An algorithm for 3D surface reconstruction of large objects using a structured light pattern ranging system is presented. Highly accurate industrial inspection applications have been constrained by the limited range resolution and accuracy of current ranging devices and techniques. To overcome the limited range resolution, the ranging sensor uses a small field-of-view and multiple views. The proposed algorithm fuses surface data patches from the views to construct a large object surface. The algorithm also increases the accuracy of the reconstructed object with efficient numerical analysis and pre-processing. Experimental results show that the algorithm and the current sensor setup can reconstruct an object for inspection applications with the accuracy of approximately 1 mil (2.54μm ) tolerance.

In a prototype for monitoring hot steel wire different technologies are integrated to achieve a robust, flexibly configurable and scalable imaging system. It is designed as a distributed system with private network and Tuplespace communication implementable on a LINUX Server. Intelligent cameras grab and process the image data. For real time communication between the cameras and standard industrial I/O-modules (IEC-61131) MODBUS/TCP messaging is applied. A switch with integrated firewall makes services available to the supervisory control system. Results are available as XML-logfiles. The image processing defines the upper and lower edges of the material by minimum/maximum filtering of the y-gradient. Dual Grassmanian coordinates are used to fit two parallel lines to the edge points by singular value decomposition. This gives the distance between the lines and the confidence interval of each measurement simultaneously, whereas latter is used to reject poor data. Changes of the distance are analysed computing local central moments. Presently, 12 images per second are acquired. The application is able to detect spontaneous rotation of the wire around the axis of rolling directly at the rolling stands and treats also poor images (due to steam of cooling water). It indicates resulting defects, which may go undetected otherwise.

This paper describes the research approach to identify the contaminants which have similar color to the background wool and their removal from wool in real time. First, different light source is sought for getting the high contrast image between wool and contaminants. Second, different CCD detector including infrared camera, monochrome area scan camera was tried for identification of white contaminants. Relative infrared theory and spectral theory are also presented. Third, different image processing algorithms including threshold in HSV color space, local adaptive threshold, region-growing algorithm and their comparisons are presented. The combination of local adaptive threshold and global threshold algorithms can well identify most of white contaminants. At last, a research approach on contaminant removal from wool by the image processing algorithm in real time is presented. Both software and hardware approach are reported.

The traditional method for the evaluation of cashmere quality is subjective and low in accuracy. In this paper, a computer vision system is presented for the objective identification and classification of pigmented fibres, which consists of a web maker, an image acquisition system and a computer for image processing. The techniques of fibre preparation, image acquisition and the development of suitable algorithm together with software for removal of the background fibres and counting of pigmented fibres, are described in detail.

The use of laser technology to scan hardwood log surfaces for defects holds great promise for improving processing efficiency and the value and volume of lumber produced. External and internal defect detection to optimize hardwood log and lumber processing is one of the top four technological needs in the nation’s hardwood industry. The location, type, and severity of defects on hardwood logs are the key indicators of log quality and value. These visual cues provide information about internal log characteristics and products for which the log is suitable. We scanned 162 logs with a high-resolution industrial four-head laser surface scanner. The resulting data sets contain hundreds of thousands of three-dimensional coordinate points. The size of the data and noise presented special problems during processing. Robust regression models were used to fit geometric shapes to the data. The estimated orthogonal distances between the fitted model and the log surface are converted to a two-dimensional image to facilitate defect detection. Using robust regression methods and standard image processing tools we have demonstrated that severe surface defects on hardwood logs can be detected using height and contour analyses of three-dimensional laser scan data.

The NATM, a widely used tunnel excavation method, requires precise periodical monitoring of deformations especially at fault zones, which tends to hamper traffics with conventional measurement means. In this paper vision metrology was applied to tunnel profile measurement with a view to developing a new method. Two hundred of Retro-targets are placed on a one-meter spacing lattice at a tunnel site of 7m in diameter and 15m in longitude, and 66 images were taken to cover the target field. The object space coordinates of targets obtained by bundle adjustment were compared with ones obtained by high-precision total station observation. The root mean square (RMS) of differences of coordinates was 0.548mm, which is precise enough for monitoring deformations for the NATM.

Recently developed mining machines are capable of cutting different profiles. Cutting desired profiles opens new fields of application for these machines. The precision of the profile, which is cut, depends on the kinematics of the machine and its calibration. The dimensions of the profiles up to 10 m wide and 5 m high make it difficult to calibrate and even measure. This paper presents an image processing system, which was developed to solve this problem. An ultra-bright infrared LED was mounted on the primary calibration point of the machine. The 2-R manipulator (i.e. the cutting arm) is moved so as to generate the desired outer profile. The 2-R kinematics of the machine result in the calibration point moving along the surface of a torus. The imaging system acquires a sequence of images, each of them captures the machine in one point along the profile. This delivers a 2-D central projection of the 3-D motion. The inverse projection is determined using projective geometry. The true position of the calibration points is determined by applying the inverse projection, which is then compared to the desired position. Measurements of a mining machine and a comparison with the desired profile are presented.

In this study, we propose an original method for a 3D reconstruction of the relief of a textured rough surfaces. This 3D reconstruction is obtained through the elaboration of a photometric model, which takes into account camera and light source positions according to the plan of the rough surface. The proposed model expresses the gray level on the image according to the local relief variations. Three images of the same relief obtained under different angles of lighting are used to reconstruct the altitude map of the rough surface. The effectiveness of this method was checked by comparing the extracted relief to its corresponding relief obtained from a mechanical device method using autofocus laser sensor. This photometric model display good results in simulation experience and will be applied on real photographic images of road covering surface in order to study its wear level and its adherence.

Range imaging based on fast light-sectioning techniques is used to acquire the three dimensional shape of a steel block with its embedded flaws. The aim of the paper is the computation of suitable features for describing the characteristics of sequentially acquired profiles. The features should discriminate well between flaws, the intact surface, and pseudo errors caused by inhomogeneous surface properties such as, e.g., a strongly changing reflection factor. The orthogonal distance from the planar curve to its spline approximation serves as a suitable descriptor. In addition, the multi-scale curvature is useful since the defects are characterized as local perturbations of a relatively smooth curve. Since the embedded flaws on the surface of the steel block are existent in neighboring profiles, a measure for evaluating the local perturbations with respect to subsequent profiles is presented. Therefore, a kernel is used for weighting the neighborhood of a defect. The shape of the kernel might be Gaussian, however, a special kernel was developed for emphasizing a preference direction of the flaws.

The inspection of sidewall thickness provides important information about the production processes for glass container manufacture. By monitoring the thickness profile around the perimeter of bottles in real-time, the manufacturing process can be altered to produce higher quality products. This also provides the ability to identify and remove defective products. In order to improve the speed and accuracy of inspections, a new non-contact method for acquiring thickness profiles of glass bottles that employs optical and machine vision techniques has been developed and tested. One of the fundamental laws of optics, Snell's Law, is the basic concept upon which the inspection technique relies. The thickness of a flat plane of transparent material can be determined from Snell’s Law with a single beam of light that passes through the medium, reflects off the secondary surface, and travels back to the initial surface and passes through it. Based upon this principle, a new non-contact glass thickness measurement technique has been developed and it has demonstrated good accuracy.

This paper presents a computer vision system for measuring the weight of gobs during a glass forming process, and a control strategy to correct automatically any weight deviation from a given set-point. During the formation of molten glass gobs, several noise sources can cause a deviation in the weight from a predefined reference value. Among them, there is a random white-noise disturbance caused by the lack of synchronisation of mechanical devices, the periodic disturbances due to changes in the spinning direction of the tube inside the feeder, and some long-term drifts caused by variations in temperature and viscosity of the raw glass material. The gob weight measurement system developed is based on a monochrome CCD high-resolution camera and photo-detector for synchronizing the frame acquisition. The molten glass provides the illumination, so a high contrast image is obtained with a bright object and dark background. Several image-processing algorithms are presented for reliable area estimation. Assuming that the gob is a symmetric geometry of revolution and uniform mass density, the proposed system estimates the weight of gobs with an accuracy better than ±0.75%. A learning weight control strategy is proposed based on a PI-repetitive control scheme. The weight deviation from a set point is used as a control signal to adjust the glass flow into the feeder. This regulation scheme allows effective weight control, canceling mid and long-term effects. The tracking error, ±1.5%, means a reduction of 40% when compared with a traditional PI controller.

In this paper a new automatic strain analyzer system for sheet metal stamped parts is presented. This device is based on an optic 3D scanner working with structure white light, that performs measurements with high speed and accuracy. This new device can be a very useful tool in manufacturing industry.

Serial number plays an important role in the manufacturing and managing processes to identify the industry products, and automatic recognition of the serial number has become an essential step in modern industries. The limits of manufacture condition in factories and the requirement of real-time operation, however, make the design of a practical automatic serial number recognition system a difficult task. We developed a novel automatic serial number recognition system for railway wheel products, which has been successfully used in real production lines. This system uses visual inspection approaches to automatically recognize the serial numbers in real-time. The flexibility and low cost of the system make it very suitable for the applications described in this paper.

As a prospective intelligent sensing method for Autonomous Guided Vehicle (AGV), machine vision is expected to have balanced ability of covering a large space and also recognizing details of important objects. For this purpose, the proposed hybrid machine method here combines the stereo vision method and the traditional 2D method. The former implements coarse recognition to extract object over a large space, and the later implement fine recognition about some sub-areas corresponding to important and/or special objects. This paper is mainly about the coarse recognition. In order to extract objects in the coarse recognition stage, the disparity image calculated according to stereo vision principle is segmented by two consequent steps of region expansion and convex split. Then the 3D measurement about the rough positions and sizes of extracted objects is performed according to the disparity information of the corresponding segmentation, and is used for recognizing the objects' attributes by means of pattern learning/recognition. The attribute information resulted is further used to assist fine recognition in the way of performing gaze control to input suitable image of the interested objects, or to directly control AGV's travel. In our example AGV application, some navigation-signs are introduced to indicate the travel route. When the attribute shows that the object is a navigation-sign, the 3D measurement is used to gaze the navigation-sign, in order for the fine recognition to analyze the specific meaning by means of traditional 2D method.

This paper presents an enhanced vision system of the "Selective Stereo Gradient Method" (SSGM). Its purpose is the detection of topographies of highly reflective, metallic surfaces of quickly moving metallic tokens. We call this vision system the 3-Color-SSGM. It represents a decisive improvement of the serial-SSGM. The objective is to decide from comparison of the measured characteristic surface topography with topographical data stored in a database whether the token belongs to a reference class or not. In the improved SSGM a 3 sector 120° color LED-illumination setup is used for generating a single image of a moving object. Using the spectral properties of the illumination, which matches to the special spectral characteristics of the camera, three independent images can be extracted. The comparison between these images leads to a discrimination between a real object with 3D topography and a photographic image. The experimental setup and special illumination conditions are described. The raw data images are segmented and scaled. Rotation and translation invariance of the recognition and classification process are implemented. A specimen can be classified by using statistical image analysis and template matching methods. The classification statistics results will be reported.

Scanning techniques combining laser line projection with motion are simple and efficient. But there are number of cases in which laser triangulation fails. Some have well known solutions. Other, like adverse illumination by intense white light or presence of textures make laser projection hard to distinguish, and have no specific solution. In this article, a method is presented to improve retrieving laser projection for those cases. It’s build upon two main ideas. First, using auxiliary lines to create local high frequencies. Second, Transform a high speed camera in an intensity modulation receiver. The principle is to send a periodic message in the lines intensity and try to track traces of a spatial-temporal deforming pattern in the video sequences produced by the camera. It permits two main improvements. First, when adverse illumination produce other lines, they can be discriminate by the fact they don’t send the message. Second, when adverse illumination produce a highly luminous image or when a texture diffuse a part of the laser energy, it’s sufficient to track the noise of the message. By choosing a message, it’s possible to create every type of noise in order to distinguish it between the rest of image noises.

In this paper we present steps taken to implement a content-based analysis of crack patterns in paintings. Cracks are first detected using a morphological top-hat operator and grid-based automatic thresholding. From a 1-pixel wide representation of crack patterns, we generate a statistical structure of global and local features from a chain-code based representation. A well structured model of the crack patterns allows post-processing to be performed such as pruning and high-level feature extraction. High-level features are extracted from the structured model utilising information mainly based on orientation and length of line segments. Our strategy for classifying the crack patterns makes use of an unsupervised approach which incorporates fuzzy clustering of the patterns. We present results using the fuzzy k-means technique.

Scanning electron microscope (SEM) images for semiconductor line-width measurements are generally acquired in a top-down configuration. As semiconductor dimensions continue to shrink, it has become increasingly important to characterize the cross-section, or sidewall, profiles. Cross-section imaging, however, requires the physical cleaving of the device, which is destructive and time-consuming. The goal of this work is to examine historical top-down and cross-section image pairs to determine if the cross-section profiles might be estimated by analyzing the corresponding top-down images. We present an empirical pattern recognition approach aimed at solving this problem. We compute feature vectors from sub-images of the top-down SEM images. Principal component analysis (PCA) and linear discriminant analysis (LDA) are used to reduce the dimensionality of the feature vectors, where class labels are assigned by clustering the cross-sections according to shape. Features are extracted from query top-downs and compared to the database. The estimated cross-section of the query is computed as a weighted combination of cross-sections corresponding to the nearest top-down neighbors. We report results obtained using 100nm, 180nm, and 250nm dense and isolated line data obtained by three different SEM tools.

While evaluating the performance of image processing algorithms, the starting point is often the acquired image. However, in practice, several factors, extrinsic to the actual algorithm, affect its performance. These factors depend largely on the features of the acquisition system. This paper focuses on some of the key factors that affect algorithm performance, and attempts to provide some insight into defining “optimal” system features for best performance. The system features studied in depth in the paper are camera type, camera SNR, pixel size, bit-depth and system illumination. We were primarily interested in determining the effect of each of these factors on system performance. Towards this end, we designed an experiment to measure performance on a precision measurement system using several different cameras under varying illumination settings. From the results of the experiment, we observed that the variation in performance was greater for the same algorithm under different test system configurations, than for different algorithms under the same system configuration. Using these results as the basis, we discuss at length the combination of features that contributes to an optimal system configuration for a given purpose. We expect this work to have relevance to researchers in all areas of image processing who want to optimize the performance of their algorithms when ported to actual systems.

This paper introduces a segmentation algorithm suitable for semiconductor wafer images generated by optical inspection tools. The primary application of this work is content-based region segmentation for automatic threshold selection during recipe generation in die-to-die wafer inspection. Structures associated with different functional areas lead to different levels of noise in the difference image during the defect detection process. The ability to automatically create a mask to separate the different structures and materials is necessary to determine local thresholds for each area and thus to improve the signal-to-noise ratio. A supervised segmentation based on the discrete wavelet transform is used to segment a whole die to create a mask. During the inspection, the mask is applied on the difference image, and the threshold is automatically set as a function of the noise within the region and the thresholding coefficient specific to that region. Preliminary segmentation results are very promising. The use of the segmented region in content-based threshold defect detection improves the number of defects detected, and reduces the number of false detections. This paper will show the performance of the segmentation method on optical microscope wafer images, and the subsequent improvement of the defect detection process.

In this paper, we describe a method to optimize parameters of the Canny edge detector, when it is applied to critical dimension (CD) measurements in the cross-sectional scanning electron microscope (SEM) image of the LSI. The Canny is a typical boundary extraction method that convolves the input image with the Laplacian of the Gaussian mask to suppress noise and identify the zero crossing points. The parameters to be optimized are a standard deviation of the Gaussian mask and a threshold value that is used when the zero crossing points are identified. The statistical characteristics of the noise in the SEM image are well known: the major gredient of the SEM image is shot noise. Therefore shot noise is assumed in this method. We apply the method to a model image where shot noise is added and derive the condition where the Canny is effective to measure the fine structure in the image accurately.

A common problem in optical metrology is the determination of the exact location of an edge. In practice, however, exact edge information is generally impossible to obtain. The best we can do is to locate the edge with very high precision through the use of sub-pixeling techniques. In this paper, we review several sub-pixel edge detection schemes and compare them with respect to two figures of merit - resolution accuracy and repeatability. Towards this end, we design experiments to determine the relative performance of different algorithms using a simulated system model. Finally, we verify the model accuracy by performing similar experiments on an off-line test vision system. Our experiments achieve a two-fold purpose. First, they provide a reliable indication of the relative performance of the different algorithms under similar test conditions. Next, by validating the results obtained from the simulated model with the results from the test system, they illustrate a methodology to simulate a real measurement system. This is significant because the simulated model can be used to generate data to quickly evaluate algorithms without the need to conduct expensive and time-consuming data collection experiments. We expect that this will be of considerable value to researchers in the field.

Automated image registration based on pattern recognition is a critical procedure in many applications of machine vision and is essential for accurate navigation and change detection. In this paper, an overview of the specific applications of image registration in wafer inspection is given, followed by a case study in the application of image registration for direct to digital holography (DDH) wafer inspection. A complete system of novel algorithms for holographic image capable of accepting a variety of data streams as inputs: (1) complex frequency data; (2) complex spatial data; (3) magnitude of data extracted from holograms; (4) phase data extracted from holograms; and (5) intensity-only data. This flexibility facilitates the development of faster, more reliable, and more efficient DDH processing systems, which is important in system optimization and production. In particular, the system enables the use of the full complex wavefront, which contains both reflectance and structural topology information, in the registration process. The added information contained in the wavefront can be utilized for increased robustness and computational efficiency. Both the theory and implementation of the proposed registration system are briefly described within the framework of DDH processing for wafer inspection tasks. Several examples of defect detection and wafer alignment are given with estimates of accuracy and robustness.

The near infrared spectral region (NIR) is useful in many applications. These include agriculture, the food and chemical industry, and textile and medical applications. In this region, spectral reflectance measurements are currently made with conventional spectrophotometers. These instruments are expensive since they use a diffraction grating to obtain monochromatic light. In this work, we present a multispectral imaging based technique for obtaining the reflectance spectra of samples in the NIR region (800 - 1000 nm), using a small number of measurements taken through different channels of a conventional CCD camera. We used methods based on the Wiener estimation, non-linear methods and principal component analysis (PCA) to reconstruct the spectral reflectance. We also analyzed, by numerical simulation, the number and shape of the filters that need to be used in order to obtain good spectral reconstructions. We obtained the reflectance spectra of a set of 30 spectral curves using a minimum of 2 and a maximum of 6 filters under the influence of two different halogen lamps with color temperatures Tc₁ = 2852K and Tc₂ = 3371K. The results obtained show that using between three and five filters with a large spectral bandwidth (FWHM = 60 nm), the reconstructed spectral reflectance of the samples was very similar to that of the original spectrum. The small amount of errors in the spectral reconstruction shows the potential of this method for reconstructing spectral reflectances in the NIR range.

Many devices are used to realize non-contact temperature measurements. Whenever the body to be controlled behaves as a black body, all the devices inferring the temperature from the body radiation are accurate and reliable. On the other hand, when the body exhibits a behavior different from the black body, emissivity compensation has to be done. In case of a known emissivity, the mono-wavelength system (spectral system) is mainly used. However, when the body under examination radiates as a gray body a bicolor system is more likely to be used. In our case, we present a real time multispectral imaging system based on two CCD cameras. The system is herein presented, characterized and applications such as vision control system are presented. In the present case, two CCD, which function transfer are different due to a carefully selected set of filters, acquire data in two different wavelength ranges. This system has the advantage of providing an accurate temperature measurement as well as to display it in real time.

Many manufactured good products present different market characteristics according to their pictorial aspects. Such aspects usually depends from many factors: in some cases they are related to the adopted production process, in other cases they are intrinsically linked to the handled material characteristics. Ornamental stone represent a typical example, where both the two previous mentioned factors have to be considered. Both the aspects, in fact, are of primarily importance to certify, in pictorial terms, the aesthetic attributes of slabs, tiles and, in general, of ornamental and/or dimensional stone based products. The study was mainly addressed to investigate the possibility to develop a methodology and a technique to measure the quality of the polished stone samples, evaluating at the same time, the effect of polishing on the final detected stone surface pictorial attributes. To reach such a goal a specially designed probe holder, together with computer generated spectrum analysis techniques, has been developed and measurements were made on various rock samples to quantify, independently from stone surface pictorial attributes, as color and texture, stone surface degree of polishing. The obtained results allowed to quantify the correlation existing between stone characteristics (constituting minerals, texture, structure) and surface status after different polishing actions.

In this paper the dynamic processing of interferometric fringe patterns obtained by real-time optical measurement methods like holographic interferometry is shown. A hologram of the tested component is superimposed with the hologram of the stressed component. The achieved fringe patterns vary according to the degree of stress applied. To evaluate these varying fringe patterns in real time, dynamic filtering is required. A hybrid opto-electronic system with a digital image processing and optical correlation module based on liquid-crystal spatial light modulators gives us the possibility to use dynamic filters and input images. In order to process interferometric fringes the adaptive wavelet transformation is applied. We will show two methods of dynamic filtering. Firstly a static filter is used to process varying fringe patterns. With this method changes of features in the fringe patterns can be observed correlating to changes of stress applied on the tested component. Another application of dynamic filtering uses a static input image and dynamic filters. This method is used for the classification of interferometric fringe patterns. A set of different wavelet filters is applied to the input image using the ability of the spatial light modulator to display images in video frame rates. Comparing the wavelet filters and the output images it is possible to assign the fringe patterns to a fault class.

Removing speckle noise in electronic speckle pattern interferometry (ESPI) from a single speckle fringe pattern while keeping the fringe features is a difficult problem. The spin filtering with curved surface windows proposed by the authors is successful to filter out speckle noise nearly completely from a single speckle fringe pattern. However the new filtering has a difficulty to be overcome that the speckle fringe orientation map (SFOM) depends on the processing window size which is tryout and is difficult to be derived correctly when the speckle fringe density changes considerably. In this paper we utilize the original speckle pattern sequence with one-beam setup to determine the speckle movement direction field by digital correlation methods so that the SFOM is determined from the direction field. In this way the SFOM can be derived regardless of fringe density.

Here we focus our discussion on symmetry analysis of image. A new method, Symmetric Point Pairs Sequence (SPPS), is proposed for skeletonization and applied to process complex intersections in biomedical images. Among many thinning algorithms, methods based on contour information have been explored popular recently. In many methods based on polygonal approximation of contour, Voronoi Diagram is applied to compute the Voronoi skeleton. But for each thinning algorithm, crossing region is a rather difficult problem, in which skeleton always deforms. In this paper the SPPS is described firstly. It is obtained through the Delaunay Triangulation for sampling points of the contour. In the crossing region whose structure is represented by the triangulation dual graph model, the SPPSs are merged and reconstructed. So the skeleton that isn’t deformed is obtained. We apply this method to process complex intersections in biomedical images. In the neuroscience field, a number of pathologies seem to be connected to morphological alterations of neural cells. In our experiment a lot of symmetric point pairs are displayed. The result shows this method preserve the precise the topological relation among the crossing regions, so our purpose to individualize all the real cells by different shape is reached.

We have tried to estimate the electron beam profile from a scanning electron microscope (SEM) image by using Wavelet multiresolution analysis for in-process SEM inspection. At first, an ideal secondary electron (SE) profile for step edge is calculated by using the Monte Carlo simulator. Then the SE profiles observed by the electron beam with Gaussian profiles are simulated with the electron beam diameter as a parameter. Wavelet analyzed results of SE profiles show that it is possible to estimate the size of electron beam profile from the Wavelet coefficient of SE profile. Next we apply the proposed estimation method to SEM images of test pattern. The procedure is as follows. 1) Noise included in SEM images is reduced by using the denosing method by Wavelet transform. 2) The SE profile for edge is extracted from a SEM image and is normalized. 3) The normalized SE profile is decomposed by Wavelet multiresolution analysis. 4) By comparing the obtained Wavelet coefficient of SE profile with the relation between the electron beam diameter and the Wavelet coefficient, the electron beam profile is estimated. Proposed procedure was applied to the SEM image of test pattern to obtain the beam profile.

Thermal imaging systems are usually based on IR-camera systems analyzing the infrared part of the electromagnetic spectrum to carry out temperature measurements. These systems are very accurate but also very expensive. A “low-cost” thermal-vision-system based on a standard CCD color video camera in combination with an image processing system is presented. The system was developed to visualize the temperature distribution inside a plasma-reactor. The temperature of the heat-treated work pieces lies in the range of 450°C to 650°C. By analyzing the thermal light emission of these objects a temperature map can be made to visualize and measure temperature differences of the reactor interior.

A flexible new camera calibration technique using 2D-DLT and bundle adjustment with planar scenes is proposed in this paper. The equation of principal line under image coordinate system represented with 2D-DLT parameters is educed using the correspondence between collinearity equations and 2D-DLT. A novel algorithm to obtain the initial value of principal point is put forward in this paper. The practical decomposition algorithm of exterior parameters using initial values of principal point, focal length and 2D-DLT parameters is discussed elaborately. Planar-scene camera calibration algorithm with bundle adjustment is addressed. For the proposed technique, either the camera or the planar pattern can be moved freely, and the motion need not be known. Very good results have been obtained with real data calibration. The calibration result can be used in some high precision applications, such as reverse engineering and industrial inspection.

Increased globalisation of the ornamental stone market has lead to increased competition and more rigorous product quality requirements. As such, there are strong motivators to introduce new, more effective, inspection technologies that will help enable stone processors to reduce costs, improve quality and improve productivity. Natural stone surfaces may contain a mixture of complex two-dimensional (2D) patterns and three-dimensional (3D) features. The challenge in terms of automated inspection is to develop systems able to reliably identify 3D topographic defects, either naturally occurring or resulting from polishing, in the presence of concomitant complex 2D stochastic colour patterns. The resulting real-time analysis of the defects may be used in adaptive process control, in order to avoid the wasteful production of defective product. An innovative approach, using structured light and based upon an adaptation of the photometric stereo method, has been pioneered and developed at UWE to isolate and characterize mixed 2D and 3D surface features. The method is able to undertake tasks considered beyond the capabilities of existing surface inspection techniques. The approach has been successfully applied to real stone samples, and a selection of experimental results is presented.

This paper is concerned with research into machine vision techniques for measuring the size and shape of objects. Considerable potential is considered to exist for the application of portable hand-held vision metrology systems. However, unlike a bench-mounted system, a hand held device will be subjected to movement during the measurement process. Employment of active techniques, such as a scanning laser line, usually result in measurement errors due to movements that can occur during capture of the sequence of images involved. In contrast, passive techniques such as stereo vision can operate rapidly, with two images being captured simultaneously in order to construct a three dimensional map of the object. Unfortunately, stereo vision is subject to a number of technological difficulties. These include the Correspondence Problem (i.e. the challenge of relating points in the image to points in 3D space), and the presence of relatively high levels of noise in the distance measurements for the points in the image. In this paper, a hybrid approach is described for alleviation of these problems, in a methodology that combines structured light and stereo vision techniques.

In this paper, we have proposed an application of pattern recognition technique to recognize partially destroyed objects in debris. The study employs an appearance-based eigenspace technique for investigating the representation and recognition of partially destroyed objects, which is one of the severe limitations of this technique. Since the conventional parametric eigenspace technique cannot handle the occluded or even partially destroyed objects, we propose creation of a mean-appearance for representing and recognizing them. The word mean-appearance describes a mean image set, which allows some partially destroyed objects for producing an eigenspace. In the mean image set, averaging of a few destroyed images along with non-destroyed images make a mean appearance, which has less effect of partially destroyed images. In addition, we have proposed to apply eigenspace method to measure lost goods in debris when the conventional method will not be an alternative. The proposed approach is performed using various destroyed objects and experimental results show the effectiveness of the proposed method.