QUANTITATIVE IMAGE TREATMENT FOR PDI-TYPE QUALIFICATION OF VT INSPECTIONS

QUANTITATIVE IMAGE TREATMENT FOR PDI-TYPE QUALIFICATION OF VT INSPECTIONS Matthieu TAGLIONE, Yannick CAULIER AREVA NDE-Solutions France, Intercontrôle Televisual inspections (VT) lie within a technological domain that has significantly evolved in recent years. The preponderant factor in this evolution is the adaptation of digital image acquisition and recording systems that allow calculations to be performed on the data acquired. Such calculations can minimize uncontrolled operator dependant results and improves standardization of the image quality. The ability to manipulate image data and to perform computation on this data allows a much deeper understanding of the factors to be taken into account when performing televisual inspections, and allows characterizing the performances of the system in a way that can be integrated in a Performance Demonstration Initiative (PDI) type qualification. A work aiming at proposing PDI-type qualification tests for any VT system has been realized within AREVA NDE Solutions. It is based on standards and state of the art methods, and takes into account needs from operators in France, the US and Germany. This paper discusses inspection needs-analysis and implementation of state of the art image quality measurement and enhancement methods including: Needs evaluation and operators requests: Surveys and operator interviews to determine needs within each country and to determine specific requirements from various codes and standards were performed. This provided input for a global view of the image capture and processing specifications. Type of standards used: Description of the different standards identified as providing an answer to the operator needs, with description of the items that required more reflection. Main factors to be taken into account when measuring a VT system image quality: According to the requirements, different test pattern and analysis schemes can be proposed. An example is given. Work to be performed: Implementation of this technology for VT qualification is ongoing and is being introduced piece-wise as it becomes available. There are still technical challenges to overcome and as new techniques are created, they will be part of AREVA s continuous efforts to improve inspection quality. The PDI-type qualification demonstration for VT inspection incorporating many of these methods is expected to be available performed by end of 2012. AREVA NDE-SOLUTIONS ACTIVITIES IN VT One important part of AREVA NDE-Solutions activity is aimed at developing and qualifying inspection systems dedicated to the non-destructive inspection of nuclear power plants. The expertise of AREVA NDE-Solutions lies in many surface and volumetric inspection methods, including ultrasonics (UT), eddy currents (ET), radiography (RT), thermography (IT) and last but not least visual inspection (VT). The VT inspections realized are aimed at detecting disorders that can affect components such as the vessel, vessel head, Bottom Mounted Instrumentation (BMI), Steam Generator (SG) divider plate,. Some tools used by AREVA for performing VT inspections are presented in Figure 1. 71

Figure 1 : Different VT inspection tools in use at AREVA NDE-Solutions The disorders looked for are often fatigue cracks or surface condition alterations. In such cases, the quality of the inspection systems used is paramount: it is not unusual in nuclear power plants to look for cracks with openings of 5 or 10 micrometers. A VT acquisition chain is composed of a light source, a camera, optics, a signal transmission link, acquisition electronics, a display and usually storage means. In the case of digital video, there is also a need for compression of the data. This architecture is presented on Figure 2. The quality of the image data obtained through this acquisition chain depends on each of the items listed above. The use of digital acquisition allows evaluating the characteristics of this data by performing mathematical operations on it to enhance performance capability over using direct data visualization. The use of digital image acquisition also allows more quantitative analysis of the VT system quality. Figure 2 : Digital VT inspection chain block diagram NEEDS EVALUATION AND OPERATOR REQUESTS The need for evaluating the intrinsic performances of VT systems is expressed in AREVA NDE-Solutions clients specifications, as a means to insure inspection quality. The main parameters defined by different clients were gathered in France, the US and Germany. These parameters that can be divided into different families mainly concern spatial resolution, optical parameters, sensitivity, lighting, and setup (Table 1). It can be seen from this table that the only request in all clients specifications is the visualization of a horizontal or vertical line with width equal to that of the smallest defect to be detected. This simple test is usually performed using a target comprising such a line printed in black over a white or grey background or, if the dimensions of the smallest defect allow it, comprising EDM notches. The corresponding test patterns used in France, the US and Germany are presented on Figure 3. While this verification proves that the VT system tested is able to reproduce an indication whose width is the same as the smallest defect to be detected, it is not enough to conduct a PDI-type qualification of this system. Hence it was decided to investigate the different means of measuring image quality, using digital techniques. 72

Figure 3 : Line tests patterns used in France, the US and Germany to check imaging system quality France Germany US Spatial resolution Visualize horizontal and/or vertical line(s) Measure the resolving power Read small-size characters Sensitivity Visualize N gray levels (dynamic range) Verify colour accuracy Optical parameters Evaluate the distortion of geometric figures Measure the depth of field Lighting Measure the light intensity and homogeneity Maintain performances on all surfaces Be able to modify light balance Setup Maintain performances for all incidence angles Maintain performances while moving Maintain performances on recordings Maintain performances when the carrier vibrates Table 1 : Clients requests summary STANDARDS USED The goal of this work is producing tools for PDI-type qualification of VT systems that can be used internationally. The existing ISO international standards provide many answers in the field of image quality evaluation: it was decided to use them as a basis to our developments. Most of the work performed was hence aimed at understanding the standards and interpreting them so as to derive actual performance validation procedures from them. For example, the measurement of resolving power is described in ISO 12233, the description of the way to check the dynamic range is included in ISO 15739. A list of the standards that were actually used is presented in Table 2 73

Standard Name Application ISO 14524 ISO 12233 ISO 15739 ISO 15795 ISO 9039 Photography Electronic still-picture cameras Methods for measuring optoelectronic conversion functions (OECFs) Linearization of the sensor response (needed for computation of resolving power, dynamic range, ) Measurement of the resolving power Measurement of the black and Photography Electronic still-picture cameras Resolution measurements Photography Electronic still-picture imaging Noise measurements white sensitivity Optics and optical instruments Quality evaluation of optical systems Assessing the image quality degradation due to chromatic aberrations Optics and photonics Quality evaluation of optical systems Determination of distortion Table 2 : Standards used for the project Measurement of the impact of chromatic aberrations on spatial resolution (useful for underwater systems) Measurement of the distortion of geometrical figures PARAMETERS MEASUREMENT Comparing Table 1 and Table 2, it can be seen that while the approach for measuring some parameters is described clearly and simply in the international standards, the quantitative evaluation of some other parameters requires further thought. An approach that could be used for each parameter of Table 1 is described below. Spatial resolution Measurement of resolving power This parameter can be directly measured using the approach described in ISO 12233. For colour images, especially underwater, a complementary approach can be found in ISO 15795. Visualization of lines or readability of characters Both of these parameters rely on human interpretation of an image; hence they are not directly concerned by this work. However the measurement of the system resolving power will provide a good estimation of the smallest indication detectable, and hence tell if the system will be able to visualize the lines and read the characters. Sensitivity Visualization of N gray levels This parameter can be directly measured using the approach described in ISO 15739. Verification of colour accuracy This parameter has to be measured for any application that requires colour imaging. However, it is difficult to estimate the colour accuracy of a digital imaging system for two reasons: 74

- Different colour spaces can be used for representing digital images that will all give the same colour sensation to a standard human eye. According to its configuration, each system will use one or many different colour space(s) to transmit, display and record the data: defining once and for all a procedure for measuring the final performance obtained is an significant task. - There exists a metric for the colour accuracy of a system, called the CRI (Colour Rendering Index). It is based on the comparison of a set of known colour samples against their images acquired by a digital imaging system in a certain colour space. This metric gives a good indication in most cases, but can be unsuitable when using new types of light sources, such as for example LEDs. A standardization group is working on those topics; when it is done their output will be used for completing AREVA s analysis tool set. Optical parameters Evaluation of the distortion of geometrical figures This parameter can be directly measured using the approach described in ISO 9039, as is presented on Figure 4 Measurement of the depth of field The depth of field of an optical system is linked with a degradation of its resolving power. Hence, a simple procedure for measuring the depth of field of a system consists in applying the approach described in ISO 12233 using a displacement device, and moving the test pattern in parallel to the optical axis without changing the system focus settings. The depth of field limit is found when the system resolution becomes lower than the requirements of the inspection. Lighting Measurement of light intensity and homogeneity This parameter can be directly measured using the approach described in ISO 15739. In this case, the measurements are realized using the camera: only the adequacy between the optical setup and the lighting is considered (there is no notion, for example, of how many lux the light source produces). Adaptability to all surfaces This parameter is strongly dependant on the lighting used for the inspection, as well as on the inspected surface characteristics. An approach that could be used for verifying that it is compliant with the target would be the ISO 15739 approach on different types of surfaces, representative of the inspection conditions. Ability to light balance control This parameter relies only on the optical setup chosen: the capability to realize a homogeneous or side - lighting of the scene has to be specified during the inspection system realization. The lighting settings must be reported in this case along with the result of the measurements of any other parameter. 75

Setup Maintained performances for all incidence angles The incidence angle used during the inspection depends on the actual carrier and setup used. The performances should be tested for the standard inspection angle (usually normal to the inspected surface) and the most stringent incidence angle (for example 30 to the surface normal). Maintained performance while moving and insensitivity to vibrations The evaluation of these parameters is mandatory for dynamic systems, as camera movement will produce blur on the image. It is even more important in the case of interlaced cameras, in which the movement will produce artefacts as presented on Figure 4. The movement is likely to produce a deterioration of the lateral resolution of the system, all other parameters being mainly insensitive to this displacement: hence a good test to validate the performance of the system in acquisition conditions is checking its resolution according to ISO 12233, using the carrier, both on a static image for insensitivity to vibrations and at maximum inspection speed for performance while moving. Figure 4: Left : measurement of distortion according to ISO 9039 Right : artefact due to the use of interlaced video Maintained performances on recordings This parameter is of primary importance for digital video systems. The best way to check that it is fulfilled is performing all tests at compression settings that are the same as those applied during the inspection. All standards used state that compression should be turned off if possible. However, this does not seem advisable in our case as this would not be representative of inspection conditions. EAMPLE MEASUREMENT The size of the smallest element that can be resolved using an optical system is often hard to define. As discussed above, ISO 12233 is the international standard that can answer this question, providing a quantitative evaluation of this value. This paragraph provides an overview of the test procedures that can be derived from this standard. Test chart The standard proposes a rather complex test chart as can be seen on Figure 5, but states that other charts can be used if required: this allows the use of simpler charts, dedicated to the needs of a particular application. Different results will be obtained using the procedures described in the standard: - Visual resolution (equivalent to USAF1951 test pattern resolution) - Limiting resolution (quantitative result of the same measurement) - Spatial frequency response (SFR - amplitude of the camera response versus the spatial frequency of the pattern recorded). 76

Figure 5 : ISO 12233 proposed test chart Test procedure: visual resolution and limiting resolution The evaluation of visual resolution and limiting resolution is based on the analysis of a pattern made of straight or slanted black bars spaced by a white bar with equal width, as presented on Figure 6. The bars spatial frequency increases along the pattern and the camera response exhibits lower contrast between black and white for higher spatial frequencies because of aliasing, until the point where it is impossible to see any contrast. The evaluation of visual resolution is performed by an operator who looks for the point of the pattern where he is not able to discriminate white bars from black bars. This measurement is subjective and must only be realized by well acquainted operators. The evaluation of limiting resolution is based on a quantitative analysis of the chart image. The average intensity modulation due to the alterning white and black bars (know as the Modulation Transfer Function - MTF) will be measured for each spatial frequency, and the point where aliasing leads to a contrast of 5% between the black and white bars will be called the limiting resolution. Test procedure: spatial frequency response The spatial frequency response (SFR) of a VT system is the quantitative evaluation of the evolution of the system output contrast versus the spatial frequency of the pattern observed. It can be considered as a continuous MTF measurement, and can be used to evaluate the camera sensitivity versus the size of the object observed, the same way as MTF. The test pattern used for measuring SFR is a slightly slanted (near-vertical or near-horizontal) black-to-white and white-to-black edge. The measurement procedure is detailed in ISO 12233 and will not be presented here. Two points that can be outlined are that the measurement shall be performed for multiple orientations of the target, and for at least 32 spatial frequencies below the sensor spatial sampling frequency. 77

Normalized Response Visual resolution 1 5% Ideal MTF Real MTF SFR 0 Figure 6 : Normalized MTF and SFR presented together Limiting resolution Analysis The analysis of the data presented on Figure 6 brings an answer to the question: what is the smallest defect detectable using this material? We can see on the SFR curve that for low frequencies, the response of the system is at its best: any defect with such spatial frequency will be reproduced by the imaging system at its real contrast. When the indication becomes smaller, the image contrast will become smaller and smaller. For some higher frequency, the contrast will reach 5%, which is the limiting resolution, the minimum contrast discerned by the human eye. Even lower, the contrast goes down to zero. We can hence analyze the data as follows: - For any defect with spatial frequency at which the SFR is 100%, the defect contrast will not be affected by the spatial resolution - For any defect with spatial frequency at which the SFR is below 100% and over 5%, the defect contrast will be affected by the spatial resolution. We are able to know by how much it will be affected according to its spatial frequency. - For any defect with a spatial frequency at which the SFR is below 5%, the human eye will not be able to see the defect. This particular example shows the power of such methods: we are able to measure the influence of a defect spatial frequency on its detectability. The visual resolution measured using for example the USAF1951 test pattern, is maintained: a correspondence can be realized between historic analysis methods and digital analysis methods. CONCLUSION AREVA NDE-Solutions is working at developing methods for quantifying the quality of digital images in anticipation of providing better PDI-type performance demonstrations of VT inspection systems. The parameters studied are those specified by our main clients in France, the US and Germany. The approach taken to evaluate the quality of inspection systems is based on existing ISO standards, which ensures that 78

the measurement procedures can be used internationally. A way of measuring all influential parameters identified in this work has been proposed, except for colour accuracy. An example implementation has been given in this paper with the measurement of resolving power, where both the acquisition and analysis of the data needed for measuring this parameter are presented. The proposed method allows getting quantitative information on the VT system intrinsic performances, and is compatible with historic resolving power evaluation methods which use for example the USAF1951 test pattern. All the approaches defined are now being implemented as acquisition procedures and data processing algorithms, so that they can be made available as a software package which can be used for evaluating the performances of VT inspections systems. The timeline for the development of this software package ends in 2012. This development will allow straightforward and quantitative comparison of VT systems with one another, for existing systems as well as for new systems. It will also be an aid for better quantification of system performances during R&D actions, which will allow the realization of better quality VT tools allowing faster and better quality inspections. 79