensor is presented for measurement of gloss differences of injection molded convex or concave plastic products. Increases of local gloss of the samples, and hence local wear of the tool, were observed to correlate linearly with the DOE signal by exploiting a simple image analyzing method. The measurement is based on the utilization of the fully developed speckle pattern, which grain size is much larger than the aperture of the DOE. Corroborative measurements by spectrophotometer and a commercial glossmeter confirmed the reliability of the DOE sensor. It is suggested that robust DOE sensor can be utilized for industrial inspection of non-planar plastic products.

A real-time implementation of an approximation of the support vector machine (SVM) decision rule is proposed. This method is based on an improvement of a supervised classification method using hyperrectangles, which is useful for real-time image segmentation. The final decision combines the accuracy of the SVM learning algorithm and the speed of a hyperrectangles-based method. We review the principles of the classification methods and we evaluate the hardware implementation cost of each method. We present the combination algorithm, which consists of rejecting ambiguities in the learning set using SVM decision, before using the learning step of the hyperrectangles-based method. We present results obtained using Gaussian distribution and give an example of image segmentation from an industrial inspection problem. The results are evaluated regarding hardware cost as well as classification performances.

The traditional job of a quality control inspector, whether human or a machine-vision device, is to discriminate products into acceptable and non-acceptable items. The performance of inspectors can be measured in terms of rejecting non-acceptable items (Type I error) or classifying non-acceptable items as acceptable (Type II error). Reinspection is often used afterward to remedy this situation if required. Unfortunately, all such inspections remain prone to errors. This paper uses an experimental investigation to research the various ways in which the positive attributes of human and machine-vision inspection systems can be arranged to achieve enhanced system performance in industrial inspection tasks.

We present an introduction to Multivariate Image Regression (MIR) with a selection of illustrative application studies. Generalisation from two-way multivariate calibration to the three-way regimen leads to—at least—three alternative image regression cases depending on the nature of the available Y-data: IPLS-Ydiscrim; IPLS-Ygrid; IPLS-Ytotal. A systematic image regression typology is briefly introduced. We here present the core of the principles of applied MIR. Two major MIR application studies are worked through, a food mass product industrial inspection study (IPLS-Ydiscrim) and a food product (fruit) storage stability image analytical monitoring (IPLS-Ygrid). These exemplifications are presented as archetypes, representing a much wider range of potential industrial/technological application areas. Based on simple three-channel imagery (in order to simulate many industrial systems), they nevertheless represent all higher-dimensional multivariate image cases as well, since the pertinent MIR principles and software are invariant w.r.t. any number of channels/variables employed. The present paper represents one major element of our work towards establishing a complete, stand-alone facility for Multivariate Image Regression (MIR); the second paper in this series deals with the development, implementation and extensive exemplifications of a complementary cross-validation facility.

This article describes some recent advances in elastic X-ray scatter as a means of non-invasive industrial inspection. Elastic scatter leads directly to X-ray diffraction and is dependent upon the molecular structure of the scattering medium. This makes the technique particularly sensitive to discrimination between low-Z materials and has been exploited in such areas as the detection of “soft”, or low-Z, contaminants in foodstuff, for the discrimination of oil and water and in the detection of mineral deposits. These applications are discussed together with an outline of X-ray scattering theory and an overview of the various means of exploiting elastic X-ray scatter for materials inspection and imaging.

This paper studies the application of advanced computer image processing techniques for solving the problem of automated defect detection for textile fabrics. A new defect detection scheme is proposed, which consists of an odd symmetric real-valued Gabor filter, an even symmetric real-valued Gabor filter and one smoothing filter. In developing the scheme, the Gabor filters are designed on the basis of the texture features extracted optimally from a non-defective fabric image by using a Gabor wavelet network (GWN). The performance of the proposed defect detection scheme is evaluated off-line by using a set of fabric images taken from a database consisting of a wide variety of homogeneous fabric images. The results exhibit accurate defect detection with low false alarms, thus showing the effectiveness and robustness of the proposed scheme. To evaluate the performance of the proposed defect detection scheme further, real-time tests are conducted by using a prototyped automated defect detection system. The experimental results obtained further confirm the efficiency, effectiveness and robustness of the proposed detection scheme.

Many industrial inspection tasks can be performed satisfactorily by conventional image processing systems which utilise feature extraction and related feature measuring algorithms. However, for the inspection of parameters such as texture or colour, these techniques may be inadequate. This paper outlines a new system which incorporates algorithms for coloured image processing and a logical neural network for subsequent pattern classification. Past research has indicated that logical neural nets can provide quite a good recognition performance for artefacts which are complex. However, in the area of coloured object inspection, the inspection tasks are more difficult. Furthermore, if the lighting source changes, the image of a coloured object will vary considerably and this will significantly affect the outputs of the networks. This paper presents a practical approach for colour invariant recognition and a method of adaptive sampling which utilises colour information more effectively. The system which is armed with these techniques provides a good recognition performance for artefacts which are both complex and coloured and, also, it is robust in terms of tolerance to noise and lighting.

In order for an industrial inspection system to be of utility in manufacturing it must be fast, accurate, and flexible. It would seem that the best method for creating flexible machine vision systems is to define a subclass of inspection problems where all the problems within the subclass have a number of common features. Such a subclass must be of interest to a number of manufacturers. It must also be "reasonable," to solve. Once the subclass has been selected, the next logical step is to create a device that makes performing all the needed experiments on the various problems within the class easy. This work has four major objectives. The first objective is to define a meaningful subclass of inspection problems that are (a) of interest to a number of manufacturers, and (b) represent inspection tasks that seem "reasonable" within the current state-of-the-art of computer vision. The subclass selected for this work is the longitudinal planar web inspection problem under the two-dimensional imaging restriction. The second objective of this work is to create a vehicle that will allow the types of experimentation usually associated with the development of machine vision systems to be facilitated. This vehicle created is called a "machine vision prototyper." The third objective of this work is to use the machine vision prototyper system to attack a particular planar web applications problem. The application considered is the problem of locating and identifying surface defects in surfaced hardwood lumber in a species independent manner. The fourth objective of this research is to indicate how the prototyper system can be used to attack a second planar web application problem. This application problem is the inspection of hardwood parts coming out of a molder. The utility of the machine vision prototyper system as an experimental tool is demonstrated on two of the three possible types of longitudinal planar web inspection problems. The results include the development of a machine vision system for a hardwood surfaced lumber surface feature detection problem, and a discussion of how the prototyper can be used to attack the problem of inspecting hardwood parts coming out of a molder. (Abstract shortened by UMI.)

Detection of closely spaced edges in images contaminated by noise is important in medical applications, industrial inspection, advanced TV systems, and image understanding. Laplacian-of-Gaussian (Marr-Hildreth) edge detectors are known from their tunable resolution to edge spacing. However, when they are tuned for the detection of very closely spaced edges, their noise performance deteriorates. In this paper, a class of nonlinear edge detectors suitable especially for the detection of closely spaced edges is introduced. The detectors are based on a difference between linear and median-type filters and use the principle of zero-crossings. Computer simulations indicate that the nonlinear edge detectors perform better in noisy conditions than Marr-Hildreth edge detectors of comparable resolution.

A big challenge for automated industrial inspection systems is the hundred percentin-line surface inspection. Because of material speeds up to 10 m/s and material widths of several meters the amount of data received by the sensor systems is very large. In most cases the material flow is endless and therefore it is necessary that the complete inspection system works in real-time. Nearly 25 years ago the first systems based on laser scanning technology were introduced in industry. At the moment more than one thousand laser scanning systems are working worldwide. During the last five years a competitive sensor system, the CCD camera, became commercially available. In the present paper we will give an overview and, if necessary, a closer look of the differentitems of both technologies. Advantages and limitations of both systems are discussed. This discussion is not based on pure theory but mostly on real experiences. Therefore a couple of industrial applications are shown. In conclusion we will give an outlook on future developments in sensor technology and evaluation techniques.

This paper describes a computationally efficient method for identifying weakly visible cracks in images of man-made surfaces. The method relies on a multistage nonlinear filtering process designed to yield maximum response to symmetrical fine structure. The result is a set of candidate sites which can then be verified on the basis of features obtained from the source image and candidate shape. Results are presented which demonstrate the application of the method to a difficult problem in industrial inspection.

In the search for ever more sensitive non-destructive testing (NDT) techniques to detect progressively smaller defects in increasingly complex industrial materials, environments and geometries, conventional eddy-current testing (ECT) methods, have reached their sensitivity limit. Fortunately, advanced resonance-based techniques, demonstrated in recent years, promise to help extend that limit, but, before such techniques can be implemented with confidence, a detailed understanding of their sensitivity and stability must be achieved. In this study, statistical probability-of-detection (PoD) analysis was performed to assess the sensitivity of a novel single-frequency, near electrical resonance signal enhancement (SF-NERSE) technique relative to an equivalent conventional probe operation. This study was performed on 36 real fatigue defects in Titanium 6V-4Al (Ti6-4) with defects ranging from 0.10 − 6.48 mm in surface extent. In addition, a critical evaluation of background noise stability in the SF-NERSE technique (a common concern with NERSE-based methods) was also performed to establish the viability of such a technique in relation to industrial inspection and assessment criteria. The results of this study demonstrate a sensitivity enhancement of up to 20% for the SF-NERSE method over conventional operation and, through controls, confirmed that the effect is a result of the resonance-shifting phenomenon and not just an increase in operational frequency. In addition, an examination of background noise implies that a SF-NERSE method exhibits more stable background noise than conventional excitation. This study validates the SF-NERSE technique and methodology as a viable industrial inspection technique, able to significantly improve detection capabilities.

Material attribute recognition from the visual appearance of attributes is an important problem in computer vision field. However, few works model the hierarchical relationship between material attributes and their deep-level semantics that occurs in the same image. Meanwhile, single image feature is insufficient to achieve high-quality material attribute classification. In this paper, we present methods for generating a new hierarchical material attribute representation mechanism using a new-designed feature mid-fusion algorithm and the state-of-the-art relative attribute (RA) model. The novel feature mid-fusion algorithm can improve the performance of material attribute classification. The deep-level semantics of material attributes are mined by the state-of-the-art RA model. They provide considerable useful and detailed knowledge on material attributes. We call the novel feature mid-fusion algorithm gene selection eXtreme gradient boosting (GS-XGBoost). This algorithm considers the state-of-the-art boosting idea (eXtreme gradient boosting) and the popular multi-feature fusion idea (effective range-based gene selection). To comprehensively describe material attributes, we also measure the relative degree of their deep-level semantics. A new hierarchical material attribute representation mechanism is constructed on the basis of the correctly classified material attributes and their deep-level semantics. The mechanism has two forms. One is binary attribute representation mechanism, and the other is relative attribute representation mechanism. We demonstrate the effectiveness of the proposed GS-XGBoost algorithm on two different datasets. The proposed GS-XGBoost algorithm is not an end-to-end framework but is efficient and practical for fine- and coarse-grained material attribute classification problems that can be applied in different scenarios in large-scale product image retrieval, robotics, and industrial inspection. The novel hierarchical material attribute representation mechanism will help humans or robotics accurately recognize diverse materials and their deep-level semantics. Our research contributes to not only computer science but also material science and engineering.

In the fields of customs inspection, airport security inspection, medical diagnosis, industrial inspection and arms control verification, it is often necessary to use radiography to determine whether the targeted object is contained in a closed container. By analyzing whether the specific figure of the targeted object is contained in the irradiation image of the closed container, we can determine if the container should be opened for a further manual check. In many cases, considering the time cost of radiography, the determination should be made as soon as possible, which requires fast analysis of the figure information of the irradiation image. In this paper, a fast figure determination method based on the sequential Bayesian analysis is proposed. This method applies the sequential Bayesian analysis to the irradiation image analysis, and can make a fast determination of the existence of the specific figure according to the location information of the ray particles in the irradiation image. First, the theoretical framework of this method was established in this paper. Second, the mathematical model of this method was established, and its feasibility was also studied. The research results show that, this method can make an accurate and fast determination of the existence of the targeted figure in the irradiation image. Third, the factors influencing the figure determination speed of this method was analyzed in depth. At last, a Monte Carlo simulation study of this method was carried out: the physical model of the inspection scheme of industrial materials based on the X-ray transmission imaging was established, and the figure determination method proposed by this paper can make an accurate and fast determination of the existence of hole defects in the industrial materials according to the X-ray transmission images of the industrial materials, which proves the feasibility of the application of this method in radiography.

In this study, the industrial inspection of optical thin film on touch screen panels was demonstrated using line field spectral domain optical coherence tomography. The conventional Fourier domain optical coherence tomography system requires a single scanner for two-dimensional cross-sectional images and two scanners for volumetric images. Our developed line field spectral domain optical coherence tomography has the advantage of needing only a single scanner for volumetric images, while two-dimensional cross-sectional images are obtained by the parallel acquisition of an illuminated line on a sample using an area camera. Further, the image acquisition speed was enhanced by implementing a high speed camera (340 frames per seconds) with improved quantum efficiency at near infrared region enabling two-fold frame rate. Cross-sectional and volumetric images were acquired to detect the internal sublayer defects in the optical thin films, which are difficult to observe using visual or machine vision-based inspection methods. The developed pseudo code for defect identification in optical thin films was well-utilized here for the defect inspection. The system characterization is demonstrated using United State Air Force (USAF) resolution target. The results indicate the possible application of the proposed system in touch screen panel inspection for the quality assurance of products at the consumer end.

Image preprocessing and edge detection are the two most critical steps when applying machine vision to measure workpiece dimension in industrial field. The surface of the workpiece is usually covered with interference regions, which makes edge detection difficult. Aiming at the problem, the Full Convolutional Network (FCN) model is used to detect interference regions of workpiece images, and then interference regions are processed by the directional texture repair method. To avoid the influence of workpiece surface texture after rough machining, a method based on Holistically-nested Edge Detection (HED) model for rough edge detection of workpiece images is proposed. The edge is obtained by HED, and then it is post-processed based on the non-maximum value suppression and double threshold segmentation methods in the Canny operator to obtain a refined edge image. In order to further improve the accuracy of the dimension measurement, sub-pixel level edge detection accuracy is achieved by cubic spline interpolation. Finally, the method is validated by a shaft workpiece. The measuring accuracy of the outer diameter of the workpiece can reach 0.02 mm, which can effectively meet the requirements of fast semi-finishing inspection. This research provides a common workpiece dimension measurement method in industrial field and methodological guidance for the application of deep learning in industrial inspection.

The aim of the current work was to investigate the modulation transfer function (MTF) of a non-destructive testing (NDT)/ industrial inspection CMOS sensor in conjunction with a calcium tungstate (CaWO4) thin screen, following both the IEC 62220-1:2003 and IEC 62220-1-1:2015 methods. Thin screen samples, with dimensions of 2.7x3.6 cm2 and mean coating thickness of 36.26 mg/cm2 (actual thickness: 118.9 μm estimated from scanning electron microscopy-SEM images) were extracted from an Agfa Curix universal screen and were manually coupled to the active area of a high resolution, active pixel (APS), complementary metal oxide semiconductor (CMOS) sensor. Experiments were performed using the RQA-5 beam quality, as described in the IEC series. Modulation transfer function was assessed using the slanted-edge method. The final MTF, following IEC 62220-1-1:2015 was obtained through averaging the oversampled edge spread function (ESF), using a custom-made software developed in our laboratory. MTF values were found with close agreement in the low and medium spatial frequency ranges, for both methods. Thereafter, MTFs calculated following the 62220-1:2003 protocol, were found overestimated for spatial frequencies higher than 5.7 cycles/mm. The combination of the thin calcium tungstate screen and the CMOS sensor provided very promising image resolution properties and thus could be also considered for use in CMOS based X-ray imaging devices, for various applications.

Automatic industrial inspection is critical to modern manufacturing enterprises. Due to their low cost and real-time processing speed, vision-based inspection systems are often used for this task. Deep Convolutional Neural Networks (CNNs) have been extensively applied to many computer vision tasks but require large numbers of annotated examples to train. Obtaining such annotations is expensive and time-consuming. In this study we describe a method which aims to make the best use of unannotated image data, which can often be collected easily. We propose a novel unsupervised local deep feature learning method based on image segmentation to build a network which can extract useful features from an image. The training algorithm alternates between (1) obtaining pseudo-labels by clustering the features extracted using a segmentation CNN and (2) training the CNN for feature learning using these pseudo-labels. To our knowledge, unsupervised local deep feature learning has not been addressed based on image segmentation in this way before. We demonstrate the approach on two aerospace weld inspection tasks. Our results show that the proposed unsupervised method performs almost as well as a CNN with the same architecture trained in a supervised manner.

The paper presents a number of signal processing approaches for the spectral interferometric interrogation of extrinsic Fabry-Perot interferometers (EFPIs). The analysis of attainable microdisplacement resolution is performed and the analytical equations describing the dependence of resolution on parameters of the interrogation setup are derived. The efficiency of the proposed signal processing approaches and the validity of analytical derivations are supported by experiments. The proposed approaches allow the interrogation of up to four multiplexed sensors with attained resolution between 30 pm and 80 pm, up to three times improvement of microdisplacement resolution of a single sensor by means of using the reference interferometer and noise-compensating approach, and ability to register signals with frequencies up to 1 kHz in the case of 1 Hz spectrum acquisition rate. The proposed approaches can be used for various applications, including biomedical, industrial inspection, and others, amongst the microdisplacement measurement.

To minimise a spectrometer maximally, using a micro-electro-mechanical system or planar waveguide chip, has aroused interest in both academia and industry. This is because a portable and inexpensive spectral analysis tool has tremendous potential in various applications including scientific exploration, industrial inspection, and environmental protection. Here, we design, fabricate, and characterise a Fourier-transform spectrometer (FTS) chip consisting of 101 Mach–Zehnder interferometers (MZIs) with gradually varying optical path differences on a silica planar waveguide platform, aiming to cover the entire visible band. To ensure a wide-band operation while simultaneously compressing the size of the chip, wide-bending arm waveguides that can support high-order modes in a short wavelength range are employed in the design of the MZI array. The experimental result of our FTS chip shows its ability to analyse the visible spectrum with a minimum resolution of better than 10 nm and to reproduce the original spectrum with an accuracy of less than 0.5 nm. The successful implementation of the FTS chip based on a matured and low-cost silica planar lightwave circuit process makes it quite promising to realise mass production and widespread application.

Since conventional mechanical punching technology for Flexible Printed Circuit Board (FPCB) drilling has restricted via-hole size and depth control for multi-layer circuit boards, CO 2 and Ultra Violet (UV) laser drilling technologies have been developed. However, the FPCBs for mobile phones and Personal Digital Assistants (PDAs) require smaller via-hole diameters, since the development of thinner and higher circuit density devices is demanded. Currently, UV laser systems are widely used for FPCB drilling of 75–105 μm diameter via-holes and inspectors performs quality test manually using microscopes. We developed a high-precision UV laser microfabrication system for next-generation FPCB drilling of 15 μm diameter via-holes. The degrees of the precision of the microfabricated via-holes of 15, 35, 50 and 85 μm were mean absolute error rate of 4.4, 2.2, 2.3, and 2.2 which was fully satisfied with industrial inspection specification ±10%. The drilling speed of the system of 2800 via-holes per second at stationary state was achieved. In addition, we applied modified Greedy 2-opt algorithm to find out optimal drilling path which reduced the total time of via-hole fabrication. We successfully reduced the production time by 25% compared with the result obtained in the normal Greedy 2-opt algorithm. Moreover, we designed very accurate inspection method using Canny edge detection and geometric pattern matching algorithms and successfully applied it to the Automated Optical Inspection (AOI) module for the inspections of 15 μm diameter via-hole which was required for the fabrication of high density FPCB.

Object recognition systems constitute a deeply entrenched and omnipresent component of modern intelligent systems. Research on object recognition algorithms has led to advances in factory and office automation through the creation of optical character recognition systems, assembly-line industrial inspection systems, as well as chip defect identification systems. It has also led to significant advances in medical imaging, defence and biometrics. In this paper we discuss the evolution of computer-based object recognition systems over the last fifty years, and overview the successes and failures of proposed solutions to the problem. We survey the breadth of approaches adopted over the years in attempting to solve the problem, and highlight the important role that active and attentive approaches must play in any solution that bridges the semantic gap in the proposed object representations, while simultaneously leading to efficient learning and inference algorithms. From the earliest systems which dealt with the character recognition problem, to modern visually-guided agents that can purposively search entire rooms for objects, we argue that a common thread of all such systems is their fragility and their inability to generalize as well as the human visual system can. At the same time, however, we demonstrate that the performance of such systems in strictly controlled environments often vastly outperforms the capabilities of the human visual system. We conclude our survey by arguing that the next step in the evolution of object recognition algorithms will require radical and bold steps forward in terms of the object representations, as well as the learning and inference algorithms used.

In this paper the problem of a vision system implementation for autonomous flying agents is considered in the context of industrial inspection tasks performed by unmanned aerial vehicles. A syntactic algorithm of a two-dimensional object vectorization and recognition is proposed. An algorithm of two-dimensional map recognition has been introduced as well. The algorithms have been tested by using both artificial data and real data - the satellite image. They have turned out to be effective.

Reliable identification of the discontinuities from range data is an important problem in many application domains, such as reverse engineering, object recognition, industrial inspection, and autonomous navigation. In this paper, a novel method is proposed for identifying the C0, C1, and C2 discontinuities from range data based on similarity indicators. We research the behaviors of the principal curvatures at the discontinuities and define the similarity indicator at a point. Based on the similarity indicators, the C0, C1, and C2 points are identified from range data. The similarity indicators make the threshold setting easier and more reasonable, and improve the performance of discontinuity identification. The experimental results demonstrate that the proposed method is efficient and reliable.

This paper presents an accurate stereo vision system for industrial inspection, which uses a self-calibration method based on photogrammetry. A cross-shaped calibration pattern which is portable and easy to be manufactured is designed. The cross target can be used to calibrate stereo vision systems and obtain higher measurement precision conveniently. The mathematical model of the stereo vision system with 10 distortion parameters for each camera is proposed. The feature point detection method with sub-pixel accuracy is explored. The calibration initial values are computed using the relative orientation method and the direct linear transform (DLT) method of photogrammetry. The bundle adjustment algorithm is used to optimize the calibration parameters as well as the 3D coordinates of the feature points. Experiment results show that the RMS error of the reprojection in our method is less than 0.05 pixels and the distance measurement error is 0.031 mm with a high precision scale bar which length is 221.001±0.003 mm.

This paper presents an efficient method for the detection of delamination using a combined technology of ultrasonic guided wave and electromagnetic acoustic transducer (EMAT). The material for inspection studied in this paper is a three-layered brass/copper/brass metal composite for coin stocks. We first observed a cyclic behavior of wave sensitivity to delamination width in the experiment. This observation is then confirmed with finite element simulation. We found out that the cyclic behavior is related to the phase velocity of the converted wave modes in the two subsystems after delamination. To compensate for the cyclic behavior, two wave modes are used in an industrial inspection system to ensure effective detection of delamination in the composite sample. The inspection speed achieved with this EMAT inspection system is 1.8 m2/s.

In this study, we propose a robust technique based on invariant moments - adaptive network based fuzzy inference system (IM-ANFIS). In this technique, some digital image processing methods such as noise reduction, contrast enhancement, segmentation, and morphological process are used for feature extraction stage of IM-ANFIS approach used in this study. Recently, the pattern recognition principles have come into prominence. The pattern recognition includes operation and design of systems that recognize patterns in data sets. Important application areas of pattern recognition techniques are character recognition, speech analysis, image segmentation, man and machine diagnostics and industrial inspection. The technique presented in this study enables to classify 16 different parasite eggs from their microscopic images. This proposed recognition method includes three stages. In first stage, a preprocessing subsystem is realized for obtaining unique features from the same group of patterns. In second stage, a feature extraction mechanism which is based on the invariant moments is used. In third stage, an adaptive network based fuzzy inference system (ANFIS) classifier is used for recognition process. We conduct computer simulations on MATLAB environment. The overall success rate is almost 95%.

In the performance assessment of an image processing algorithm, an image is often compared with an available reference. Measuring image similarity can be achieved in many ways; comparison algorithm varies from pixel-based mean square error method to structure-based image quality index. In this paper, we present a new feature-based approach that utilizes image phase congruency measurement to quantify the assessment of the similarities or differences between two images. Test results with standard images and industrial inspection images are presented.

Industrial Geometry aims at unifying existing and developing new methods and algorithms for a variety of application areas with a strong geometric component. These include CAD, CAM, Geometric Modelling, Robotics, Computer Vision and Image Processing, Computer Graphics and Scientific Visualization. In this paper, Industrial Geometry is illustrated via the fruitful interplay of the areas indicated above in the context of novel solutions of CAD related, geometric optimization problems involving distance functions: approximation with general B-spline curves and surfaces or with subdivision surfaces, approximation with special surfaces for applications in architecture or manufacturing, approximate conversion from implicit to parametric (NURBS) representation, and registration problems for industrial inspection and 3D model generation from measurement data. Moreover, we describe a 'feature sensitive' metric on surfaces, whose definition relies on the concept of an image manifold, introduced into Computer Vision and Image Processing by Kimmel, Malladi and Sochen. This metric is sensitive to features such as smoothed edges, which are characterized by a significant deviation of the two principal curvatures. We illustrate its applications at hand of feature sensitive curve design on surfaces and local neighborhood definition and region growing as an aid in the segmentation process for reverse engineering of geometric objects.

The normalized cross correlation (NCC) has been used extensively in machine vision for industrial inspection, but the traditional NCC suffers from false alarms for a complicated image that contains partial uniform regions. In this paper, we study the use of NCCs for defect detection in complicated images. The performance of NCCs in monochrome and color images, and the effect of image smoothing are empirically evaluated. The proposed NCC in a smoothed color image can effectively alleviate false alarms in defect detection applications.

This paper shows how line segment detection can be achieved with a minimum of computation if two masks embodying a vectorial design strategy are employed. To achieve the greatest combination of accuracy and speed, a two-stage procedure was used with the vectorial operator backed up by a template matching technique. Experimental tests verified the effectiveness of this type of line segment detector for locating insects in cereal grain images: in a set of 60 images containing 150 insects, there were no false positives and the only false negative arose from two insects which were in contact. The approach should be useful in a good many other areas ranging from industrial inspection to document processing and remote sensing--and indeed, in the many applications where line segments have to be located or thin lines have to be tracked.