A label position anomaly recognition method and system based on image feature recognition

Abstract

The application discloses a label position abnormality recognition method and system based on image feature recognition, and relates to the field of machine recognition, which comprises the following steps: determining a reference object on an object to which a label is attached; obtaining a sample reference area; obtaining an actual reference area; pairing the sample reference area and the actual reference area corresponding to the same reference object; forming an actual contour function of the label in an actual coordinate system and a sample contour function of the label in a sample coordinate system; calculating a position difference between the sample contour function and the actual contour function, and calculating a critical value of the position difference; when the position difference exceeds the critical value, the label position is recognized as abnormal, otherwise, the label position is recognized as normal. Through the setting of the area acquisition module, the area pairing module, the coordinate establishment module and the contour acquisition module, no matter whether the size of the label in the image is inconsistent or the shooting angle is inconsistent, no interference will be generated to the recognition, so that the abnormality can be accurately recognized.

Description

Technical Field

[0001] This invention relates to the field of machine recognition, specifically to a method and system for identifying label location anomalies based on image feature recognition. Background Technology

[0002] The layout of tags has a crucial impact on positioning accuracy. When deploying tags, the number, location, and density of tags should be determined reasonably based on the actual application scenario. Interference between tags should be avoided to ensure that each tag can be effectively identified by the reader.

[0003] However, image recognition is usually used for label recognition. However, image acquisition is affected by the object to which the label is attached. The label in the acquired image will be different depending on the position of the object. Therefore, the recognition of labels with abnormal positions will be greatly interfered with, affecting the accuracy of recognition. Summary of the Invention

[0004] To address the aforementioned technical problems, this paper provides a method and system for identifying label location anomalies based on image feature recognition. This technical solution resolves the issues raised in the background section.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A method for identifying label location anomalies based on image feature recognition, comprising:

[0007] Acquire at least one sample image of the label being attached, determine the reference objects on the object to which the label is attached, and there are at least three reference objects, which are set around the label.

[0008] The region where the reference object is located is determined in the sample attachment image to obtain the sample reference region;

[0009] Obtain the actual attached image of the label, identify the reference object in the actual attached image, and obtain the actual reference area;

[0010] Pair the sample reference area with the actual reference area corresponding to the same reference object;

[0011] Three actual reference regions with non-collinear centers are selected to construct the actual coordinate system, and the sample coordinate system is constructed using the sample reference regions corresponding to the selected actual reference regions.

[0012] In the actual coordinate system, the actual contour function of the label is formed; in the sample coordinate system, the sample contour function of the label is formed.

[0013] Calculate the positional difference between the sample contour function and the actual contour function, and calculate the critical value of the positional difference;

[0014] If the position difference exceeds a critical value, the label position is identified as abnormal; otherwise, the label position is identified as normal.

[0015] Preferably, determining the reference object on the object to which the label is attached includes the following steps:

[0016] Enlarge the label by focusing on its center to obtain the enlarged label area;

[0017] When the zoomed-in area of ​​a label first covers more than three objects around the label, the covered objects are used as the reference objects.

[0018] Preferably, determining the region where the reference object is located in the sample attachment image to obtain the sample reference region includes the following steps:

[0019] In the sample attachment image, the sample area of ​​the label is determined, and the sample area is magnified with the center of the sample area to obtain the magnified sample area;

[0020] When the sample magnification area first covers more than three objects around the label, the area where the covered objects are locked is identified as the sample reference area.

[0021] Preferably, the step of identifying the reference object in the actual attached image to obtain the actual reference region includes the following steps:

[0022] In the actual attached image, determine the actual area of ​​the label, and then magnify the actual area using the center of the actual area to obtain the magnified actual area;

[0023] When the actual magnified area first covers more than three objects around the labels, the area where the covered objects are locked is identified as the actual reference area.

[0024] Preferably, the process of pairing the sample reference region and the actual reference region corresponding to the same reference object includes the following steps:

[0025] Both the sample reference region and the actual reference region are modeled using a random coordinate system.

[0026] At least one sample point is uniformly selected on the contour line of the sample reference area, and the sample point is fitted to obtain the sample fitting function.

[0027] At least one actual point is uniformly selected on the contour line of the actual reference area, and the actual point is fitted to obtain the actual fitting function.

[0028] The curvature at the sample points is calculated using the sample fitting function and the mean value is taken to obtain the sample curvature.

[0029] The curvature at the actual point is calculated using the actual fitting function and the mean value is taken to obtain the actual curvature.

[0030] The sample reference region with the largest sample curvature is used as the target sample reference region, and the actual reference region with the largest actual curvature is used as the target actual reference region.

[0031] The sample reference area is numbered sequentially, starting from the target sample reference area; the actual reference area is numbered sequentially, starting from the target actual reference area.

[0032] Pair the sample reference regions with the same number with the actual reference regions.

[0033] Preferably, the step of selecting three non-collinear actual reference regions to construct the actual coordinate system includes the following steps:

[0034] The selected actual reference areas are respectively designated as the first actual reference area, the second actual reference area, and the third actual reference area;

[0035] The center of the first actual reference region is used as the actual origin, the line connecting the centers of the first and second actual reference regions is used as the actual x-axis, and the line connecting the centers of the first and third actual reference regions is used as the actual y-axis, thus forming the actual coordinate system.

[0036] In the actual coordinate system, the distance between the centers of the first actual reference region and the second actual reference region is a unit distance on the actual x-axis, and the distance between the centers of the first actual reference region and the third actual reference region is a unit distance on the actual y-axis.

[0037] Preferably, the step of constructing the sample coordinate system using the sample reference region corresponding to the selected actual reference region includes the following steps:

[0038] The sample reference regions corresponding to the first actual reference region, the second actual reference region, and the third actual reference region are respectively used as the first sample reference region, the second sample reference region, and the third sample reference region;

[0039] The center of the first sample reference area is used as the origin, the line connecting the centers of the first and second sample reference areas is used as the x-axis, and the line connecting the centers of the first and third sample reference areas is used as the y-axis. These are combined to form the sample coordinate system.

[0040] In the sample coordinate system, the distance between the centers of the first and second sample reference regions is a unit distance on the sample x-axis, and the distance between the centers of the first and third sample reference regions is a unit distance on the sample y-axis.

[0041] Preferably, the process of forming the actual contour function of the label in the actual coordinate system and the sample contour function of the label in the sample coordinate system includes the following steps:

[0042] In the actual coordinate system, the coordinates of the actual points are obtained, and the actual contour function is fitted. In the sample coordinate system, the coordinates of the sample points are obtained, and the sample contour function is fitted.

[0043] Obtaining the coordinates of an actual point involves the following steps:

[0044] Draw lines through the actual points that are parallel to the actual x-axis and the actual y-axis, intersecting the actual x-axis and the actual y-axis at the actual x-point and the actual y-point, respectively. The coordinates of the actual points are (a / b, c / d), where a is the value of the actual x-point on the actual x-axis, c is the value of the actual y-point on the actual y-axis, b is the unit distance on the actual x-axis, and d is the unit distance on the actual y-axis.

[0045] Obtaining the coordinates of sample points involves the following steps:

[0046] Draw lines parallel to the sample x-axis and sample y-axis through the sample points, intersecting the sample x-axis and sample y-axis at sample x points and sample y points, respectively. The coordinates of the sample points are (e / f, g / h), where e is the value of sample x point on the sample x-axis, g is the value of sample y point on the sample y-axis, f is the unit distance on the sample x-axis, and h is the unit distance on the sample y-axis.

[0047] Preferably, the step of calculating the positional difference between the sample contour function and the actual contour function, and calculating the critical value of the positional difference, includes the following steps:

[0048] The difference function is obtained by subtracting the sample contour function from the actual contour function.

[0049] The total range is obtained by taking the union of the range of values ​​of the independent variable of the sample profile function and the range of values ​​of the independent variable of the actual profile function.

[0050] Integrating the difference function over the overall range yields the positional difference.

[0051] At least one sample attached image is randomly combined to obtain at least one sample attached image group;

[0052] The sample function is obtained by subtracting the sample contour functions from the sample-attached image group;

[0053] The combined range is obtained by taking the union of the ranges of the independent variables of the contour functions of two samples in the sample-attached image group.

[0054] Integrating the sample function over the overall range yields the difference sample value;

[0055] The maximum value among at least one gap sample value is used as the critical value of the location gap.

[0056] A label location anomaly detection system based on image feature recognition, used to implement the aforementioned label location anomaly detection method based on image feature recognition, includes:

[0057] The reference setting module acquires at least one sample attachment image of the label and determines a reference object on the object to which the label is attached. The number of reference objects is at least three, and the reference objects are set around the label.

[0058] The region acquisition module determines the region where the reference object is located in the sample attachment image to obtain the sample reference region, acquires the actual attachment image of the label, and identifies the reference object in the actual attachment image to obtain the actual reference region.

[0059] A region pairing module, which pairs sample reference regions and actual reference regions that correspond to the same reference object;

[0060] The coordinate establishment module selects three non-collinear actual reference regions to construct an actual coordinate system, and uses the sample reference region corresponding to the selected actual reference region to construct a sample coordinate system.

[0061] The contour acquisition module forms the actual contour function of the label in the actual coordinate system and the sample contour function of the label in the sample coordinate system.

[0062] The gap acquisition module calculates the positional difference between the sample contour function and the actual contour function, and calculates the critical value of the positional difference.

[0063] The anomaly identification module identifies the label position as abnormal when the position difference exceeds a critical value; otherwise, it identifies the label position as normal.

[0064] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0065] By setting up a region acquisition module, a region matching module, a coordinate establishment module, and a contour acquisition module, during recognition, coordinates are established in the same way in both the sample attachment image and the actual attachment image through the identification of a reference object. Thus, the coordinates in the sample attachment image and the actual attachment image are consistent, and label position anomalies can be identified based on these coordinates. According to the processing of this method, regardless of the inconsistent size of the labels in the image or the inconsistent shooting angle, it will not interfere with the recognition, thereby enabling accurate identification of anomalies. Attached Figure Description

[0066] Figure 1 This is a flowchart illustrating the label location anomaly identification method based on image feature recognition of the present invention.

[0067] Figure 2 This is a schematic diagram illustrating the process of determining the reference object on the object to which the label is attached according to the present invention.

[0068] Figure 3 This is a schematic diagram of the process of determining the region where the reference object is located in the sample attachment image to obtain the sample reference region according to the present invention.

[0069] Figure 4 This is a schematic diagram illustrating the process of identifying a reference object in an actual attached image to obtain the actual reference area according to the present invention.

[0070] Figure 5 This is a schematic diagram illustrating the process of pairing sample reference regions and actual reference regions corresponding to the same reference object according to the present invention.

[0071] Figure 6 A schematic diagram illustrating the process of constructing the actual coordinate system by selecting three non-collinear actual reference regions for this invention;

[0072] Figure 7 This is a schematic diagram illustrating the process of constructing a sample coordinate system using the sample reference region corresponding to the selected actual reference region according to the present invention.

[0073] Figure 8 This is a schematic diagram illustrating the process of calculating the positional difference between the sample contour function and the actual contour function, and calculating the critical value of the positional difference according to the present invention. Detailed Implementation

[0074] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0075] Reference Figure 1 As shown, a label location anomaly identification method based on image feature recognition includes:

[0076] Acquire at least one sample image of the label being attached, determine the reference objects on the object to which the label is attached, and there are at least three reference objects, which are set around the label.

[0077] The region where the reference object is located is determined in the sample attachment image to obtain the sample reference region;

[0078] Obtain the actual attached image of the label, identify the reference object in the actual attached image, and obtain the actual reference area;

[0079] Pair the sample reference area with the actual reference area corresponding to the same reference object;

[0080] Three actual reference regions with non-collinear centers are selected to construct the actual coordinate system, and the sample coordinate system is constructed using the sample reference regions corresponding to the selected actual reference regions.

[0081] In the actual coordinate system, the actual contour function of the label is formed; in the sample coordinate system, the sample contour function of the label is formed.

[0082] Calculate the positional difference between the sample contour function and the actual contour function, and calculate the critical value of the positional difference;

[0083] If the position difference exceeds a critical value, the label position is identified as abnormal; otherwise, the label position is identified as normal.

[0084] In actual image acquisition, the position of the tagged object may not be fixed. The label may not be facing the lens directly during shooting, and the distance may also be different. In addition, the angle may be rotated. The image of the marker used to determine the label position will change, making it difficult to use as a basis for identification. All of these will make it impossible to use standard comparison methods to identify abnormal label positions. Therefore, in this solution, corresponding steps are set to overcome these problems.

[0085] Reference Figure 2 As shown, determining the reference object on the object to which the label is attached includes the following steps:

[0086] Enlarge the label by focusing on its center to obtain the enlarged label area;

[0087] When the zoomed-in area of ​​a label first covers more than three objects around the label, the covered objects are used as the reference objects.

[0088] First, the reference object is determined, but this determination differs from the usual method. This is because, during image acquisition, the position of the object to which the label is attached varies, making it impossible to obtain the reference object based on contour recognition. Instead, the reference object needs to be determined based on the relative relationship between the label and the objects surrounding it. Although there is an error in label attachment, it is very small. Therefore, it can be assumed that the relative positions of the surrounding objects and the label are consistent. Thus, regardless of the shooting situation, as the label is gradually enlarged, the coverage of the objects surrounding more than three labels is consistent at the beginning. Therefore, the same method can be used to determine the sample reference area and the actual reference area.

[0089] Reference Figure 3 As shown, determining the location of the reference object in the sample attachment image and obtaining the sample reference region includes the following steps:

[0090] In the sample attachment image, the sample area of ​​the label is determined, and the sample area is magnified with the center of the sample area to obtain the magnified sample area;

[0091] When the sample magnification area first covers more than three objects around the label, the area where the covered objects are locked is identified as the sample reference area.

[0092] Reference Figure 4 As shown, the identification of the reference object in the actual attached image to obtain the actual reference area includes the following steps:

[0093] In the actual attached image, determine the actual area of ​​the label, and then magnify the actual area using the center of the actual area to obtain the magnified actual area;

[0094] When the actual magnified area first covers more than three objects around the labels, the area where the covered objects are locked is identified as the actual reference area.

[0095] Reference Figure 5 As shown, pairing the sample reference region and the actual reference region corresponding to the same reference object includes the following steps:

[0096] Both the sample reference region and the actual reference region are modeled using a random coordinate system.

[0097] At least one sample point is uniformly selected on the contour line of the sample reference area, and the sample point is fitted to obtain the sample fitting function.

[0098] At least one actual point is uniformly selected on the contour line of the actual reference area, and the actual point is fitted to obtain the actual fitting function.

[0099] The curvature at the sample points is calculated using the sample fitting function and the mean value is taken to obtain the sample curvature.

[0100] The curvature at the actual point is calculated using the actual fitting function and the mean value is taken to obtain the actual curvature.

[0101] The sample reference region with the largest sample curvature is used as the target sample reference region, and the actual reference region with the largest actual curvature is used as the target actual reference region.

[0102] The sample reference area is numbered sequentially, starting from the target sample reference area; the actual reference area is numbered sequentially, starting from the target actual reference area.

[0103] Pair the sample reference regions with the same number with the actual reference regions.

[0104] Even after the sample reference area and the actual reference area are determined, they still cannot be paired because the object to which the label is attached may rotate. Therefore, the average curvature of the sample reference area and the actual reference area is calculated. When the image is transformed, the curvature will change, but the relative size between the curvatures will not change. That is, under the same shooting conditions, the object with greater curvature will definitely have greater curvature in the image. Therefore, the initial numbering positions of the sample reference area and the actual reference area are determined in this way, and thus the two are matched.

[0105] Reference Figure 6 As shown, constructing the actual coordinate system by selecting three non-collinear actual reference regions includes the following steps:

[0106] The selected actual reference areas are respectively designated as the first actual reference area, the second actual reference area, and the third actual reference area;

[0107] The center of the first actual reference region is used as the actual origin, the line connecting the centers of the first and second actual reference regions is used as the actual x-axis, and the line connecting the centers of the first and third actual reference regions is used as the actual y-axis, thus forming the actual coordinate system.

[0108] In the actual coordinate system, the distance between the centers of the first actual reference region and the second actual reference region is a unit distance on the actual x-axis, and the distance between the centers of the first actual reference region and the third actual reference region is a unit distance on the actual y-axis.

[0109] Since the sample attachment image and the actual attachment image may be acquired in different ways, directly performing coordinate modeling will not result in the same coordinates being the same point. Therefore, it is impossible to identify anomalies. It is necessary to perform the same coordinate modeling for both, so that the same coordinates correspond to the same point in the established coordinate system, thereby enabling the identification of anomalies based on the determined coordinates.

[0110] The coordinate system is established based on the fact that, under each shooting condition, if the label placement is completely error-free, the relative positional relationship between identical points in the image remains unchanged. Therefore, taking an actual attached image as an example:

[0111] A coordinate system is established by selecting the centers of the first, second, and third actual reference regions, whose centers are not collinear. Simultaneously, the unit distances on the actual x-axis and y-axis are determined. This allows us to determine the relative position of each point in the actual attached image with respect to the centers of the first, second, and third actual reference regions. Similarly, in the sample attached image, a coordinate system is established using the centers of the corresponding regions of the first, second, and third actual reference regions. Therefore, the result is the relative position of each point with respect to the centers of the first, second, and third sample reference regions. Since the first, second, and third actual reference regions correspond to the first, second, and third sample reference regions, and their unit distances are also corresponding, positions with the same coordinates correspond to the same point, thus enabling anomaly identification.

[0112] Reference Figure 7 As shown, constructing a sample coordinate system using the sample reference region corresponding to the selected actual reference region includes the following steps:

[0113] The sample reference regions corresponding to the first actual reference region, the second actual reference region, and the third actual reference region are respectively used as the first sample reference region, the second sample reference region, and the third sample reference region;

[0114] The center of the first sample reference area is used as the origin, the line connecting the centers of the first and second sample reference areas is used as the x-axis, and the line connecting the centers of the first and third sample reference areas is used as the y-axis. These are combined to form the sample coordinate system.

[0115] In the sample coordinate system, the distance between the centers of the first and second sample reference regions is a unit distance on the sample x-axis, and the distance between the centers of the first and third sample reference regions is a unit distance on the sample y-axis.

[0116] In the actual coordinate system, the actual contour function of the label is formed. In the sample coordinate system, the sample contour function of the label is formed by the following steps:

[0117] In the actual coordinate system, the coordinates of the actual points are obtained, and the actual contour function is fitted. In the sample coordinate system, the coordinates of the sample points are obtained, and the sample contour function is fitted.

[0118] Obtaining the coordinates of an actual point involves the following steps:

[0119] Draw lines through the actual points that are parallel to the actual x-axis and the actual y-axis, intersecting the actual x-axis and the actual y-axis at the actual x-point and the actual y-point, respectively. The coordinates of the actual points are (a / b, c / d), where a is the value of the actual x-point on the actual x-axis, c is the value of the actual y-point on the actual y-axis, b is the unit distance on the actual x-axis, and d is the unit distance on the actual y-axis.

[0120] Obtaining the coordinates of sample points involves the following steps:

[0121] Draw lines parallel to the sample x-axis and sample y-axis through the sample points, intersecting the sample x-axis and sample y-axis at sample x points and sample y points, respectively. The coordinates of the sample points are (e / f, g / h), where e is the value of sample x point on the sample x-axis, g is the value of sample y point on the sample y-axis, f is the unit distance on the sample x-axis, and h is the unit distance on the sample y-axis.

[0122] Reference Figure 8 As shown, calculating the positional difference between the sample contour function and the actual contour function, and calculating the critical value of the positional difference, includes the following steps:

[0123] The difference function is obtained by subtracting the sample contour function from the actual contour function.

[0124] The total range is obtained by taking the union of the range of values ​​of the independent variable of the sample profile function and the range of values ​​of the independent variable of the actual profile function.

[0125] Integrating the difference function over the overall range yields the positional difference.

[0126] At least one sample attached image is randomly combined to obtain at least one sample attached image group;

[0127] The sample function is obtained by subtracting the sample contour functions from the sample-attached image group;

[0128] The combined range is obtained by taking the union of the ranges of the independent variables of the contour functions of two samples in the sample-attached image group.

[0129] Integrating the sample function over the overall range yields the difference sample value;

[0130] The maximum value among at least one gap sample value is used as the critical value of the location gap.

[0131] Here, we need to obtain a critical value for the positional difference. Since the error fluctuates to some extent, we choose the maximum value of the difference sample as the critical value, because the fluctuation caused by the sample being attached to the image is acceptable.

[0132] A label location anomaly detection system based on image feature recognition, used to implement the aforementioned label location anomaly detection method based on image feature recognition, includes:

[0133] The reference setting module acquires at least one sample attachment image of the label and determines a reference object on the object to which the label is attached. The number of reference objects is at least three, and the reference objects are set around the label.

[0134] The region acquisition module determines the region where the reference object is located in the sample attachment image to obtain the sample reference region, acquires the actual attachment image of the label, and identifies the reference object in the actual attachment image to obtain the actual reference region.

[0135] A region pairing module, which pairs sample reference regions and actual reference regions that correspond to the same reference object;

[0136] The coordinate establishment module selects three non-collinear actual reference regions to construct an actual coordinate system, and uses the sample reference region corresponding to the selected actual reference region to construct a sample coordinate system.

[0137] The contour acquisition module forms the actual contour function of the label in the actual coordinate system and the sample contour function of the label in the sample coordinate system.

[0138] The gap acquisition module calculates the positional difference between the sample contour function and the actual contour function, and calculates the critical value of the positional difference.

[0139] The anomaly identification module identifies the label position as abnormal when the position difference exceeds a critical value; otherwise, it identifies the label position as normal.

[0140] Furthermore, this solution also proposes a storage medium on which a computer-readable program is stored. When the computer-readable program is invoked, it executes the aforementioned label position anomaly recognition method based on image feature recognition.

[0141] It is understandable that the storage medium can be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; an optical medium, such as a DVD; or a semiconductor medium, such as a solid-state drive (SSD).

[0142] In summary, the advantages of this invention are as follows: by setting up a region acquisition module, a region matching module, a coordinate establishment module, and a contour acquisition module, during recognition, coordinates are established in the same way in both the sample attachment image and the actual attachment image through the identification of the reference object. Thus, the coordinates in the sample attachment image and the actual attachment image are consistent, and label position anomalies can be identified based on these coordinates. According to the processing of this method, regardless of the inconsistent size of the labels in the image or the inconsistent shooting angle, it will not interfere with the recognition, thereby enabling accurate identification of anomalies.

[0143] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for identifying label location anomalies based on image feature recognition, characterized in that, include: Acquire at least one sample image of the label being attached, determine the reference objects on the object to which the label is attached, and there are at least three reference objects, which are set around the label. The region where the reference object is located is determined in the sample attachment image to obtain the sample reference region; Obtain the actual attached image of the label, identify the reference object in the actual attached image, and obtain the actual reference area; Pair the sample reference area with the actual reference area corresponding to the same reference object; Three actual reference regions with non-collinear centers are selected to construct the actual coordinate system, and the sample coordinate system is constructed using the sample reference regions corresponding to the selected actual reference regions. In the actual coordinate system, the actual contour function of the label is formed; in the sample coordinate system, the sample contour function of the label is formed. Calculate the positional difference between the sample contour function and the actual contour function, and calculate the critical value of the positional difference; If the position difference exceeds the threshold, the label position will be identified as abnormal; otherwise, the label position will be identified as normal. The process of pairing the sample reference region and the actual reference region corresponding to the same reference object includes the following steps: Both the sample reference region and the actual reference region are modeled using a random coordinate system. At least one sample point is uniformly selected on the contour line of the sample reference area, and the sample point is fitted to obtain the sample fitting function. At least one actual point is uniformly selected on the contour line of the actual reference area, and the actual point is fitted to obtain the actual fitting function. The curvature at the sample points is calculated using the sample fitting function and the mean value is taken to obtain the sample curvature. The curvature at the actual point is calculated using the actual fitting function and the mean value is taken to obtain the actual curvature. The sample reference region with the largest sample curvature is used as the target sample reference region, and the actual reference region with the largest actual curvature is used as the target actual reference region. The sample reference area is numbered sequentially, starting from the target sample reference area; the actual reference area is numbered sequentially, starting from the target actual reference area. Pair the sample reference regions with the same number with the actual reference regions.

2. The label location anomaly identification method based on image feature recognition according to claim 1, characterized in that, The process of determining the reference object on the object to which the label is attached includes the following steps: Enlarge the label by focusing on its center to obtain the enlarged label area; When the zoomed-in area of ​​a label first covers more than three objects around the label, the covered objects are used as the reference objects.

3. The label location anomaly identification method based on image feature recognition according to claim 2, characterized in that, The process of determining the region where the reference object is located in the sample attachment image to obtain the sample reference region includes the following steps: In the sample attachment image, the sample area of ​​the label is determined, and the sample area is magnified with the center of the sample area to obtain the magnified sample area; When the sample magnification area first covers more than three objects around the label, the area where the covered objects are locked is identified as the sample reference area.

4. The label location anomaly identification method based on image feature recognition according to claim 3, characterized in that, The process of identifying the reference object in the actual attached image to obtain the actual reference region includes the following steps: In the actual attached image, determine the actual area of ​​the label, and then magnify the actual area using the center of the actual area to obtain the magnified actual area; When the actual magnified area first covers more than three objects around the labels, the area where the covered objects are locked is identified as the actual reference area.

5. The label location anomaly identification method based on image feature recognition according to claim 1, characterized in that, The process of constructing the actual coordinate system by selecting three non-collinear actual reference regions includes the following steps: The selected actual reference areas are respectively designated as the first actual reference area, the second actual reference area, and the third actual reference area; The center of the first actual reference region is used as the actual origin, the line connecting the centers of the first and second actual reference regions is used as the actual x-axis, and the line connecting the centers of the first and third actual reference regions is used as the actual y-axis, thus forming the actual coordinate system. In the actual coordinate system, the distance between the centers of the first actual reference region and the second actual reference region is a unit distance on the actual x-axis, and the distance between the centers of the first actual reference region and the third actual reference region is a unit distance on the actual y-axis.

6. The label location anomaly identification method based on image feature recognition according to claim 5, characterized in that, The process of constructing the sample coordinate system using the sample reference region corresponding to the selected actual reference region includes the following steps: The sample reference regions corresponding to the first actual reference region, the second actual reference region, and the third actual reference region are respectively used as the first sample reference region, the second sample reference region, and the third sample reference region; The center of the first sample reference area is used as the origin, the line connecting the centers of the first and second sample reference areas is used as the x-axis, and the line connecting the centers of the first and third sample reference areas is used as the y-axis. These are combined to form the sample coordinate system. In the sample coordinate system, the distance between the centers of the first and second sample reference regions is a unit distance on the sample x-axis, and the distance between the centers of the first and third sample reference regions is a unit distance on the sample y-axis.

7. The label location anomaly identification method based on image feature recognition according to claim 6, characterized in that, The process of forming the actual contour function of the label in the actual coordinate system and the sample contour function of the label in the sample coordinate system includes the following steps: In the actual coordinate system, the coordinates of the actual points are obtained, and the actual contour function is fitted. In the sample coordinate system, the coordinates of the sample points are obtained, and the sample contour function is fitted. Obtaining the coordinates of an actual point involves the following steps: Draw lines through the actual points that are parallel to the actual x-axis and the actual y-axis, intersecting the actual x-axis and the actual y-axis at the actual x-point and the actual y-point, respectively. The coordinates of the actual points are (a / b, c / d), where a is the value of the actual x-point on the actual x-axis, c is the value of the actual y-point on the actual y-axis, b is the unit distance on the actual x-axis, and d is the unit distance on the actual y-axis. Obtaining the coordinates of sample points involves the following steps: Draw lines parallel to the sample x-axis and sample y-axis through the sample points, intersecting the sample x-axis and sample y-axis at sample x points and sample y points, respectively. The coordinates of the sample points are (e / f, g / h), where e is the value of sample x point on the sample x-axis, g is the value of sample y point on the sample y-axis, f is the unit distance on the sample x-axis, and h is the unit distance on the sample y-axis.

8. The label location anomaly identification method based on image feature recognition according to claim 7, characterized in that, The calculation of the positional difference between the sample contour function and the actual contour function, and the calculation of the critical value of the positional difference, includes the following steps: The difference function is obtained by subtracting the sample contour function from the actual contour function. The total range is obtained by taking the union of the range of values ​​of the independent variable of the sample profile function and the range of values ​​of the independent variable of the actual profile function. Integrating the difference function over the overall range yields the positional difference. At least one sample attached image is randomly combined to obtain at least one sample attached image group; The sample function is obtained by subtracting the sample contour functions from the sample-attached image group; The combined range is obtained by taking the union of the ranges of the independent variables of the contour functions of two samples in the sample-attached image group. Integrating the sample function over the overall range yields the difference sample value; The maximum value among at least one gap sample value is used as the critical value of the location gap.

9. A label location anomaly identification system based on image feature recognition, used to implement the label location anomaly identification method based on image feature recognition as described in any one of claims 1-8, characterized in that, include: The reference setting module acquires at least one sample attachment image of the label and determines a reference object on the object to which the label is attached. The number of reference objects is at least three, and the reference objects are set around the label. The region acquisition module determines the region where the reference object is located in the sample attachment image to obtain the sample reference region, acquires the actual attachment image of the label, and identifies the reference object in the actual attachment image to obtain the actual reference region. A region pairing module, which pairs sample reference regions and actual reference regions that correspond to the same reference object; The coordinate establishment module selects three non-collinear actual reference regions to construct an actual coordinate system, and uses the sample reference region corresponding to the selected actual reference region to construct a sample coordinate system. The contour acquisition module forms the actual contour function of the label in the actual coordinate system and the sample contour function of the label in the sample coordinate system. The gap acquisition module calculates the positional difference between the sample contour function and the actual contour function, and calculates the critical value of the positional difference. The anomaly identification module identifies the label position as abnormal when the position difference exceeds a critical value; otherwise, it identifies the label position as normal.

Images