DE102007063473A1 - Motor vehicle body'lacquered surface evaluating method, involves comparing brightness and color values to reference brightness value and reference color value, respectively, to determine brightness and color deviations for each sub-area - Google Patents

Abstract

The method involves generating an image and producing a set of sub-areas in the image. A weighted brightness value and a weighted color value are generated for image points of each sub-area. The brightness values are compared to a reference brightness value of a color control model and the color values are compared to a reference color value of a color control model for determining a brightness deviation and a color deviation for each sub-area. The determined deviations are compared to a predetermined threshold value, and a lacquered surface and the sub-areas are graphically represented. An independent claim is also included for an arrangement for quality control of a surface.

Description

The The present invention relates to a method of quality control a variety of surfaces, in particular relates to present invention a method for quality control of colored surfaces of bodies and their attachment and installation parts for motor vehicles.

For example, methods for quality control of surfaces are off DE 101 03 555 A1 and WO 2004/010124 A1 known.

DE 101 03 555 A1 describes a method for assessing a color layer, in particular the surface impression of the color layer on a viewer, which allows a quick assessment of the color layer with little technical and programmatic effort and recognizes any color deviations or other errors. In this method, two-dimensional information is extracted from the five-dimensional (x, y, Y, I, Q) information recorded by means of an optical receiver, which measures the color and / or brightness curve of the received light in a predetermined geometric region of the color layer represents.

WO 2004/010124 A1 discloses a simulation method in which the perception of the human eye is understood. One degree of freedom of the viewing human eye is the viewing angle. The angle-dependent contributions of color perception can also be detected by any adjustment of the viewing angle. On the basis of these angle-dependent contributions, a material-specific characteristic curve can be created, which can be used, for example, in painting processes in the automotive industry to evaluate the risks of individual work steps.

The object of the present invention is to provide an improved simulation of the perception of the human eye for the evaluation of a plurality of surfaces, in particular paint layers of body parts of a series production. In particular, larger areas (> 500 cm ² ) can be evaluated with little effort, but quickly and reliably. Another object of the present invention is to automatically recognize the necessity of interfering with the production process of the surfaces to be evaluated (eg, the painting process) so as to be able to respond to trends (color trends) in a timely manner during the production process. A further object of the present invention is to automatically detect and eliminate or interpolate technical and / or statistical outliers in the measured value acquisition and to supply only the remaining measured values to a further evaluation.

These Tasks are according to the invention by the features of claim 1. Preferred embodiments The invention are contained in the subclaims.

• – sequentielles Lackieren einer Vielzahl von Oberflächen,
• – jeweils nachfolgendes Einstrahlen von Licht auf jede der lackierten Oberflächen,
• – Abbilden mindestens eines Teils des von der jeweiligen Oberfläche reflektierten und/oder gestreuten Lichts auf einen zweidimensionalen, ortsaufgelösten Empfänger, und
• – Erzeugen eines zweidimensionalen Bildes (mit einer Vielzahl von Bildpunkten) für jede lackierte Oberfläche, wobei jedem Bildpunkt neben seinen geometrischen Koordinaten weiterhin Farb-Koordinaten und Helligkeits-Koordinaten zugeordnet werden,
• – für jede lackierte Oberfläche eine Vielzahl zweidimensionaler Unterbereiche gebildet wird, die innerhalb des zweidimensionalen Bildes der jeweils zu untersuchenden Oberfläche liegen,
• – aus den Bildpunkten jedes Unterbereichs wird jeweils ein gewichteter Helligkeitswert, mindestens ein gewichteter Farbwert und/oder ein aus Helligkeitswert und mindestens einem Farbwert generierter gewichteter Gesamtfarbwert erzeugt,
• – die gewichteten Helligkeitswerte und/oder die gewichteten Farbwerte und/oder die gewichteten Gesamtfarbwerte werden jeweils mit einem Referenzwert verglichen und daraus wird eine Helligkeitsabweichung und/oder mindestens eine Farbabweichung und/oder ein Gesamtfarbabstand ermittelt,
• – die ermittelte Helligkeitsabweichung und/oder die mindestens eine Farbabweichung und/oder der Gesamtfarbabstand mindestens eines Unterbereichs wird chronologisch für die Vielzahl lackierter Oberflächen in einer Datenreihe erfasst,
• – aus den mindestens zwei letzten Elementen der Datenreihe wird ein Trend für die Helligkeitsabweichung und/oder die Farbabweichung und/oder den Gesamtfarbabstand ermittelt,
• – die Helligkeitsabweichung und/oder die Farbabweichung und/oder der Gesamtfarbabstand wird für eine vorgegebene Anzahl von Oberflächen vorausberechnet,
• – und die vorausberechnete Helligkeitsabweichung und/oder die vorausberechnete Farbabweichung und/oder der vorausberechnete Gesamtfarbabstand wird mit einem ersten Grenzwert verglichen.

The method according to the invention comprises the following method steps:

• Sequential painting of a plurality of surfaces,
• Each succeeding irradiation of light on each of the painted surfaces,
• Imaging at least a portion of the light reflected and / or scattered by the respective surface onto a two-dimensional, spatially resolved receiver, and
• - Generating a two-dimensional image (with a plurality of pixels) for each painted surface, wherein each pixel in addition to its geometric coordinates continue to be assigned color coordinates and brightness coordinates,
• For each painted surface, a multiplicity of two-dimensional subregions is formed which lie within the two-dimensional image of the respective surface to be examined,
• In each case a weighted brightness value, at least one weighted color value and / or a weighted total color value generated from the brightness value and at least one color value are generated from the pixels of each subarea,
• The weighted brightness values and / or the weighted color values and / or the weighted total color values are each compared with a reference value and from this a brightness deviation and / or at least one color deviation and / or a total color distance is determined,
• The determined brightness deviation and / or the at least one color deviation and / or the total color distance of at least one sub-area is recorded chronologically for the plurality of painted surfaces in a data series,
• A trend for the brightness deviation and / or the color deviation and / or the total color distance is determined from the at least two last elements of the data series,
• The brightness deviation and / or the color deviation and / or the total color distance is determined for a predetermined number of surfaces precalculated,
• And the predicted brightness deviation and / or the predicted color deviation and / or the predicted total color distance is compared with a first limit value.

Out The result of the comparison can be modified for Derived the manufacturing process of the painted surfaces become. To determine the trend are preferably between 2 and 20 of the last elements of the data series used. Preferably is determined from the ascertained brightness deviations and the determined Color deviations for each sub-area one total color distance determined and for this total color distance one each Data set for the variety of bodies recorded.

The Idea of the invention is the quality of a variety similar and painted with the same color body parts to examine their color fidelity or their distance from a reference model. Here are the painted body parts in situ and successively examined and evaluated during the manufacturing process to check for any trends (color trends) during to be able to react quickly to the production process. Such Trends can be made, for example, by successive bonding a nozzle, but also by other unwanted influences occur on the devices used for production.

According to the invention, the visual impression of a larger area of preferably more than 100 cm ² (more preferably more than 500 cm ² and more preferably more than 1000 cm ² ) can be represented by a few averaged (or weighted) color or brightness values and easily evaluated. That is, preferably a painted surface of greater than 100 cm ² (more preferably greater than 500 cm ² and more preferably greater than 1000 cm ² ) is imaged onto the two-dimensional, spatially resolved receiver.

Around corresponding tendencies of color deviation from the reference model being able to react very quickly, it is intended the chronological or chronological development of the color deviation (Brightness deviation, color deviation or total color distance) too determine and the color deviation for a given number of bodies based on the temporal evolution into the future to forecast. It is considered according to the invention, that the survey (by irradiation of light and imaging to a receiver) and evaluation of the actual production (Painting) is always delayed. Therefore, the color deviation becomes preferably for at least such a number of bodies predicts that within series production in between time Paint and rating have been re-painted. Then corresponds the predicted color deviation namely at least the in-situ expected color deviation of the currently painted Body. The specified number of bodies to be predicted can be determined by an operator. It is alternative but also possible, this predetermined number in advance in the Implement system and thus provide as system parameters.

Around to be able to intervene in production on time, should the brightness deviations, color deviations or total color distances not only be compared to an absolute criterion, but It is provided according to the invention, each value (Brightness deviation, color deviation or total color distance) of Data series (i.e., for a variety of measurements Body parts) for one or more clusters a trend value to be assigned, taking into account at least one (preferably a plurality) possibly existing preceding and / or subsequent element of the data series is formed. For the value of the first measured bodywork can only be the trend value from the determined value (brightness deviation, color deviation or total color distance) and possibly subsequent values become. For later to be measured bodies can this trend value is a (weighted) average of a number of previous ones and / or subsequent elements of the data series. Out of the constantly progressive data series for brightness deviation, Color deviation or total color distance thus becomes at least one cluster (preferably each cluster) assigned a further data series, from their course on corresponding trends regarding brightness deviation, Color deviation or total color distance can be closed. Farther should be the trend of the trend value at the time of valuation in the future, d. H. on the actual painting process be predicted within serial production (the actual evaluated later).

In order to be able to make an intervention in the production in good time, the predicted data series of the trend values (predicted color deviation) as well as a permissible spread are compared with a first limit value, the so-called tolerance limit. If this tolerance limit is allowed by the sum of If the spread and the corresponding trend value are reached or exceeded, it is assumed that a surface (not yet measured or not yet rated but possibly already painted) will lie or lie outside of the set tolerances, and the production process will be modified (interruption, if applicable) and initiate a fault analysis).

The allowable spread is determined such that a given Percentage of the determined differences (between measured value such as brightness deviation, Color deviation or total color distance and assigned trend value) within the permitted spread. The percentage can be specified by a user. It is an alternative, however also possible to use this percentage (% security) Determination of the permitted spread in advance in the system too implement and thus provide as system parameters.

The Tolerance limit (first limit) can be selected from a color control model (For example, a multiple (real number) of the scatter of the color control model). Such a control model can be provided in advance by having multiple series bodies (or their parts), in particular 10 to 100, preferably 35 to 65 such series bodies (or their parts), whose Surfaces were classified subjectively as flawless, with regard to their brightness (Y) and color coordinates (I, Q) and from this for the individual clusters a respective one If (reference) mean value is formed. This mean value represents then the corresponding reference color value or the corresponding Reference brightness value for the respective cluster. This procedure makes it possible, for example, to find out the cloudiness of a motor vehicle. The single one Body is thus defined in its current situation with the Color control model compared.

A painted surface in the context of the present invention is a surface that one or more on top of each other lying paint layers may have. The color-determining layer does not have to be the top layer for multi-layer coatings Rather, the latter may, for example, also be a clearcoat. The color-determining layer in the sense of the present invention is therefore the coating layer in the case of multi-layer coatings, the intended, final color of the painted Essentially determined.

Next the tolerance limit based on the color control model can Further limit values (intervention limit, warning limit) based on required Tolerances are defined, resulting from the production process of the overall vehicle or objectives of the company depict. These limits can then be used corresponding color trends before reaching the tolerance limit signal. The tolerance limit becomes regular by spread of the color control model - this is the state of technology related to the product "painted body" - multiplied with a real number in the range between 1.5 and 10 (preferably 2-4); Limit of intervention and warning limit also multiplied by the spread of the color control model with a real number (the multiplier for the intervention limit is less than the tolerance limit and the multiplier for the warning limit is lower than for the intervention limit) selected. The multipliers can by a Be set operator. Alternatively, it is also possible to implement these multipliers in advance in the system and thus to be provided as system parameter.

Preferably the individual subareas are disjoint to each other. hereby can the visual perception of the surface to be examined particularly efficient with a small number of subregions be represented. Of course it is also possible that individual clusters overlap but lead this to redundant data, so that thereby the number of processed Data would be increased unnecessarily.

Preferably, the subregions have an area between 0.1 cm ² and 100 cm ² (more preferably 1 cm ² and 25 cm ^2, and more preferably 5 cm ² and 15 cm ² ). The area of a sub-area (cluster) in the context of the present invention is understood to be the actual geometric area of the imaged sub-area on the surface to be examined (painted body part) and not the area of the image on the two-dimensional, spatially resolved receiver. The area of the image on the two-dimensional, spatially-resolved receiver is regularly smaller than the actual area of a sub-area.

On the one hand, the number of parameters to be evaluated is reduced to an efficient level by the size selection according to the invention for the individual clusters; on the other hand, the spatial progression of color / brightness deviations for the entire (to be examined) surface or almost the entire (to be examined) surface can be represented become. The representation of the spatial course of color / brightness deviations is important, for example, because a certain amount of color / brightness deviation tions in peripheral areas of a component can be significantly more relevant than in the middle, since the component must harmonize in the edge region, for example, to other components and therefore a certain degree of color / brightness deviation can be perceived by a viewer more sensitive - especially with different surface normals of joined Sharing in the adjacent area.

The Number of subregions is preferably between 5 and 3000 (more preferably between 15 and 500 more preferably between 25 and 200). The location and size of each Subareas may preferably be pre-set by a user become. This has the advantage of having areas for the optical perception by a viewer are not relevant (for example, faces in a larger area) can be suppressed accordingly beforehand since the clusters are preferably localized only in areas, that for the visual perception by a viewer are relevant. Furthermore, it is possible for the optical perception is considered to be particularly relevant Define areas (sub-areas) and only these sub-areas (Prio partial areas).

When Color coordinates are preferably those of the RGB color space, the CIELab color space or the DIN99 color space. It is alternative also possible to use other color coordinates. From the Color coordinates become a weighted brightness value and at least a weighted color value that determines a measure of Display the brightness and color of the cluster. This allows the variety the pixels on the spatially resolved receiver be transformed into a variety of clusters, with a cluster represents a particular geometric section. The weighted brightness value is preferably an arithmetic one Average of the brightness values of all pixels of the cluster, and the weighted color value is preferably an arithmetic mean the color values of all pixels of the cluster. It is also an alternative possible to add a weighted brightness value / color value which deviates from the arithmetic mean. Essential is, however, that at least 10% (preferably at least 50%, especially preferably at least 95%) of the pixels of the cluster in the weighted Include brightness value / color value.

by virtue of this transformation becomes the number of color values to evaluate significantly reduced without the essential information about the spatial course of color / brightness deviations get lost.

Preferably Subareas will always be in the same way within the two-dimensional Image of the respective surface is positioned. preferably, are used in the same way shaped and painted bodies. As a result, the color / brightness values of individual clusters are different Body conclusions on color / brightness trends to, so that an error analysis can be made promptly.

In a preferred embodiment of the invention it provided that the clusters at least 30% (more preferably at least 50% more preferably at least 80%) of the area to investigative surface on the spatially resolved Receiver is covered, cover. As a result, according to the invention, the overall impression particularly realistic due to the large number of sub-areas represents.

at the acquisition of data series of different body parts a series (for at least one cluster - this one Cluster data series can also be used for all clusters it is intended to be technical outliers (For example, by a wrong body part, failure of the Flash or late shutter release such that the body part is no longer in the given Position is, may be conditional) from the data series to eliminate. These are brightness deviations, color deviations and / or Total color distances, the amount exceed a second predetermined limit, advance (i.e., before determining the trend value), with the second Limit value greater in size as the first limit. The second limit is preferably chosen such that clusters with a total color spacing greater than the second limit as outside the tolerance limits determined by measurement / production errors become.

By the elimination of technical outliers is on the one hand reduces the amount of data to be processed and continues to ensure that corresponding averages (= trend values), which for a long-term analysis (trend analysis) or a prognosis underlying scenarios (wrong body part, flash failure or too late Triggering the recording so that the body part not in the default position) is not distorted become.

In a further preferred embodiment of the invention, it is provided (preferably additionally) statistical outliers of a data series (for example, brightness deviation, color deviation or total color distance) of different body parts of adjacent elements (previous and / or subsequent measurement values) of the data series to avoid the falsification of trends.

To it is provided the numerical value of the brightness deviation (or the color deviation or the total color distance) those elements of a data series of a sub-area, the amount exceed a third predetermined limit, through replace a numerical value resulting from the brightness deviation (or the color deviation or the total color distance) at least one adjacent element of the data series of the subarea is formed.

Preferably, before irradiating light, a vehicle body part is painted and at least 100 cm ^{2 of} the painted surface of the vehicle body part (more preferably more than 500 cm ² and more preferably more than 1000 cm ² ) is imaged on the two-dimensional, spatially resolved receiver. This makes it possible to very realistically evaluate a very large surface by evaluating only one image (for example, a CCD camera).

Preferably become the surface to be examined as well as all Subareas are shown graphically, with subregions whose Sum of predicted trend value and the determined allowable spread greater than the first threshold, and subregions, their total color difference (or brightness deviation and at least one color deviation) smaller than the first limit value is marked differently in the graph, for example, by different colors (for example Sum of predicted trend value and the determined permissible value Spread width greater than the tolerance limit (first Limit value) is displayed in red and the sum of the predicted trend value and the determined allowable spread is less than Tolerance limit and greater than the limit of intervention turns yellow and all other clusters turn green shown). This is an assessment by a User, whether the deviations due to their spatial Arrangement lead to a sorting of the body part, especially easy. Such an assessment by a User is regularly from other points of view, for example, taking into account further with the same or similar color painted parts in the later Environment of the body part (when installed), with regular deviations be considered more relevant in borderline areas than in the center a body part, because the sensitivity of a viewer for color differences in border areas regularly is increased.

alternative However, it is also possible only a part of the sub-areas to represent according to predetermined criteria, for example, only preselected Subregions (so-called prio subfaces) or just the or the "worst" sub-areas of a to be evaluated surface, that is those or those subareas with the largest Total color distance (or brightness deviation and at least a color deviation).

Preferably a smoothing of the elements of the data series is performed. Preferably, the smoothing by means of 2-parameter method and / or pre-post averaging.

The Forecast / trend analysis described above can be both real-time respectively. In this case, of course, only previously determined deviations (that is previously painted and evaluated bodies of the same cluster) in the evaluation incorporated. Alternatively, it is possible the above described trend analysis after the metrological detection of a Variety of (same painted bodies) perform. Then both previously and subsequently determined Deviations (ie the same cluster from before / after painted and evaluated bodies) are included in the rating. From such a trend analysis can very quickly conclusions be drawn to which process engineering (of a User, if necessary parameters to be set) to a noticeable Trend in the color / brightness gradients, so that a faster response / correction is possible.

The Invention will be described below with reference to FIGS Embodiments explained in more detail become. Show it:

1a a flowchart of a preferred embodiment of the method according to the invention;

1b an enlarged view of the subdivision of a painted, evaluated surface in a plurality of sub-areas;

2a Different display options for brightness / color deviations and prognosis for brightness / color deviations of a data series of a plurality of body parts for a selected sub-area;

2 B the representation of the total color distance for 100 body parts and the prediction of the total color distance for 5 body parts for a selected sub-area;

3 a representation of a data set of the total color distance of a plurality of body parts for a selected sub-area for the detection of technical and statistical outliers;

4a a photographic representation of a proper body part (reference);

4b - 4c photographic representations for the evaluation of body parts in different situations;

5 a representation of the data series 3 after elimination technical outliers;

6a - 7b a schematic representation for the detection and elimination of statistical outliers;

8th a representation of the data series 3 after elimination of technical outliers and interpolation of statistical outliers;

9 a representation of the data series 3 after the elimination of technical and statistical outliers;

10a C is a schematic representation of the trend abstraction of successive measured values and

11a - 12b a further schematic representation of the trend abstraction and the evaluation of the individual measured values.

1a shows a flowchart of a preferred embodiment of the method according to the invention. On the one hand, the object of the present invention is to monitor color and brightness scattering (deviations from a reference model) of series production, that is to say a multiplicity of similarly painted body parts, and to present it in a clear manner to a user on the basis of preset criteria. From this representation, a user can either remove immediately or at least quickly infer whether a particular painted body part of the series meets the criteria for visual perception. On the other hand, tendencies (trends) of color and brightness scattering over the entire series (or at least) of a part of the series are to be determined and predicted, ie by a prognosis of the color and brightness scattering, the delay between coating and measuring / evaluating the surface should be compensated so that in-situ conclusions can be drawn on compliance or non-compliance with production parameters during painting.

The The invention relates to the evaluation and presentation of a series In particular, on a variety of identical body parts, with painted in the same color. The corresponding painted Body parts should then by means of the invention Be evaluated, the technical characteristics of the evaluation in particular the visual perception by a viewer as possible to simulate realistically.

According to the invention, as in 1a (top window), a painted body part - which in the exemplary embodiment has a non-paintable opening for a fuel cap - preferably completely on a two-dimensional receiver - such as a CCD camera - imaged and this image in a variety of sub-areas (also called "cluster" Compared to the measurement at individual measuring points, the (preferably complete) imaging of the body part (at least 30%, preferably at least 50% and more preferably at least 80% of the area of the body part visible from a camera position should be imaged on the receiver). the advantage that not only a non-representative section of the body part is detected, which can lead to a significantly different rating compared to the visual perception of the human eye (which regularly covers the entire body part).

1b shows an enlarged view of the subdivision into a plurality of subregions. It is intended to form a color and brightness mean value for each cluster and to assign this data to the cluster. This means, for example, that each cluster has two location coordinates (x, y) and three color coordinates according to RGB color space, CIELab color space, DIN99 color space or similar (which already include a brightness value). However, comparable spatial coordinates and color coordinates obtainable by appropriate transformation can also be assigned. The color coordinates also comprise, at least to some extent, information about the brightness of the surface to be evaluated. The isolated clusters in the upper area of the 1b are used to adjust the image brightness and check the system stability. For example, it is possible to measure whether the cameras were moving or the amount of light in the flash was too high. However, these isolated clusters are not used to determine the brightness / color deviations of the color control model. By assigning a color and brightness mean value for each or predefined cluster, a larger area can be represented by a small number of data (color and brightness values) according to the invention. According to the invention, these individual cluster-related data are compared with corresponding data of a color control model and the corresponding deviations are stored. In a preferred embodiment of the method according to the invention, the ascertained brightness and color deviations are calculated to a so-called total color distance.

Color differences can in principle be determined by vector addition of the ΔL, ΔC and ΔH values between the reference color and the color specimen. ΔE = √ ΔL² + ΔC² + ΔH²

The CIE builds on the L * a * b color space in their 2001 additionally released CIEDE2000 color difference formula corrections to the underlying ΔL-, ΔC- and ΔH precedes by a further distortion of the color space a consideration of the influence of color saturation and to achieve a correction of the alignment of the tolerance ellipses. As total color distance but can also equivalent Sizes are used.

This total color distance represents a weighted value from the corresponding brightness and color deviations. In order to be able to carry out a quick analysis of a painted body part by a user, it is provided according to a preferred embodiment of the invention, initially technical outliers (see 1a under 1.), then to interpolate statistical outliers (in 1a : small windows next to 1.-3 .: statistical outliers are indicated by a cross) and below a trend value 10 . 14 (solid line 10 in the small window next to 2nd and 3rd in 1a and solid line 14 in 2 B ). This trend value 10 . 14 is preferably obtained by averaging a plurality (preferably between 1 and 20) (possibly existing) previous and subsequent values 17 (Total color difference 20 from the color control model). After that, the scatter of the given total color distances becomes 17 around the trend value 10 . 14 determined and represented graphically ( 2 B and points and crosses in the small window next to 3rd and 5th in 1a ).

Thereafter, the tolerance and intervention limits are set. For this purpose, first a security in% (see in 1a under 10.), for example by an operator or as a system parameter. Furthermore, a prognosis range for the determination of a tolerance limit and possibly also a prognosis range for the determination of an intervention limit (see 1a under 4.), for example by an operator or as a system parameter.

For example, it may be required that at least 95% of the bodies measured are within a permissible range 15 . 16 ( 2 B ) should lie. This then becomes apparent from the total color distances measured up to that point 17 in-situ the allowable spread 15 . 16 determined and also graphically represented (dashed line 15 . 16 above and below the trend value line 14 ). The permissible spread 15 . 16 is preferably approximated by a distribution function of the scattering about the trend line 10 . 14 (preferably Gaussian approximation or polynomial approximation).

The prognosis width for the determination of a tolerance limit is the number of bodies to be predicted in order to decide on the suitability of painted body parts. This number (prognosis margin of the tolerance limit) is preferably greater than the number of painted bodies, which are painted in the period between painting and evaluation of a body, so that the production process may be interrupted or modified in good time before painting a body outside the tolerance limit. The prognosis range for the determination of an intervention limit is the number of bodies to be predicted in order to determine the necessity of an intervention in to decide the production process. This number (prognosis size of the intervention limit) is preferably greater than the number of painted bodies that are painted in the period between painting and evaluation of a body, so that the production process can be interrupted or modified in good time before painting a body outside the tolerance limit, if necessary.

Furthermore, a tolerance limit 12 and an intervention limit 13 determines, for example, the tolerance limit 12 as a 3-fold spread of the color control model and the intervention limit 13 set as double the width of the color control model (see also 2 B ).

Furthermore, it is provided, the trend line 14 ( 2 B ) respectively. 10 ( 2a and in the small window next to 5th and 6th in 1a ) to forecast or interpolate. In addition, there will be a mean increase in a large number of past trend trend values 14 ( 2 B ) and based on the last trend value 14 (Abscissa = 0) pre-calculated for a given number of bodies (prognosis = 5). This gives a prediction of how the total color separation (or brightness / color deviations) will develop with the trend remaining the same. From this, a prognosis can be made as to whether the bodies lying within the permitted spread width - which have not yet been measured / evaluated - are still within the tolerance limit 12 or the intervention limit 13 lie. On the basis of this assessment, it is possible to intervene in the production process in good time if an unacceptable shift in the total color distance (or the brightness / color deviations) sets in, ie if the sum of the predicted total color distance and the permitted spread is greater than or equal to the tolerance limit 12 (or the intervention limit 13 ).

In 2a are the predicted values for the trend line 10 in the windows 7 shown. In 1a are the predicted values for the trend line 10 (and the allowable spread 15 . 16 ) in the tel area 19 whereas the actual readings are in the subrange 18 are shown.

After that, the compliance probability becomes the tolerance limit 12 determines, that is, whether the sum of predicted total color distance and allowable spread of a given number (prognosis distance for the tolerance limit) of not yet evaluated bodies within the tolerance limit 12 lies. If this is not the case, the corresponding subarea (cluster) is classified as out of order and represented accordingly (for example red) (see below in FIG 1a ). In such a case, preferably the production process is interrupted or modified.

The determination of the compliance probability can be carried out by known mathematical methods, for example by determining a relative ratio between a) an indefinite integral of the measured value dispersion (around the trend line 10 . 14 ) and b) through the (upper and lower) tolerance limits 12 particular integral considering the location.

Furthermore, it is possible the compliance probability of the intervention limit 13 ( 2 B ), that is, whether the sum of predicted total color distance and allowable spread of a given number (prognosis range for the intervention limit) of not yet evaluated bodies within the intervention limit 13 lies. If this is not the case, the corresponding sub-area (cluster) will be displayed accordingly (eg yellow) (bottom in 1a ). In this case, an operator (or a system routine) can intervene in the production process, since it already appears that the color values tend to develop such that a timely departure from the tolerance limit 12 is to be expected. An intervention in the production process can be, for example, an interruption of the series coating and checking of the means of production (nozzles, color mixing, etc.).

The above analysis between the measured brightness and color values and the reference brightness and color values (color control model) incl. Trend analysis, determination of the permitted spread, forecast and comparison with tolerance limit 12 (possibly also intervention limit 13 ) is preferably made for all clusters (subregions) of each body part. Alternatively, however, it is possible to perform the corresponding comparison of the total color distances (or brightness and color distances) with the first limit value only for a certain number of clusters of a body part. Another aspect of the present invention is not only to evaluate large painted areas in a simple system, but also to be able to efficiently detect changes in brightness and color values within a series of painted surfaces, with the evaluation of color and brightness variations (or the total color distances) of a visual perception realized by a viewer should come as close as possible and continue to be able to detect trends in color and brightness changes as quickly as possible.

For this purpose, it is provided according to the invention to make the corresponding division of the two-dimensional image for each body part of a series to be evaluated in each case the same sub-areas, so that the corresponding sub-areas of a series (a plurality of successively painted bodies) can be compared and evaluated accordingly (ie , the location coordinates and the extent of the clusters are the same for all bodies of a series). For this purpose, according to the invention, the brightness and color differences (for example the total color distance) of a cluster for a multiplicity of painted surfaces are recorded in a data series, wherein the individual elements of the data series should preferably be sorted chronologically according to the time of painting. Such a representation is for example in 2a for different color / brightness criteria, showing a series of 100 painted parts (see abscissa from -100 to 0). In particular, by ΔY in window 1 the brightness deviation of the individual bodies no. -100 to no. 0 for a preselected (always the same) cluster is represented by the corresponding color control model. From the brightness deviations recorded over the series, corresponding mean values, which are determined, for example, from a large number of preceding and following measured values, can be represented for each measured value. Likewise, corresponding tolerance limits, which are likewise obtained or generated from the multiplicity of measured values, can be displayed. In a similar way, windows 2a with ΔI a color deviation red-green, window 3 with ΔQ a color deviation yellow-blue, window 4 ΔE represents an overall color deviation. In window No. 5 the color deviations are Cartesian and in windows 6 is illustrated to illustrate a brightness deviation as a bar graph.

In the 2a shown trend lines 10 (or weighted gradients) can allow a user conclusions on the compliance or non-compliance of production parameters, so leaving the tolerance limit 12 within a series can be detected in time. The limit of intervention serves as indicators 13 ( 2 B ). In the windows 1 - 4 of the 2a the abscissas show one consecutive number of the painted bodies and the ordinates each the determined values for brightness deviation / color deviation or total color distance. In the fields of 7 become the values of trendlines 10 is approximated to the future, preferably by determining a mean increase of a predetermined number of the last determined values of the trend lines 10 and updating the values of the trend lines 10 taking into account the average increase.

To the in 2a shown mean value lines 10 To be able to determine properly, it is provided according to the invention to eliminate technical outliers and statistical outliers from a corresponding data series and not to be considered in the averaging (or weighting). 3 Fig. 11 shows a data series of the total color distance ΔE (ordinates) of a plurality of body parts (abscissa) for a selected sub-area for detection of technical and statistical outliers. (technical outlier: indicated by a cross and statistical outliers by a border).

A technical outlier can be a number of scenarios that may not be caused by a faulty finish, but by other circumstances. By way of illustration is in 4a the photograph of a correctly painted body part shown. In 4b For example, a photograph of a body part painted with a wrong color is shown, in which case (which is to be classified as a technical outlier) the total color distance of a preselected cluster exceeds a second limit. The same as in 4c represented, for example, be the case when the triggering of the camera (two-dimensional spatially resolved receiver) takes place so late that automatically moving in a production line body is no longer (or not yet) at the predetermined position for photographic detection (also technical outlier ). Another case is in 4d shown, where it is the failure of the flash, so that a corresponding photograph can only be made incorrectly (also technical outliers). In 4e is another example, but it is not a technical outlier. There, a photograph of a painted body part is shown, which was done under the influence of (unwanted) extraneous light. Here, however, the total color distance of the corresponding cluster has remained below a second limit, so that this total color distance in the data series, see 3 , continues to be considered (no elimination). The second limit should be adjusted according to the above scenarios. The second limit value is preferably used as the overall color difference ΔE = 10.

In 5 will be after 3 shown measured values after elimination of the technical outlier.

Another aspect of the method according to the invention consists in the interpolation of statistical Aus lure (see 3 bordered measured values with 2 <ΔE <10). Statistical outliers are measured values which are not classified as technical outliers, that is, whose total color distance is, for example, smaller than a second limit, but whose deviations deviate so significantly from the other measured values, so that these are not random deviations can (here, for example, the total color distance is greater than a third threshold value). The advantage of interpolating these statistical outliers is that the corresponding trend analysis (for example, those in the 2a . 11a and 12a shown statistical center line) is not unnecessarily distorted. The third limit value is preferably used as the total color difference with ΔE <10 and ΔE> 2.

In 6a to 7b an example for the interpolation of statistical outliers is presented. In particular, the determination of the third limit value from a plurality of measured values is possible in different ways. In 6a to 7b On the abscissa is a multitude of painted bodies and on the ordinates the respectively determined values for the total color distance are plotted.

In the exemplary embodiment, the determination of the third limit value is to be demonstrated on the basis of 5 known measured values of one series, ie 5 known total color distances for 4 different bodies:
x ₁ = -8; x ₂ = 6; x ₃ = 6.4; x ₄ = 10.25; x ₅ = 35

First the median of the measured values is formed: x ~ = 6.4

Then the median of the absolute deviations (MaD) of x ~:
| x ₁ - x ~ | = 14.4; | x ₂ - x ~ | = 0.4; | x ₃ - x ~ | = 0; | x ₄ - x ~ | = 3.85; | x ₅ - x ~ | = 28.6; MaD = 3.85

Finally, the upper and lower outlier limits are calculated using the following table Sample size N g (n, α) Gavies / Gather Grubbs Dixon g (n, 5%) g (n, 1%) g (n, 10%) g (n, 5%) g (n, 10%) g (n, 5%) g (n, 1%) 3 33.59 161.15 1,148 1,153 0.886 0.941 0.988 4 10.12 24.71 1,425 1,463 0.679 0.765 0,889 5 11.77 28.84 1.602 1,672 0.557 0.642 0,780 6 8.20 15.40 1,729 1.822 0.482 0.560 0.698 7 8.32 13.75 1,828 1,938 0.434 0.507 0.637 8th 7.30 11.87 1,909 2,032 0.479 0.554 0.683 9 7.38 11.48 1,977 2,110 0.441 0.512 0,635 10 6.81 10.20 2,036 2,176 0.409 0.477 0.597 11 6.97 10.96 2,088 2,234 0.517 0.576 0.679 12 6.51 9.27 2,134 2,285 0,490 0.546 0.642 13 6.71 10.53 2,175 2.331 0,467 0.521 0.615 14 6.30 8.70 2,213 2,371 0.492 0.546 0,641 15 6.49 9.95 2,247 2,409 0.472 0.525 0.616 16 6.15 8.32 2,279 2,443 0,454 0.507 0.595 17 6.29 9.25 2,309 2,475 0.438 0,490 0,577 18 6.03 8.05 2,335 2,504 0.424 0,475 0.561 19 6.10 8.49 2,361 2,532 0.412 0.462 0.547 20 5.92 7.84 2,385 2,557 0.401 0,450 0.535 21 5.92 7.70 2,408 2,580 0.391 0,440 0.524 22 5.83 7.67 2,429 2,603 0,382 0,430 0.514 23 5.74 6.92 2,448 2,624 0.374 0.421 0,505 24 5.75 7.53 2,467 2,644 0.367 0.413 0,497 25 5.59 6.15 2,486 2,663 0,360 0.406 0.489 Table: Critical factors g (n, α) for the methods of Davies / Gather, Grubbs and Dixon are calculated:
g _u (n, α) = x - g (n, α) · MaD; g _o (n, α) = x ~ + g (n, α) · MaD;
g _u (5.1%) = 6.4 - 1.76 x 3.85 = -0.376; g _o (5.5%) = 6.4 + 1.76 × 3.85 = 13.176

The However, determination of the third limit value may also be after others Criteria are made.

8th shows a representation of the data series 6a . 6b after the elimination interpolation of the statistical outliers, the circles in 8th represent the interpolated values.

9 shows a schematic representation of the measured values of 3 After elimination of both technical and statistical outliers In 3 . 7 and 9 a series of 250 painted body parts was rated. In contrast to technical outliers, according to the invention, statistical outliers are not completely eliminated since, in addition to additive outliers, it is also possible to identify jumps or trend changes in the course of the measured values as outliers. Interpolation, rather than complete elimination of the outliers, provides an adequate response of the monitoring system without the need for a special outlier evaluation. The effects of a wrong decision (unrecognized outliers) are thus minimized.

Another aspect of the present invention is the so-called trend abstraction, which in the 10a to 12b is shown schematically. The idea is to compare the corresponding measured values of a cluster or several clusters (measured values are determined brightness deviations, determined color deviations or the determined total color distance) for a series of a plurality of In order to illustrate trends for each measured value, a weighted average, which is formed from a large number of preceding or succeeding elements of the data series, is likewise represented graphically (as a line). For example, the 2-parameter method according to Holt can be used for the corresponding averaging, in which initially optimal parameters are determined and, if appropriate, a smoothing of the weighted average results (pre-post averaging, golden ratio) is carried out.

10a . 10b . 10c show by way of example the result of the method of exponential smoothing which simultaneously determines the basic period value and the trend of a measured value curve and smooths it by means of 2 different parameters. Optimum results are selected from pre-post averaging. It provides significantly better results than trend abstraction by averaging.

11a shows a representation of the data series 9 after performing the 2-parameter procedure with the optimal parameters. 11b shows the scatter of the elements of the data series 11a ,

12a shows a representation of the data series 9 after performing the 2-parameter procedure and the pre-post averaging. 12b shows the scatter of the elements of the data series 12a ,

1

color deviation brightness

2

color deviation Red Green

3

color deviation Yellow blue

4

Total color difference

5

color deviation Red-green and color deviation yellow-blue Cartesian

6

color deviation Brightness as a bar graph

7

forecast update

10

trend value

12

tolerance limit

13

control limit

14

trend value (calculated trend)

15

spread

16

spread

17

reading (Total color difference)

18

review (over 10 bodies)

19

forecast (5 bodies)

20

Total color difference

.DELTA.I

color deviation Red Green

.DELTA.Q

color deviation Yellow blue

.DELTA.Y

color deviation brightness

AE

Total color difference

QUOTES INCLUDE IN THE DESCRIPTION

This list The documents listed by the applicant have been automated generated and is solely for better information recorded by the reader. The list is not part of the German Patent or utility model application. The DPMA takes over no liability for any errors or omissions.

Cited patent literature

• - DE 10103555 A1 [0002, 0003]
• - WO 2004/010124 A1 [0002, 0004]

Claims (14)

A method of evaluating a plurality of painted surfaces, comprising the steps of: sequentially painting a plurality of surfaces, respectively irradiating light onto each of the painted surfaces; imaging at least a portion of the light reflected and / or scattered by the respective surface onto one another two-dimensional, spatially resolved receivers, and - generating a two-dimensional image for each painted surface, wherein each pixel next to its geometric coordinates color coordinates and brightness coordinates are still assigned, characterized in that - for each painted surface, a plurality of two-dimensional sub-areas is formed, are within the two-dimensional image of each surface to be examined, - from the pixels of each sub-area in each case a weighted brightness value, at least one weighted color value and / or a Helligk the weighted brightness values and / or the weighted color values and / or the weighted total color values are each compared with a reference value and from this a brightness deviation (ΔY) and / or at least one color deviation (ΔI, ΔQ ) and / or a total color difference (ΔE) is determined, - the determined brightness deviation (ΔY) and / or the at least one color deviation (ΔI, ΔQ) and / or the total color separation (ΔE) of at least one subregion chronologically for the plurality of painted surfaces in one Data series is acquired, - a trend for the brightness deviation (ΔY) and / or the color deviation (ΔI, ΔQ) and / or the total color separation (ΔE) is determined from the at least last two elements of the data series, - the brightness deviation (ΔY) and / or the color deviation (ΔI, ΔQ) and / or the total color separation (ΔE) for a given anza hl is predicted from surfaces taking into account the trend, and - the predicted brightness deviation (ΔY) and / or the predicted color deviation (ΔI, ΔQ) and / or the predicted total color separation (ΔE) is compared with a first limit value. Method according to claim 1, characterized in that that to determine the trend between 2 and 20 of the last elements the data series are used. Method according to one of the preceding claims, characterized in that as a trend the average increase in Brightness deviation (ΔY) and / or the color deviation (ΔI, ΔQ) and / or the total color difference (ΔE) used for a given number of surfaces becomes. Method according to one of the preceding claims, characterized in that each element of at least assigning a trend value to a data series, the trend value considering at least one existing previous one and / or subsequent element of the data series is formed. Method according to claim 4, characterized, that - the difference between each element of at least a data series and the associated trend value are still determined an allowable spread is determined such that a predetermined percentage of the determined differences within the permissible spread, and for each Element of the data series the sum of associated trend value and the determined allowable spread with a given first limit value is compared. Method according to one of the preceding claims, characterized in that the subregions are mutually disjoint areas and / or the area of the subregions is between 1 cm ² and 25 cm ² . Method according to one of the preceding claims, characterized in that the number of subregions between 15 and 1500 is. Method according to one of the preceding claims, characterized in that the plurality of two-dimensional subregions always in the same way within the two-dimensional image of the respective surface is positioned. Method according to one of the preceding claims, characterized in that the brightness deviations (ΔY), the at least one color deviation (ΔI, ΔQ) and / or the total color distances (ΔE) of the plurality of surfaces are detected for all subregions as a data series. Method according to claims 4-9, characterized that those elements of a data series of a subsection before Calculating the trend and / or the trend value are sorted out, their brightness deviation (ΔY), color deviation (ΔI, ΔQ) or total color difference (ΔE) in terms of amount exceed a second predetermined limit, wherein the second limit is greater in magnitude as the first limit. Method according to one of claims 4-10, characterized in that the numerical value of the brightness deviation (ΔY), the color deviation (ΔI, ΔQ) or the total color distance (ΔE) of those elements of a Data series of a sub-area, the amount exceed a third predetermined limit Calculation of the trend and / or the trend value by a numerical Value is interpolated, taking into account the Brightness deviation (ΔY), the color deviation (ΔI, ΔQ) or the total color difference (ΔE) of at least one adjacent one Elements of the data series of the sub-area is formed. Method according to claim 11, characterized in that that third limit is smaller in magnitude than that second limit value and / or the third limit value is greater than the first limit. Method according to one of the preceding claims, characterized in that in each case an area of at least 100 cm ^{2 of} the painted surface is imaged on the two-dimensional, spatially resolved receiver. Method according to one of the preceding claims, characterized in that at least 50% of the painted surface is assigned to the plurality of sub-areas.