Industrial Internet data management method and system based on artificial intelligence

Abstract

The present invention relates to the field of artificial intelligence technology, and in particular to an industrial Internet data management method and system based on artificial intelligence. The method comprises: obtaining medicine bottle information, determining an image to be analyzed corresponding to each medicine bottle according to an angle influence coefficient and a proportion of inclined text; determining a label state of the image to be analyzed according to a text complexity coefficient and a label deviation value; determining a region division method according to the label state to determine a division optimization method or a combined division according to a character distribution reference value; determining a label information analysis method according to a label fuzzy state to determine a label information qualification degree according to text similarity or to determine an analysis optimization method according to the number of affected character groups and a coefficient of affected character group distribution; determining whether the quality of the medicine bottle is qualified according to a color difference and a label information qualification degree; the present invention can improve the recognition effect of drug labels, thereby improving the detection qualification degree of drugs.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an industrial Internet data management method and system based on artificial intelligence. Background Art

[0002] Industrial Internet is a concept that deeply integrates information technologies such as the Internet, Internet of Things, big data, cloud computing, and artificial intelligence with industrial scenarios. It aims to collect, exchange, analyze, and optimize industrial data by connecting machines, equipment, sensors, and personnel. When pharmaceutical production is combined with the Industrial Internet, the Industrial Internet can be effectively combined with various mechanical equipment to improve industrial production efficiency and reduce manpower use. When applied to pharmaceutical sorting, it is necessary to perform visual extraction of the character information on the pharmaceutical packaging to match and record the information. Due to the uneven quality of pharmaceutical packaging, there are problems with text serialization and uneven character distribution. It is difficult to recognize the information characters of pharmaceuticals, which makes it impossible to meet actual use needs. Therefore, how to improve the recognition effect of character information on pharmaceutical packaging and thereby improve the production efficiency of the pharmaceutical industry is a technical problem that needs to be urgently solved by technicians in this field.

[0003] Chinese patent publication number CN117911668A discloses a drug information recognition method and device, including: obtaining initial medicine bottle image information taken from multiple angles; preprocessing the initial medicine bottle image information to obtain reference medicine bottle image information; using a scale-invariant feature conversion strategy to perform image stitching on the reference medicine bottle image information to obtain target medicine bottle image information; performing character recognition on the target medicine bottle image information to obtain a medicine bottle code; querying a preset medicine information archive according to the medicine bottle code to obtain drug information, obtaining initial medicine bottle image information by taking pictures from multiple angles, and thus identifying the initial medicine bottle image information. It can be seen that the above technical solution has the following problems: it does not take into account the complexity of the medicine bottle character information, does not adaptively adjust the character area recognition method according to the actual situation, and leads to poor label information recognition effect when the text is serialized and the characters are unevenly distributed, which in turn leads to poor drug detection effect. Summary of the invention

[0004] To this end, the present invention provides an industrial Internet data management method and system based on artificial intelligence to overcome the problem that the prior art does not take into account the complexity of the character information of the medicine bottle, does not adaptively adjust the character area recognition method according to the actual situation, and leads to poor label information recognition effect when the text is serialized and the characters are unevenly distributed, which in turn leads to poor drug detection effect.

[0005] To achieve the above objectives, the present invention provides an industrial Internet data management method based on artificial intelligence, comprising:

[0006] Obtain the information of the medicine bottles, and determine the image to be analyzed corresponding to each medicine bottle based on the angle influence coefficient and the proportion of tilted text;

[0007] Determine the label status of the image to be analyzed according to the text complexity coefficient and the label deviation value;

[0008] Determining the area division method according to the label status is to determine the division optimization method or combined division according to the character distribution reference value, and the division optimization method is the associated impact area division or the associated character division;

[0009] Determine the label information analysis method according to the label fuzzy state, determine the label information qualification according to the text similarity or determine the analysis optimization method according to the number of affected character groups and the distribution coefficient of affected character groups;

[0010] The analysis and optimization method is to determine the eligibility of the label information according to the comprehensive evaluation value or to determine the eligibility of the label information according to the character group matching degree and fuzzy sensitivity coefficient of the replacement label information;

[0011] The quality of the medicine bottle is determined based on the color difference and the conformity of the label information.

[0012] Furthermore, for a number of label images corresponding to a medicine bottle, any label image whose angle influence coefficient is less than a preset angle influence coefficient and whose tilted text ratio is less than a preset tilted text ratio is selected and recorded as the image to be analyzed.

[0013] Furthermore, the label status is determined according to the text complexity coefficient and the label deviation value, and the label status includes:

[0014] The first label state in which the character complexity coefficient is greater than or equal to the preset character complexity coefficient or the label deviation value is greater than or equal to the preset label deviation value;

[0015] A second label state in which the character complexity coefficient is less than a preset character complexity coefficient and the label deviation value is less than a preset label deviation value.

[0016] Further, the area division method is determined according to the tag status;

[0017] In the first label state, the area division method is to determine the division optimization method according to the character distribution reference value;

[0018] In the second label state, the area division method is combined division.

[0019] The combination division includes:

[0020] For each character line in the image to be analyzed, a combination analysis is performed in the order from near to far from the bottle cap of the medicine bottle. When a single character line is combined and analyzed, the character line is recorded as a target character line, and other character lines other than the target character line that are not recorded in the character combination are recorded as reference character lines;

[0021] Recording a set of reference character lines and target character lines located above the difference reference character line as a character combination;

[0022] And continue to perform combination analysis on the character rows that are not recorded as character combinations until all character rows are recorded as character combinations;

[0023] The boxed area corresponding to each character combination is used as the label information area;

[0024] The difference between the difference reference character line and the target character line is greater than the preset character difference and is the reference character line closest to the bottle cap of the medicine bottle;

[0025] The framed area is the smallest rectangle that can contain all characters corresponding to a single character combination.

[0026] Further, determining a division optimization method according to the character distribution reference value;

[0027] If the character distribution reference value is greater than or equal to the preset character distribution reference value, the division optimization method is the associated impact area division;

[0028] If the character distribution reference value is less than the preset character distribution reference value, the division optimization method is the associated character division.

[0029] Among them, similarity detection is performed on each character according to the reading order. When similarity detection is performed on a single character that is not included in the frame selection area, the character is recorded as the target character, and the characters with the same font and font size as the target character and the target character are recorded as associated characters. The minimum rectangle that can contain each associated character is recorded as the frame selection area.

[0030] And continue to perform similarity detection on the characters not recorded in the framed area until all characters are recorded in the framed area, and each framed area is recorded as a label information area.

[0031] Furthermore, the association impact area division includes:

[0032] Identify the impact area corresponding to each character line, and perform association analysis on each impact area in the order from near to far from the bottle cap. When performing association analysis on a single impact area, record the impact area as the target impact area, and record other impact areas outside the target impact area that are not recorded in the box selection area as reference impact areas;

[0033] If there is no missing text paragraph in the target impact area, the target impact area is recorded as the framed area;

[0034] If there are missing text paragraphs in the target influence area, the reference influence areas located above the reference influence area without missing text paragraphs and the target influence area are recorded as associated influence areas, and the minimum rectangle that can contain each associated influence area is recorded as the frame selection area;

[0035] And continue to perform association analysis on the impact areas that are not recorded in the framed area until all the impact areas are recorded in the framed area, and each framed area is recorded as a label information area;

[0036] The affected area is a minimum rectangle that can contain all characters corresponding to a single character row and has one side with the same length as the reference area;

[0037] The reference area is a rectangular area with the smallest area that can contain all characters in a tag information.

[0038] Further, a tag information analysis method is determined according to the tag fuzzy state;

[0039] In the first label fuzzy state, the label information analysis method is to determine the label information qualification according to the text similarity, wherein the label information qualification is positively correlated with the text similarity.

[0040] In the second label fuzzy state, the label information analysis method is to determine the analysis optimization method according to the number of affected character groups and the distribution coefficient of affected character groups;

[0041] The tag fuzzy state includes a first tag fuzzy state in which no fuzzy character group exists in the tag information, and a second tag fuzzy state in which an fuzzy character group exists in the tag information.

[0042] Further, the analysis and optimization method is determined according to the number of affected character groups and the distribution coefficient of affected character groups;

[0043] If the number of affected character groups is greater than or equal to the preset number of affected character groups or the distribution coefficient of affected character groups is less than the preset distribution coefficient of affected character groups, the analysis optimization method is to determine the label information qualification according to the comprehensive evaluation value, wherein the label information qualification is negatively correlated with the comprehensive evaluation value;

[0044] If the number of affected character groups is less than the preset number of affected character groups and the affected character group distribution coefficient is greater than or equal to the preset affected character group distribution coefficient, the analysis optimization method is to determine the label information qualification according to the character group matching degree and fuzzy sensitivity coefficient of the replacement label information;

[0045] The influencing character group is an influencing character group whose replaceable difficulty is greater than a preset replaceable difficulty.

[0046] Further, whether the quality of the medicine bottle is qualified is determined based on the color difference and the qualification of the label information;

[0047] If the color difference is greater than or equal to the preset color difference or the label information qualification is less than the preset label information qualification, the quality of the medicine bottle is unqualified;

[0048] If the color difference is less than the preset color difference and the label information qualification is greater than or equal to the preset label information qualification, the quality of the medicine bottle is qualified.

[0049] The present invention also provides an industrial Internet data management system based on artificial intelligence, comprising:

[0050] Data collection module, used to obtain medicine bottle information;

[0051] An image selection module, which is connected to the data acquisition module and is used to determine the image to be analyzed corresponding to each medicine bottle according to the angle influence coefficient and the proportion of inclined text;

[0052] A mode judgment module, which is connected to the data acquisition module and the image selection module respectively, and is used to determine the label state of the image to be analyzed according to the text complexity coefficient and the label deviation value, and determine the area division mode according to the label state, which is to determine the division optimization mode or the combined division according to the character distribution reference value, and the division optimization mode is the associated influence area division or the associated character division;

[0053] A quality judgment module, which is connected to the mode judgment module, and is used to determine the label information analysis mode according to the label fuzzy state, which is to determine the label information qualification according to the text similarity or to determine the analysis optimization mode according to the number of affected character groups and the distribution coefficient of affected character groups, and the analysis optimization mode is to determine the label information qualification according to the comprehensive evaluation value or to determine the label information qualification according to the character group matching degree and fuzzy sensitivity coefficient of the replaced label information;

[0054] The qualified judgment module is connected to the data acquisition module and the quality judgment module respectively, and is used to determine whether the quality of the medicine bottle is qualified according to the color difference and the qualified degree of the label information.

[0055] Compared with the prior art, the beneficial effect of the present invention lies in that, in the technical scheme of the present invention, the degree of missing characters and the degree of text inclination in each label image are effectively reflected through the angle influence coefficient and the proportion of tilted text, and then the image to be analyzed corresponding to each medicine bottle is determined according to the angle influence coefficient and the proportion of tilted text, which can improve the recognition effect of characters, avoid the problem of inaccurate text recognition effect caused by poor quality of label image shooting, improve the recognition effect of drug labels, and thus improve the qualified degree of drug inspection.

[0056] Furthermore, in the present invention, the label status of the image to be analyzed is determined according to the text complexity coefficient and the label deviation value, and the text complexity coefficient and the label deviation value are used to effectively reflect the degree of text confusion and seriality, and then different area division methods are adaptively selected according to the label status, so that the selection of area division method is more in line with the actual application scenario, avoiding the defects of the prior art that the complexity of the medicine bottle character information is not considered, the character area is not divided, and the label information recognition effect is poor when the text is serialized and the characters are unevenly distributed, thereby improving the accuracy of label recognition, and then improving the accuracy of drug quality detection.

[0057] Furthermore, the present invention effectively reflects the degree of difference in distribution of different types of characters through character distribution reference values, and then adaptively selects different division optimization methods according to the character distribution reference values, so that the selection of division optimization method is more in line with the actual application scenario, avoiding the defect of poor character information recognition effect due to large differences in character distribution, and improving the accuracy of drug detection.

[0058] Furthermore, in the present invention, a label information analysis method is determined according to the label fuzzy state, and the label fuzzy state is used to effectively reflect whether there are ambiguous characters in the label information. Then, different label information analysis methods are adaptively selected according to different label fuzzy states, so that the selection of label information analysis method is more in line with the actual application scenario, avoiding the defect of low accuracy in determining the label information qualification due to the presence of ambiguous characters in the label information, and improving the accuracy of drug detection.

[0059] Furthermore, the present invention effectively reflects the difficulty of replacing fuzzy character groups by influencing the number of character groups and the coefficient of influencing the distribution of character groups, and then adaptively selects different analysis optimization methods according to the number of influencing character groups and the coefficient of influencing the distribution of character groups, so that the selection of analysis optimization methods is more in line with the actual application scenario, avoiding the defect of low accuracy in determining the qualification of label information due to the large degree of influence of fuzzy character groups in label information, thereby improving the accuracy of drug detection.

[0060] Furthermore, in the present invention, whether the quality of the medicine bottle is qualified is determined based on the color difference and the qualification of the label information. The color difference and the qualification of the label information can effectively reflect the qualification degree of the medicine, which can ensure the accuracy and consistency of the medicine packaging, improve the accuracy of the quality inspection of the medicine, and further optimize the production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a schematic diagram of the industrial Internet data management method based on artificial intelligence of the present invention;

[0062] Figure 2This is a flow chart of the present invention for determining a region division method according to a tag state;

[0063] Figure 3 This is a flow chart of the present invention for determining a tag information analysis method according to a tag fuzzy state;

[0064] Figure 4 This is a module connection diagram of the industrial Internet data management system based on artificial intelligence of the present invention. DETAILED DESCRIPTION

[0065] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0066] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0067] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0068] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0069] See also Figures 1 to 3 As shown, the present invention provides an industrial Internet data management method based on artificial intelligence, comprising:

[0070] Obtain the information of the medicine bottles, and determine the image to be analyzed corresponding to each medicine bottle based on the angle influence coefficient and the proportion of tilted text;

[0071] Determine the label status of the image to be analyzed according to the text complexity coefficient and the label deviation value;

[0072] Determining the area division method according to the label status is to determine the division optimization method or combined division according to the character distribution reference value, and the division optimization method is the associated impact area division or the associated character division;

[0073] Determine the label information analysis method according to the label fuzzy state, determine the label information qualification according to the text similarity or determine the analysis optimization method according to the number of affected character groups and the distribution coefficient of affected character groups;

[0074] The analysis and optimization method is to determine the eligibility of the label information according to the comprehensive evaluation value or to determine the eligibility of the label information according to the character group matching degree and fuzzy sensitivity coefficient of the replacement label information;

[0075] The quality of the medicine bottle is determined based on the color difference and the conformity of the label information.

[0076] The application scenario of the present invention is the quality inspection of medicine bottle packaging when combining pharmaceutical production with the industrial Internet. The medicine bottle information in the present invention includes several label images corresponding to each medicine bottle. The label image is an image of a complete medicine bottle with a medicine bottle label. In the present invention, several historical records are correspondingly set. Any historical record records the angle influence coefficient, the proportion of inclined text, the text complexity coefficient, the label deviation value and the character difference value in at least one historical process of medicine bottle quality inspection, and each historical record corresponds to a qualified mark, which records whether the quality inspection result of the medicine bottle meets user requirements, and the qualified mark can be recorded manually.

[0077] Specifically, for several label images corresponding to a medicine bottle, any label image whose angle influence coefficient is less than a preset angle influence coefficient and whose tilted text ratio is less than a preset tilted text ratio is selected and recorded as the image to be analyzed.

[0078] Among them, for a single label image, the angle influence coefficient = (area of ​​the medicine bottle area in the label image - area of ​​the text area in the label image) / area of ​​the medicine bottle area in the label image, the text area is the area of ​​the smallest rectangle that can contain all characters in a label image, the medicine bottle area is the area of ​​the smallest rectangle that can contain the entire medicine bottle in a label image, the tilted text ratio = tilted text area in the label image / text area in the label image, the tilted text area is the sum of the reference areas corresponding to each tilted text, the reference area is the area of ​​the smallest rectangle that can contain a single tilted text, the tilted text is caused by the tilt angle between the camera lens and the label plane or the camera calibration problem, etc., the tilted text can be identified by machine vision and deep learning networks, which is easy for technicians in this field to understand and will not be elaborated on.

[0079] The values ​​of the preset angle influence coefficient and the preset tilted text ratio can be determined by the user according to the actual application scenario. The higher the user's demand for the quality inspection effect of the medicine bottle, the smaller the values ​​of the preset angle influence coefficient and the preset tilted text ratio are. A value of a preset angle influence coefficient and a preset tilted text ratio is provided, and the average value of the angle influence coefficient corresponding to the historical records that can meet the user's needs is recorded as the preset angle influence coefficient, and the average value of the tilted text ratio corresponding to the historical records that can meet the user's needs is recorded as the preset tilted text ratio.

[0080] Specifically, the label status is determined according to the text complexity coefficient and the label deviation value. The label status includes:

[0081] The first label state in which the character complexity coefficient is greater than or equal to the preset character complexity coefficient or the label deviation value is greater than or equal to the preset label deviation value;

[0082] A second label state in which the character complexity coefficient is less than a preset character complexity coefficient and the label deviation value is less than a preset label deviation value.

[0083] Wherein, the text complexity coefficient = the number of glyphs + the font influence value, the number of glyphs is the total number of glyphs in an image to be analyzed, the glyphs include fonts of several different sizes, including but not limited to size 4, size 4 and size 3, the fonts include but not limited to bold, regular and Song, the font influence value is the number of overlapping glyph areas, the glyph area is the smallest rectangle that can contain all characters corresponding to the same glyph, the label deviation value is the number of influence areas with missing text paragraphs, the missing text paragraphs are the incomplete character lines in an influence area, and the character lines are all the characters in the same line;

[0084] The values ​​of the preset text complexity coefficient and the preset label deviation value can be determined by the user according to the actual application scenario. The higher the user's demand for the quality detection effect of the medicine bottle, the smaller the values ​​of the preset text complexity coefficient and the preset label deviation value are. Provide a value of the preset text complexity coefficient and the preset label deviation value. The detection area is divided into historical records of combined division. The average value of the text complexity coefficient corresponding to the historical records that can meet the user's needs is recorded as the preset text complexity coefficient, and the average value of the label deviation values ​​corresponding to the historical records that can meet the user's needs is recorded as the preset label deviation value.

[0085] Specifically, the region division method is determined according to the label status;

[0086] In the first label state, the area division method is to determine the division optimization method according to the character distribution reference value;

[0087] In the second label state, the area division method is combined division.

[0088] The combination division includes:

[0089] For each character line in the image to be analyzed, a combination analysis is performed in the order from near to far from the bottle cap of the medicine bottle. When a single character line is combined and analyzed, the character line is recorded as a target character line, and other character lines other than the target character line that are not recorded in the character combination are recorded as reference character lines;

[0090] Recording a set of reference character lines and target character lines located above the difference reference character line as a character combination;

[0091] And continue to perform combination analysis on the character rows that are not recorded as character combinations until all character rows are recorded as character combinations;

[0092] The boxed area corresponding to each character combination is used as the label information area;

[0093] The difference between the difference reference character line and the target character line is greater than the preset character difference and is the reference character line closest to the bottle cap of the medicine bottle;

[0094] The framed area is the smallest rectangle that can contain all characters corresponding to a single character combination.

[0095] Among them, the character difference is the absolute value of the difference in the number of characters corresponding to the two character rows. The value of the preset character difference can be determined by the user according to the actual application scenario. The higher the user's demand for the quality inspection effect of the medicine bottle, the smaller the value of the preset character difference is. A preset character difference is provided, and the average value of the character difference corresponding to the historical records that can meet the user's needs is recorded as the preset character difference;

[0096] The confirmation method above is to record the minimum rectangle that can contain the difference character row as the reference minimum rectangle, and the straight line where the longest side of the reference minimum rectangle is farthest from the medicine bottle cap is the baseline, the top is above the baseline, and the long side is the longer side of the reference minimum rectangle.

[0097] Specifically, the division optimization method is determined according to the character distribution reference value;

[0098] If the character distribution reference value is greater than or equal to the preset character distribution reference value, the division optimization method is the associated impact area division;

[0099] If the character distribution reference value is less than the preset character distribution reference value, the division optimization method is the associated character division.

[0100] Among them, similarity detection is performed on each character according to the reading order. When similarity detection is performed on a single character that is not included in the frame selection area, the character is recorded as the target character, and the characters with the same font and font size as the target character and the target character are recorded as associated characters. The minimum rectangle that can contain each associated character is recorded as the frame selection area.

[0101] And continue to perform similarity detection on the characters not recorded in the framed area until all characters are recorded in the framed area, and each framed area is recorded as a label information area.

[0102] The character distribution reference value is the maximum value of the distribution coefficients corresponding to each glyph. The distribution coefficient is confirmed in the following manner: for a single glyph, each character corresponding to the glyph is recorded as a feature character, and the maximum value of the reference distance corresponding to each feature character is recorded as the distribution coefficient. For a single feature character, the feature character is recorded as a target feature character, and other feature characters other than the target feature character are recorded as reference feature characters. The shortest distance from the target feature character to each reference feature character is detected, and the maximum value of each shortest distance is recorded as the reference distance. The shortest distance corresponding to two feature characters is the length of the line connecting the center points corresponding to the two characters, and the center point is the center of the circumscribed circle of a character.

[0103] The value of the preset character distribution reference value can be determined by the user according to the actual application scenario. The smaller the value of the preset character distribution reference value, the greater the user's demand for associated impact area division. A value of the preset character distribution reference value is provided, and the detection division optimization method is the historical record of associated character division. The average value of the character distribution reference value corresponding to the historical record that can meet the user's needs is recorded as the preset character distribution reference value.

[0104] Specifically, the association impact area division includes:

[0105] Identify the impact area corresponding to each character line, and perform association analysis on each impact area in the order from near to far from the bottle cap. When performing association analysis on a single impact area, record the impact area as the target impact area, and record other impact areas outside the target impact area that are not recorded in the box selection area as reference impact areas;

[0106] If there is no missing text paragraph in the target impact area, the target impact area is recorded as the framed area;

[0107] If there are missing text paragraphs in the target influence area, the reference influence areas located above the reference influence area without missing text paragraphs and the target influence area are recorded as associated influence areas, and the minimum rectangle that can contain each associated influence area is recorded as the frame selection area;

[0108] And continue to perform association analysis on the impact areas that are not recorded in the framed area until all the impact areas are recorded in the framed area, and each framed area is recorded as a label information area;

[0109] The affected area is a minimum rectangle that can contain all characters corresponding to a single character row and has one side with the same length as the reference area;

[0110] The reference area is a rectangular area with the smallest area that can contain all characters in a tag information.

[0111] Specifically, the label information analysis method is determined according to the label fuzzy state;

[0112] In the first label fuzzy state, the label information analysis method is to determine the label information qualification according to the text similarity, wherein the label information qualification is positively correlated with the text similarity.

[0113] In the second label fuzzy state, the label information analysis method is to determine the analysis optimization method according to the number of affected character groups and the distribution coefficient of affected character groups;

[0114] The tag fuzzy state includes a first tag fuzzy state in which no fuzzy character group exists in the tag information, and a second tag fuzzy state in which an fuzzy character group exists in the tag information.

[0115] Among them, the label information contains several influencing character groups, and a single influencing character group is a combination of characters corresponding to a single keyword in the label information; the fuzzy character group is a character group that is blurred and cannot be recognized due to stains and other reasons. The present invention is provided with a standard label library, and the standard label library has several standard labels, and the standard label is a text with several character groups; the text similarity is confirmed in a way that the text corresponding to the label information is detected with the text matching degree of each standard label in the standard label library, and the maximum value of each text matching degree is recorded as the text similarity, and the text matching degree is confirmed in a way that each character group in the text corresponding to the label information is converted into a text vector, and the present invention adopts TF-IDF technology to convert the text vector; for a single label information, the cosine value of the angle between the reference vector corresponding to the label information and each standard vector is calculated, and the text similarity is the maximum value of the cosine value of each angle; the reference vector is the sum of the text vectors corresponding to each character group in a label information, and the standard vector is the sum of the text vectors corresponding to each character group in a single standard label.

[0116] Specifically, the analysis and optimization method is determined according to the number of affected character groups and the distribution coefficient of affected character groups;

[0117] If the number of affected character groups is greater than or equal to the preset number of affected character groups or the distribution coefficient of affected character groups is less than the preset distribution coefficient of affected character groups, the analysis optimization method is to determine the label information qualification according to the comprehensive evaluation value, wherein the label information qualification is negatively correlated with the comprehensive evaluation value;

[0118] If the number of affected character groups is less than the preset number of affected character groups and the affected character group distribution coefficient is greater than or equal to the preset affected character group distribution coefficient, the analysis optimization method is to determine the label information qualification according to the character group matching degree and fuzzy sensitivity coefficient of the replacement label information;

[0119] The influencing character group is an influencing character group whose replaceable difficulty is greater than a preset replaceable difficulty.

[0120] The present invention includes a first analysis condition and a second analysis condition. The first analysis condition is that the number of affected character groups is greater than or equal to the preset number of affected character groups or the distribution coefficient of affected character groups is less than the preset distribution coefficient of affected character groups. The second analysis condition is that the number of affected character groups is less than the preset number of affected character groups and the distribution coefficient of affected character groups is greater than or equal to the preset distribution coefficient of affected character groups. Under the second analysis condition, the label information qualification degree = character group matching degree - fuzzy sensitivity coefficient. It should be noted that the label information qualification degree under the second analysis condition is always greater than the label information qualification degree under the first analysis condition.

[0121] The number of affected character groups is the total number of affected character groups, the affected character group distribution coefficient is the average value of the minimum reference distances corresponding to each affected character group, for a single affected character group, the affected character group is recorded as the target affected character group, and other affected character groups outside the target affected character group are recorded as reference affected character groups, the shortest length from the target affected character group to each reference affected character group is detected, and the minimum value of each shortest length is recorded as the minimum reference distance; for two affected character groups, the shortest length corresponding to the two affected character groups is the length of the line connecting the reference points corresponding to the two affected character groups, and the reference point is the center of the circumscribed circle of the first character of an affected character group;

[0122] The values ​​of the preset number of influencing character groups and the preset distribution coefficient of influencing character groups can be determined by the user according to the actual application scenario. The smaller the value of the preset number of influencing character groups and the larger the value of the preset distribution coefficient of influencing character groups, the greater the user's need to determine the qualification of label information according to the comprehensive evaluation value. A value of the preset number of influencing character groups and the preset distribution coefficient of influencing character groups is provided, and the historical records of determining the qualification of label information according to the comprehensive evaluation value are detected, and the average value of the number of influencing character groups corresponding to the historical records that can meet the user's needs is recorded as the preset number of influencing character groups, and the average value of the distribution coefficient of influencing character groups corresponding to the historical records that can meet the user's needs is recorded as the preset distribution coefficient of influencing character groups;

[0123] Comprehensive evaluation value = number of affected character groups + distribution coefficient of affected character groups; the confirmation method of the first and second affecting conditions for replaceable difficulty is different. For an affecting character group in a tag information, the affecting character group is recorded as the first target affecting character group. The first affecting condition is that there is an ambiguous character group in the adjacent character group corresponding to the first target affecting character group, and the second affecting condition is that there is no ambiguous character group in the adjacent character group corresponding to the first target affecting character group. The adjacent character group is a character group that is located in the same text line as the first target affecting character group and has no other character group between it and the first target affecting character group.

[0124] Under the first influencing condition, the replaceable difficulty is 1; under the second influencing condition, the replaceable difficulty is positively correlated with the number of pre-replacement character groups. It should be noted that under the second influencing condition, the replaceable difficulty is greater than 1; the confirmation method of the pre-replacement character group is to record the character group located between and adjacent to two adjacent character groups and having a number of appearances greater than a preset number of appearances in the historical records as the pre-replacement character group. The value of the preset number of appearances can be determined by the user according to the actual application scenario, and a value of the preset number of appearances is provided, and the preset number of appearances is 20 times; the character group with the largest number of appearances in the historical records and located between and adjacent to two adjacent character groups is recorded as the replacement character group; replacing the label information is the text after each influencing character group in a label information is replaced with the replacement character group;

[0125] The character group matching degree is to detect the unit matching degree between the text corresponding to the replacement label information and each standard label in the standard label library, and the maximum value of each unit matching degree is recorded as the character group matching degree. The unit matching degree is confirmed in the following way: for any standard label, the standard label is recorded as the target standard label, the number of the same character groups in the text corresponding to the replacement label information and the text corresponding to the target standard label is recorded as A, and the number of character groups corresponding to the target standard label is recorded as B, and the unit matching degree = A / B; the fuzzy sensitivity coefficient is the sum of the number of pre-replacement character groups corresponding to each influencing character group in a label information;

[0126] The value of the preset replaceable difficulty can be determined by the user according to the actual application scenario. The higher the user's demand for the quality inspection effect of the medicine bottle, the smaller the value of the preset replaceable difficulty. A value of the preset replaceable difficulty is provided, and the average value of the replaceable difficulty corresponding to the historical records that can meet the user's needs is recorded as the preset replaceable difficulty.

[0127] Specifically, the quality of the medicine bottle is determined based on the degree of color difference and the degree of conformity of the label information;

[0128] If the color difference is greater than or equal to the preset color difference or the label information qualification is less than the preset label information qualification, the quality of the medicine bottle is unqualified;

[0129] If the color difference is less than the preset color difference and the label information qualification is greater than or equal to the preset label information qualification, the quality of the medicine bottle is qualified.

[0130] The color difference is the difference between the color of the bottle cap and the bottle body, which can be measured by a colorimeter. This is easy to understand for those skilled in the art and will not be described in detail.

[0131] The values ​​of the preset color difference and the preset label information qualification can be determined by the user according to the actual application scenario. The higher the user's demand for the quality inspection effect of the medicine bottle, the smaller the value of the preset color difference and the larger the value of the preset label information qualification. The values ​​of the preset color difference and the preset label information qualification are provided, and the historical records of qualified medicine bottle quality are inspected. The average value of the color difference corresponding to the historical records that can meet the user's needs is recorded as the preset color difference, and the average value of the label information qualification corresponding to the historical records that can meet the user's needs is recorded as the preset label information qualification.

[0132] See also Figure 4 As shown, it is a module connection diagram of the industrial Internet data management system based on artificial intelligence of the present invention. The present invention also provides an industrial Internet data management system based on artificial intelligence, including:

[0133] Data collection module, used to obtain medicine bottle information;

[0134] An image selection module, which is connected to the data acquisition module and is used to determine the image to be analyzed corresponding to each medicine bottle according to the angle influence coefficient and the proportion of inclined text;

[0135] A mode judgment module, which is connected to the data acquisition module and the image selection module respectively, and is used to determine the label state of the image to be analyzed according to the text complexity coefficient and the label deviation value, and determine the area division mode according to the label state, which is to determine the division optimization mode or the combined division according to the character distribution reference value, and the division optimization mode is the associated influence area division or the associated character division;

[0136] A quality judgment module, which is connected to the mode judgment module, and is used to determine the label information analysis mode according to the label fuzzy state, which is to determine the label information qualification according to the text similarity or to determine the analysis optimization mode according to the number of affected character groups and the distribution coefficient of affected character groups, and the analysis optimization mode is to determine the label information qualification according to the comprehensive evaluation value or to determine the label information qualification according to the character group matching degree and fuzzy sensitivity coefficient of the replaced label information;

[0137] The qualified judgment module is connected to the data acquisition module and the quality judgment module respectively, and is used to determine whether the quality of the medicine bottle is qualified according to the color difference and the qualified degree of the label information.

[0138] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0139] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An industrial Internet data management method based on artificial intelligence, characterized in that: include: Obtain the information of the medicine bottles, and determine the image to be analyzed corresponding to each medicine bottle based on the angle influence coefficient and the proportion of tilted text; Determine the label status of the image to be analyzed according to the text complexity coefficient and the label deviation value; Determining the area division method according to the label status is to determine the division optimization method or combined division according to the character distribution reference value, and the division optimization method is the associated impact area division or the associated character division; Determine the label information analysis method according to the label fuzzy state, determine the label information qualification according to the text similarity or determine the analysis optimization method according to the number of affected character groups and the distribution coefficient of affected character groups; The analysis and optimization method is to determine the eligibility of the label information according to the comprehensive evaluation value or to determine the eligibility of the label information according to the character group matching degree and fuzzy sensitivity coefficient of the replacement label information; Determine whether the quality of the medicine bottle is qualified based on the color difference and label information qualification; The area division method is determined according to the label status, including: If the label state is the first label state where the text complexity coefficient is greater than or equal to the preset text complexity coefficient or the label deviation value is greater than or equal to the preset label deviation value, the area division method is to determine the division optimization method according to the character distribution reference value; If the label state is the second label state where the text complexity coefficient is less than the preset text complexity coefficient and the label deviation value is less than the preset label deviation value, the area division method is combined division. The combination division includes: For each character line in the image to be analyzed, a combination analysis is performed in the order from near to far from the bottle cap of the medicine bottle. When a single character line is combined and analyzed, the character line is recorded as a target character line, and other character lines other than the target character line that are not recorded in the character combination are recorded as reference character lines; Recording a set of reference character lines and target character lines located above the difference reference character line as a character combination; And continue to perform combination analysis on the character rows that are not recorded as character combinations until all character rows are recorded as character combinations; The boxed area corresponding to each character combination is used as the label information area; The difference between the difference reference character line and the target character line is greater than the preset character difference and is the reference character line closest to the bottle cap of the medicine bottle; The frame selection area is the smallest rectangle that can contain all characters corresponding to a single character combination; Determine the division optimization method based on the character distribution reference value, including: If the character distribution reference value is greater than or equal to the preset character distribution reference value, the division optimization method is the associated impact area division; If the character distribution reference value is less than the preset character distribution reference value, the division optimization method is the associated character division. Among them, similarity detection is performed on each character according to the reading order. When similarity detection is performed on a single character that is not included in the frame selection area, the character is recorded as the target character, and the characters with the same font and font size as the target character and the target character are recorded as associated characters. The minimum rectangle that can contain each associated character is recorded as the frame selection area. Continue to perform similarity detection on the characters not recorded in the framed area until all characters are recorded in the framed area, and record each framed area as a label information area; The character distribution reference value is the maximum value of the distribution coefficients corresponding to each glyph, and the distribution coefficient is confirmed in the following manner: for a single glyph, each character corresponding to the glyph is recorded as a feature character, and the maximum value of the reference distance corresponding to each feature character is recorded as the distribution coefficient; for a single feature character, the feature character is recorded as a target feature character, and other feature characters other than the target feature character are recorded as reference feature characters, and the shortest distance from the target feature character to each reference feature character is detected, and the maximum value of each shortest distance is recorded as the reference distance; The influence character group distribution coefficient is the average value of the minimum reference distances corresponding to each influence character group. For a single influence character group, the influence character group is recorded as the target influence character group, and other influence character groups outside the target influence character group are recorded as reference influence character groups. The shortest length from the target influence character group to each reference influence character group is detected, and the minimum value of each shortest length is recorded as the minimum reference distance. The influencing character group is an influencing character group whose replaceable difficulty is greater than a preset replaceable difficulty; The comprehensive evaluation value = the number of affected character groups + the distribution coefficient of affected character groups; The character group matching degree is the maximum value of the unit matching degrees of the texts corresponding to the standard tags in the standard tag library and the replacement tag information. The unit matching degree is confirmed by recording any standard tag as the target standard tag, recording the number of identical character groups in the text corresponding to the replacement tag information and the text corresponding to the target standard tag as A, and recording the number of character groups corresponding to the target standard tag as B. The unit matching degree = A / B; The fuzzy sensitivity coefficient is the sum of the number of pre-replacement character groups corresponding to each influencing character group in a tag information; The confirmation method of the pre-replacement character group is to record a character group located between two adjacent character groups and adjacent to the two adjacent character groups, which has a number of appearances greater than a preset number of appearances in the historical record, as the pre-replacement character group.

2. The industrial Internet data management method based on artificial intelligence according to claim 1 is characterized in that: For several label images corresponding to a medicine bottle, any label image whose angle influence coefficient is less than the preset angle influence coefficient and whose tilted text ratio is less than the preset tilted text ratio is selected and recorded as the image to be analyzed.

3. The industrial Internet data management method based on artificial intelligence according to claim 1 is characterized in that: The association impact area division includes: Identify the impact area corresponding to each character line, and perform association analysis on each impact area in the order from near to far from the bottle cap. When performing association analysis on a single impact area, record the impact area as the target impact area, and record other impact areas outside the target impact area that are not recorded in the box selection area as reference impact areas; If there is no missing text paragraph in the target impact area, the target impact area is recorded as the framed area; If there are missing text paragraphs in the target influence area, the reference influence areas located above the reference influence area without missing text paragraphs and the target influence area are recorded as associated influence areas, and the minimum rectangle that can contain each associated influence area is recorded as the frame selection area; And continue to perform association analysis on the impact areas that are not recorded in the framed area until all the impact areas are recorded in the framed area, and each framed area is recorded as a label information area; The affected area is a minimum rectangle that can contain all characters corresponding to a single character row and has one side with the same length as the reference area; The reference area is a rectangular area with the smallest area that can contain all characters in a tag information.

4. The industrial Internet data management method based on artificial intelligence according to claim 3 is characterized in that: Determine the label information analysis method according to the label fuzzy state; In the first label fuzzy state, the label information analysis method is to determine the label information qualification according to the text similarity, wherein the label information qualification is positively correlated with the text similarity; In the second label fuzzy state, the label information analysis method is to determine the analysis optimization method according to the number of affected character groups and the distribution coefficient of affected character groups; The tag fuzzy state includes a first tag fuzzy state in which no fuzzy character group exists in the tag information, and a second tag fuzzy state in which an fuzzy character group exists in the tag information.

5. The industrial Internet data management method based on artificial intelligence according to claim 4 is characterized in that: Determine the analysis and optimization method based on the number of affected character groups and the distribution coefficient of affected character groups; If the number of affected character groups is greater than or equal to the preset number of affected character groups or the distribution coefficient of affected character groups is less than the preset distribution coefficient of affected character groups, the analysis optimization method is to determine the label information qualification according to the comprehensive evaluation value, wherein the label information qualification is negatively correlated with the comprehensive evaluation value; If the number of affected character groups is less than the preset number of affected character groups and the affected character group distribution coefficient is greater than or equal to the preset affected character group distribution coefficient, the analysis optimization method is to determine the label information qualification based on the character group matching degree and fuzzy sensitivity coefficient of the replacement label information.

6. The industrial Internet data management method based on artificial intelligence according to claim 5 is characterized in that: Determine whether the quality of the medicine bottle is qualified based on the color difference and label information qualification; If the color difference is greater than or equal to the preset color difference or the label information qualification is less than the preset label information qualification, the quality of the medicine bottle is unqualified; If the color difference is less than the preset color difference and the label information qualification is greater than or equal to the preset label information qualification, the quality of the medicine bottle is qualified.

7. A management system using the artificial intelligence-based industrial Internet data management method according to any one of claims 1 to 6, characterized in that: include: Data collection module, used to obtain medicine bottle information; An image selection module, which is connected to the data acquisition module and is used to determine the image to be analyzed corresponding to each medicine bottle according to the angle influence coefficient and the proportion of inclined text; A mode judgment module, which is connected to the data acquisition module and the image selection module respectively, and is used to determine the label state of the image to be analyzed according to the text complexity coefficient and the label deviation value, and determine the area division mode according to the label state, which is to determine the division optimization mode or the combined division according to the character distribution reference value, and the division optimization mode is the associated influence area division or the associated character division; A quality judgment module, which is connected to the mode judgment module, and is used to determine the label information analysis mode according to the label fuzzy state, which is to determine the label information qualification according to the text similarity or to determine the analysis optimization mode according to the number of affected character groups and the distribution coefficient of affected character groups, and the analysis optimization mode is to determine the label information qualification according to the comprehensive evaluation value or to determine the label information qualification according to the character group matching degree and fuzzy sensitivity coefficient of the replaced label information; The qualified judgment module is connected to the data acquisition module and the quality judgment module respectively, and is used to determine whether the quality of the medicine bottle is qualified according to the color difference and the qualified degree of the label information.

Images