Prologue

An overview is provided of how basic chromatic principles (Jones, G. R. et al„ 2008) have been further developed into new domains, along w'ith practical examples of their deployment. The emphasis is on the practicalities of the method for diverse applications.

As technology in general advances, there are extensive ways in which it may be deployed for measurement and to produce an extensive amount of data. The acquisition of such large amounts of data often contains much unnecessary data which effectively constitutes a high level of noise and so masks the required data whilst inflating cost. Many approaches are available which seek to address the problem in different ways - for example, through the use of artificial intelligence.

The chromatic approach may be regarded as addressing this problem from a different perspective. It recognises the undesirable need for extra measurement capability alone and recognises the fact that information comes from monitoring, which should not be confused with measurement but is the means whereby diagnosis is made. The approach enables emerging conditions to be recognised without the need for an excessive amount of measured data.

The present book shows that there are different ways of deploying chromaticity for particular applications. It provides examples which can form the basis of yet more future developments.

It illustrates how the use of chromatic monitoring has evolved from only the use of optical fibre- based sensing (Jones, G. R. et al„ 2008) through the use of PC-based systems using VDU screen illumination as the optical source and a miniature camera for detection, leading to the production of a self-contained, controlled unit for use at remote sites without the need for a PC (Chapter 2).

Advances have been made in the domain of monitoring mechanical vibrations (Chapter 3) and in a variety of environmental monitoring, ranging from marine water monitoring and power from wind production to assessing the environmental impact of various electrical insulation gases (Chapter 4). Advances have also been made in adapting chromaticity for multidimensional monitoring such as elderly care and fish monitoring, three-phase power-line monitoring and time-frequency and time- wavelength monitoring, as well as a combination of different factors for assessing the degradation of high-voltage transformer oils and different gases for high current interruption.

The techniques described have the potential for deployment in several other applications in a cost- effective and convenient-to-use manner (e.g. mobile phone neonate in vivo jaundice monitoring).

Section I: Basic Chromatic Principles

A brief summary of the basic concepts of the chromatic monitoring approach is presented which has enabled chromatic monitoring concepts to be further developed, leading to their deployment for a variety of different applications, ranging from preliminary medical monitoring via environmental issues to electric power system monitoring.

Overview of Chromatic Monitoring

C. R. Jones and J. W. Spencer

Introduction

This chapter summarises the concept of chromaticity and its basic attributes for advancing its capabilities beyond those already established (Jones et al., 2008a). It indicates how chromaticity is related to pure measurement and the diagnosis of a condition. It explains how chromaticity quantifies information in terms of signal properties via an optimum of only three parameters. Approaches for extracting values of these three parameters via chromatic processors are described, and the manner in which these values may be displayed via chromatic maps and calibration graphs is indicated. The concepts of primary and secondary chromatic monitoring are explained, and the manner in which continuous and discrete signals can be addressed to check for emerging information about complex conditions is indicated. These aspects form the basis from which more recent advances have been made.

Monitoring and Chromaticity

Monitoring may be regarded as assessing the condition of a system or component (Jones et al., 2008a). It can be regarded as a means for connecting measurement with diagnosis (Figure 1.1). Measurement (which is reduction in nature) provides accurate quantification, whereas diagnosis involves distinguishing between conditions. Thus, monitoring provides a means for interconnecting measurement and diagnosis and dealing with complexity (Jones et al.. 2008a).

One characteristic of complexity is emergence, that is, dealing with unexpected events. Chromaticity addresses emergence via cross-correlation with processors whose responses overlap (non-orthogonal) rather than being discrete. An example of such non-orthogonal processors (Jones et al., 2008a) is shown in Figure 1.2 with three processors (R, G, B) whose responses overlap in the measurement domain. The distribution of a signal (U) addressed by the three R. G. В processors is also shown in Figure 1.2.

An example of such monitoring is the capability of the human vision system to distinguish between different optical colours (e.g., Billmeyer and Saltzman. 1981) with only three non-orthogonally

Monitoring in relation to measurement and diagnosis

FIGURE 1.1 Monitoring in relation to measurement and diagnosis.

Three non-orthogonal processors (R. G. B) superimposed upon a signal (U)

FIGURE 1.2 Three non-orthogonal processors (R. G. B) superimposed upon a signal (U).

responsive processors (R, G. B), where R addresses the longer optical wavelengths, G the middle wavelengths and В the shorter wavelengths.

Examples of Basic Chromatic Parameters

The outputs of the three non-orthogonal processors (Ro, Go, Bo) may be treated mathematically to produce various chromatic parameters, as shown by Jones et al. (2008a). Details of two particular examples, which have been shown to be useful for widespread monitoring applications, are shown in Table 1.1. These parameters quantify various features of a signal. One set of such parameters, which can be calculated from Ro, Go, Bo, are H, L, S, defined by the mathematical formulae given in Table 1.1.

L is the strength of the signal, which is the effective area under the signal envelope (Figure 1.3). H is the hue, which indicates the dominant part of the signal. 5 is the saturation, which quantifies the difference between the maximum and minimum levels of the R, G, В components of the signal (and so also indicates the spread of the signal).

An alternative set of chromatic parameters that have been shown to be useful for monitoring applications (Jones et al., 2008a) are X, Y, Z. These three parameters represent the relative magnitudes of the outputs from each processor, Ro, Go, Bo, as indicated in Table 1.1. As such, they provide an approximate indication of the signal distribution.

These two sets of chromatic parameters may be used to form chromatic maps on which a particular signal may be represented. Investigations (Jones et al., 2008a) have shown that a particularly useful map for monitoring applications is a polar diagram of L versus H (Figure 1.4a). This provides an indication of signal strength (L) as the radial coordinate and dominant range of the signal (H) as the azimuthal coordinate. More details of the signal R, G, В structure are provided by a Cartesian map

TABLE 1.1

Mathematical Definitions of Some Conventional Chromatic Parameters (Jones et al., 2008a)

Parameter

Expression

Name

Physical Meaning

HLS

L

(Ro + Go + Bo)/3

Strength

Signal Area

H

Hue

Dominant Part

120-120r/(g + r) b = min

r = Ro-min(Ro. Go. Bo)

240 - 120g/(g + b) r = min

g = Go-min(Ro. Go. Bo)

360 - 120b/(b = r) g = min

b = Bo-min(Ro. Go. Bo)

S

[(m+) - (m-)] / [(m+) + (m-)]

Saturation

Spread

(m+) = max(Ro.Go.Bo) (m-) = min(Ro.Go.Bo)

XYZ:

X

Ro/3L

R Proportion

Y

Go/3L

G Proportion

Z

Bo/3L

В Proportion

X + Y + Z=1:X = Y = Z = 0.33;X = 1 - ZwhenY = 0.

S = 0 —> Uniform Signal; S = 1 —► Monochromatic.

Physical meaning of various chromatic parameters (a) signal strength (L). hue (H) and saturation (S). (b) relative magnitudes X, Y. Z

FIGURE 1.3 Physical meaning of various chromatic parameters (a) signal strength (L). hue (H) and saturation (S). (b) relative magnitudes X, Y. Z.

of X versus Z. which provides an indication of the relative magnitudes of the R, G, В components and hence distribution of the signal (Figure 1.4b). (It should be noted that the XYZ chromatic map incorporates values of X as well as Y and Z by virtue of X + Y +Z= 1.) Taken together, these two maps provide a quantification of the signal, distinguishing features of signal strength, dominant region and proportion of signal strength spread over the R, G. В zones.

Chromatic maps (a) H

FIGURE 1.4 Chromatic maps (a) H : L polar map. (b) X : Y : Z Cartesian map.

Some additional chromatic parameters are available from colour science which can be adapted for some particular monitoring applications (Jones et al„ 2008b). Two of these additional sets of parameters (HSV and Lab) are defined in Appendix 1 A.

Effect of Different Chromatic Processor Profiles

The extent and shape of the chromatic R, G. В profiles may be varied to provide the most appropriate information extraction (Jones et al„ 2008a) for particular applications. Figure 1.5a (i) shows three R, G, В Gaussian profiles overlapping at their midpoints. With such processor profiles, the value of the dominant range (H) for a monochromatic signal displaced along the signal axis is shown in Figure 1.5a (ii). The resultant H distribution shows some non-linear variation, thus leading to a moderately variable sensitivity.

Figure 1.5b (i) shows three R, G, В Gaussian profiles with only minimal overlaps, whilst Figure 1.5b (ii) shows the corresponding H variation for a monochromatic signal shifted along the signal axis. This has a highly non-linear variation of H, implying a highly non-linear sensitivity.

Effect of R. G, В profile shape and overlap on processing,

FIGURE 1.5 Effect of R. G, В profile shape and overlap on processing, (a) Gaussian profiles, midpoint overlaps (i) profiles (ii) sensitivity, (b) Gaussian profiles of R, G, B, minimal overlaps (i) profiles (ii) sensitivity, (c) Linear profiles, midpoint overlaps (i) profiles (ii) sensitivity. (Jones, G.R. et al. (2008a) Chromatic Monitoring of Complex Conditions. CRC Press. ISBN 978-1-58488-988-5. figures 2.2.2 (a), (b), (c) and 2.2.3 (a), (b), (c) combined.)

Figure 1.5c (i) shows three R, G, В processors with linear variation profiles overlapping at their midpoints. Figure 1.5c (ii) shows a highly linear variation of H for this case so that the sensitivity is relatively constant throughout the overall range.

An appreciation of these properties is important for the further development of chromatic monitoring.

Chromaticity with Discrete Rather Than Continuous Signals

Discrete signal processing involves addressing a series of separate signals (rather than a single continuous one) with three non-orthogonal processors (Jones et al., 2008a). Such discrete signals may correspond to outputs from a number of different sensors operating in parallel (Figure 1.6). The signals may be divided into three groups, each group corresponding to different measured parameters (e.g„ set 1 being a series of temperatures and corresponding to R, set 2 being a series

Discrete signals addressed by three non-orthogonal processors

FIGURE 1.6 Discrete signals addressed by three non-orthogonal processors.

of electric currents and corresponding to G, set 3 being a series of gas pressures and corresponding to B). Each group is addressed by one of three non-orthogonal processors (R. G, B). The chromatic parameters derived from the outputs of R. G. В provide a comparison of the various sensor outputs so that the relative magnitudes of the groups can be defined. An example of such a system is the monitoring of different gases dissolved in transformer oils, which can indicate different causes of the oil degradation (Jones et al.. 2008a).

Various Chromatic Representations

Consideration of previous research (Jones et ah, 2008a) shows that chromatically monitored data may be represented in different ways in order to convey information in the most appropriate manner for different purposes. Three main ways in which this may be achieved are via chromatic trend mapping, chromatic parameter-based calibration graphs and secondary chromaticity. These provide the basis of future developments and applications of chromatic monitoring.

Chromatic Trend Mapping

The H : L and X, Y, Z chromatic maps (Figure l .4a and b) provide a convenient means for classifying monitored signals. A test result shown on such maps indicates the condition of the monitored system according to the values of the chromatic map coordinates (H. L or X, Y, Z). Variation in the condition of the system is indicated via the locus of such measured coordinate values. Figure 1.7a shows an example of such a variation on an XYZ Cartesian chromatic map, whilst Figure 1.7b shows a variation on an H : L polar chromatic map. This illustrates the power of such chromatic mapping for providing a rapid and convenient impression of trends in a complex system.

As a result, the H : L map provides an indication of how the strength (L) and dominant value (H) vary, whilst the X, Y, Z map gives an indication of the approximate distribution of the relative magnitudes of three of the signal's components.

Chromatic Parameter-Based Calibration Graphs

Many applications require a quantification of an effect in order to indicate how far from a critical level the condition may be. A chromatic parameter can be derived to provide such quantification by calibration with known conditions. A first step in the choice of the most appropriate chromatic parameter is checking for a required trend on the XYZ and H : L chromatic maps [Figure 1.7a (i)

Representation of system trends on chromatic maps (i) and calibration graphs (ii). (a) X. Y, Z chromatic Cartesian map and Z/Y calibration graph, (b) H

FIGURE 1.7 Representation of system trends on chromatic maps (i) and calibration graphs (ii). (a) X. Y, Z chromatic Cartesian map and Z/Y calibration graph, (b) H : L chromatic polar map and L or H calibration graph.

and b (i)]. A more precise calibration curve may then be produced by more detailed monitoring tests with regard to that chromatic parameter.

Examples of such chromatic parameter calibration curves are shown in Figure 1.7a (ii) and b (ii) with the chromatic parameters Z/Y as a function of impurity level [Figure 1.7a (ii)] and H or L [Figure 1.7b (ii)] as a function of temperature.

Choices may be made based upon which chromatic parameter might provide the best trend. For example, consideration may be given to which parameter shows the most linear variation over the range of the monitored physical condition. Alternatively, the parameter providing the highest sensitivity within a given range may be preferred. Furthermore, a combination of chromatic parameters may be preferred, for example, Z/Y=B/G.

Secondary Chromatic Monitoring

There are many cases where the time variation of a complex condition may need to be observed and quantified, for example, degradation of a liquid with age (Elzagzoug et al„ 2014). In such a case, the condition may be chromatically tracked at selected intervals of time (tl, t2, t3), as shown in Figure 1.8a. This is referred to as primary chromatic monitoring. A suitable chromatic parameter from primary chromatic monitoring is then chosen and tracked as a function of time. Secondary chromatic monitoring may then be undertaken in the time domain using this parameter and secondary chromatic maps produced. Various trends may then be observed in the resulting secondary chromatic maps (Figure 1.8b).

Secondary chromatic processing, (a) Primary processing step, (b) Secondary processing step

FIGURE 1.8 Secondary chromatic processing, (a) Primary processing step, (b) Secondary processing step.

Summary and Recent Developments

Chromatic techniques have already been applied for monitoring a variety of complex conditions, including electric power equipment, medical diagnosis and environmental aspects (Jones et al., 2008a). From these studies, it is apparent that chromatic techniques have potential for further monitoring developments. The present book describes progress made in such extrapolation of chromatic monitoring.

Details of the chromatic methodologies described in this chapter are based upon preferences demonstrated by previously used investigations (Jones et al., 2008a). Consideration has mainly been made of the widely used chromatic parameters HLS and XYZ plus their properties and chromatic maps. The significance of such maps has been indicated for providing a rapid and convenient impression of trends in a complex system. Such maps can also be used for identifying an appropriate chromatic parameter for use as a calibration factor in the more detailed quantification of the variation of a physical condition. Secondary chromatic monitoring has been indicated to have the potential to provide a further level of chromatic analysis, such as the time variation, of a different domain chromatic condition. Discrete chromatic processing has been used for monitoring an array of different signals.

Further developments of chromatic monitoring have occurred for many areas of application, ranging from the optical domain (e.g., combining different optical properties) via the acoustical domain to the processing of a multiplicity of complex discrete components. The chromatic monitoring approach has been extended for use with different instrumentation and devices in the quest for improved efficiency plus convenient and economic monitoring. For example, such developments have involved the use of mobile phone cameras and visual display units as optical sources and devices for pre-data transmission processing. These various examples are considered in the chapters which follow.

Appendix 1A: ADDITIONAL CHROMATIC MAPS: HSV, LAB

There are transformations other than the HLS and XYZ w'hich are well established in colour science and which can have potential for some particular chromatic monitoring applications (Jones et al., 2008b). Two such transformations are the H, S, V and L, a, b methods. The mathematical formulations of these transformations are presented in Table l A.l for convenient comparison with the H, L, S and X, Y, Z transformations given in Table 1.1.

TABLE 1A.1

Mathematical Definitions of Some Additional Colour Science Parameters (Rogers, 1985; Ainouz et al., 2006)

Parameter

Expression

Physical Meaning

HSV

V

max(R. G. B)

Highest output

s

(V - min)/V

Relative spread of V min = min (R. G, B)

H

cb - eg 2 + cr - cb 4 + eg - cr

V = R dominant

V = G dominant

V = В dominant

cr = (V - R)/(V - min) eg = (V - G)/(V - min) cb =( V - B)/(V - min) H = H*60 H < 0 H = H + 360

Lab

L

116(G/Gn)*l/3 - 16

Normalised mid processor

a

500[(R/Rn)*l/3 - (G/Gn)* 1/3]

Long —> mid difference

b

200[(G/Gn)*l/3 - (B/Bn)* 1/3]

Mid —> short difference

Rn. Gn, Bn references

For R/Rn. G/Gn. B/Bn > 0.008856

References

Ainouz. S„ Zallet, J., de Martino. A. and Colledt. C. (2006) Physical interpretation of polarisation-encoded images by colour preview. Opt. Express 14(13), 5916-5927.

Billmeyer. F.W. and Saltzman. M. (1981) Principles of Color Technology. John Wiley, New York.

Elzagzoug. E., Jones, G.R., Deakin. A.G. and Spencer, J.W. (2014) Condition monitoring of high voltage transformer oils using optical chromaticity. Meas. Sci. Techno! 25. 065205.

Jones, G.R., Deakin A.G. and. Spencer. J.W. (2008a) Chromatic Monitoring of Complex Conditions. CRC Press, ISBN 978-1-58488-988-5.

Jones, G.R.. Pavlova, P. and Spencer, J.W. (2008b) Other Chromatic Processing Algorithms, Chapter 3, Chromatic Monitoring of Complex Systems. CRC Press. ISBN 978-1-58488-988-5.

Rogers, D. (1985) Procedural Elements for Computer Graphic. McGraw-Hill, New York.

 
Source
< Prev   CONTENTS   Source   Next >