HU217732B - Method and device for qualifying a physical object or determining characteristic thereof - Google Patents

Abstract

The present invention relates to a method of geometric, high data reduction with which a physical object, such as a sample, quality, or property of a given feature of a particular feature, can be characterized by two coordinates (x0, y0) of a quality point (Q) defined as a center of gravity by sorting in a polar coordinate system. In determining the quality point (Q) according to the invention, the physical object characteristics of the examined object are weighted by their distance from the origin of the polar coordinate system, preferably by linear or square weighting. The present invention, on the other hand, defines the physical object of a device or determines its property, comprising a device for determining the values of one or more of the features of an object, and a data processing unit for polar coordinate system sensor for the specified characteristic values, which data processing unit in the apolar coordinate system determines a quality point (Q) based on characteristic values based on characteristic values equipped with tools. According to the invention, the data processing unit has qualitative means (Q) for determining the characteristic values with respect to their distance from the origin of the apolar coordinate system. Conclusions about the quality of the quality point (Q), and possibly the distance from a reference product quality point, to the quality or quality of the physical object may be drawn for classification or sorting. ŕ

Description

The invention relates to a method and apparatus for qualifying or determining a property of a physical object.

By measuring the evolution of a first physical property on a material while changing the value of another physical property, we obtain a change in the first physical property as a function of the second physical property, which is generally the property of the material. Such a function is formed, for example, by various spectra, such as an optical characteristic (transmission or reflection) of a substance as a function of the wavelength of the illumination radiation. Spectra are also obtained by plotting the complex impedance of the test substance as a function of frequency, or the radiation power of a filament as a function of temperature, or the X-ray intensity of a crystalline material as a function of the angle of reflection.

Spectra are carriers of important information about the properties (composition, quality) of the test substance, however, obtaining and interpreting useful information usually requires sophisticated mathematical statistical methods. Most such spectral acquisition and estimation instruments and methods only become capable of determining their properties from their spectra after a "learning" process (calibration) of a large number of measurements on a sample army of known properties (composition or quality characteristics).

Changes in physical characteristics as a function of another physical characteristic (e.g., wavelength, frequency, temperature, angle, time) are generally described as a continuous and non-ruptured function. Spectra are usually such continuous, non-bursting functions that are generally available in the form of an analog electrical signal through sensors and transmitters of physical characteristics. Digital computers commonly used in signal processing and evaluation, however, require digital signals. Therefore, continuous spectra should be converted to digital signals by sampling. Thus, the spectrum is then represented by an ordered set of dependent variables measured at the discrete values of the physical characteristic that plays the role of the independent variable. With today's measurement techniques, it is not uncommon for a single spectrum to be represented by up to a thousand spectral values. Spectrum processing refers to operations on these thousands of data.

The relationship between the characteristics (composition, quality) of the test substance and some of the spectral data of the sample cannot yet be described in most cases in the form of mathematical equations or equations. Therefore, they resort to the empirical determination of relationships. This requires a large number of samples whose properties (such as composition) need to be known. The data should cover the expected range of values and include the spectra of each sample with thousands of data.

Both the determination of property characteristic data and the measurement of spectral values are costly and time consuming. All data is error prone. This is due to the inaccuracy of chemical definitions, spectrum noise, drift, value scattering, temperature, particle size, compactness, etc. due to its interfering effects. Although some of the errors can be reduced by transformations on the spectrum (eg smoothing, derivation, Fourier transform), the determination of the relation (calibration) requires the use of mathematical-statistical methods.

Of these mathematical-statistical methods, the multiple linear regression (MLR) method, which has several variants, is the relatively simplest and most widely used. The stepwise method determines the correlation and standard deviation between the spectral values of the samples measured at the discrete values of the independent variable and the given property characteristics, and selects the best-performing independent variable from this value and searches similarly for the next best variable independent variable value and so on. For a spectrum of 750 spectral values, assuming 50 samples, this would mean a few minutes of computation on a standard-built IBM-compatible 386 PC. With the "stepwise" method, it is not certain that the independent variable pair or triple that gives the best result is actually calculated because the baseline is fixed and the result is very dependent on the baseline. In order to determine the truly best independent variable value pair, the so-called "best", to study all combinations, over a quarter of a million correlations have to be made for a spectrum of 750 spectral values, which can be done overnight in the above PC configuration, and defining the best three requires one week of computer time. Between 'stepwise' and 'best' there is the 'iterative' method which gives better results to the 'stepwise' method, but it is not certain that the result of the 'best' method will be achieved; this is, of course, more time-consuming for the "stepwise" mode person. The disadvantage of using the multiple linear regression method, although it has achieved many good results, is that it only uses spectral values for a few discrete independent variables to determine the property characteristic (s) sought, whereas other spectral values carry valuable information.

In order to shorten the computation time, the learning process (calibration) and to obtain more complete information from the spectrum, new mathematical-statistical methods were needed. One solution is to transform the spectra into the Fourier domain and to consider only the first 20-50 terms of the Fourier series coefficients. This means a significant reduction in data, and omitting higher coefficients only causes the noise to be cut off. With the Fourier transform, spectral data can be reduced by an order of magnitude such that all original spectral data plays a role in the formation of this compressed data.

The principal component regression (PCR) method provides another procedure with other results. This method measures the spectral values of the samples on the principal component2

It transforms it into the coordinate system of the sequences and then uses only a few of the principal components. This data reduction also solves problems due to collinearity in spectral values and allows for the elimination of non-sample populations as well as erroneous measurements.

The partial least squares regression (PLS) method takes into account, besides the correlation between the spectral values measured on the samples, the undesired differences in the data characterizing the properties of the samples.

Finally, we mention the neural network (NN) method, which is well-suited for nonlinear relationships, although it is very time consuming. The disadvantage of the neural network method is that it is difficult to infer physical content from it.

Previously, we have outlined some ways to extract information from the spectra, illustrating the difficulty and time-consuming nature of the learning process. While the Fourier transformation method is a mathematical data reduction solution, the MLR, PCR, and PLS methods are more physical, physico-chemical data reduction, and the NN method is a biological data reduction model.

A geometric approach for large-scale data reduction, other than the foregoing, has been proposed in which the measured spectral values are plotted instead of the normal Cartesian coordinate system, and the resulting spectral points are assigned equal masses, J. Kaffka, L. S. Gyarmati: Qualitative (Comparative) Analysis by Near Infrared Spectroscopy, Proceedings of the Third International Conference on Near Infrared Spectroscopy, June 25-29, 1990, Brussels, Belgium, p. 135-144], the polar coordinate system is called the "quality plane", the center of gravity is the "quality point" and the vector drawn from the origin of the coordinate system to the quality point is called the "quality vector". According to this method, the quality, for example the composition, of the materials characterized by their spectra can be characterized by this quality point or quality vector.

The present invention is an improvement on the above quality plane certification process. During the examination and practical testing of the method we made the following insights.

It has been found that in the polar coordinate system, for similar masses assigned to the spectral points, the effect of the spectral values closer to the origin is relatively large because they are compacted around the origin and thus play a significant role in shaping the center of gravity. However, the absorption peaks characteristic of the composition of the product are generally far from the origin of the polar coordinate system, and the absorption valleys, which are barely representative of the composition, are located near the origin. This means that the uninteresting spectral values are emphasized. We have recognized that it is possible to determine a higher quality point for qualification by assigning to each spectral point masses that are weighted at a distance from the origin rather than equal masses. The position of the center of gravity thus determined, i.e. the quality point, is more influenced by the characteristic spectral values.

The invention thus provides, on the one hand, a method of qualifying or defining a property of a physical object by determining the values of one or more characteristics of the object, and sequencing the characteristic values in a polar coordinate system to determine at least one quality point quality point position is used to qualify an object or determine its property. According to the invention, in defining the at least one quality point, the characteristic values are weighted by their distance from the origin of the polar coordinate system.

The process of the present invention achieves that high value characteristics, such as the absorption peaks of a spectrum, that are particularly characteristic of the physical object under study play a greater role in determining the quality point or quality points. Thus, the quality point (s) according to the invention is more representative of the physical object than the quality point obtained by the known method.

The weighing according to the invention can be done in several ways. It is preferable to use linear weighting, for example to determine at least one quality point by considering the center of gravity of the characteristic curve or the adjacent value points of the characteristic in the plane of the polar coordinate system.

The role of important characteristic values can be further enhanced by considering the center of gravity of the area of the surface enclosed by the characteristic curve in the plane of the polar coordinate system, i.e. the sheet cut along the spectrum, which corresponds to a square weighting. In this embodiment, the determination of the at least one quality point is based on the center of gravity of the area bounded by the characteristic curve and the line joining the start and end points of the curve, or the polygonal area obtained by joining the adjacent value points of the characteristic.

The method according to the invention allows for several classifications. In one embodiment, it is determined whether the quality point falls within a predetermined range or in a plurality of predefined ranges in the plane of the polar coordinate system.

A physical object can also be qualified by determining the distance between the quality point and a point corresponding to a reference quality in the plane of the polar coordinate system.

The method of the invention is suitable not only for classifying physical objects but also for determining their properties. This can be accomplished according to the invention by the polar coordinate system3

In the plane of the agent, lines for specific values of one or more properties are defined, and the value of said one or more properties of the object is determined by the position of the quality point relative to these lines.

You can also determine the value of the physical object property from the coordinate values of the quality point by a linear relation, where the constant of the linear relation can be obtained by regression calculation.

The method of the invention can be applied to any characteristic of any physical object, provided that the values of the properties are quantifiable and quantifiable. The characteristics may be, for example, values of a spectrum. However, it is also possible that the characteristics are quantities of different dimensions. When applying the method according to the invention, the characteristic values must be ordered in any, but always the same way, and one or more quality points in the polar coordinate system must be determined on the basis of the characteristic values so ordered. For a given rating or measurement, only the quality points obtained by the same sorting can be compared.

The method according to the invention is advantageous when the characteristics of the physical object under investigation are given as a function of a single parameter, for example a physical property. The characteristic values can then be sorted according to the values of the parameter, that is, the value range of the parameter corresponds to the angular range of the polar coordinate system, thus determining one or more quality points in the polar coordinate system.

Optionally, the value range of the parameter in the polar coordinate system is set to an angle of 360 °, and the alignment is performed such that the mean values of the two value ranges of the characteristic in the polar coordinate system are approximately 90 ° or 270 ° in the polar coordinate system. be apart. This provides valuable information on the shape of the spectrum plotted in the polar coordinate system, for example, for the composition of the test substance.

It may also be advantageous to map the value range of the parameter to two or more angle ranges, to define a quality point for each angle range, and to qualify the object or determine its property based on the two or more quality points.

The invention further provides an apparatus for classifying or determining a property of a physical object, comprising a device for determining values of one or more characteristics of the object and a data processing unit sequentially defining said characteristic values in a polar coordinate system based on at least one center of gravity. means for determining a quality point, wherein according to the invention, the data processing unit has means for determining a quality point that takes into account characteristic values by distance from the origin of the polar coordinate system.

A preferred embodiment of the apparatus according to the invention is that, after determining each characteristic value, the quality point determining means calculates a component of the quality point coordinates, adds them to the sum of the previously calculated components and obtains the coordinates after determining the last characteristic value. have defining assets in proportion to amounts.

With this solution, the device continuously generates the appropriate coordinate value after each characteristic value is determined, so that previous values are not stored and the coordinate value is immediately available at the end of the procedure. Thus, even with a large number of characteristic values, such as more than a thousand, the final result can be quickly obtained with relatively low computing capacity.

BRIEF DESCRIPTION OF THE DRAWINGS The invention will now be described, by way of example only, with reference to the accompanying drawings, in which:

Figure 1 is a reflection spectrum of a physical object in the near infrared (NIR) range, a

Fig. 2 is a graph in polar coordinate system of Fig. 1, a

FIG. 3 is a diagram illustrating one embodiment of the qualification method of the present invention based on the spectrum of FIG.

Figure 4 is a diagram illustrating another embodiment of the rating method of the present invention based on the spectrum of Figure 2;

Figure 5 is a diagram illustrating a further embodiment of the rating method of the present invention based on the spectrum of Figure 2;

Figure 6 is a flowchart of an algorithm applicable to the method of the invention illustrated in Figure 3, a

Fig. 7 is a flowchart of the algorithm applicable in the method of the invention illustrated in Figs. 4 and 5, a

Figure 8 is a simplified block diagram of an embodiment of the apparatus of the invention, a

Figure 9 is a schematic drawing illustrating a method of spectral acquisition based on interactivity of the physical object under investigation in Figure 8;

Fig. 10 is a schematic diagram illustrating the transmittance spectral acquisition of a physical object in the apparatus of Fig. 8, and

Figure 11 is a schematic diagram illustrating the spectral acquisition of the physical object under investigation in the apparatus of Figure 8.

In the drawings, like reference numerals are used to denote elements having the same or same function.

Figure 1 shows the reflectance NIR intensity spectrum of a physical object under investigation, in this case a goose liver sample, in the Cartesian coordinate system at a wavelength of 1320-1800 nm. The λ _0, λ _1, ... _λ ί, ... λ _[2 wavelength values V _o, spectral values (relative numbers are dimensionless) tar4

The case where i = 0, l, ... k, and thus the number of spectral values (k + 1) = 13.

Figure 2 shows the spectrum of Figure 1 is shown a polar coordinate system where they are assessing the V _o, V '... V ₁₂ spectral values of the angle. The wavelength range 1320-1800 nm corresponds to the 0-360 ° angle range of the polar coordinate system, so the S = 40 nm wavelength range means an angle of = 30 °. The plane of the polar coordinate system is called the quality plane, in this plane the Q quality point is characterized by x ₀ and y ₀ coordinates. The vector drawn from the origin of the polar coordinate system to the Q quality point is called the quality vector.

The Q quality point describes the quality (composition) of the physical object being examined, for example a material sample. Each material sample has a Q quality point on the quality plane. However, the reverse is not always true, and several samples of different grades of material can produce the same Q quality point. The likelihood that two quality points defined on the basis of different spectra coincide is extremely small, but in principle this is not excluded. In this case, two quality planes can be used, for example, to equate the measured wavelength range to an angle range of 2x360 °. And the probability is practically zero that the Q1 and Q2 quality points determined from the spectral values of two different quality material samples in two different wavelength ranges.

Figures 3, 4 and 5 illustrate the determination of the x ₀ and y ₀ coordinates of the Q quality point Q in accordance with the invention. In all three cases the quality point Q Xo ₀ and y coordinates are determined from the sum of a finite series.

According to Fig. 3, the spectrum represented by the polar coordinate system according to the invention is given by k 77 = k cos ia - cos (i - 1) a J + sin ia -V ^ _, ^ ^k y ₀ = -Y y _b . assigning masses linearly weighted with the spectral values to the points corresponding to the spectral values obtained, and determining their center of gravity, which is taken to be the Q quality point.

The x ₀ and y ₀ coordinates of the Q quality point can be determined from the measured spectral values by the sum of the following finite series:

^k χθ = -V Xj, where Xj = Vj ^{1 2} cos ία (1) k + ^k

Yo ⁼ U "? Zy, where ^ = ¾ ² k +1 i = o sin ία (2) and where Vj is the spectral value at the i-th wavelength, k + 1 is the number of spectral values, [°] = 360 / k. If the wavelength range (λ _ηΐ3χ -λ _γηίη ), then the wavelength range used for spectrum _{acquisition is} S = (k _max- k _min ) / k. If we assign to the points corresponding to the spectral values squares weighted with the spectral values, Vj ^{3 is} replaced by Vj ^{2 in the} context of (1) and (2).

Figure 4 illustrates another embodiment of the process of the invention. Here, the center of gravity of the straight lines connecting the points corresponding to the spectral values is the Q quality point. Thus, the absorption peaks characteristic of the composition are given a greater role in determining the Q quality point. In practice, this means that the spectral values are linearly weighted by their distance from the origin. The Xo and y ₀ coordinates of the Q quality point can be determined from the sum of the following finite sequences:

y ^k 'x ₀ = - £ x, in which k í ₌ i sin

Vj cos ia + V {j_Q cos (i - l) a (3) y, = ^ [ _Vj cos ia - cos (i - l) a J + [v, sin ia - sin (i - l) a J x

Vj sin ia + V (j_,) sin (i - l) a (4) where the notations correspond to the notations used in equations (1) and (2).

Figure 5 illustrates a further embodiment of the process of the invention. Here, the center of gravity of the area bounded by the straight lines joining the points corresponding to the spectral values, i.e., the center of gravity of the sheet cut along the spectrum, is considered to be the Q quality point. In practice, this means taking into account the square values weighted by their distance from the origin. The x ₀ and y ₀ coordinates of the Q quality point can be determined from the sum of the following finite sequences:

li x ₀ = - £ ^x j in which ki ₌ i

VjV, j ,, sin α V cos ia + V _fi cos íi-í) a _X j = —- _x - (5) ^k y ₀ = τΣΥί 'in which kj = j

Vj V / j „sin is Vj sin ia + V _(i sin (i - l) is _y . = -1-χ-i_2-, (6) where the notations correspond to the notations used in equations (1) and (2) .

Note that the methods of Figures 3, 4 and 5 result in a Q quality point different from each other as shown in (1) - (2), (3) - (4) and (5) - (6).

EN 217 732 Also visible from dependencies. Of course, there is only one way to use a particular certification.

6 is a flow chart showing a preferred computation algorithm for generating the x ₀ and y ₀ coordinate values according to (1) and (2). After the step 29 (START) occurs in step 30 the index i, and Xo is set to ₀ and y coordinates of the initial value to zero. In step 31, it is determined whether the value of i is less than or equal to k, where k + 1 is the number of spectral values. If yes (Y), step 32 is measured Vj spectrum value corresponding to the index i, and to determine at step 33 is included in the ith spectral values x _f value is calculated, and it is added to the Xj previously stated value, and in step 34 the value of y ₍ for the ith spectral value is calculated and added to the previously determined value of yj.) In step 35, the value of i is incremented, and the program returns to step 31 and the cycle repeats to the next spectral value. the answer is not (N), i.e. i is greater than k, then all spectral values have been measured and taken into account, so in step 36, x _{ i = Zx _i / (ki-1) and y ₀ = Zy ₁ / (k + 1) coordinate values stored and displayed by the data processing unit.

The flowchart of Fig. 7 shows a preferred computation algorithm for generating the coordinate values x 1 and y _{0 according to} (3) and (4) and (5) and (6). After step 29 (START), in step 30 the values of the index i and the coordinates Xq and y ₀ are set to an initial zero. In step 31, it is determined whether i is less than or equal to k, where k + 1 is the number of spectral values. If so (Y), the spectral value Vj for index i is measured or determined in step 32 and then examined in step 37 for the value of i not zero. If i = 0, that is, the first spectral value, no calculation is made, but in step 35 the value of i is incremented. If i + 0, then xj for the i th spectral value is calculated in step 33 and added to the previously determined value of X j, and y 34 for the i th spectrum value is calculated and added to yj until then. Then, in step 35, the value of i is incremented, and the program returns to step 31, and the cycle is repeated for the next spectral value. If the answer in step 31 is not (N), i.e. i is greater than k, then all spectral values are already measured and taken into account, so in step 36, x _{) = Zx _i / k and y ₀ = Zy, / k are calculated. coordinate values that are stored or displayed by the data processing unit.

When the method of the present invention is applied based on features of the physical object under investigation that are already available and stored in the data processing unit, FIGS.

In the flow chart of FIG. 7, step 32 is omitted.

Figure 8 illustrates an apparatus according to the invention with an example of a composition measuring apparatus operating in the NIR (i.e., near infrared) wavelength range of electromagnetic radiation. The NIR wavelength range is particularly suitable for the fast, non-destructive determination of the composition of matter and the concentration of components in the material, since in this range most organic materials have absorption bands, i.e. the information is rich in spectrum. In addition, the most transparent materials in this range, such as transmission or reflection measurements, can overcome difficulties due to inhomogeneity due to the higher layer thickness and greater penetration depth.

The apparatus illustrated in FIG. 8 does not contain any rotating or moving parts and does not require sampling, so it has a very long life. The radiation is provided by a series of 1 LED providing cold light, the LEDs of which are successively energized by a series of 2 LED driver units. The LED driver unit 2 is controlled by a microprocessor data processing unit 24 via 3 I / O units as described below. In the data processing unit 24, 20 displays (monitors or LCDs) are connected to 4 CPUs via 18 buses via 21 RAMs, 22 ROMs and 19 I / O units. The bus 18 also has a 3 I / O unit and a 17 A / D converter. The spectral distribution and bandwidth of the light emitted by the LEDs generally does not provide enough monochromatic radiation, so we provide transmission narrow grids with 5 optical grids and 6 slots. Monochromatic radiation of sufficiently narrow bandwidth may be provided in a manner different from that illustrated, for example by means of a reflection optical grid or other type of monochromator, such as an optoacoustic filter tunable by an electrical control signal. The pre-optical grid 5 collimator lens 7 and the subsequent imaging lens 8 perform collimation (parallelization) and mapping functions. The monochromatic beam emanating from the slit 6 is divided into lenses 9, which divides the beam into two halves of a transparent glass, preferably at an angle of 45 °. One smaller portion is directed to a reference detector 11, while the other larger portion serves to illuminate the Physical object 13, such as a material sample, through 12 fiber optics. The radiation reflected from or transmitted through the physical object 13 is transmitted to another fiber optic measuring detector 15. The I _R signal of the All Reference detector and the I signal of the measuring detector 15 are applied to a twin logarithmic amplifier input, which outputs a signal proportional to the I / I _R ratio of the signals. The output signal of the twin logarithmic amplifier 16 passes through the A / D converter 17 to the CPU 4 for processing, which displays the result, for example, a quality vector characteristic of the physical object 13 being examined, on the display 20. This is done in the following way.

Upon power-up, the CPU 4 receives a current pulse at 0, for example, from a thirteen-membered row of LEDs 1, emitting a beam 7 which parallels the rays and passes it onto an optical grid 5 which, through the lens 8, λ ₀ emits wavelength monochromatic radiation. This is the monochromatic su6

As described above, the irradiation passes partly to the reference detector 11, and partly to the physical object 13 by reflection and passage through the fiber optics 12 and 14, and then to the measuring detector 15.

The geometric relationship between the physical object 13 and the fiber optics 12, 14 may vary according to reflection (Figure 8), interactivity (Figure 9), transmittance (Figure 10) or, for transparent liquids, e.g., the reflection surface 23 behind the measuring vessel with the help of - transflectance (Figure 11). For the measurement of interactivity, the fiber optics 12 and 14 are connected side by side on the same surface of the physical object 13 and no reflecting surface is disposed on the opposite surface of the physical object 13.

The I and I _R signals are logarithmic to the twelve logarithmic amplifiers and form the logarithmic difference. Thus the log I-log I _k = log (I / I _k) j _ob according _{ektum λ0} (7) of the correlation signals I and _R¹ coefficient appears at the output 16 of the dual logarithmic amplifier. In this way, the error caused by the change in the intensity of the LED radiation can be eliminated. The A / D converter 17 digitizes this signal and divides the CPU 4 into a log (I / I _R ) _SI standard obtained from a standard (reference) material (usually ceramic) stored in ROM 22 at the same wavelength λ ₀ , with λο. The resulting spectral value V _o belongs to the wavelength λ ₀ . By generating the quotient of the signals from the physical object 13 and the standard material by the output of the twin logarithmic amplifier 16, the error due to non-uniform changes in time between the all reference detector and the measuring detector 15 can be eliminated. As the characteristics of the reference detector 11 and the measuring detector 15 change slowly, it is sufficient to measure and store the spectral values of the standard material at greater intervals, such as monthly.

The CPU 4 then calculates the value of the first term of the series for λθ using the formulas given for Xj and y _r , for example, in the flowchart of Figure 6. It was the first cycle.

In the next cycle, the CPU 4 receives a pulse at the command of the serial number 1, resulting in a monochromatic radiation of wavelength λ) passing through the two channels as described above in the slot 6. Is obtained as log (I / I _k) j _{ob ektum} value appearing at the output 16 of the dual logarithmic amplifier divided by log (I / I _{k) λ ι standard} values measured in the standard materials, λ, V, spectral value associated wavelength. The CPU 4 calculates the value of the second term of the series for λ, using the formulas given for x and yj, and adds it to the value calculated in the previous cycle.

Then the process continues until the last LED in the example, number 12, is operated, until the last member of the series x, yj for X ₁₂ , then the total sum of the series, the coordinates of minőségθ and y ₀ Up to 20 displays. Of course, the Q quality point of the reference product can be simultaneously displayed on the display 20, and even the distance Ορθ _ΐ3Γ described below can be displayed. In the example above, we used thirteen spectral values, but it could be hundreds or thousands, the process is the same, but it is repeated several times.

During the practical testing of the process according to the invention, the following discoveries were made. It is easy for those skilled in the art to apply the method of generating large amounts of data reduction according to the invention for the sake of clarity and ease of interpretation. In fact, factories often do not require precise knowledge of all the composition data (concentration values) of a product during production, and are satisfied that the quality of a reference product (a good sample) would be better than a certain amount. In this case, we can describe the extent of the difference between the quality points of the reference product and the product under investigation. This can be determined as follows:

Xor wherein _e and f y _(lref the reference product quality Q _ref-point coordinates, x ^ and yovizsg and coordinates of the quality point Q UUT.

We can therefore define a circle around the Q _ref quality point of the reference product within which the product under consideration is acceptable for Q quality points, and if the Q quality point falls outside this range, it should be considered as a waste. The procedure can be used for grading or sorting.

Distances such as the Euclidean and Mahalanobis distances have already been defined to measure the differences between the spectra. These differences give a good indication but do not say anything about the reason for the difference so that the two products tested at the same distance from the reference sample may still be very different. In contrast, in the quality plane, the cause of the difference can also be inferred from the position of the quality point of the product under investigation. For example, if the quality of the foie gras examined differs from the quality of the reference foie gras in the same direction as the one in the polar coordinate system, the fat absorption peak shows that the fat of the test foie gras is higher than the fat of the reference foie gras. For this purpose, the wavelength range corresponding to 360 ° should be chosen such that the directions of the absorption peaks of the two most important components in the polar coordinate system are approximately 90 ° or 270 °.

During the practical testing of the process according to the invention, the following additional advantages have been found. The method is insensitive to spectrum noise. The plus and minus differences of high frequency noise compensate for the location of the Q quality point, so there is no need to “smooth” the spectra.

The method is also insensitive to changes in particle size of the physical object under investigation and, in the case of transmission measurements, to changes in thickness. These changes affect the spectrum in the polar coordinate system as if it were magnified or reduced while the center of gravity remains in place.

HU 217 732 A

The process does not require any learning process, no calibration, only the knowledge of the spectrum of the reference product is required to qualify a product. The procedure operations can be performed very quickly because only the sum is calculated (wavelengths of the previous series members) calculated by adding the values of the members of the series for calculating the coordinates of the Q quality point and adding to the previous one, so only one sum is measured. add and the result is ready.

The process of the present invention, which is essentially qualitative, can be further developed to be quantitative. There are two ways to do this. One method involves empirically defining contour lines for a given property of a physical object, such as equilibrium lines for the concentration of one of its components. When defining contour lines for each value of a given property, it is advisable to select an angle range corresponding to the value range of the parameter so that it corresponds to a value range in which the property in question has characteristic effects. With this in mind, it may be advisable to apply multiple quality planes to quantify multiple properties of a physical object and apply multiple quality points accordingly.

Alternatively, the concentration of a given property, such as a component K in a given concentration range, can be calculated by linear approximation as follows:

K = A + BxQ + Cy ₀ , (9) where Χο and y _{0 are} the coordinates of Q and the coefficients A, B and C can be determined by regression.

The method of the invention is applicable to any type of spectrum (optical, mechanical, electrical, etc.) and any transform thereof, and to any set of characteristics of a physical object that can be quantified and quantified.

Claims (19)

PATENT CLAIMS A method for qualifying or defining a property of a physical object, comprising determining the values of one or more of the object properties, and sequentially defining the characteristic values in a polar coordinate system based on the characteristic values at least one quality point based on the position of the at least one quality point qualifying or defining an attribute of an object, characterized in that, when determining at least one quality point, the characteristic values are weighted by their distance from the origin of the polar coordinate system. Method according to claim 1, characterized in that the at least one quality point is determined by linearly weighting the characteristic values by their distance from the origin. A method according to claim 1, characterized in that the at least one quality point is determined by weighting the characteristic values by their square distance from the origin. Method according to claim 1, characterized in that the determination of the at least one quality point is performed by taking into account the center of gravity of the characteristic curve or the adjacent value points of the characteristic in the plane of the polar coordinate system. Method according to claim 1, characterized in that the determination of the at least one quality point is performed by joining in the plane of the polar coordinate system the area bounded by the characteristic curve and the line joining the start and end points of the curve, or we consider the center of gravity of the polygonal area. 6. A method according to any one of claims 1 to 6, characterized in that the object is classified as falling within a predetermined range in the plane of polar coordinate system. 7. A method according to any one of claims 1 to 6, characterized in that the object is classified as falling within one or more of two or more predetermined ranges in the plane of the polar coordinate system. 8. A method according to any one of the preceding claims, characterized in that, for classifying the object, the distance between the quality point and a point corresponding to a reference quality is determined in the plane of the polar coordinate system. 9. A method according to any one of claims 1 to 6, characterized in that the property of the object is determined by determining lines belonging to given values of one or more properties in the plane of the polar coordinate system, and determining the value of said one or more properties of the object a. 10. A method according to any one of claims 1 to 5, characterized in that the value of the property of the object is determined from the coordinates of the quality point by a linear relation, wherein the constants of the linear relationship are swallowed by regression calculation. 11. A method according to any one of the preceding claims, characterized in that the characteristics of the object under investigation are determined in function of a parameter and the ordering is performed in accordance with the values of the parameter. The method of claim 11, wherein the parameter range in the polar coordinate system is mapped to an angle range of 360 °, and the alignment is performed such that the mean values of the two value ranges of the characteristic in the polar coordinate system are approximate to the assay. 90 ° or 270 ° apart. HU 217 732 A 13. The method of claim 11, wherein the value range of the parameter in the polar coordinate system is mapped to two or more angles, defining a quality point for each angle region, and qualifying the object or determining its property based on the two or more quality points. 14. Apparatus for classifying or defining a property of a physical object, comprising a device for determining the values of one or more characteristics of the object, and a data processing unit for sequencing said characteristic values in a polar coordinate system with means for determining at least one quality point characterized in that the data processing unit (24) has means for determining a quality point which takes into account characteristic values by their distance from the origin of the polar coordinate system. 15. The apparatus of claim 14, characterized in that after determining the quality point determining means for each characteristic value of the quality points x ₀ and y ₀ coordinates of a \ and y _s cos i - V ^ _,) cos (i - l) aj + | ^ V _] sin ia -V ^^ sin (i - l) a] x Vj cos ia + V ^ cos (il) has means for calculating its component, adding these to the sum of the previously calculated components, and determining the coordinates a and y 'after determining the last characteristic value as Lxj and Ly ₍ proportional to the sums). 16. Apparatus according to claim 15, characterized in that the xj component of the quality point determining means xj is represented by the relation Xj = Vj ² cos ia, the y coordinate y ₀ and the component yj = Vj ² sin ia. having calculating means, where Vj is the angle value in the polar coordinate system between the i th characteristic value and the consecutive characteristic values. 17. The apparatus of claim 15, wherein said quality point determining means of the quality point χθ coordinates x _s component of Xj = Vj ³ cos ia relationship _y0 coordinate yj component and calculating on the basis of the set yj = Vj ³ sin ia context having means wherein V; the angle between the i th characteristic and the consecutive characteristic values in the polar coordinate system. 18. The apparatus of claim 15, wherein the quality point determining means has an X ₍₎ coordinate of the quality point x; the component yj of the relation, and the component yj of the y ₀ coordinate is / i ^{cos 3a_} \ ii) ^cos (il) aj + [ ^V ' ^{S3n 3a_} \ ii) ^s * ⁿ 0 ^_ O ^a ] ^x Vj sin ία + ν ₍₍ _η sin (il) has means for computing the relation, where Vj and ν ₍ ί_η are the ith and (il) -th values and the angle between successive values is the polar coordinate -System. 19. The apparatus of claim 15, wherein the quality point determining means has an x coordinate of the quality point Xq _; component of VjV (j_,) sin a Vj cos ia + V ^ cos (i-l) a X = 3 χ2 context, y ₀ y coordinate; and its component is VjV.j has sin calculating tools Vj sinia + V _(il , sin (il) _y = -L2-χ-2 3, where V; and V ₍ j_ |) are the i-th and ( il) -the characteristic value and the angle between successive characteristic values in the polar coordinate system.