A digital display instrument indication value pattern processing method

Abstract

The application discloses a digital display instrument indicating value pattern processing method, acquires a group of images containing a digital display instrument indicating value pattern by using an image acquisition device, calculates the image element feature quantity, takes the element feature quantity difference of two images as an image difference and classifies, the category number at least includes three categories, then divides the image difference according to a single digit and obtains the image difference geometric feature and image difference feature value of the single digit. The character geometric feature and character feature value of the indicating value are obtained by using the method, or the template geometric feature and template feature value of a standard template are matched, so that the corresponding digit in the indicating value pattern is obtained. The method does not need to carry out binaryzation processing on the indicating value pattern by using a threshold value, and is especially suitable for low-contrast indicating value patterns and compatible with non-low-contrast patterns.

Description

Technical Field

[0001] This invention relates to the field of image processing, and more specifically, to a method for processing patterns of digital display instrument indication values. Background Technology

[0002] In existing technologies, an increasing number of testing / inspection / calibration / verification devices and methods acquire the display graphics of the digital instrument being tested, preprocess and binarize the acquired indication value pattern graphics, and then convert them into instrument readings (or indication values) represented by ASCII encoding to generate corresponding test reports / inspection reports / calibration certificates / verification certificates. Their common feature is that the indication value pattern is processed within the same image plane.

[0003] However, some of the displays on the instruments being tested are low-contrast displays, mainly battery-powered LCD displays. Due to the low contrast of the displays, cluttered backgrounds, and the imaging of the camera on the display surface, the acquired images contain uneven gray levels, which significantly affects subsequent processing. It is extremely difficult to perform digital character recognition after binarizing such low-contrast indicator value patterns. Therefore, avoiding binarization of low-contrast images and recognizing the digital characters of the digital display instrument's indicator values ​​is a problem that needs to be solved.

[0004] The novelty of this invention compared to simple image processing lies in the following: During testing / detection / calibration / verification, patterns of (multiple) digital instrument indication values ​​before and after changes can be easily captured to create indication value map differences and derive the numerical values ​​before and after the changes. In the specific processing of the indication value map differences, it is preferable to avoid using thresholds to divide the map differences (similar to binarization). Instead, the amplitude characteristics of the map differences are used to extract "feature seeds" from (multiple) map differences. This not only utilizes the spatial correlation of the map differences (and / or images) to expand the range of the "feature class" near the "feature seed," but also utilizes the periodicity of the map differences (and / or images) in the time domain to expand the range of the "feature class." Correlation operations are performed in the three-dimensional space of the spatial and temporal domains. This method is particularly suitable for low-contrast indication value patterns and is compatible with non-low-contrast patterns.

[0005] For specialized projects in the specific field of testing / inspection / calibration / verification, by taking full advantage of the model of the instrument being tested as a preset condition and the display of the instrument being tested as prior information, the area containing the digital display instrument indication value pattern can be easily and very simply segmented. The "preset condition" and "prior information" that can be used include the following features.

[0006] (1) When the image acquisition device acquires a set of images containing the indication values ​​of the digital display instrument, the display type of the digital display instrument is known and associated with the model of the digital display instrument, the geometric dimensions of the display window are fixed with the geometric relationship of the digital display instrument, and the optical characteristics of the display are fixed (e.g., ① Character high brightness type - LED segment type / dot matrix type / pattern type, LCD display using backlight illumination, etc.); ② Character low brightness type - passive reflective LCD segment type / dot matrix type / character type / digital display instrument indication value pattern type, and display of low brightness foreground in high brightness background, etc.); ③ Character color difference type - using color LCD image display, etc.).

[0007] (2) Under stable influence conditions, it is necessary to obtain the digital display instrument indication value at intervals. The indication value has a certain regularity and is often evaluated by indicators such as average value and coefficient of variation.

[0008] (3) Change the amount of influence to obtain a new indicator value;

[0009] (4) Cumulative quantity measuring instruments such as dose equivalent meters, electricity meters, tap water meters, and natural gas meters (hereinafter collectively referred to as: cumulative quantity measuring instruments) have an indication value that is proportional to the time of the applied influence quantity under certain conditions;

[0010] (5) The geometric position of the image acquisition device and the instrument display can remain unchanged during the testing / inspection / calibration / verification process;

[0011] (6) The indicator value pattern will show alternating visible disappearances in certain areas;

[0012] (7) The testing laboratory is relatively fixed and the same type of instrument being tested will appear multiple times. By learning the background of the laboratory, the characteristics and geometric relationships of the instrument being tested from the previous images, it is possible not only to "circle" the pattern of the digital instrument indication value in the image, or to remove the image outside the display window, but also to obtain the "character geometric features" of the digital characters of the indication value.

[0013] The above features provide sufficient data resources for the digital display instrument indication value pattern processing method, laying a solid foundation for the present invention. During testing / inspection / calibration / verification, the pattern before and after the change of the digital display instrument indication value can be easily captured to create an indication value difference and obtain the value before and after the change of the indication value. This method can avoid binarizing the digital display instrument indication value pattern. Summary of the Invention

[0014] The purpose of this invention is to provide a method for processing digital display instrument indication value patterns, which is applicable to processing low-contrast indication value patterns in the fields of testing / inspection / calibration / verification. After processing by this method, the numerical value of the digital display instrument indication value represented by ASCII encoding can be obtained.

[0015] The method for processing the indication value pattern of a digital display instrument includes the following steps:

[0016] S1, Use an image acquisition device to acquire a set of images containing a pattern of digital display instrument indication values, wherein the set of images includes at least 2 images, and the sequence number of the images arranged in the acquisition order is called the image number;

[0017] Images are usually arranged in the order of acquisition, and the nth image is denoted by Pn. n is often referred to as the image sequence.

[0018] S2, calculates the primitive features of the image;

[0019] The primitive feature quantities are parameters describing the features of primitives, including the optical feature quantities of the pixels contained in the primitive; the primitive is the name of a set of pixels; usually, this type of pixel has some common characteristics, such as becoming visible or invisible simultaneously in the time domain, and / or being adjacent in the spatial domain, etc.; the optical feature quantities of the pixel are functions composed of RGB; the functions composed of RGB include optical feature quantities that can be converted from RGB to grayscale (i.e., grayscale is a function composed of RGB, the same below), or can be converted from RGB to luminance, or can be converted from RGB to color saturation, or can be converted from RGB to chroma, or expressed in various components.

[0020] S3, calculate the graph difference, which is the difference in primitive feature quantities at the same position in two images;

[0021] S4, classify the map differences, with at least three categories;

[0022] The three categories can be specifically represented by feature class 0, feature class 1, and feature class -1, or they can be specifically represented by A, B, and C, respectively. The representation method is not restricted.

[0023] In addition, there may be categories such as unknown feature class (represented by × feature class). For ease of description, the map difference is represented by × feature class before it is classified. When the map difference classification is completed, there may be no feature class 1, or / and -1 feature class.

[0024] S5, segment the map difference according to the individual numbers in the indicator value pattern;

[0025] S6, obtain the geometric features and feature values ​​of the map difference for a single number, wherein the feature values ​​include at least three possible values.

[0026] Different representation methods can be used for geometric features of image difference. For example, intuitively, for equivalent light-emitting LEDs, the image difference feature value can be represented by three values: "unchanged", "lit", and "off". More generally, when the applicability and reasonableness of the expression are extended to include non-light-emitting LCDs, the image difference feature value can be represented by three values: "unchanged", "visible", and "hidden". Abstractly, the image difference feature value can be represented by three values: integers "0", "1", and "-1". More practically, the image difference feature value can be represented by positive integer codes "1", "2", and "0" containing 0, etc. (the aforementioned integer values ​​+ 1), etc.

[0027] Of course, there may also be fourth and fifth states that we don't need to worry about for the time being, such as "unrecognized state", "unassigned state", "unrecognized state", "NULL state", etc.

[0028] Furthermore, in step S4, when classifying the map difference, the three categories are specifically represented by feature class 0, feature class 1, and feature class -1, respectively. The map difference value corresponding to feature class -1 is negative, and the map difference value corresponding to feature class 1 is positive.

[0029] Furthermore, the classification of map differences specifically includes any one, two, or all of Y41A, Y41B, and Y41C; among them,

[0030] Y41A is a feature class where the difference value of the graph is within the range of the first positive threshold pT1 and the first negative threshold nT1, where nT1≤0 and pT1≥0.

[0031] nT1 = -pT1 indicates that the map difference is marked as a feature class of 0 using positive and negative symmetric thresholds; Convention: when the threshold pT1 = nT1 = 0, the map difference = 0 satisfies "the map difference value is in the interval between the first positive threshold pT1 and the first negative threshold nT1";

[0032] Y41B is a feature class where the difference value is not lower than the second positive threshold pT2, and pT2 ≥ pT1;

[0033] The map difference value of feature class 1 is positive, also known as positive map difference, but not all map difference values ​​that are positive are marked as feature class 1;

[0034] Y41C is a feature class where the difference value does not exceed the second negative threshold nT2, and nT2≤nT1;

[0035] The graph difference value of the -1 feature class is negative, also known as negative graph difference, but not all graph difference values ​​that are negative are marked as the -1 feature class.

[0036] Before the map difference is classified, it is represented by the × feature class. When nT2≠nT1 or / and pT1≠pT2, the map difference value of the × feature class may be distributed between (nT2, nT1) and / or (pT1, pT2).

[0037] pT2 = pT1 indicates that a single threshold is used to classify positive image differences, nT2 = nT1 indicates that a single threshold is used to classify negative image differences, nT2 = -pT2 indicates that a symmetrical positive and negative threshold is used to classify image differences, pT2 = pT1 = -nT2 = -nT1 indicates that a symmetrical positive and negative single threshold is used to classify image differences, and nT2 = nT1 = pT2 = pT1 = 0 is used to classify image differences in high-brightness images (e.g., LEDs), that is: image difference values ​​> 0 are marked as feature class 1, and image difference values ​​< 0 are marked as feature class -1.

[0038] Furthermore, this includes arranging the feature micro-components in the image sequence into a feature micro-component VS image sequence curve and marking the feature micro-components between feature class 1 and feature class -1 on the curve as feature class 0. The feature micro-components are the difference in primitive feature quantities at the same position in two adjacent images in the image sequence.

[0039] Furthermore, it includes enhancing the graph difference for non-zero feature classes, wherein the enhancement process includes increasing the absolute value of the graph difference.

[0040] Furthermore, it also includes any one, two, or all of the processing methods from Y43A, Y43B, and Y43C; among which,

[0041] Y43A performs dilation processing based on feature class;

[0042] Y43B performs shrinkage processing based on feature class;

[0043] Y43C removes isolated noise from feature classes in the image difference.

[0044] Furthermore, step S4 also includes classifying the map differences using evaluation metrics of adjacent map differences. These evaluation metrics include any one, any two, any three, any four, any five, any six, or all of Y45A, Y45B, Y45C, Y45D, Y45E, Y45F, and Y45G.

[0045] Y45A uses the average difference between adjacent maps as the evaluation metric;

[0046] Y45B uses the median difference between adjacent maps as the evaluation metric;

[0047] Y45C uses the moving smoothing result of adjacent map differences as the evaluation metric;

[0048] Y45D uses the difference between adjacent maps as a function with weighting coefficients as the evaluation metric;

[0049] Y45E uses the distance between adjacent map differences as a function with weighting coefficients as the evaluation metric;

[0050] Y45F uses the number of identical feature classes in adjacent map differences as the evaluation metric;

[0051] Y45G uses the sign of the difference between adjacent graph values ​​as an evaluation metric.

[0052] Furthermore, the specific meaning of "adjacent map difference" includes any one, any two, or all of Y45X, Y45Y, and Y45Z, wherein...

[0053] Y45X represents adjacent coordinates in the spatial domain;

[0054] Y45Y represents adjacent graph sequences in the time domain;

[0055] Y45Z is the difference in primitive features at the same location in two sequentially adjacent images in the time domain.

[0056] Furthermore, the character geometric features and character feature values ​​of the indicator value are derived using the image difference geometric features and image difference feature values, and then the numbers corresponding to the indicator value pattern are derived based on the character geometric features and character feature values.

[0057] Furthermore, the character series of the digital display instrument indication value pattern is classified into any one of the Y61, Y62, and Y63 character series. Template characters are created for this character series. The difference between two template character patterns is used as a standard template and described by template geometric features and template feature quantities. The geometric features and feature values ​​of the difference obtained in S6 are matched with the geometric features and feature quantities of the template to obtain the corresponding numbers in the indication value pattern.

[0058] Y61 represents a character series composed of seven segments of digits 0 to 9. The pattern of characters 0 to 9 can be equivalently divided into seven segments. The same segment has the same optical features at the same time. The display of characters 0 to 9 can be achieved by changing the combination of optical features of the segments.

[0059] Y62 represents a character series consisting of Arabic numerals 0 to 9, and the patterns of characters 0 to 9 are Arabic numerals;

[0060] Y63 represents a character series composed of character patterns in the same digital display instrument indication value.

[0061] The character series includes those appearing in the same digital display instrument indication value pattern.

[0062] Furthermore, all template character pattern differences in the character series are enumerated as standard templates and described using template geometric features and template feature quantities.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] This invention acquires a set of at least two images containing a pattern of digital display instrument indication values ​​using an image acquisition device, calculates the image difference using the primitive features of the images, and classifies the image difference into three "feature classes". When classifying image differences, it is preferable to avoid using only a threshold to divide the image differences (similar to the binarization of grayscale images). Instead, the amplitude characteristics of the image differences are used to extract "feature seeds" from (multiple) image differences. This not only utilizes the spatial correlation of the image differences (and / or images) to expand the range of "feature classes" near the "feature seeds," but also utilizes the periodicity of the image differences (and / or images) in the time domain to expand the range of "feature classes." Correlation operations are performed in the three-dimensional space of the spatial and temporal domains, thus avoiding the difficulty of binarizing low-contrast patterns using a threshold. After processing by this method, the indicator value can be further identified and represented by ASCII encoding. This method is essential for "low-contrast" patterns in the fields of testing / inspection / calibration / verification, and the benefits are obvious. At the same time, it is compatible with non-"low-contrast" patterns, and the identified indicator value can be verified for consistency with the indicator value identified by other methods. Attached Figure Description

[0065] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:

[0066] Figure 1 This is a flowchart illustrating the method for processing the indication value pattern of a low-contrast digital display instrument.

[0067] Figure 2 The diagram shows the black and white contrast of the indicator value patterns "000" and "1" of a "low contrast" digital display instrument after binarization with different thresholds. In the diagram, (201), (202), and (203) are all "000" indicator value patterns, and (204), (205), and (206) are all "1" (leading "0" hidden) indicator value patterns. In the diagram, (201) and (204) are black and white contrast diagrams after binarization of the indicator value patterns using the same higher threshold, (202) and (205) are black and white contrast diagrams after binarization of the indicator value patterns using the same medium threshold, and (203) and (206) are black and white contrast diagrams after binarization of the indicator value patterns using the same lower threshold.

[0068] Figure 3The diagram shows the black-and-white contrast after binarizing the image difference with different thresholds. In the diagram, (301) shows that when the threshold is high, the image is incomplete but there is only a small amount of "residual noise". (302) and (303) are black-and-white contrast diagrams showing that the "residual noise" in the image gradually increases when the threshold is gradually reduced.

[0069] Figure 4 This is a comparative schematic diagram of the changes in the frequency distribution of the graph differences of X135, X136, and X137 during the process of classifying graph differences. In the figure, (401) is a schematic diagram of the frequency distribution of the graph differences of X135, X136, and X137, (402) is a schematic diagram of the frequency distribution of the graph differences of X135, X136, and X137 after setting the absolute value of the graph differences below the threshold to zero, and (403) is a schematic diagram of the frequency distribution of the graph differences of X135, X136, and X137 after further classifying the × feature class into the 0 feature class and setting it to zero using the evaluation metric.

[0070] Figure 5 The diagram shows the curve of the absolute value |D| of the feature micro-components calculated using a rectangular primitive with a width of 10 on the X-axis. The horizontal axis represents the sequence number (n), and the vertical axis represents the absolute value |D| of the feature micro-components. Legends X = 1, X = 11, ..., X = 51 indicate the position of the rectangular primitives. In the diagram, (501) is a schematic diagram of the curve of the absolute value |D| of the feature micro-components VS the curve of the hundreds digit character in the indicator value pattern, (502) is a schematic diagram of the curve of the absolute value |D| of the feature micro-components VS the curve of the tens digit character in the indicator value pattern, and (503) is a schematic diagram of the curve of the absolute value |D| of the feature micro-components VS the curve of the individual digit character in the indicator value pattern.

[0071] Figure 6 The diagram shows how to establish a standard template using two template characters "0" and "3" from the seven-segment digital series. In the diagram, (601) is a schematic diagram of each segment of the template characters "0" and "3" from the seven-segment digital series, where ■ represents a visible segment; (602) is a schematic diagram of the standard template, where ■ represents "a hidden segment becomes a visible segment", and the number 1 in the box indicates that its template feature quantity is "1", ◇ represents "a visible segment becomes a hidden segment", and □ represents "a visible segment with no change", and the number 0 in the box indicates that its template feature quantity is "0".

[0072] Figure 7 This is a schematic diagram of the recognition process of two numbers to be identified in the indication value of a digital display instrument. In the figure, (701) is a schematic diagram of the image acquisition device acquiring the indication value pattern and changing it into another indication value pattern. "?" indicates "unknown"; (702) is a schematic diagram of the geometric features of the indication value image difference and the representation of the image difference feature value; (703) is a schematic diagram of the matching and matching to obtain the recognition result.

[0073] Figure 8 This is a schematic diagram illustrating the effect of dilation and contraction operations on the image difference.

[0074] Figure 9 The diagram shows how to calculate the height and width of a character from the difference between the graphs. In the diagram, (9A) is a diagram showing how to calculate the height and width of a character from the difference between any one character, and (9B) and (9C) are diagrams showing how to calculate the height and width of a character from the difference between two characters.

[0075] Figure 10 is a schematic diagram of the changes in the effect of classifying the map difference. In the figure, “■” represents feature class 1, “□” represents feature class 0, and blank “” represents feature class ×. Figure 10(a) is a schematic diagram of the classification effect when “map difference = 0” is only classified into feature class 0. Figure 10(b) is a schematic diagram of the classification effect after classifying feature class 0 when the absolute value of map difference is lower than the threshold. Figure 10(c) is a schematic diagram of the classification effect when the × feature class is further classified into feature class 0 using the evaluation metric.

[0076] Figure 11 The diagram shows the blending of character grayscale and background grayscale in the "low contrast" indicator value pattern; Figure (11A) is a schematic diagram of the distribution curve of the RGB mean of the background after removing the area containing the "088" stroke of the "000" indicator value pattern; Figure (11B) is a schematic diagram of the distribution curve of the grayscale of the unit digit "0" pattern of the indicator value pattern "000", where the black solid line symbol "A" represents the distribution curve of the grayscale of the entire unit digit "0" pattern, and the black dotted line symbol "T" represents the distribution curve of the grayscale of the entire unit digit "0" pattern. “” represents the gray distribution curve at the top 1 / 4 of the unit digit “0” pattern, where the black hollow line “B” represents the gray distribution curve at the bottom 1 / 4 of the unit digit “0” pattern; (11C) is a schematic diagram of the gray distribution curve at the top 1 / 4 of the unit digit “0” pattern and the gray distribution curve divided into “T1” and “T2” on the left and right sides; (11D) is a schematic diagram of the gray distribution curve at the bottom 1 / 4 of the unit digit “0” pattern and the gray distribution curve divided into “B1” and “B2” on the left and right sides.

[0077] Figure 12 According to Figure 11 The background of the concave dot pattern “000” with “088” segment removed from the digital display instrument indication value pattern “000” shown in Figure (11A) is a binary black and white schematic diagram. In the figure, ■ represents the primitive below the threshold, □ represents the primitive above the threshold, and “” represents the space where “088” is removed.

[0078] Figure 13This diagram illustrates the calculation of the average value of the difference in primitive features at the same location in two sequentially adjacent images in the time domain, within a three-dimensional space comprised of adjacent coordinates in the spatial domain and sequentially adjacent images in the time domain. In the diagram, n-1, n, and n+1 represent sequentially adjacent images in the time domain; Dn-1 represents a 3×3 feature differential matrix with adjacent coordinates in the spatial domain of (n-1) time domain; Dn represents a 5×5 feature differential matrix with adjacent coordinates in the spatial domain of (n) time domain; Dn+1 represents a 3×3 feature differential matrix with adjacent coordinates in the spatial domain of (n+1) time domain; and => indicates the calculation of the average value in three-dimensional space, with a result of 0.9.

[0079] Figure 14 This diagram illustrates the classification of image differences using a multi-threshold method. Figure (14A) shows the distribution of local primitive features and image differences along the Y-axis of the "low contrast" indicator patterns "000" and "1" (leading "0" has been hidden). Legendary figures BX12, BX13, and BX14 represent the primitive features at the locations of the "stroke e" and "stroke f" of the hundreds digit of the indicator pattern "000". Legendary figures AX12, AX13, and AX14 represent the primitive features at the locations of the "stroke e" and "stroke f" of the hundreds digit of the indicator pattern "1" (leading "0" has been hidden). Legendary figures AX12-BX12, AX13... -BX13 and AX14-BX14 are the corresponding map differences; (14B) is a schematic diagram of obtaining "feature seeds" using a threshold. In the figure, "0 feature seeds" are marked as T0 and "1 feature seeds" are marked as T1; (14C) is a schematic diagram of classifying the map differences using the evaluation of adjacent map differences. In the figure, the regions (Y1~Y42) and (Y51~Y91) are adjacent to "1 feature seeds" and are marked as feature class 1. In the figure, T0A represents being adjacent to "0 feature seeds" and marked as feature class 0. Y-axis 71 is marked as feature class 0 in (14B) but is subsequently corrected to feature class 1 by adjacent map differences. In (14C), it is marked as T1A.

[0080] Figure 15This is a schematic diagram of classifying image differences using a multi-threshold method. Figure (15A) shows the distribution of local primitive features and image differences along the Y-axis of the "low contrast" indicator patterns "000" and "1" (leading "0" has been hidden). The horizontal axis is the Y-axis. Legendary symbols BX148, BX149, and BX150 represent the primitive features at the locations of "segment b" and "segment c" of the units digit of the indicator pattern "000". A148, A149, and A150 are the primitive feature quantities at the locations of "segment b" and "segment c" of the unit digit character of the indicator value pattern "1". The legend AX148-BX148, AX149-BX149, and AX150-BX150 are the corresponding map differences. Figure (15B) is a schematic diagram of obtaining "feature seeds" using a threshold. Figure (15C) is a schematic diagram of classifying the map differences using the evaluation values ​​of adjacent map differences.

[0081] Figure 16 for Figure 14 and Figure 15 A schematic diagram showing the positions of X12, X13, X14, X148, X149, X150, where X and Y indicate the directions of the coordinate axes; (16A) is... Figure 14 A schematic diagram showing the positions of X12, X13, and X14; (16B) is... Figure 15 A schematic diagram showing the positions of X148, X149, and X150.

[0082] Figure 17 The diagram shows the application of the "Geometric Feature Extraction Orifice Plate (17A)" to a seven-segment digital character. In the diagram, (17A) is a schematic diagram of the "Geometric Feature Extraction Orifice Plate (17A)" and the "Extraction Hole (b)", "Extraction Hole (c)", "Extraction Hole (e)" and "Extraction Hole (f)", (17B) is a schematic diagram of a seven-segment digital character, and (17C) is a schematic diagram of the effect of the orifice plate on the seven-segment digital character.

[0083] Figure 18 This is a schematic diagram of the "Geometric Feature Extraction Orifice Plate (18A)" applied to a seven-segment digital character. In the figure, (18A) is a schematic diagram of the "Geometric Feature Extraction Orifice Plate (18A)" and the "Extraction Orifice (a)", "Extraction Orifice (d)" and "Extraction Orifice (g)". In the figure, (18B) is a schematic diagram of a seven-segment digital character. In the figure, (18C) is a schematic diagram of the orifice plate applied to a seven-segment digital character.

[0084] Figure 19 This is a schematic diagram of the remaining images after the Arabic numerals 0 to 9 have been removed by the "removal panel 0" to "removal panel 9" respectively.

[0085] Figure 20This is a schematic diagram illustrating the extraction of the digits "0" and "1" from a "character series composed of character patterns in the same digital instrument indication value" using a "geometric feature extraction orifice plate (20A)". The "geometric feature extraction orifice plate (20A)" has two through holes (K1) and (K2). (20B) represents the digits... A schematic diagram; (20C) shows the application of this orifice plate to digital... (20D) is a schematic diagram of the effect of the orifice plate applied to the number "1"; (20E) is a schematic diagram of the effect of the orifice plate applied to the number "1".

[0086] Figure 21 This is a schematic diagram showing the effect of classifying map differences using weighted coefficients as evaluation metrics. In the figure, (21A) shows the effect before classification, and (21B) shows the effect after classification. Figure 24 The diagram shows the effect of the weight coefficients in (24C) on the classification of map differences; ■ represents feature class 1, □ represents feature class ×, and “” (space) represents feature class 0.

[0087] Figure 22 The diagram shows the difference classification of six digital instrument indication values. In the diagram, ■ represents feature class 1, □ represents feature class -1, "" (space) represents feature class 0, and "·" represents feature class ×. (22A) is a diagram showing the difference classification when the number "69" jumps to "75". (22B) is a diagram showing the difference classification when the number "75" jumps to "80". (22C) is a diagram showing the difference classification when the number "80" jumps to "86". (22D) is a diagram showing the difference classification when the number "86" jumps to "91". (22E) is a diagram showing the difference classification when the number "91" jumps to "96".

[0088] Figure 23 To derive the character geometric features and character feature values ​​of the digits in the indicator value pattern using image difference geometric features and image difference feature values, and to obtain a schematic diagram of the digits corresponding to the indicator value pattern; where the "indicator value pattern" is the seven-segment digit series indicator value pattern "X" (P) to be identified. n ), "Y" (P n+1 ), "Z" (P) n+2The image difference feature value is represented by three values: "0", "1" and "-1". In the figure, ■ represents the segment "from hidden to visible", and the image difference feature value is "1"; ◇ represents the segment "from visible to hidden", and the image difference feature value is "-1" ("-1" is not marked in the figure); □ represents the segment "no change", and the image difference feature value is "0"; (23A) is a schematic diagram of the indicator value pattern "X" changing to the indicator value pattern "Y", and "?" in the figure indicates to be identified; (23D) is a schematic diagram of the indicator value pattern "Y" changing to the indicator value pattern "Z", and "?" in the figure indicates to be identified; (23B) is a schematic diagram of the image difference geometric features and image difference feature values ​​when the indicator value pattern "X" changes to the indicator value pattern "Y"; (23C) is based on Figure 23 (23B) A schematic diagram showing the transformation of the indicator value pattern "X1" into the segment values ​​of the indicator value pattern "Y1" is derived, in which... — Indicates that the segment value is “unknown”; (23E) is a schematic diagram of the geometric features and feature values ​​of the difference when the indicator value pattern “Y” changes to the indicator value pattern “Z”; (23F) is a schematic diagram of the segment values ​​of each stroke when the indicator value pattern “Y” changes to the indicator value pattern “Z” based on the figure (23E).

[0089] Figure 24 The diagram shows how to classify map differences using weighted coefficients as evaluation metrics. In the diagram, (24A) shows the density distribution curves of map differences along the X and Y axes obtained by projecting map differences within the same map sequence onto the X and Y axes respectively before classification. (24B) shows the density distribution curves of map differences along the X and Y axes obtained by projecting map differences within the same map sequence onto the X and Y axes respectively after classification. (24C) shows the diagram of constructing weighted coefficients using the density distribution curves of the X and Y axes without thresholds. (24D) shows the density distribution curves of the evaluation metrics along the X and Y axes obtained by projecting the evaluation metrics onto the X and Y axes respectively.

[0090] Figure 25 This is a schematic diagram comparing the feature quantity vs. the sequence curve and the feature differential component vs. the sequence curve after defining the primitive using stroke segments. In Figures (1a) to (1K2), the Y-axis represents the primitive feature quantity, and the X-axis represents the sequence (n). Figures (1a) to (1K2) correspond to the primitive feature quantities of stroke segments a to K2, respectively. In Figures (2a) to (2K2), the Y-axis represents the feature differential component D, and the X-axis represents the sequence (n). Figures (2a) to (2K2) correspond to the feature differential components of stroke segments a to K2, respectively.

[0091] Figure 26 This is a schematic diagram showing the characteristics of the R+G+B value of the units digit of the equivalent meter reading after the fourth irradiation.

[0092] Figure 27 This is a schematic diagram showing the characteristics of the R+G+B value of the units digit of the equivalent meter reading after the 5th irradiation. Detailed Implementation

[0093] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0094]

Example 1

[0095] The embodiments of the present invention will be further described in detail below with reference to the examples, and the flowchart is shown below. Figure 1 As shown; the method for processing the indication value pattern of a digital display instrument is characterized by including the following steps:

[0096] S1, Use an image acquisition device to acquire a set of images containing a pattern of digital display instrument indication values, wherein the set of images includes at least 2 images, and the sequence number of the images arranged in the acquisition order is called the image number;

[0097] Image acquisition equipment – ​​a device, apparatus, component, or instrument, such as a camera, CCD image array, or CMOS image array, used to convert optical images into digital data.

[0098] Images – a collection of data generated by image acquisition devices, including images, graphics, photographs, and images, graphics, photographs converted from video data streams, etc., are all simply referred to as "images". The position of an image in the group is called the image order. Images are usually arranged in the order of acquisition, with the nth image denoted as Pn and the mth image denoted as Pm. Often, n and m are also referred to as the image order. Preferably, keyframes in the camera video stream are used as the acquired images to avoid the possibility of character confusion in the image synthesized between two keyframes. For example, the numbers 2 and 1 in a keyframe can be combined to form a detectable garbled character "6", while the numbers 6 and 1 in a keyframe can be combined to form an undetectable erroneous character "8", etc. More preferably, images are acquired from keyframes at intervals; keyframes with the same content are discarded to reduce the amount of data. For example, the display value pattern of a digital instrument usually updates the display content approximately once per second, while the number of keyframes is higher than this value.

[0099] The digital display instrument indication value pattern—specifically refers to the "character pattern" of the indicated value, abbreviated as "indication value pattern," which is the focus of image processing in this embodiment. The "indication value pattern" can be a pattern from a light-emitting LED segment / dot-matrix / geometric display, a passive reflective LCD segment / dot-matrix / geometric display, or a backlit LCD segment / dot-matrix / geometric display, etc.

[0100] For ease of consistent description, it is agreed that: on the white paper, "black" corresponds to the visible state of the "indicator value pattern", and the "lit" state (i.e., visible state) of the "luminous" indicator value pattern is transformed to be equivalent to "black" on the white paper (RGB = 0,0,0).

[0101] The example image used in this embodiment is a "low-contrast" digital instrument indicator pattern. The characters in the pattern are "000" and "1" (leading "0" is hidden). To meet the requirement of using black and white images in the accompanying drawings of the embodiment document, six black and white images are used as comparative illustrations to describe "low contrast". The digital instrument indicator pattern uses the RGB average value as "grayscale", and then binarizes it using different thresholds. The binarized black and white illustration is shown below. Figure 2 As shown; Figure 2 (201), (202), and (203) are "000" indicator symbols. Figure 2 Patterns (204), (205), and (206) are "1" (leading "0" hidden) indicator values. Figure 2 (201) and (204) are black and white schematic diagrams after binarizing the indicator value pattern using the same high threshold. Figure 2 (202) and (205) are black and white schematic diagrams after binarizing the indicator value pattern using the same moderate threshold. Figure 2 (203) and (206) are black and white schematic diagrams after binarizing the indicator value pattern using the same lower threshold.

[0102] The above comparative diagrams illustrate that the grayscale of the indicator value pattern blends with the grayscale of the background, making it difficult to binarize the "low contrast" indicator value pattern. For a detailed description of the characteristics of the "low contrast" digital display instrument indicator value pattern used in this embodiment, please refer to the "Postscript" at the end of this embodiment.

[0103] When the acquired image does not contain only the pattern of the digital instrument indicator value, it is preferable to "segment, mark, or crop" the pattern of the digital instrument indicator value. The marking includes using coordinates to represent the position of the digital instrument indicator value pattern, or using the subscript of an array to represent the position of the digital instrument indicator value pattern in the array. When only the pattern of the digital instrument indicator value is contained, the impact of background changes on subsequent processing can be ignored, thereby improving the signal-to-noise ratio of the data in subsequent processing. For example, after removing the image outside the display window of the digital instrument indicator value in the sequence image as the background, the impact of background changes caused by people walking on the subsequent processing can be reduced.

[0104] In the "Test / Inspection / Calibration / Verification" process, the model of the instrument being tested is a necessary information to determine. Prior information associated with the model may include the type of display, the optical characteristics of the display, the geometry / size of the display window and its positional features on the instrument being tested, the text features, pattern line features, and pattern color features on the instrument being tested, etc., and this information is stable over a long period of time.

[0105] (1) On the one hand, prior information can be fully utilized to “circle” (“segment out, mark out, or crop out”) the pattern of the digital instrument indication value in the image using geometric relationships, or to remove the image outside the display window (specifically, “segment out, mark out, or crop out”, etc. can be used).

[0106] (2) On the other hand, the digital instrument readings usually change during “testing / inspection / calibration / verification”. This feature can be used to identify and locate the “changing digital instrument reading pattern” (i.e. the focal point of image processing). Then, the display window usually has visible color or grayscale boundaries on the panel, so as to “circle” the pattern of the digital instrument reading in the image, or remove the image outside the display window.

[0107] (3) The testing laboratory is relatively fixed and the same type of instrument being tested will appear multiple times. Learn the background of the laboratory, the characteristics and geometric relationships of the instrument being tested from previous images, so as to "circle" the pattern of the digital instrument indication value in the image, or remove the image outside the display window;

[0108] (4) Use the methods (1), (2), and (3) above in combination to "circle" the pattern of the digital instrument indication value in the image, or remove the image outside the display window.

[0109] S2, calculates the primitive features of the image;

[0110] The primitive feature quantities are parameters describing primitive features, including the optical features of the pixels contained in the primitive. The primitive feature quantities are expressed by a generalized formula as follows:

[0111] T n (i A ,j B )=f(X n (i,j)) (1)

[0112] In the formula: T n (i A ,j B —Image P with sequence number n n Located in (i A ,j B The primitive features of a point;

[0113] f(X n (i,j))——A function of the optical features of the pixels contained in the primitive.

[0114] As f(X) n The (i,j) function is a special case where the primitive features are calculated by accumulating, averaging, weighted accumulating, or weighted averaging the optical features of the pixels contained in the primitive. Assuming the primitive contains w×h pixels, let (i,j) represent the pixel position, and let X represent the optical feature of the pixel at point (i,j). n (i,j) indicates that the primitive feature quantity is the total accumulation of the optical features of the contained pixels as shown in formula (2); the primitive feature quantity is the average value of the optical features of the contained pixels as shown in formula (3).

[0115]

[0116]

[0117] In addition, primitive features can also be calculated using weighting coefficients, as shown in formula (4);

[0118]

[0119] In the formula: W n (i,j) — Generalized weight function, for example: W n The (i,j) weight function can be a distance weight function (e.g., Gaussian filtering), a gray-scale weight function, or a combination of distance weight function and gray-scale weight function (e.g., bilateral filtering), etc.

[0120] Beyond these complex functions, primitive features include the optical features of the pixels contained within the primitive.

[0121] The primitive is the name of a set of pixels; usually, these pixels have some common characteristics, such as becoming visible or invisible at the same time in the temporal domain, or / and being adjacent in the spatial domain, etc.

[0122] A primitive is the basic unit for image processing in this embodiment; a primitive consists of at least one pixel, and an image consists of at least one primitive. When a primitive consists of one pixel, the primitive is a pixel.

[0123] The optical features of the pixel are functions composed of RGB;

[0124] The functions composed of RGB include optical characteristic quantities that can be converted from RGB to grayscale (i.e., grayscale is a function composed of RGB, the same below), or can be converted from RGB to luminance, or can be converted from RGB to color saturation, or can be converted from RGB to chroma, or expressed in various components.

[0125] For example: Computers typically acquire images using the "RGB color mode," and the image P... nThe optical feature X of the pixel located at point (i,j) n (i,j) is a function composed of RGB values, which can be expressed by a generalized formula as X. n (i,j)=f(R,G,B,n,i,j), which can be converted to grayscale, luminance, color saturation, or chroma, or optical characteristic quantities represented by various components through RGB;

[0126] As a special case of the function f(R,G,B,n,i,j), the optical characteristic quantity can be expressed by formula (5):

[0127] X n (i,j)=k R ·R n (i,j)+k G ·G n (i,j)+k B ·B n (i,j) (5)

[0128] In the formula: R n (i,j),G n (i,j),B n (i,j) represent the image P. n The red, green, and blue components located at point (i,j), k R ,k G ,k B These are their respective weighting factors, but k R ,k G ,k B Not all of them are simultaneously "zero"; when k R =0.299,k G =0.587,k B =0.114 means that grayscale is used as the optical characteristic of a pixel.

[0129] When using pure blue (RGB:0,0,255) as the background and displaying pure green (RGB:0,255,0) or pure red (RGB:255,0,0) numbers, there is a very strong visual "parallax". However, the image difference calculated using (R+G+B) as the optical feature is zero. Therefore, it is advisable to use the chromaticity H in the HSI model as the optical feature of the pixel. Similarly, for pixels with visible color difference but the same gray level, it is also advisable to use the chromaticity H in the HSI model as the optical feature of the pixel.

[0130] A primitive consists of at least one pixel. For example, primitives can usually be represented by a "3×3" matrix, a "3×5" matrix, or a "5×9" matrix, depending on actual needs. They can also be represented by circular or rhomboid matrices, or any other arbitrary shape. For example, primitives can be defined by a cutout template, or primitives can be learned from the camera data stream and simultaneously become visible or "invisible".

[0131] A 1×1 primitive consists of one pixel; a primitive is a pixel. Two adjacent primitives can share the same number of pixels. For example, two 3×3 matrix primitives occupy only a 5×3 matrix in the image. A pixel in an image can be not contained within a primitive.

[0132] (1) Using multiple pixels to construct primitives reduces computational load and can also reduce the "residual noise" after image difference calculation, and may even increase the value of image difference—this is particularly useful for processing low-contrast images; for example: Figure 25 After defining primitives (i.e., segments) simultaneously in the time domain as "visible / invisible", Figures (2a) to (2K2) are curves of feature differential components versus sequence. The feature differential components are as high as 1500 to 3000, while the "residual noise" is much lower than 1000.

[0133] (2) For example: use rectangular primitives to quickly search for the position of numerical characters in the indicator value pattern, such as Figure 5 As shown, the absolute values ​​of the characteristic differential components of the hundreds, tens, and units digits in the indicator value pattern. |D| VS plot sequence curve.

[0134] The above (1) and (2) illustrate that different types of primitives can be defined simultaneously in an image and processed differently, and a pixel can belong to different types of primitives.

[0135] When constructing primitive features, weighting coefficients can be used as needed to selectively highlight key features.

[0136] S3, calculate the graph difference, which is the difference in primitive feature quantities at the same position in two images;

[0137] Suppose in a set of images, there are two images P in a generalized sense. n and image P m The coordinates (i) at the same position (where m≠n) A ,j B The primitive features of the point are respectively represented by T n (i A ,j B ) and T m (i A ,j B ) indicates that it is located at coordinate (i A ,j B The difference between points is represented by D. n,m(i A ,j B If ) indicates that:

[0138] D n,m (i A ,j B ) = T m (i A ,j B )-T n (i A ,j B (m≠n) (6)

[0139] D n,m (i A ,j B ) represents image P n With image P m The difference is therefore also called the map difference;

[0140] Preferably: when m = n + 1, i.e. the graphs are adjacent, the graph difference is represented by D. n,n+1 (i A ,j B ) represents, or is abbreviated as D n (i A ,j B ),but:

[0141] D n,n+1 (i A ,j B ) = T n+1 (i A ,j B )-T n (i A ,j B )or

[0142] D n (i A ,j B ) = T n+1 (i A ,j B )-T n (i A ,j B )

[0143] Under these conditions, D n,n+1 (i A ,j B (or D) n (i A ,j B Also known as the feature differential component, the feature differential component is the difference in the feature quantity of primitives at the same position in two adjacent images.

[0144] A series of characteristic differential components are calculated from adjacent images according to the graph sequence (n = 1, 2, 3, ...). The characteristic differential components are used as the Y-axis and the graph sequence (n) is used as the X-axis. This is called the "characteristic differential component VS graph sequence curve".

[0145] The above explanation states that the graph difference is a generalized definition, while the characteristic differential component is a narrow definition of the graph difference when m = n + 1.

[0146] Example 3-1: A black-and-white comparison diagram showing the absolute value of the difference between the "000" indicator pattern and the "1" indicator pattern after binarization with different thresholds. Figure 3 As shown, when the threshold is high, the graph is incomplete but there is only a small amount of "residual noise", as shown in Figure (301); Figure (302) and (303) are comparative diagrams showing that the "residual noise" in the graph gradually increases as the threshold is gradually reduced.

[0147] from Figure 3 and Figure 2 It can be preliminarily concluded that although the original image of the digital display instrument indication value pattern is affected by low contrast and environmental imaging on the display window surface, which affects the grayscale uniformity of the original image, the impact on the image difference is reduced.

[0148] Example 3-2: Thousands of cumulative dose digital display instrument indication value patterns were acquired. In n=26 and n=27, the "image acquisition device" changed the indication value pattern "000" to "1" (leading "0" was hidden). In n=875 and n=876, the indication value pattern changed from "99" (leading "0" was hidden) to "100". The absolute value |D|VS curve of the characteristic differential component was calculated using a rectangular primitive with a width of 10 on the X-axis. Figure 5 As shown, the horizontal axis is the graphic sequence (n), and the vertical axis is the absolute value of the feature component |D|. The legends X=1, X=11, ..., X=51 indicate the position of the rectangular graphic element. In the figure, (501) is a schematic diagram of the absolute value of the feature component |D| VS the graphic sequence curve containing the hundreds digit character in the indicator value pattern, (502) is a schematic diagram of the absolute value of the feature component |D| VS the graphic sequence curve containing the tens digit character in the indicator value pattern, and (503) is a schematic diagram of the absolute value of the feature component |D| VS the graphic sequence curve containing the units digit character in the indicator value pattern.

[0149] from Figure 5 As can be seen, when using characteristic micro-components, the pattern of the cumulative dose digital display instrument indication value changes. This pattern can be used to verify the consistency of the identification results and to correct errors. It can even be used to assist in identification.

[0150] S4, classify the map differences, with at least three categories;

[0151] The three categories can be specifically represented by feature class 0, feature class 1, and feature class -1, or they can be specifically represented by A, B, and C, respectively. The representation method is not restricted.

[0152] In addition, there may be categories such as unknown feature class (represented by × feature class). For ease of description, the map difference is represented by × feature class before it is classified. When the map difference classification is completed, there may be no feature class 1, or / and -1 feature class.

[0153] S5, segment the map difference according to the individual numbers in the indicator value pattern;

[0154] S6, obtain the geometric features and feature values ​​of the map difference for a single number, wherein the feature values ​​include at least three possible values.

[0155] (6.1) Geometric characteristics of map differences

[0156] For example, the geometric features of the seven-segment digital series are consistent with the geometric features of the seven-segment digital characters. The distribution position of the character strokes "top, top right, bottom right, bottom, bottom left, top left, center" is represented by (segment a, segment b, segment c, segment d, segment e, segment f, segment g), which is conventionally abbreviated as (a, b, c, d, e, f, g) or (abcdefg).

[0157] When "segment a of a seven-segment digital code" is mentioned, people generally think of the "upper horizontal" segment, and when "segment g of a seven-segment digital code" is mentioned, people generally think of the "middle horizontal" segment. These are specific examples of the geometric features of the image difference and the geometric features of the character in the embodiments of this invention.

[0158] To extract the feature values ​​of the image differences in the seven-segment digital series "top right, bottom right, bottom left, top left" (corresponding to the stroke segments b, c, e, and f of the seven-segment digital characters), a method was designed as follows: Figure 17 The “Geometric Feature Extraction Orifice Plate (17A)” shown in the figure has “Extraction Hole (b)”, “Extraction Hole (c)”, “Extraction Hole (e)” and “Extraction Hole (f)” used to extract the image difference feature values ​​of “top right, bottom right, bottom left and top left”, respectively. The schematic diagram of the seven-segment digital character is shown in Figure (17B), and the schematic diagram of the effect of using this orifice plate on the seven-segment digital character is shown in Figure (17C).

[0159] In addition, in order to extract the "upper, middle, and lower" (corresponding to the stroke segments a, g, and d of the seven-segment digital series image difference feature values), a method was designed as follows: Figure 18The “Geometric Feature Extraction Orifice Plate (18A)” shown in the figure extracts the image difference feature values ​​of the “upper, middle and lower” stroke segments using “extraction hole (a)”, “extraction hole (d)” and “extraction hole (g)”. The schematic diagram of the seven-segment digital character is shown in Figure (18B). The schematic diagram of the effect of using this orifice plate on the seven-segment digital character is shown in Figure (18C).

[0160] Each "extraction hole" corresponds to one segment of the seven-segment digital character. The position and size of the "extraction hole" are described using relative quantities, which allows the image difference containing the segment to pass through while leaving a tolerance for geometric errors, and minimizes interference from other segments as much as possible. Figure 17 and Figure 18 The "Geometric Feature Extraction Well Plate" has a small tilt angle, which improves its tolerance while avoiding the increased computational burden caused by tilt correction for each image. The "Adaptive Extraction Well Plate," formed by merging the 1-class or -1-class features from multiple image differences after performing multiple dilation / shrink operations according to their categories, with the "Geometric Feature Extraction Well Plate," exhibits adaptive characteristics and a higher signal-to-noise ratio. Figure 12 The space where "088" is removed is an adaptive "geometric feature extraction plate". Obviously, by merging the space where "8" is removed with the space where "0" is removed on the left, we can get the space where "888" is removed.

[0161] The "extraction hole" can also be modified into a line that passes through the pen segment and is called a "detection line".

[0162] For example, the geometric features of the difference in the Arabic numeral series are described by 10 "removal plates". The specific process is to use bold Arabic numeral characters 0 to 9 as "removal plates", which are represented by "removal plate 0" to "removal plate 9" respectively.

[0163] For example: To extract the digits "0" and "1" from the "character series composed of character patterns in the same digital display instrument indication value", where the digit "0" has a slash in the center. The diagram for the number "0" is shown in Figure (20B), and the diagram for the number "1" is shown in Figure (20D). The following designs are also provided. Figure 20 The "Geometric Feature Extraction Orifice Plate (20A)" shown has two through holes (K1) and (K2). This orifice plate is used to extract digital features. The effect diagram is shown in Figure (20C), and the effect diagram of the orifice plate applied to the number "1" is shown in Figure (20E).

[0164] The "Geometric Feature Extraction Plate" / "Removal Plate" is merely a schematic diagram for illustration. In practice, the position of the "extraction hole" / "removal plate" is often represented by the coordinates of an array. The processing in the software involves data operations on the array.

[0165] The geometric features of the map difference can also be obtained from the following methods:

[0166] In the "Test / Inspection / Calibration / Verification" process, the model number of the instrument being tested is a necessary information to determine. Prior information associated with the model number may include the type of display, the optical characteristics of the display, the geometry / size of the display window and its positional features on the instrument being tested, the text features, pattern line features, and pattern color features on the instrument being tested, etc.

[0167] ① Using prior information, geometric relationships are used to obtain the geometric features of all characters in the indicator value pattern of the display instrument;

[0168] ② After the camera is pointed at the instrument being tested, it learns the "character geometric features" of the indicated value from the changing indicator pattern;

[0169] ③ Use all the methods in (6.1) to obtain the geometric features of the map difference.

[0170] (6.2) Map difference eigenvalues ​​— Assigning values ​​to the geometric features of map differences

[0171] Another quantitative parameter describing the geometric characteristics of map difference is defined as the map difference feature value, which includes at least three possible values. The geometric characteristics of map difference can be assigned values ​​based on statistical indices of the map difference within the "extraction well".

[0172] Different representation methods can be used for geometric features of image difference. For example, intuitively, for equivalent light-emitting LEDs, the image difference feature value can be represented by three values: "unchanged", "lit", and "off". More generally, when the applicability and reasonableness of the expression are extended to include non-light-emitting LCDs, the image difference feature value can be represented by three values: "unchanged", "visible", and "hidden". Abstractly, the image difference feature value can be represented by three values: integers "0", "1", and "-1". More practically, the image difference feature value can be represented by positive integer codes "1", "2", and "0" containing 0, etc. (the aforementioned integer values ​​+ 1), etc.

[0173] Of course, there may also be fourth and fifth states that we don't need to worry about for the time being, such as "unrecognized state", "unassigned state", "unrecognized state", "NULL state", etc.

[0174] For example, the "g segment" of the difference characteristic value in Figure (702) is determined to be "1" using the following statistical indicators:

[0175] ① The sum of the map differences in the "extraction well (g)" of the "geometric feature extraction well plate (18A)" is higher than the set lower limit of positive numbers;

[0176] ② The number of feature class 1 in the "extraction hole (g)" of the "geometric feature extraction plate (18A)" is higher than the set lower limit, for example, the proportion of feature class 1 is higher than 50% (the total proportion of other feature classes is lower than 50%).

[0177] Similarly: by using the "extraction hole" (e) and "extraction hole" (f) of the "geometric feature extraction hole plate (17A)", the "e segment" and "f segment" of the map difference feature value in Figure (702) are determined to be "-1"; by using the "extraction hole" (a), "extraction hole" (b), "extraction hole" (c) and "extraction hole" (d) of the "geometric feature extraction hole plate (17A)" and "geometric feature extraction hole plate (18A)", the "a segment", "b segment", "c segment" and "d segment" of the map difference feature value are determined to be "0".

[0178] In summary, the geometric features and feature values ​​of the difference in Figure (702) can be expressed as "Δa=0,Δb=0,Δc=0,Δd=0,Δe=﹣1,Δf=﹣1,Δg=1".

[0179] The strokes (a, b, c, d, e, f, g) together constitute the geometric features of the image difference or the geometric features of the seven-segment digital characters. The geometric features of the image difference or the geometric features of the seven-segment digital characters have strokes (a, b, c, d, e, f, g), and each stroke has numerical attributes: ① The image difference feature quantity represents the state change (Δ) of the stroke; ② The numerical attributes of the character geometric features are broadly referred to as character feature values, and the numerical attributes of the character geometric features of the seven-segment digital characters can be specifically referred to as "segment values".

[0180] When using the "Geometric Feature Extraction Orifice Plate" to quantize the image difference into image difference feature values, if we define "unchanged" = 0, "lit" = 1, and "extinguished" = -1 to represent the three values, then the change states of each segment when the character "0" jumps to "3" are: Δa = "unchanged", Δb = "unchanged", Δc = "unchanged", Δd = "unchanged", Δe = "extinguished", Δf = "extinguished", Δg = "lit"; if we define "unchanged" = 0, "visible" = 1, and "hidden" = -1 to represent the three values, then the change states of each segment when the character "0" jumps to "3" are: Δa = 0, Δb = 0, Δc = 0, Δd = 0, Δe = -1, Δf = -1, Δg = 1.

[0181] For example: After removing the Arabic numerals 0 to 9 from the "Cutout Panel 0" to "Cutout Panel 9" respectively, the remaining image is as follows: Figure 19As shown in the figure, it can be concluded that there is no residual image only when the Arabic numeral matches the "removal board" number. That is, after the image difference of the Arabic numeral series is removed by "removal board 0" to "removal board 9" respectively, if feature class 1 disappears or feature class -1 disappears, then the Arabic numeral corresponding to the "removal board" number exists in the image difference. The image difference feature value can be defined as {the image difference feature value of removal board A is "-1 removal board"} and {the image difference feature value of removal board B is "1 removal board"} respectively, corresponding to the Arabic numeral A jumping to the Arabic numeral B.

[0182] (6.3) Use the geometric features and feature values ​​of the map difference to identify the numbers represented by the map difference;

[0183] Example 6-1: In a seven-segment digital series, if the image difference feature value corresponding to the "extraction hole" (c) of the "geometric feature extraction hole plate (17A)" is determined to be "1", the conclusion "the number '2' becomes the number '?' (here, '?' represents the numbers 1, 3, 4, 5, 6, 7, 8, 9, 0)" is immediately obtained; this uses both the image difference geometric feature {"extraction hole" (c)} and the condition {image difference feature value is "1"}. Similarly, if in a seven-segment digital series, if the image difference feature value corresponding to the "extraction hole" (c) of the "geometric feature extraction hole plate (17A)" is determined to be "-1", the conclusion "the number '?' becomes the number '2'" is immediately obtained; here, the number "2" is the number represented by the identified image difference, and both the image difference geometric feature {"extraction hole" (c)} and the condition {image difference feature value is "-1"} are used.

[0184] Example 6-2: The map difference feature value corresponding to the "extraction hole" (K2) of the "geometric feature extraction orifice plate (20A)" is determined to be "1", and the conclusion "the number '?' becomes the number '0'" can be immediately obtained. This uses both the map difference geometric feature {"extraction hole" (K2)} and the condition {map difference feature value is "1"}. The following examples are similar and will not be repeated. Similarly, the map difference feature value corresponding to the "extraction hole" (K2) is determined to be "-1", and the conclusion "the number '0' becomes the number '?'" can be immediately obtained.

[0185] Example 6-3: When the map difference feature value corresponding to the "extraction hole" (K2) of the "geometric feature extraction orifice plate (20A)" is determined to be "0", if the map difference feature value corresponding to the "extraction hole" (K1) is determined to be "1", then the conclusion is "the number '?' becomes the number '1'". If the map difference feature value corresponding to the "extraction hole" (K1) is determined to be "-1", then the conclusion is "the number '1' becomes the number '?'".

[0186] Example 6-4: The map difference feature value corresponding to the "extraction hole" (K2) of the "geometric feature extraction plate (20A)" is determined to be "1", and the map difference feature value corresponding to the "extraction hole" (K1) is determined to be "0". The conclusion that "the number '1' becomes the number '0'" can be drawn immediately.

[0187] Example 6-5: The map difference feature value corresponding to the "extraction hole" (K2) of the "geometric feature extraction plate (20A)" is determined to be "-1", and the map difference feature value corresponding to the "extraction hole" (K1) is determined to be "0". The conclusion that "the number '0' becomes the number '1'" can be drawn immediately.

[0188] Example 6-6: The difference between the Arabic numeral A and the Arabic numeral B is removed by "Removal Panel A" and the -1 feature class disappears. After being removed by "Removal Panel B", the 1 feature class disappears. "Removal Panel" is the geometric feature of the difference, and "-1 feature class disappears" or "1 feature class disappears" is the feature value of the difference.

[0189] In summary, Examples 6-1 to 6-6 involve using both geometric features and feature values ​​of the map difference when identifying the numbers represented by the map difference.

[0190] (6.4) Divide the graph difference by individual numbers;

[0191] In the "Test / Inspection / Calibration / Verification" process, the model number of the instrument being tested is a necessary information to determine. Prior information associated with the model number may include the type of display, the optical characteristics of the display, the geometry / size of the display window and its positional features on the instrument being tested, the text features, pattern line features, and pattern color features on the instrument being tested, etc.

[0192] ① On the one hand, prior information can be fully utilized to segment the map difference into individual numerical characters using geometric relationships;

[0193] ②On the other hand, image processing methods are used to segment the image difference into individual numeric characters;

[0194] ③When the camera is pointed at the instrument being tested, it learns the individual digits of the indicated value from the changing indicator pattern;

[0195] ④ Use methods ①, ②, and ③ above to segment the map difference into individual numeric characters.

[0196] From the graph difference statistics, geometric features of the graph difference are obtained, and further, the geometric features of the "stroke segments" (position, length, width, etc.) are statistically determined; for example, the height and width of characters are statistically determined from the graph difference of a seven-segment digital series, such as... Figure 9 As shown, from Figure 9In (9A), the height and width of a character can be obtained by using any one of the three character image differences, and the "geometric features" of the "stroke segment" can be further statistically determined; while (9B) and (9C) use the two character image differences to obtain the height and width of the character without any error, thus laying the foundation for using the "geometric feature extraction plate".

[0197] Furthermore, in step S4, when classifying the map difference, the three categories are specifically represented by feature class 0, feature class 1, and feature class -1, respectively. The map difference value corresponding to feature class -1 is negative, and the map difference value corresponding to feature class 1 is positive.

[0198] Furthermore, the classification of map differences specifically includes any one, two, or all of Y41A, Y41B, and Y41C; among them,

[0199] Y41A is a feature class where the difference value of the graph is within the range of the first positive threshold pT1 and the first negative threshold nT1, where nT1≤0 and pT1≥0.

[0200] nT1 = -pT1 indicates that the map difference is marked as a feature class of 0 using positive and negative symmetric thresholds; Convention: when the threshold pT1 = nT1 = 0, the map difference = 0 satisfies "the map difference value is in the interval between the first positive threshold pT1 and the first negative threshold nT1";

[0201] Y41B is a feature class where the difference value is not lower than the second positive threshold pT2, and pT2 ≥ pT1;

[0202] The map difference value of feature class 1 is positive, also known as positive map difference, but not all map difference values ​​that are positive are marked as feature class 1;

[0203] Y41C is a feature class where the difference value does not exceed the second negative threshold nT2, and nT2≤nT1;

[0204] The graph difference value of the -1 feature class is negative, also known as negative graph difference, but not all graph difference values ​​that are negative are marked as the -1 feature class.

[0205] Before the difference is classified, it is represented by the × feature class. When nT2≠nT1 or / and pT1≠pT2, the difference value of the × feature class may be distributed between (nT2, nT1) or / and (pT1, pT2). Compared with image binarization, the term "tri-(multi-)-featureization / tri-(multi-)-valued" here is inaccurate. "Tri-" should be "multi-". "Tri-valued" is inaccurate. The feature classes are not all classified according to the amplitude. Between pT1 and pT2, there is a mixture of 0 feature classes and 1 feature classes. It is possible that: the difference is small but it is adjacent to the "1 feature seed" and is classified as the 1 feature class, and the difference is large but it is adjacent to multiple "0 feature seeds" and is classified as the 0 feature class.

[0206] pT2 = pT1 indicates that a single threshold is used to classify positive image differences, nT2 = nT1 indicates that a single threshold is used to classify negative image differences, nT2 = -pT2 indicates that a symmetrical positive and negative threshold is used to classify image differences, pT2 = pT1 = -nT2 = -nT1 indicates that a symmetrical positive and negative single threshold is used to classify image differences, and nT2 = nT1 = pT2 = pT1 = 0 is used to classify image differences in high-brightness images (e.g., LEDs), that is: image difference values ​​> 0 are marked as feature class 1, and image difference values ​​< 0 are marked as feature class -1.

[0207] "Not lower than or equal to the second positive threshold pT2" (i.e., ≥) includes "Not lower than the second positive threshold pT2" (i.e., >), and similarly, "Not more than or equal to the second negative threshold nT2" (i.e., ≤) includes "Not more than the second negative threshold nT2" (i.e., <).

[0208] Preferably, by utilizing the amplitude characteristics of the map difference, a threshold is used to select "feature seeds" from the map difference with high confidence. The range of the same "feature class" is expanded in the vicinity of the "feature seed" by the spatial correlation with the "feature seed". The range of "feature class" can also be expanded by utilizing the periodicity of the map difference (and / or image) in the time domain.

[0209] Specifically: Map differences that satisfy nT1 ≤ map difference ≤ pT1 are labeled as feature class 0, also known as being classified into feature class 0. The specific values ​​of the thresholds nT1 and pT1 are usually determined based on the range of map difference values ​​and the probability of misclassifying the map difference. For example:

[0210] (1) Marking “map difference = 0” as feature class 0 results in an extremely low probability of misclassifying the map difference, i.e., extremely high confidence; preferably, marking “map difference = 0” as “feature seed 0”.

[0211] (2) When using 1×1 pixel grayscale for primitive features, the grayscale value is 0 to 255 and the map difference value ranges from -255 to 255. The map difference that satisfies "-1≤map difference≤1" is marked as feature class 0. The probability of misclassifying the map difference is low, that is, the confidence level is high.

[0212] (3) When using 1×1 pixel grayscale for primitive features, the map difference that satisfies “﹣2≤map difference≤2” is marked as a 0 feature class, that is, the map difference is regarded as “0”. The probability of misclassifying the map difference is low, that is, the confidence is high, but the probability of misclassifying the map difference is higher than that of (2).

[0213] Similarly: when using the (R+G+B) value of 1×1 pixel for primitive features, the (R+G+B) value is 0 to 765, and the map difference value ranges from -765 to 765. Map differences that satisfy "-3≤map difference≤3" are marked as feature class 0, and the probability of incorrect classification is low.

[0214] Preferably, the zero-feature class is included in the subsequent classification process according to "map difference = zero".

[0215] Preferably, a 0 feature class with a higher confidence level is obtained by using a narrower threshold range as the "0 feature seed". For non-"low contrast" graphics, marking the "graphic difference value in the range of the first positive threshold pT1 and the first negative threshold nT1" as the 0 feature class is not necessary but compatible. The software executes this code without any other side effects except for increasing the time consumption.

[0216] Figure 10 is a schematic diagram of the effect changes during the classification process of map difference. In the figure, “■” represents feature class 1, “□” represents feature class 0, and blank “” represents feature class ×. Figure 10(a) is a schematic diagram of the classification effect when “map difference = 0” is only classified into feature class 0. Figure 10(b) is a schematic diagram of the classification effect after the absolute value of map difference is below the threshold and classified into feature class 0. Figure 10(c) is a schematic diagram of the classification effect when the × feature class is further classified into feature class 0 using the evaluation quantity. In this case, the × feature class and the 1 feature class constitute obvious character features. This schematic diagram demonstrates the effect of finding “0 feature seed” and expanding the range of the same “0 feature class” in its neighborhood through the spatial correlation with the “0 feature seed”.

[0217] Map differences not lower than a threshold pT2 are labeled as feature class 1, also known as classifying to feature class 1. The specific value of the threshold pT2 is usually determined based on the maximum value of the map difference (denoted by "MAX(map difference)", where MAX(map difference) > 0 and is not lower than a preset value) and the probability of misclassifying the map difference. For example:

[0218] (1) When the threshold pT2 = MAX(map difference), the map difference that meets this condition is marked as feature class 1, and the probability of misclassifying the map difference is extremely low.

[0219] (2) When using the grayscale of a single pixel as the primitive feature quantity, when the threshold pT2 = MAX(map difference) - 1, the map difference that meets this condition will be marked as feature class 1, and the probability of misclassifying the map difference is low.

[0220] (3) When using the grayscale of a single pixel as the primitive feature quantity, when the threshold pT2 = MAX(map difference) - 2, the map difference that meets this condition will be marked as feature class 1, and the probability of misclassifying the map difference is low.

[0221] In graph difference analysis, there can be cases where there is no 1-feature class.

[0222] Preferably, a higher threshold pT2 is used to obtain a 1-feature class with higher confidence as a "1-feature seed".

[0223] like Figure 14As shown in Figure (14B), a "0 feature seed" with higher confidence is obtained by using a smaller threshold pT1, marked as T0 in the figure; a "1 feature seed" with higher confidence is obtained by using a larger threshold pT2, marked as T1 in the figure; regions (Y1~Y42) and (Y51~Y91) in the figure are marked as feature class 1 based on their proximity to the "1 feature seed", as shown in Figure (14B). Figure 14 As shown in (14C), features adjacent to the "0 feature seed" are marked as 0 feature class, labeled T0A in the figure; at Y-axis 71, features marked as 0 feature class in Figure (14B) are subsequently corrected to 1 feature class by adjacent map differences, labeled T1A in Figure (14C). Figure (15B) is a schematic diagram of obtaining "feature seeds" using a threshold. In the region of Figure (15B), the threshold pT2 results in no 1 feature class. Figure (15C) is a schematic diagram of classifying the map differences using the evaluation metric of adjacent map differences. In the figure, regions (Y11~Y21) and (Y81~Y91) are marked as 1 feature class using features adjacent to the "1 feature seed".

[0224] Similarly: Map differences not exceeding a threshold nT2 are labeled as feature classes of -1, also known as being classified into the -1 feature class. The specific value of the threshold nT2 is usually determined based on the minimum value of the map difference (denoted by "MIN(map difference)", where MIN(map difference) < 0 and is not higher than a preset value) and the probability of misclassifying the map difference. For example:

[0225] (1) When the threshold nT2 = MIN (map difference), the map difference that meets this condition is marked as the -1 feature class, and the probability of misclassifying the map difference is extremely low.

[0226] (2) When using the grayscale of a single pixel as the primitive feature quantity, when the threshold nT2 = MIN(map difference) + 1, the map difference that meets this condition is marked as the -1 feature class, and the probability of misclassifying the map difference is low.

[0227] (3) When using the grayscale of a single pixel as the primitive feature quantity, when the threshold nT2 = MIN(map difference) + 2, the map difference that meets this condition will be marked as the -1 feature class, and the probability of misclassifying the map difference is low.

[0228] In graph difference analysis, there can be cases where there is no -1 feature class.

[0229] Preferably, a low threshold nT2 is used to obtain a -1 feature class with a high confidence level as the "-1 feature seed".

[0230] Figure 4This is a comparative schematic diagram of the changes in the frequency distribution of the graph differences of X135, X136, and X137 during the process of classifying graph differences. In the figure, (401) is a schematic diagram of the frequency distribution of the graph differences of X135, X136, and X137, (402) is a schematic diagram of the frequency distribution of the graph differences of X135, X136, and X137 after setting the absolute value of the graph differences below the threshold to zero, and (403) is a schematic diagram of the frequency distribution of the graph differences of X135, X136, and X137 after further classifying the × feature class into the 0 feature class and setting it to zero using the evaluation metric.

[0231] Preferably: when classifying map differences using different methods, the classification method is labeled on the classification results; the confidence levels of classification results obtained from map differences by different methods and different iteration numbers may be inconsistent.

[0232] The difference in the map can be labeled as (or classified into) a feature class, or it can be represented by another variable, or the value of the difference can be set to a value that is normally impossible.

[0233] For example, in a grayscale system with 1×1 pixels as primitives, the absolute value of the difference usually does not exceed 255. Any value greater than 255 can be used to represent feature class 1, or / and any value less than -255 can be used to represent feature class -1.

[0234] Specifically: ">255" represents a feature class, which is further subdivided into "1000" representing "1 feature seed" in the feature class, "900" representing the feature class marked by method 1 (derived from "1 feature seed"), "800" representing the feature class marked by method 2, "700" representing the feature class marked by method 3, and so on.

[0235] Similarly: use "<-255" to represent the -1 feature class, which is further subdivided into "-1 feature seed" in the -1 feature class using "-1000", "-900" to represent the -1 feature class marked by method 1 (derived from "-1 feature seed"), "-800" to represent the -1 feature class marked by method 2, "-700" to represent the -1 feature class marked by method 3, ...

[0236] The map differences distributed between (nT2, nT1) and (pT1, pT2) are classified in the spatial domain according to their adjacent correlation and in the temporal domain, and the existence of × feature classes is allowed.

[0237] Furthermore, it includes arranging the feature micro-components in the image sequence into a feature micro-component VS image sequence curve and marking the feature micro-components between feature class 1 and feature class -1 on the curve as feature class 0. The feature micro-components are the difference in primitive feature quantities at the same position in two adjacent images in the image sequence.

[0238] The digital display instrument's indicated value segment pattern (or segment pixel) alternates between "visible" and "invisible" in the time domain. Therefore, the characteristic micro-component VS sequence curve shows alternating appearances of feature class 1 and feature class -1, such as... Figure 25 As shown in Figures (2a) to (2K2).

[0239] Furthermore, it includes enhancing the graph difference for non-zero feature classes, wherein the enhancement process includes increasing the absolute value of the graph difference.

[0240] Furthermore, it also includes any one, any two, or all of the processing methods of Y43A, Y43B, and Y43C; wherein, Y43A is to perform expansion processing according to feature class; for example, specifically to perform expansion processing on the 0 feature class, or on the 1 feature class, or on the -1 feature class;

[0241] Y43B performs shrinkage processing based on feature class; for example, specifically shrinking the 0 feature class, or shrinking the 1 feature class, or shrinking the -1 feature class.

[0242] Figure 8 This is a schematic diagram of the effect of dilation and contraction operations on the map difference. Usually, dilation is performed only on the × feature class, while contraction does not process the "feature seed".

[0243] Y43C removes isolated noise from feature classes in the image difference.

[0244] Furthermore, step S4 also includes classifying the map differences using evaluation metrics of adjacent map differences. These evaluation metrics include any one, any two, any three, any four, any five, any six, or all of Y45A, Y45B, Y45C, Y45D, Y45E, Y45F, and Y45G.

[0245] Y45A uses the average difference between adjacent maps as the evaluation metric;

[0246] Y45B uses the median difference between adjacent maps as the evaluation metric;

[0247] Y45C uses the moving smoothing result of adjacent map differences as the evaluation metric;

[0248] Y45D uses the difference between adjacent maps as a function with weighting coefficients as the evaluation metric;

[0249] The evaluation metric is expressed by the generalized formula as follows:

[0250] J n (i A j B )=f(T m (i, j)) (7)

[0251] In the formula: m=n, n±1, n±2,…, i=i A i A ±1, i A ±2, ..., j = j B j B ±1,j B ±2, ... represent adjacent ranges.

[0252] The generalized formula for the evaluation metric, which includes the difference between adjacent graphs as a weighting coefficient, is expressed as:

[0253] J n (i A j B )=f(T m (i, j), W T (m)) (8)

[0254] In the formula: m=n, n±1, n±2,…, i=i A i A ±1, i A ±2, ..., j = j B j B ±1,j B ±2, ... represent adjacent ranges; W T (m) is the map difference weighting coefficient.

[0255] Y45E uses the distance between adjacent map differences as a function with weighting coefficients as the evaluation metric;

[0256] The generalized formula for the evaluation metric, which includes the distance between adjacent graph differences as a weighting coefficient, is expressed as:

[0257] J n (i A j B )=f(T m (i, j), W r (m)) (9)

[0258] In the formula: m=n, n±1, n±2,…, i=i A i A ±1, i A ±2, ..., j = j B j B ±1,j B ±2, ... represent adjacent ranges; W r (m) represents the distance weighting coefficient of the graph difference (the Gaussian filter value is a specific case of it).

[0259] The generalized formula for the evaluation metric, which includes both Y45D and Y45E weighting coefficients, is expressed as:

[0260] Jn (i A j B )=f(T m (i, j), W T (m), W r (m)) (10)

[0261] Bilateral filtering is a specific special case of it.

[0262] Y45F uses the number of identical feature classes in adjacent map differences as the evaluation metric;

[0263] For example, the number of zero feature classes in 3×3 adjacent map differences in a specific spatial domain is used as an evaluation metric. If the number is greater than 4, it is marked as a zero feature class. Similarly, if the number of 1 feature classes in 3×3 adjacent map differences is greater than 4, it is marked as a l feature class. If the number of -l feature classes in 3×3 adjacent map differences is greater than 4, it is marked as a -1 feature class.

[0264] Y45G uses the sign of the difference between adjacent graph values ​​as an evaluation metric.

[0265] For example: specifically, "adjacent to feature class 1 and the graph difference value is positive" is used as the evaluation metric, and the evaluation metric = "True" is marked as feature class l. To avoid boundary overflow, the evaluation metric is preferably "adjacent to feature class l and the graph difference value satisfies not less than the threshold pTl". More preferably, the processing object is only feature class × (that is: feature class 0 with graph difference ≥ pTl is used as the boundary to avoid overflow).

[0266] For example, specifically, "adjacent to the -1 feature class and the graph difference value is negative" is used as the evaluation metric; if the evaluation metric = "True", then it is marked as a -1 feature. To avoid boundary overflow, the evaluation metric is preferably "adjacent to the 1 feature class and the graph difference satisfies not higher than the threshold nT1". More preferably, the processing object is only the × feature class (that is, the 0 feature class with graph difference ≤ nT1 is used as the boundary to avoid overflow).

[0267] The evaluation quantity can be represented by "True / False", or by positive logic "I / 0", or by negative logic "0 / 1".

[0268] For example: in the spatial domain m = n + 1, the graph differences of 3×3 adjacent graphs are respectively, D n (i A +1, j B +1), D n (i A j B +1), D n (i A -1,j B +1) and Dn (i A +1, j B ), D n (i A -1,j B The average, median, or moving smoothed result is used as the evaluation metric; when nT1≤evaluation metric≤pT1, it is marked as feature class 0; when the evaluation metric≥pT2, it is marked as feature class 1; and when the evaluation metric≤nT2, it is marked as feature class -1.

[0269] Furthermore, the specific meaning of "adjacent map difference" includes any one, any two, or all of Y45X, Y45Y, and Y45Z, wherein...

[0270] Y45X represents adjacent coordinates in the spatial domain;

[0271] Y45Y represents adjacent graph sequences in the time domain;

[0272] Y45Z is the difference in primitive features at the same location in two sequentially adjacent images in the time domain.

[0273] (1) For example: with two images P n and image P m (where m≠n) the map difference is represented by map difference D. n,m (i A ,j B ), in the spatial domain with D n,m (i A ,j B The coordinate differences between adjacent graphs are as follows:

[0274] D n,m (i A -1,j B -1),D n,m (i A -1,j B ),D n,m (i A -1,j B +1)

[0275] D n,m (i A ,j B -1),D n,m (i A ,j B ),D n,m (i A ,j B +1)

[0276] D n,m (i A +1,j B -1),Dn,m (i A +1,j B ),D n,m (i A +1,j B +1)

[0277] (2) In the time domain, the graphs are adjacent in sequence, for example:

[0278] D n,m-1 (i A ,j B ),D n,m (i A ,j B ),D n,m+1 (i A ,j B )

[0279] D n-1,m (i A ,j B ),D n,m (i A ,j B ),D n+1 ,(i A ,j B )

[0280] (3) Map difference uses the difference in primitive features at the same position in two sequentially adjacent images in the time domain; that is, m = n + 1, and map difference is expressed as D. n,n+1 (i A ,j B If ) indicates that:

[0281] D n,n+1 (i A ,j B ) = T n+1 (i A ,j B )-T n (i A ,j B )

[0282] The graph difference D under the condition m=n+1 n,n+1 (i A ,j B Also known as the characteristic differential component and abbreviated as D n (i A ,j B That is, the feature differential component is the difference in the feature quantities of primitives at the same position in two adjacent images, expressed by the following formula:

[0283] D n (i A ,j B ) = Tn+1 (i A ,j B )-T n (i A ,j B )

[0284] The characteristic differential component is a special case of the graph difference when m = n + 1. Under this condition, the adjacent characteristic differential components in the time domain are:

[0285] D n-1 (i A ,j B ),D n (i A ,j B ),D n+1 (i A ,j B )

[0286] Equivalent to D when arranged in diagram order n,n+1 (i A ,j B The difference between adjacent maps is D. n+1,n+2 (i A ,j B ),D n-1,n (i A ,j B ).

[0287] For example: for the characteristic differential component D n (i A ,j B In the adjacent three-dimensional space composed of the spatial domain and time domain described in Y45X and Y45Y, there are:

[0288] ①D in the n-1 time domain n-1 (i A ,j B Within the spatial domain:

[0289] D n-1 (i A +1,j B -1),D n-1 (i A ,j B -1),D n-1 (i A -1,j B -1)

[0290] D n-1 (i A +1,j B +1),D n-1 (i A ,j B +1),D n-1(i A -1,j B +1) and

[0291] D n-1 (i A +1,j B ),D n-1 (i A ,j B ),D n-1 (i A -1,j B );

[0292] ②D in the n-time domain n (i A ,j B Within the spatial domain:

[0293] D n (i A +1,j B -1),D n (i A ,j B -1),D n (i A -1,j B -1)

[0294] D n (i A +1,j B +1),D n (i A ,j B +1),D n (i A -1,j B +1) and

[0295] D n (i A +1,j B ),D n (i A -1,j B );

[0296] ③D in the n+1 time domain n+1 (i A ,j B Within the spatial domain:

[0297] D n+1 (i A +1,j B -1),D n+1 (i A ,j B -1),D n +1(i A -1,jB -1)

[0298] D n+1 (i A +1,j B +1),D n+1 (i A ,j B +1),D n+1 (i A -1,j B +1) and

[0299] D n+1 (i A +1,j B ),D n+1 (i A ,j B ),D n+1 (i A -1,j B ).

[0300] For example: Figure 13 As shown, the image difference uses the difference in primitive features at the same position in two adjacent images, also known as the feature differential. In the time domain, the corresponding time points n-1, n, and n+1 are taken for adjacent images. In the spatial domain of image number n-1, a 3×3 matrix with adjacent coordinates is taken. In the spatial domain of image number n, a 5×5 matrix with adjacent coordinates is taken. In the spatial domain of image number n+1, a 3×3 matrix with adjacent coordinates is taken. The value of the center point of image number n is the average value of the feature differential contained in the three-dimensional space (the value is expressed by => to indicate that the average value is calculated in the three-dimensional space). The result is 0.9.

[0301] Furthermore, the character geometric features and character feature values ​​of the indicator value are derived using the image difference geometric features and image difference feature values, and then the numbers corresponding to the indicator value pattern are derived based on the character geometric features and character feature values.

[0302] Furthermore, the character series of the digital display instrument indication value pattern is classified into any one of the Y61, Y62, and Y63 character series. Template characters are created for this character series. The difference between two template character patterns is used as a standard template and described by template geometric features and template feature quantities. The geometric features and feature values ​​of the difference obtained in S6 are matched with the geometric features and feature quantities of the template to obtain the corresponding numbers in the indication value pattern.

[0303] Y61 represents a character series composed of seven segments of digital code 0 to 9. The feature is that the pattern of characters 0 to 9 can be equivalently divided into seven segments. The same segment has the same optical features at the same time. The display of characters 0 to 9 is achieved by changing the combination of optical features of the segments.

[0304] Y62 represents a character series consisting of Arabic numerals 0 to 9, characterized in that the patterns of characters 0 to 9 are Arabic numerals;

[0305] Y63 represents a character series composed of character patterns in the same digital display instrument indication value;

[0306] The character series includes those appearing in the same digital display instrument indication value pattern.

[0307] Furthermore, all template character pattern differences in the character series are enumerated as standard templates and described using template geometric features and template feature quantities.

[0308] Repeat the above process for each digit in the digital display instrument's indicator value pattern to obtain all the characters of the indicator value.

[0309] The following is a detailed description of the "low contrast" digital instrument indication pattern feature used in this embodiment.

[0310] In this embodiment, the low-contrast digital display indicator shows a pattern where "000" transitions to "1" (leading "0" is culled). The pixel optical features use the (RGB) average value—referred to as "grayscale"—and are then binarized using different grayscale thresholds, resulting in a black-and-white illustration. Figure 2 As shown; from Figure 2 It can be concluded that the grayscale of the digit character pattern blends with the grayscale of the background. The three thresholds in the image cannot binarize the "low-contrast" digit pattern. Figure 11 This can be confirmed by (11A) and (11B).

[0311] Figure 11 Image (11A) is a schematic diagram of the RGB mean distribution curve of the background after removing the area containing the "088" segment from the "000" pattern, which is a low-contrast indicator value. The background image is divided into two parts according to the concave points in the distribution curve (as shown by the dotted lines in the image). Figure 12 As shown, Figure 12 The symbols marked with ■ represent elements below the threshold, and those marked with □ represent elements above the threshold.

[0312] Figure 11 (11B) shows the grayscale distribution curve of the unit digit "0" pattern, where the black solid line "A" represents the grayscale distribution curve of the entire unit digit "0" pattern. This curve is consistent with... Figure 11The RGB mean distribution curves of the background shown in Figure A overlap and are all distributed within the interval (51, 101). The black dotted line "T" represents the grayscale distribution curve at the upper 1 / 4 of the unit digit "0" pattern, while the black hollow line "B" represents the grayscale distribution curve at the lower 1 / 4 of the unit digit "0" pattern. From the figure, it can be seen that the grayscale distribution curves at the upper and lower 1 / 4 of the unit digit "0" pattern are significantly different. Figure 15 A mountain peak curve appears in the Y coordinate (Y11~Y42) range where the "stroke c" of the unit digit character of the indicator value pattern "000" is located;

[0313] Figure 11 The middle (11C) shows the gray level at 1 / 4 of the pattern of the unit digit "0" of the indicator value, as well as the distribution curves of the gray levels divided into the left and right parts "T1" and "T2". It can be seen from the figure that the distribution curves of the gray levels in the left and right parts "T1" and "T2" are different.

[0314] Figure 11 The middle (11D) shows the gray level at the bottom 1 / 4 of the pattern of the unit digit "0" of the indicator value, as well as the distribution curves of the gray levels divided into the left and right parts "B1" and "B2". It can be seen from the figure that the distribution curves of the gray levels of the left and right parts "B1" and "B2" are different.

[0315] From the above, we can conclude that:

[0316] ① The RGB mean distribution curve of the unit digit "0" pattern of the "low contrast" indicator value pattern "000" overlaps with the RGB mean distribution curve of the background in the interval (51, 101).

[0317] ②The gray distribution curves at the upper 1 / 4 and lower 1 / 4 of the pattern for the units digit "0" show obvious differences;

[0318] ③ There are also slight differences in the gray distribution curves of the left and right parts of the upper 1 / 4 and the left and right parts of the lower 1 / 4 of the "0" pattern.

[0319] Figure 12 according to Figure 11 The concave dot pattern "000" of the digital display instrument is shown in Figure (11A) as a background binary black and white diagram after removing the area where the "088" segment is located. In the figure, ■ represents primitives below the threshold and □ represents primitives above the threshold. From this figure, it can be inferred that the gray distribution curves of the upper 1 / 4 and lower 1 / 4 of the pattern of the unit "0" character are different; the "segment c" of the unit "0" character may have "bright" reflection; the "segment d" of the hundreds "0" character is less bright than the "segment e".

[0320] Figure 14A is a schematic diagram showing the distribution of local primitive feature quantities and their differences along the Y-axis of the "low contrast" indicator patterns "000" and "1" (leading "0" has been hidden). Legendary figures BX12, BX13, and BX14 represent the primitive feature quantities at the locations of the "stroke e" and "stroke f" segments of the hundreds digit of the indicator pattern "000". Legendary figures AX12, AX13, and AX14 represent the primitive feature quantities at the locations of the "stroke e" and "stroke f" segments of the hundreds digit of the indicator pattern "1" (leading "0" has been hidden). Legendary figures AX12-BX12, AX13-BX13, and AX14-BX14 represent the corresponding differences. Figure 14 A can be preliminarily summarized as follows: the brightness of the hundreds digit "0" character "segment e" (Y1~Y42) is lower than that of "segment f" (Y51~Y91); the image difference of "segment e" (Y1~Y42) is more obvious; the influence of the non-uniformity of the original image between Y coordinates (Y21~Y42) on the image difference is reduced; and the image difference interference of "segment f" (Y51~Y91) is more obvious.

[0321] Figure 15 A is a schematic diagram showing the distribution of local primitive feature quantities and their differences along the Y-direction of the "low contrast" indicator value patterns "000" and "1" (leading "0" has been hidden). The horizontal axis is the Y-direction. Among them, legends BX148, BX149, and BX150 are the primitive feature quantities at the locations of "segment b" and "segment c" of the unit value character of the indicator value pattern "000". Legends A148, A149, and A150 are the primitive feature quantities at the locations of "segment b" and "segment c" of the unit value character of the indicator value pattern "1". Legends AX148-BX148, AX149-BX149, and AX150-BX150 are the corresponding differences.

[0322] from Figure 15 A can be preliminarily concluded that: "segment d" located between Y coordinates (Y11~Y21) is on a slope where gray-level non-uniformity increases (actually, it is the image of a white object in the background on the digital display instrument), but the impact of non-uniformity on image difference is reduced. "Segment a" is located between Y coordinates (Y81~Y91).

[0323] The above leads to the preliminary conclusion that due to low contrast, a cluttered background, and the camera's imaging on the monitor surface, the acquired images contain uneven gray levels, which significantly affects subsequent processing. However, using image difference can reduce the impact of unevenness.

[0324] Figure 16 for Figure 14 and Figure 15A schematic diagram showing the positions of X12, X13, X14, X148, X149, and X150, with arrows and X and Y symbols indicating the origin and direction of the coordinate axes; (16A) is... Figure 14 A schematic diagram showing the positions of X12, X13, and X14; (16B) is... Figure 15 A schematic diagram showing the positions of X148, X149, and X150.

[0325]

Example 2

[0326] This embodiment is basically the same as Embodiment 1, except that it has the following specific preferred features:

[0327] In step S5, the image difference is segmented according to the individual digits in the indicator value pattern. After obtaining the image difference geometric features and image difference feature values ​​of the individual digits in step S6, the character geometric features and character feature values ​​of the indicator value are derived using the image difference geometric features and character feature values. Then, the digits corresponding to the indicator value pattern are derived based on the character geometric features and character feature values. For example, the geometric features of the seven-segment digital characters are the same as the image difference geometric features. They are represented by (segment a, segment b, segment c, segment d, segment e, segment f, segment g) according to the distribution position of the character segments "top, top right, bottom right, bottom, bottom left, top left, center", and are abbreviated as (a,b,c,d,e,f,g) or (abcdefg). In this embodiment, the character feature values ​​are represented by "segment values" ("visible" / "hidden" / "×").

[0328] Figure 23 To derive the character geometric features and character feature values ​​of the digits in the indicator value pattern using image difference geometric features and image difference feature values, and to obtain a schematic diagram of the digits corresponding to the indicator value pattern; where the "indicator value pattern" is the seven-segment digit series indicator value pattern "X" (P) to be identified. n ), "Y" (P n+1 ), "Z" (P) n+2 In this embodiment, the graph difference feature value is represented by three values: "0", "1", and "-1"; in the graph...

[0329] ■——Indicates that the stroke segment "changes from hidden to visible", and the graph difference feature value is "1";

[0330] ◇—— indicates that the line segment "changes from visible to hidden", and the feature value of the difference is "-1" ("-1" is not marked in the figure);

[0331] □ — Represents "no change" in the stroke segment, with a graph difference characteristic value of "0";

[0332] (23A) is a schematic diagram of the indicator value pattern "X" changing into the indicator value pattern "Y". In the figure, "?" indicates the one to be identified;

[0333] (23D) is a schematic diagram showing the change of the indicator value pattern "Y" to the indicator value pattern "Z". In the figure, "?" indicates that it is to be identified.

[0334] (23B) is a schematic diagram of the geometric features and feature values ​​of the map difference when the indicator value pattern "X" changes to the indicator value pattern "Y";

[0335] (23C) is based on Figure 23 (23B) A schematic diagram showing the transformation of the indicator value pattern "X1" into the segment values ​​of the indicator value pattern "Y1" is derived, in which... — This indicates that the stroke value is "unknown", and the stroke value b = ×, f = ×;

[0336] The value of segment X1 is "Hidden, ×, Visible, Hidden, Hidden, ×, Hidden";

[0337] The value of segment Y1 is "visible, ×, hidden, visible, visible, ×, visible";

[0338] (23E) is a schematic diagram of the geometric features and feature values ​​of the difference in the graph when the indicator value pattern "Y" changes to the indicator value pattern "Z". In the graph difference feature value of segment b = "-1", it means that the segment becomes "hidden", and it can be inferred that the segment was "visible" before. The graph difference feature value of segment f = "1" means that the segment becomes "visible", and it can be inferred that the segment was "hidden" before.

[0339] (23F) is a schematic diagram of the changes in the indicator value pattern "Y" to the indicator value pattern "Z" based on Figure (23E). The segment values ​​of the indicator value pattern "Y" are "visible, visible, hidden, visible, visible, hidden, visible". According to the conventional coding rules of 7-segment numbers, the indicator value pattern "Y" should be the value "2". The segment values ​​of the indicator value pattern "Z" are "visible, hidden, visible, visible, hidden, visible, visible". According to the conventional coding rules of 7-segment numbers, the indicator value pattern "Z" should be the value "5" - here, the corresponding number in the indicator value pattern is obtained by "the geometric features and character feature values ​​of the number".

[0340] If the indicator value pattern "Y" is the value "2", then the segment b = visible_unchanged f = hidden_unchanged is obtained. Therefore, the segment value of the indicator value pattern "X" is "hidden, visible, visible, hidden, hidden, hidden, hidden, hidden". According to the conventional coding rules of 7-segment codes, the indicator value pattern "X" should be the value "1".

[0341] In summary: The geometric features of the seven-segment digital series characters are abstracted into "segments" arranged in (a, b, c, d, e, f, g). When the character feature value is represented by "segment value" ("1" / "0"), "deriving the character geometric features and character feature values ​​of the numbers in the indicator value pattern" means that the character geometric features and character feature values ​​of the indicator value patterns "X", "Y", and "Z" correspond to "0,1,1,0,0,0,0", "1,1,0,1,1,0,1", and "1,0,1,1,0,1,1" respectively. "Based on the character geometric features and character feature values, the corresponding numbers in the indicator value pattern are derived," thus obtaining the indicator values ​​"X", "Y", and "Z" corresponding to 1, 2, and 5 respectively.

[0342] Repeat the above process for each digit in the digital display instrument's indicator value pattern to obtain all the characters of the indicator value.

[0343] This embodiment also reveals that: when using a camera as an image acquisition device during testing / detection / calibration / verification, preferably, keyframes in the camera video stream after discarding identical content are used as the acquired images, arranged sequentially in the group of images. The alternating appearance of the indicator value pattern in a visible / hidden manner ensures that feature class 1 and feature class -1 on the feature micro-component VS sequence curve will always alternate, such as... Figure 25 As shown in Figures (2a) to (2K2), the character feature values ​​(segment values) of the numbers in the digital display instrument indication value pattern can be tracked in real time, thereby realizing the numerical recognition of the digital display instrument indication value pattern in real time.

[0344]

Example 3

[0345] This embodiment is basically the same as Embodiment 1, except that it has the following specific preferred features:

[0346] In step S5, the image difference is segmented according to the individual digits in the indication value pattern. After obtaining the geometric characteristics and feature values ​​of the image difference for each individual digit in step S6, the character series of the digital display instrument indication value pattern can be classified into any one of the Y61, Y62, and Y63 character series using prior information. Template characters are then created for this character series, using the image difference between two template characters as a standard template and describing it using template geometric characteristics and template feature values, such as... Figure 6 The template characters shown in 601 are 0 and 3, and their template geometric features and template feature values ​​are shown in 602. The geometric features and feature values ​​of the single digit obtained in S6 are (e.g.) Figure 7 As shown in Figure 702) and template geometric features and template feature quantities (such as... Figure 6 Matching (as shown in Figure 602) yields the corresponding number in the indicator value pattern (e.g., ...). Figure 7 (As shown in Figure 703);

[0347] Y61 represents a character series composed of seven-segment digits 0-9. Its characteristic is that the pattern of characters 0-9 can be equivalently divided into seven segments, with each segment having the same optical characteristics at the same time. The display of characters 0-9 is achieved by changing the combination of the optical characteristics of the segments; this is abbreviated as the seven-segment digit series, such as... Figure 18 As shown in 18B;

[0348] Y62 represents a character series composed of Arabic numerals 0 to 9, characterized in that the pattern of characters 0 to 9 is Arabic numerals (1, 2, 3, 4, 5, 6, 7, 8, 9, 0); abbreviated as Arabic numeral series;

[0349] Y63 represents a character series composed of character patterns in the same digital display instrument indication value, such as... Figure 20 As shown in 20B and 20D; the character series includes those characterized by appearing in the same digital display instrument indication value pattern.

[0350] (1) Template geometric features

[0351] Example ① For the seven-segment digital series, the template geometric features are the same as the geometric features of the seven-segment digital characters. They are usually abstracted into template geometric features and template character geometric features according to the relative positions of the character segments such as "top, top right, bottom right, bottom, bottom left, top left, and center", rather than represented by specific geometric dimensions or pixel dimensions. They are conventionally represented by (segment a, segment b, segment c, segment d, segment e, segment f, and segment g), and are conventionally abbreviated as (a, b, c, d, e, f, g) or (abcdefg).

[0352] Example ② The template geometric features of the Arabic numeral series use bold Arabic numeral characters 0 to 9 as geometric features “removal plates”, which are represented by “removal plate 0” to “removal plate 9” respectively. That is, the template geometric features of the Arabic numeral series are described by “removal plate 0” to “removal plate 9”.

[0353] Example ③ Regarding the "character series composed of character patterns in the same digital display instrument indication value", some numeric characters have special characteristics. In this example, the numeric character "0" has a slash in the center. The schematic diagram is shown in Figure (20B), and the schematic diagram of the number "1" is shown in Figure (20D). The design is as follows: Figure 20 The “Geometric Feature Extraction Perforated Plate (20A)” shown is used as a template geometric feature. This perforated plate has two through holes (K1) and (K2). This perforated plate is used to extract characters. The schematic diagram of the effect is shown as (20C) in the figure. The schematic diagram of the effect of using this orifice plate on the character "1" is shown as (20E) in the figure. The patterns taken from the indicated values of the same digital display instrument can also be the series of lowercase Roman numerals (ⅰ, ⅱ, ⅲ, ⅳ, …, ⅸ), the series of uppercase Roman numerals (Ⅰ, Ⅱ, Ⅲ, Ⅳ, …, Ⅹ, Ⅺ, Ⅻ), the series of Chinese numerical values (one, two, three, four, five, six, seven, eight, nine, ten), the series of Chinese uppercase numerical values (壹, 贰, 叁, 肆, 伍, 陆, 柒, 捌, 玖, 拾), the series of uppercase English numerical values (ONE, TWO, THREE, …), the series of lowercase English numerical values, and the series of uppercase and lowercase English numerical values, etc. Among them, ⅳ, Ⅳ, ONE, etc. are regarded as one character, and so on; of course, there must be additional processing steps between "ⅰ, ⅱ, ⅲ" / "Ⅰ, Ⅱ, Ⅲ" / "one, two, three".

[0354] In Example ④, to solve the font differences of Arabic numerals, when using the character series composed of "Arabic numerals" in the indicated value of the same digital display instrument, the template characters are learned from the pattern of the indicated value of the digital display instrument. After bolding the template characters, they are used as the geometric feature "抠除板" (removed plate), which are represented by "Removed Plate 0" to "Removed Plate 9" respectively.

[0355] Summary: In the above examples, the same description method is used for the template geometric features and the template character geometric features. Usually, the same description method is also used for the template geometric features and the geometric features of the figure difference.

[0356] (2) Template feature quantity

[0357] Another quantitative description parameter for the template geometric features is conventionally named as the template feature quantity; it is an assignment to the template geometric features — representing a certain meaning.

[0358] Example SSL3-1: The schematic diagram of establishing a standard template with two template characters "0" and "3" in the seven-segment digital series is as Figure 6 shown. In the figure, (601) is the schematic diagram of each pen segment of the two template characters "0" and "3" in the seven-segment digital series. ■ represents the visible pen segment, and the pen segment value is represented by 1. The pen segment value of the "invisible" pen segment is represented by 0; (602) is the schematic diagram of the standard template. ■ is for "the hidden pen segment becomes a visible pen segment", and the number 1 in the box indicates that its template feature quantity is "1". ◇ is for "the visible pen segment becomes a hidden pen segment", and its template feature quantity is "-1" ( "-1" is not marked in the figure). □ is for "the visible pen segment without change", and the number 0 in the box indicates that its template feature quantity is "0".

[0359] The above uses the template characters "0" and "3" to simulate the change of the待识别数字 (to-be-recognized number) from "0" to "3" (represented by 0 ═〉 3) in the indicated value of the digital display instrument. The established standard template corresponds to the figure difference of the two to-be-recognized numbers in the indicated value of the digital display instrument.

[0360] Example SSL3-2: A schematic diagram illustrating the identification process of two digits to be recognized in the indication value of a digital display instrument, such as... Figure 7 As shown in the figure, (701) is a schematic diagram of the image acquisition device acquiring an indication value pattern and changing it into another indication value pattern, where "?" represents "unknown" or "to be identified"; (702) is a schematic diagram of the image difference being segmented by individual digit characters and represented by the geometric features and feature values ​​of the image difference; (703) is a schematic diagram of the image difference being segmented by individual digit characters and represented by the geometric features and feature values ​​of the image difference; Figure 6 The matching of the numbers in the middle (602) yields a schematic diagram of the identification of the numbers, and the result is "0═>3".

[0361] Example SSL3-1 use Figure 6 This demonstrates the process of creating a standard template using two template characters, for example, SSL3-2. Figure 7 It was revealed that the digit to be identified in the indication value pattern of the digital display instrument is matched with the geometric features and feature values ​​of the image difference (as shown in Figure 702) and the geometric features and feature quantities of the template (as shown in Figure 602) to obtain the corresponding digit in the indication value pattern (as shown in Figure 703).

[0362] For example, when the template geometric features of the Arabic numeral series are described by “elimination plate 0” to “elimination plate 9”, there is no residual image only when the Arabic numeral matches the “elimination plate” number; “elimination plate” is the template geometric feature, and “-1 feature class disappears” or “1 feature class disappears” is the template feature quantity.

[0363] Preferably, the geometric features and feature values ​​of the map difference are represented by codes to facilitate matching processing by computer software. For example, in Figure (602), the geometric features and feature values ​​of the digital template are represented as: "a=0,b=0,c=0,d=0,e=-1,f=-1,g=1", which are arranged from left to right (abcdefg), and abbreviated as "0,0,0,0,-1,-1,1". The geometric features and feature values ​​of the map difference in Figure (702) have the same code; the computer software uses this code to query the corresponding number in the indicator value pattern (that is, the geometric features and feature values ​​of the map difference match the geometric features and feature values ​​of the template).

[0364] Adding the number 1 to the template feature value changes it to a positive integer containing 0, which is more in line with coding conventions, and it is written as "1111002".

[0365] Furthermore, all template character pattern differences in the character series are enumerated as standard templates and described using template geometric features and template feature quantities.

[0366] Repeat the above process for each digit in the digital display instrument's indicator value pattern to obtain all the characters of the indicator value.

[0367]

Example 4

[0368] This embodiment is basically the same as Embodiment 1, except that it has the following specific preferred features:

[0369] S1, use an image acquisition device to acquire a set of images containing the indication value pattern of the digital display instrument, wherein the set of images includes 6 images (the images are from the original record data of the "2240" test), and the sequence number of the images arranged in the order of acquisition is called the image number;

[0370] Preferably, the image only includes the pattern of the digital display instrument indication value. The position (Left, Right, Top, Bottom) of the pattern of the digital display instrument indication value in the array can be marked by the subscripts (XL, YT) — (XR, YB) of the two-dimensional array (X, Y). The subsequent processing focuses on this area.

[0371] S2, calculate the primitive feature quantities in the image;

[0372] S3, calculate the graph difference, which is the difference in primitive feature quantities at the same position in two images; the feature differential component is the difference in primitive feature quantities at the same position in two images with adjacent graph sequences;

[0373] For the six images in the "2240" test, stroke segments (a, b, c, d, e, f, g, K1, K2) were defined. These nine stroke segments were then defined as nine primitives, with the primitive feature being the sum of (R+G+B) pixels contained within the stroke segment. Figure 25 As shown, in Figures (1a) to (1K2), the Y-axis represents the primitive feature quantity, and the X-axis represents the image sequence (n). Figures (1a) to (1K2) correspond to the primitive feature quantities of stroke a to stroke K2, respectively. In Figures (2a) to (2K2), the Y-axis represents the feature differential component D, and the X-axis represents the image sequence (n). Figures (2a) to (2K2) correspond to the feature differential components of stroke a to stroke K2, respectively. The feature differential components of these nine primitives reflect the change in the optical properties of the stroke with the image sequence (n). In the time domain, the feature differential component VS image sequence curve shows alternating changes.

[0374] Using multiple pixels to construct primitives reduces computational load and decreases the "residual noise" after image difference calculation. It may also increase the value of image difference, which is particularly useful for processing low-contrast images. When constructing primitive features, weighting coefficients can be used as needed to selectively highlight key features.

[0375] S4, classifying the map difference, characterized in that the number of categories includes at least three categories, which are specifically represented by feature class 0, feature class 1, and feature class -1 respectively;

[0376] Furthermore, when classifying the map differences, an evaluation metric is used for classification. Specifically, Y45A takes the average value of adjacent “5×5” map differences in the spatial domain as the evaluation metric, marks the map differences at the corresponding positions that satisfy -6 < evaluation metric < 6 as the 0 feature class, and preferably sets the map differences of the 0 feature class to the number “0”.

[0377] Furthermore, Y45D uses the difference between adjacent maps as a function with weighting coefficients as an evaluation metric; specifically, it projects the difference between maps onto the X-axis and Y-axis respectively to obtain the density distribution curves of the difference between maps along the X-axis and Y-axis, such as... Figure 24 As shown in Figure (24B), Tx and Ty in the figure represent the density of the graph difference along the X-axis and Y-axis, respectively.

[0378] To compare the processing results, the map difference before classification was projected onto the X and Y axes respectively to obtain the density distribution curves of the map difference along the X and Y axes, as shown below. Figure 24 As shown in Figure (24A), Tx and Ty in the figure represent the density of the map difference along the X-axis and Y-axis, respectively;

[0379] The weighting coefficient function W is constructed using density distribution curves along the X and Y axes without thresholds. n (i A j B Specifically: the absolute value of the density distribution curve of the map difference along the X-axis is taken and a value of "1" is added to it, which is then used as the weighting coefficient Wx along the X-axis. The absolute value of the density distribution curve of the map difference along the Y-axis is taken and a value of "1" is added to it, which is then used as the weighting coefficient Wy along the Y-axis. The product of the weighting coefficient Wx along the X-axis and the weighting coefficient Wy along the Y-axis is used as the function W of the weighting coefficient. n (i A j B That is: W n (i A j B )=(1+|T X (i A j B )|)*(1+|T Y (i A j B )|), as shown in the diagram Figure 24 As shown in Figure (24C), Wx and Wy represent the weighting coefficients along the X-axis and Y-axis, respectively.

[0380] Weight function W n (i A j B ) and graph difference T n (i A j B The product of ) is used as the evaluation metric J n (i A j B ), that is: Jn (i A j B ) = W n (i A j B )*T n (i A j B The evaluation metric is used to classify map differences.

[0381] Projecting the evaluation quantities onto the X-axis and Y-axis respectively yields the density distribution curves of the evaluation quantities along the X-axis and Y-axis, as shown in Figure (24D). In the figure, Jx and Jy represent the distribution functions of the evaluation quantities along the X-axis and Y-axis, respectively.

[0382] Comparing Figure (24D) and Figure (24B), the distribution density of the difference is greatly enhanced, with a maximum enhancement value of up to 100 times along the X-axis, while the distribution density of the difference is relatively weakened in areas with low distribution density due to the smaller enhancement effect.

[0383] Preferably, the graph difference of non-zero feature classes is enhanced, and the enhancement process includes increasing the absolute value of the graph difference whose absolute value is greater than a certain value; for example, replacing the graph difference within the same graph sequence with the above evaluation quantity is actually enhancing the graph difference of non-zero feature classes.

[0384] Figure 21 This is a schematic diagram showing the effect of classifying map differences using weighted coefficients as evaluation metrics. In the figure, (21A) shows the effect before classification, and (21B) shows the effect after classification. Figure 24 The diagram in (24C) shows the effect of classifying the map difference using the weight coefficients. ■ represents feature class 1, □ represents feature class ×, and “” (space) represents feature class 0.

[0385] Constructing the function W of the weight coefficients n (i A j B At that time, you can also:

[0386] ① Use a function with weighting coefficients that has a "compression-expansion" function, such as a function that takes the product of the absolute values ​​of the density distribution curves along the X and Y axes as the coefficients:

[0387] W n (i A j B )=|T X (i A j B )|*|T Y (i A j B )|

[0388] When W n (iA j B When W < 1, i.e., the "compression" function, when W n (i A j B When ) > 1, it means the "extended" function.

[0389] ② Use a function with weighting coefficients that have a "threshold-expansion" function, such as taking the absolute value of the density distribution curves of the map difference along the X and Y axes and setting the curves below the threshold to the value "1".

[0390] S5, segment the map difference according to the individual numbers in the indicator value pattern;

[0391] S6, yields the geometric features and feature values ​​of the map difference for a single digit;

[0392] Example SSL4-1

[0393] Furthermore, the character geometric features and character feature values ​​of the indicator value are derived using the image difference geometric features and image difference feature values, and then the numbers corresponding to the indicator value pattern are derived based on the character geometric features and character feature values.

[0394] Specifically: Figure 22 The diagram shows the difference classification of six digital instrument indication values. In the diagram, ■ represents feature class 1, □ represents feature class -1, “” (space) represents feature class 0, and “·” represents feature class ×. (22A) is a diagram showing the difference classification when the number “69” jumps to “75”; (22B) is a diagram showing the difference classification when the number “75” jumps to “80”; (22C) is a diagram showing the difference classification when the number “80” jumps to “86”; (22D) is a diagram showing the difference classification when the number “86” jumps to “91”; and (22E) is a diagram showing the difference classification when the number “91” jumps to “96”.

[0395] The test involved irradiating a certain equivalent meter with 0.05 mSv on a standard device, repeated 5 times. The actual readings were displayed in mSv. For simplicity, the decimal point was omitted. The initial indication value "69" (0.69, the same below) and the final indication value "96" were visually confirmed on the equivalent meter display before and after the test. However, the image captured by the camera after the 5th irradiation appeared to be "94" visually, but (22D) clearly indicated the digit "1" in the units place (see "Example 6-2", "Example 6-3" and...). Figure 20(A) K1 and K2 through holes) and (22E) clearly indicate that the units digit "1" has disappeared, plus corroborating evidence (22B) clearly indicates that the units digit "0" appears and (22C) clearly indicates that the units digit "0" has disappeared (see "Example 6-4" and "Example 6-5"). Based on the conclusion of the method of the present invention, it is consistent with the "visual inspection" of the equivalent instrument display, and the coefficient of variation of the response value (increase in reading) of the equivalent instrument after five irradiations is 10%. If the "visual inspection" of the image captured by the camera is used as the result, the coefficient of variation of the response value is 24%, so the "visual inspection" of the image captured by the camera is incorrect. In this test, four instruments under test were irradiated at the same time, and their response values ​​provided mutual corroboration and supervision.

[0396] The following example uses the processing of single-digit numbers. Using geometric features and feature values ​​(represented in this embodiment by Δa, Δb, Δc, Δd, Δe, Δf, Δg, ΔK1, ΔK2, "visible" / "hidden" / "unchanged"), the character geometric features and character feature values ​​(represented in this embodiment by stroke value letters a, b, c, d, e, f, g, K1, K2, "1" / "0") of the digit are obtained, thus yielding the numerical value represented by the indicator pattern, as detailed below:

[0397] Figure 22 The K2 through hole in the middle (22B) was judged to be "visible" (i.e., the image difference characteristic value ΔK2 = "visible") clearly indicated that the unit digit of the equivalent meter reading after the second irradiation was "0", that is, the segment value was "a = 1, b = 1, c = 1, d = 1, e = 1, f = 1, g = 0, K1 = 1, K2 = 1";

[0398] From the image difference characteristic values ​​Δb = visible, Δe = visible, ΔK1 = visible, ΔK2 = visible, and Δg = hidden in (22B), the segment value of the unit digit of the equivalent meter reading after the first irradiation is "a = 1, b = 0, c = 1, d = 1, e = 0, f = 1, g = 1, K1 = 0, K2 = 0"; using the conventional seven-segment digital character encoding-decoding rules, the unit digit of the equivalent meter reading after the first irradiation is "5";

[0399] From the difference feature value in (22A), only Δb = hidden, and the rest remain unchanged. Therefore, the value of segment b of the unit digit of the equivalent meter reading before irradiation is "1", and the segment value is "a=1,b=1,c=1,d=1,e=0,f=1,g=1,K1=0,K2=0". Using the conventional seven-segment digital character encoding and decoding rules, the unit digit of the equivalent meter reading before irradiation is "9" (consistent with visual observation).

[0400] Similarly: From the difference feature values ​​in (22C), only G = visible and Δb = hidden, ΔK1 = hidden and ΔK2 = hidden, then the unit digit of the equivalent meter reading after the 3rd irradiation is "a = 1, b = 0, c = 1, d = 1, e = 1, f = 1, g = 1, K1 = 0, K2 = 0"; using the conventional seven-segment digital character encoding-decoding rules, the unit digit of the equivalent meter reading after the 3rd irradiation is "6";

[0401] From the image difference characteristic values ​​ΔK1 = visible and ΔK2 = blanking in (22D), we can directly obtain that the units digit of the equivalent meter reading after 4 irradiations is "1". At the same time, there are also redundant image difference characteristic value information Δa = blanking, Δb = visible, Δd = blanking, Δe = blanking, Δf = blanking, Δg = blanking; C = unchanged, and the c segment value is 1 from (22C).

[0402] From (22E), the image difference characteristic values ​​are Δa = visible, Δb = hidden, Δd = visible, Δe = visible, Δf = visible, Δg = visible; ΔK1 = hidden; C = unchanged. From (22C), the value of segment c is quoted as 1. The segment values ​​are "a = 1, b = 0, c = 1, d = 1, e = 1, f = 1, g = 1, K1 = 0, K2 = 0". Using the conventional seven-segment digital character encoding-decoding rules, the units digit of the equivalent meter reading after the 5th irradiation is "6" (which is inconsistent with the visual result "94").

[0403] Example SSL4-2

[0404] The above process can also classify the character series of digital display instrument indication value patterns into "Y63 represents the character series composed of character patterns in the same digital display instrument indication value". Template characters are created for this character series, and the difference between two template character patterns is used as a standard template and described by template geometric features and template feature quantities. Specifically, the template geometric features are represented by "a,b,c,d,e,f,g,K1,K2". Some template geometric features and template feature quantities are shown in Table 1. In the table, "*" means "not needed".

[0405] The geometric features and feature values ​​of the individual digits are matched with the geometric features and feature values ​​of the template, specifically by looking up Table 1, to obtain the corresponding digits in the indicator value pattern.

[0406] Table 1 (Partial) Logical Encoding Table of Template Geometric Features and Template Feature Quantities

[0407]

[0408]

[0409] Example SSL4-3

[0410] By combining the methods described in Examples SSL4-1 and SSL4-2, the corresponding numbers in the indicator value pattern can be obtained.

[0411] In this embodiment, the grayscale of the digit character pattern in the six images is blended with the grayscale of the background interference, making it very difficult to segment the digit character pattern using a threshold. Therefore, a schematic diagram of the R+G+B value characteristics of the units digit of the equivalent meter reading after the fifth irradiation, which was manually identified as "94" by visual inspection, is presented. Figure 27 As shown in the diagram, the R+G+B value characteristics of the units digit of the equivalence meter reading after the fourth irradiation are illustrated. Figure 26 As shown, Figure 26 and 27 In the middle: R+G+B values ​​below 210 are marked with a box (equivalent to background imaging interference on the instrument display window), and R+G+B values ​​between 210 and 235 are marked with an underline, mainly representing the grayscale of character patterns.

[0412] Repeat the above process for each digit in the digital display instrument's indicator value pattern to obtain all the characters of the indicator value.

[0413] This embodiment also reveals that: when a camera is used as an image acquisition device during testing / detection / calibration / verification, the alternating appearance of visible and hidden elements causes the alternating appearance of feature class 1 and feature class -1 on the feature micro-component VS sequence curve, thereby tracking the character feature value (pen segment value) of the number in the digital display instrument indication value pattern in real time, and thus realizing the numerical recognition of the digital display instrument indication value pattern in real time.

[0414] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for processing the pattern of indicated values ​​of a digital display instrument, characterized in that... Includes the following steps: S1, Use an image acquisition device to acquire a set of images containing a pattern of digital display instrument indication values, wherein the set of images includes at least 2 images, and the sequence number of the images arranged in the acquisition order is called the image number; S2, calculate the primitive feature quantity of the image, wherein the primitive feature quantity is a parameter describing the primitive feature, including the optical feature quantity of the pixels contained in the primitive, wherein the primitive is the name of the set of pixels, and the optical feature quantity of the pixel is a function composed of RGB. S3, calculate the graph difference, which is the difference in primitive feature quantities at the same position in two images; S4. Classify the map difference, with at least three categories. Specifically, the three categories are represented by feature class 0, feature class 1, and feature class -1, where the map difference value corresponding to the feature class -1 is negative, and the map difference value corresponding to the feature class 1 is positive. The map difference classification specifically includes any one, two, or all of Y41A, Y41B, and Y41C. Specifically, Y41A is marked as feature class 0 when the map difference value is within the range of the first positive threshold pT1 and the first negative threshold nT1, where nT1≤0 and pT1≥0; Y41B is marked as feature class 1 when the map difference value is not lower than the second positive threshold pT2, where pT2≥pT1; and Y41C is marked as feature class -1 when the map difference value does not exceed the second negative threshold nT2, where nT2≤nT1. S5, segment the map difference according to the individual numbers in the indicator value pattern; S6, obtain the geometric features and feature values ​​of the map difference for a single number, wherein the feature values ​​include at least three possible values.

2. The method according to claim 1, characterized in that... This includes arranging feature micro-components in image order to form a feature micro-component VS image order curve, and marking the feature micro-components between feature class 1 and feature class -1 on the curve as feature class 0. The feature micro-components are the differences in primitive feature quantities at the same position in two adjacent images.

3. The method according to claim 1, characterized in that... This includes enhancing the graph difference for non-zero feature classes, wherein the enhancement process includes increasing the absolute value of the graph difference.

4. The method according to claim 1, characterized in that... It also includes any one, any two, or all of the processing methods from Y43A, Y43B, and Y43C; among which, Y43A performs dilation processing based on feature class; Y43B performs shrinkage processing based on feature class; Y43C removes isolated noise from feature classes in the image difference.

5. The method according to claim 1, characterized in that... Step S4 further includes classifying the map differences using evaluation metrics of adjacent map differences. These evaluation metrics include any one, any two, any three, any four, any five, any six, or all of Y45A, Y45B, Y45C, Y45D, Y45E, Y45F, and Y45G. Y45A uses the average difference between adjacent maps as the evaluation metric; Y45B uses the median difference between adjacent maps as the evaluation metric; Y45C uses the moving smoothing result of adjacent map differences as the evaluation metric; Y45D uses the difference between adjacent maps as a function with weighting coefficients as the evaluation metric; Y45E uses the distance between adjacent map differences as a function with weighting coefficients as the evaluation metric; Y45F uses the number of identical feature classes in adjacent map differences as the evaluation metric; Y45G uses the sign of the difference between adjacent graph values ​​as an evaluation metric.

6. The method according to claim 5, characterized in that... The specific meaning of "adjacent map difference" includes any one, any two, or all of Y45X, Y45Y, and Y45Z. Y45X represents adjacent coordinates in the spatial domain; Y45Y represents adjacent graph sequences in the time domain; Y45Z is the difference in primitive features at the same location in two sequentially adjacent images in the time domain.

7. The method according to claim 1, characterized in that... The character geometric features and character feature values ​​of the indicator value are obtained using the geometric features and feature values ​​of the image difference, and then the numbers corresponding to the indicator value pattern are obtained based on the character geometric features and character feature values.

8. The method according to claim 1, characterized in that... The character series of the digital display instrument indication value pattern is classified into any one of the Y61, Y62, and Y63 character series. Template characters are created for this character series. The difference between two template character patterns is used as a standard template and described by template geometric features and template feature quantities. The geometric features and feature values ​​of the difference obtained in S6 are matched with the geometric features and feature quantities of the template to obtain the corresponding numbers in the indication value pattern. Y61 represents a character series composed of seven segments of digits 0 to 9. The pattern of characters 0 to 9 can be equivalently divided into seven segments. The same segment has the same optical features at the same time. The display of characters 0 to 9 can be achieved by changing the combination of optical features of the segments. Y62 represents a character series consisting of Arabic numerals 0 to 9, wherein the patterns of characters 0 to 9 are Arabic numerals; Y63 represents a character series composed of character patterns in the same digital display instrument indication value.

Images