Technologies currently used for cotton contaminant assessment suffer from some fundamental limitations. These limitations result in the misassessment of cotton quality and may have a serious impact on the evaluation of the economic value of the cotton crop. This paper reports on the recent advances in the use of a 3D x-ray microtomographic system that employs image processing and pattern recognition techniques to accurately detect and classify trash present in cotton. The proposed method offers an attractive alternative to existing trash evaluation technologies, because of its ability to produce 3D representations of the samples, to robustly segment the trash from its background, and to accurately classify the contaminant types. This procedure could have a serious impact on the process control technologies (cotton lint cleaning), and indeed on the economic value of cotton.
There has been a long development of sugar crystal analysis techniques. Initially crystals were manually passed through various increasingly finer sieves so that one could manually calculate what percentage of crystals and crystal masses lay in various size groups. Later microscopes were used on small samples to take pictures of crystals so that they could be sized manually at higher degree of accuracy. In order to increase the accuracy, image processing are being used to analyze the pictures taken under microscope. The main concern is to analyze crystals with width greater than 50 micrometers. The ideal crystal is roughly square and has a width of approximately 120 micrometers. There is then a need to separate crystals into two main classes: the class of crystals that have to be considered for the analysis and those that will be rejected. This classification process involves: the enhancement of the quality of the image, the binarization of the image, the extraction of the connected components, the features extraction from each connected component and the characterization of the classes. During this process, there is more often a lost of information and in some case an intrusion of noise. These can have as result some misclassifications. These misclassifications can be caused by touching crystals or overlapping crystals that are treated as single crystal. These can also be due to the fact that edges of crystals are not well extracted. In this paper we present a method to alleviate those misclassifications using mathematical morphology and a combination of binarization and edge detection. This method gives better classification. Some results are presented.
Reconstruction of depth information is a common task in industrial
inspections as measurement of mechanical components or for surface
inspection. This paper presents a new approach for a real-time
self calibrating stereo based system. The concept is based on
calculating the difference-map of registered stereo images that
contains qualitative depth information. The principle is valid for
a variety of applications; a prototype for a 3D code reader is
presented in detail. Experimental results from a laboratory
prototype demonstrate the full functionality of the system
In order to meet the needs of many diverse industrial 3D inspection tasks, INO has developed a new concept for the design of a smart and modular 3D laser profiler. This stand-alone sensor which we call Smart Laser Profiler (SLP) is composed of a laser line projector, collection optics, a high frame rate camera and a digital signal processor (DSP). The on-board DSP is the key to this technology. The SLP sensor has been designed to be both compact and rugged and it is enclosed in a water resistant NEMA 4 class housing that is easy to install on a production line. The Smart Laser Profiler has several preprogrammed functions on the DSP that implement basic shape analysis algorithms like volume measurement and shape conformance. For more complex shape analysis, the sensor can transfer the raw 3D profiles to a PC through a high-speed communication link. The present article will describe both the unique hardware, electronics and optical architecture of the sensor and the software tools that were developed.
The design of edges is very important for many components. In
this paper we therefore present a light-sectioning based
measurement head, which is suitable for the edge inspection of
different workpieces. Beyond the design we also present a new
calibration technique for its camera. The calibration is mainly
based on several perspective projections, which are successively
executed. In each step, the linear system of homogeneous
equations is solved by using singular value decomposition. Each
mapping is therefore obtained in the least squares sense. Because
of the novel design of the calibration device, a high number of
reference points can be used for the description of these
mappings. The inspection of a workpiece detail implicates a large amount of data, some of which is useless. To extract the data essential for the fitting routines, a special correlation/regression based template matching is proposed. After the description of the
segmentation process we propose a measurement progression, which
enables us to obtain a fast and easy perspective correction of the
three-dimensional light sectioning data. Finally, a fitting method
is presented. Based on singular value decomposition, the data is
fitted to the corresponding form of the fillet or chamfer. As the
fit is done in the least squares sense, one can obtain statistical
information out of the decomposition process.
Camera calibration is an important step in machine vision for dimensional measurement. Based on analyzing the photogrammetric calibration and self-calibration methods, a new planar way is proposed in this paper for calibrating the camera for the 2D objects. Induced some non-linear distortion factors, a fast algorithm is adopted to solve the complicated non-linear equations in calibration model. With this algorithm, a concatenation technique is used to reconstruct a 3D object from two 2D planar photos. In the second part of this paper, a calibration method using only one plane, which can be moved on some simple equipment, is deduced for 3D object dimensional measurement. Mathematic model and its transformation process are discussed in detail. This method can be used especially for profile inspection of machine parts in industrial working-field. An application to inspect the profile of train wheel is given in this paper. Experimental results are given to show the parameters of camera system and the measuring accuracy.
This work aims at detecting defects on metallic industrial parts with streaked surface. The orientation of those parallel streaks is totally random. The searched defects are scratch and lack of machining. A specific machine vision system has been designed to deal with the particular inspected surface features. One image is acquired with an annular lighting in bright field and six images are acquired with a rotating lighting in dark field. A particular image processing is applied on the six images in order to get one image that represents all the revealed imperfections. A thresholding processing is then applied on this image in order to segment the imperfections. A trained classification, created with well known typical objects of each class, is performed. The classification has to recognize the different defects and the small imperfections that are not defects. The decision phase is used to know if the defects are acceptable, and therefore if the inspected part is acceptable. Some acceptability rules are defined for every defect class. The developed machine vision system has been implemented on an experimental industrial production line and it gives 2% of sub-detection and 16% of over-detection.
Recently specialized robots were introduced to perform the task of
inspection and repair in large cylindrical structures such as
ladles, melting furnaces and converters. This paper reports on the
image processing system and optical servoing for one such a robot.
A panoramic image of the vessels inner surface is produced by
performing a coordinated robot motion and image acquisition. The
level of projective distortion is minimized by acquiring a high
density of images. Normalized phase correlation calculated via the
2D Fourier transform is used to calculate the shift between the
single images. The narrow strips from the dense image map are then
stitched together to build the panorama. The mapping between the
panoramic image and the positioning of the robot is established
during the stitching of the images. This enables optical feedback.
The robots operator can locate a defect on the surface by
selecting the area of the image. Calculation of the forward and
inverse kinematics enable the robot to automatically move to the
location on the surface requiring repair. Experimental results
using a standard 6R industrial robot have shown the full
functionality of the system concept. Finally, were test
measurements carried out successfully, in a ladle at a temperature
of 1100° C.
In many industrial processes knowledge of the two-dimensional thermal distribution is of great importance. Conventional infrared based systems (MIR, FIR) provide very accurate results, but their quality also comes at high cost, and moreover these systems cannot always be properly applied in every case, e.g. due to problems concerning IR-radiation absorption through certain IR-blocking materials such as inspection windows . We present a “low cost” NIR thermal imaging device based on a grayscale CCD camera used in combination with image processing software applied to the thermal imaging of heated metal parts in a plasma reactor. The aim of this work is to measure the temperature distribution of objects at relatively low temperatures of approx. 350 °C and below by applying image processing techniques, assuming constancy of temperature for a few seconds. Special emphasis is put on the influence of the emission factor, which plays an important role in the field of non-contact temperature measurements, especially when thermo-chemically processed surfaces are considered. In addition, the noise characteristics of the imaging system have to be taken into account to ensure reproducible results. The underlying imaging model and a camera characterization procedure based on the 'Photon Transfer Technique' are presented which are used to adjust the relevant parameters to predict the measurement limits of such systems.
This paper proposes a comparative survey on techniques of vision based on invisible structured lighting. We have classified them in three distinct families: InfraRed Structured Light (IRSL), Imperceptible Structured Light (ISL) and Filtered Structured Light (FSL). For each of them, definition, minimal configuration and main applications found in the literature are given. Then, we compare them regarding to several criteria: required equipment, light pattern coding, color analysis, texture analysis, motion analysis, security, use in non-controlled environment. The description is IRSL, ISL and FSL sensors will permit to sum up these techniques; the comparison will permit to evaluate performances and efficiency of each of them. We think that this study could be useful to researchers that are looking for a compromise between stereovision and structured light vision, combining the processing tools extent of the former with the point matching reliability and simplicity of processing of the latter.
Recent advances in technology have made light emitting diodes (LEDs) viable in a number of applications, including vehicle stoplights, traffic lights, machine-vision-inspection, illumination, and street signs. This paper presents the results of comparing images taken by a videoscope using two different light sources. One of the sources is the internal metal halide lamp and the other is a LED placed at the tip of the insertion tube. Images acquired using these two light sources were quantitatively compared using their histogram, intensity profile along a line segment, and edge detection. Also, images were qualitatively compared using image registration and transformation. The gray-level histogram, edge detection, image profile and image registration do not offer conclusive results. The LED light source, however, produces good images for visual inspection by an operator. The paper will present the results and discuss the usefulness and shortcomings of various comparison methods.
The recent introduction of high dynamic range CMOS-cameras with logarithmic response to light intensity, justify a serious evaluation of the technology as an alternative technology for laser profiling. This paper presents a series of comparative tests of a high quality CCD-camera and a high-dynamic range CMOS-camera. Standard gray scale charts are used to verify the intensity response and the signal to noise ratio at different f-stops. It is shown that the high dynamic range of the CMOS-sensor makes the camera suitable for differential image laser profiling. Furthermore, the cross-section of steel rods and wires are observed to verify the industrial applicability
of the different standards. Both, material at room temperature and red-hot glowing steel bars were measured. The advantages and disadvantages for each technology are shown on the basis of these tests. Finally, a laser profiler was manufactured with the CMOS-camera and successfully implemented in a steel-mill.
Reliable and productive manufacturing operations have depended on people to quickly detect and solve problems whenever they appear. Over the last 20 years, more and more manufacturing operations have embraced machine vision systems to increase productivity, reliability and cost-effectiveness, including reducing the number of human operators required. Although machine vision technology has long been capable of solving simple problems, it has still not been broadly implemented. The reason is that until now, no machine vision system has been designed to meet the unique demands of complicated pattern recognition. The ZiCAM family was specifically developed to be the first practical hardware to meet these needs. To be able to address non-traditional applications, the machine vision industry must include smart camera technology that meets its users’ demands for lower costs, better performance and the ability to address applications of irregular lighting, patterns and color. The next-generation smart cameras will need to evolve as a fundamentally different kind of sensor, with new technology that behaves like a human but performs like a computer. Neural network based systems, coupled with self-taught, n-space, non-linear modeling, promises to be the enabler of the next generation of machine vision equipment. Image processing technology is now available that enables a system to match an operator’s subjectivity. A Zero-Instruction-Set-Computer (ZISC) powered smart camera allows high-speed fuzzy-logic processing, without the need for computer programming. This can address applications of validating highly variable and pseudo-random patterns. A hardware-based implementation of a neural network, Zero-Instruction-Set-Computer, enables a vision system to “think” and “inspect” like a human, with the speed and reliability of a machine.
This paper presents various applications of machine vision systems. These systems are used at four strategic points in a company manufacturing pipes for the nuclear industry. For each system, the vision problematic is presented including the industrial constraints, then, the proposed solution is detailed (acquisition conditions, image processing algorithms...), finally, the implementation on the industrial line is described and results are discussed. The first system used in the R&D department controls tube deformation under high pressure and high temperature conditions. The second vision system deals with the surface inspection of outer part as well as inner part of the tubes for scratches as well as oxidation mark detection. After the lamination, tubes are heated to release the mechanical constraints which took place during the lamination process. During the heating, oxidation may occur. Based on color analysis, a machine vision system was developed to measure the oxidation time. Once manufactured, tubes are thoroughly cleaned by air propulsed plugs and packaged in boxes. A system which detects any missing or occluded tubes was realized. The results show that the nuclear industry can take important benefits from machine vision systems. The four validated and implemented applications give satisfactory results and are currently used in the factory.
In this paper, we describe the inspection of coated particle nuclear fuel using optical microscopy. Each ideally spherical particle possesses four coating layers surrounding a fuel kernel. Kernels are designed with diameters of either 350 or 500 microns and the other four layers, from the kernel outward, are 100, 45, 35, and 45 microns, respectively. The inspection of the particles is undertaken in two phases. In the first phase, multiple particles are imaged via back-lighting in a single 3900 x 3090 image at a resolution of about 1.12 pixels/micron. The distance transform, watershed segmentation, edge detection, and the Kasa circle fitting algorithm are employed to compute total outer diameters only. In the second inspection phase, the particles are embedded in an epoxy and cleaved (via polishing) to reveal the cross-section structure of all layers simultaneously. These cleaved particles are imaged individually at a resolution of about 2.27 pixels/micron. We first find points on the kernel boundary and then employ the Kasa algorithm to estimate the overall particle center. We then find boundary points between the remaining layers along rays emanating from the particle center. Kernel and layer boundaries are detected using a novel segmentation approach. From these boundary points, we compute and store layer thickness data.
Materials surface characteristics can be investigated analyzing their spectral response when properly energized by a suitable source. When the source is represented by a light spectra of known characteristics the surface material response can thus be evaluated adopting a spectrophotometric approach. The analyses of the detected spectra can give useful information concerning the material characteristics and/or surface properties and status. In this perspective digital spectrophotometry can be considered as one of the basic techniques to characterize materials. The application of such a technique is usually confined in “high-tech” environments and can results quite expensive and difficult to apply in industrial “on-line” processes.
In this paper recent advances in imaging spectrophotometry devices and techniques are presented with reference to the possibility of recognizing different glass fragments (cullets) according to their color and, most important thing, the presence of transparent polluting fragments, that is the distinction between glass from ceramic glass fragments. The work is specifically addressed on the spectral characterization of different ceramic glass products in order to define suitable inspection strategies to preliminary identify and then sort such a class of materials inside recycling plants to perform the required separation between useful (glass) and polluting (ceramic glass) materials.
The exposure of metallic structures to rust degradation during their operational life is a known problem and it affects storage tanks, steel bridges, ships, etc. In order to prevent this degradation and the potential related catastrophes, the surfaces have to be assessed and the appropriate surface treatment and coating need to be applied according to the corrosion time of the steel. We previously investigated the potential of image processing techniques to tackle this problem. Several mathematical algorithms methods were analyzed and evaluated on a database of 500 images. In this paper, we extend our previous research and provide a further analysis of the textural mathematical methods for automatic rust time steel detection. Statistical descriptors are provided to evaluate the sensitivity of the results as well as the advantages and limitations of the different methods. Finally, a selector of the classifiers algorithms is introduced and the ratio between sensitivity of the results and time response (execution time) is analyzed to compromise good classification results (high sensitivity) and acceptable time response for the automation of the system.
This paper presents a classification work performed on industrial parts using artificial vision, SVM and a combination of classifiers. Prior to this study, defect detection was performed by human inspectors. Unfortunately, the time involved in the inspection procedure was far too long and the misclassification rate too high. Our project consists in detecting anomalies under manufacturer production and cost constraints as well as in classifying the anomalies among twenty listed categories. Manufacturer’s specifications require a minimum of ten inspections per second without a decrease in the quality of the produced parts. This problem can be solved with a classification system relying on a real-time machine vision. To fulfill both real time and quality constraints, two classification algorithms and a tree based classification method were compared. The first one, Hyperrectangle based, has proved to be well adapted for real-time constraints. The second one, based on Support Vector Machine (SVM), is more robust, more complex and more greedy regarding the computing time. Finally, naïve rules were defined, to build a decision tree and to combine it with one of the previous classification algorithms.
In this paper a system for web surface inspection is described.
It has three parts: an image acquisition part, a defect detection
part, and a defect classification part. The self-organizing maps
(SOMs) are used both in defect detection and in defect classification
which makes the system adaptable to different types of surfaces and
defects. Our main focus is on defect classification where a generic
content-based image retrieval (CBIR) system called PicSOM is
utilized. The PicSOM uses tree-structured SOMs (TS-SOMs) and
relevance feedback. It is trained with the feature sets of the
defects in the database. For defect description, features from the
MPEG-7 standard (the homogeneous texture, the color structure, and
the edge histogram) are used and for the shape description our own
shape feature set is applied. Results indicate that the system works
with a high level of success.
The paper presents an advanced solution for capturing the height of an object in addition to the 2D image as it is frequently desired in machine vision applications. Based upon the active fringe projection methodology, the system takes advantage of a series of patterns projected onto the object surface and observed by a camera to provide reliable, accurate and highly resolved 3D data from any scattering object surface. The paper shows how the recording of a projected image series can be significantly accelerated and improved in quality to overcome current limitations. The key is ALP - a metrology dedicated hardware design using the Discovery 1100 platform for the DMD micromirror device of Texas Instruments Inc. The paper describes how this DMD technology has been combined with latest LED illumination, high-performance optics, and recent digital camera solutions. The ALP based DMD projection can be exactly synchronized with one or multiple cameras so that gray value intensities generated by pulse-width modulation (PWM) are recorded with high linearity. Based upon these components, a novel 3D measuring system with outstanding properties is described. The “z-Snapper” represents a new class of 3D imaging devices, it is fast enough for time demanding in-line testing, and it can be built completely mobile: laptop based, hand-held, and battery powered. The turnkey system provides a “3D image” as simple as an usual b/w picture is grabbed. It can be instantly implemented into future machine vision applications that will benefit from the step into the third dimension.
In recent years, three-dimensional measurement in machine vision is applied to various places, such as a production line. When applying to a production line, in order to raise its reliability and cost performance, it is indispensable to measure the three-dimensional position of an object at high speed and with high precision. Then, in order to measure the position of an object, we propose the new
three-dimensional measurement technique which combined the single
light stripe method and the relative stereo method in this paper.
Further, we will show that this technique can measure the
three-dimensional position of all the objects projected in the stereo images at high speed and with high precision. Finally, we show results of an evaluation experiment for the measurement technique.