Label position abnormity identification method and system based on image feature identification

Abstract

The invention discloses a label position abnormity identification method and system based on image feature identification, and relates to the field of machine identification, and the method comprises the steps: determining a reference object on an object to which a label is attached; obtaining a sample reference area; obtaining an actual reference area; matching the sample reference region and the actual reference region corresponding to the same reference substance; forming an actual contour function of the label in the actual coordinate system, and forming a sample contour function of the label in the 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; and when the position difference exceeds a critical value, identifying the label position as abnormal, otherwise, identifying the label position as normal. Through the arrangement of the region acquisition module, the region pairing module, the coordinate establishment module and the contour acquisition module, no matter the sizes of the labels in the image are inconsistent or the shooting angles are inconsistent, the identification is not interfered, so that the abnormity is accurately identified.

Description

Technical Field

[0001] The present invention relates to the field of machine recognition, and more particularly to a method and system for identifying abnormal label positions based on image feature recognition. Background Art

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

[0003] However, when identifying labels, image recognition is usually used. Since image acquisition is affected by the object to which the label is attached, and the position of the object is different, the situation of the label in the acquired image will also be different. Therefore, there will be significant interference in the identification of abnormal label positions, affecting the accuracy of identification. Summary of the Invention

[0004] To solve the above technical problems, a method and system for identifying abnormal label positions based on image feature recognition are provided, and the technical solution solves the problems raised in the above background art.

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

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

[0007] Obtain at least one sample attachment image of the attached label, determine the reference objects on the object to which the label is attached, and the number of reference objects is at least three, and the reference objects are arranged around the label;

[0008] Determine the area where the reference object is located in the sample attachment image to obtain the sample reference area;

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

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

[0011] Select three actual reference areas with non-collinear centers to construct an actual coordinate system, and use the sample reference areas corresponding to the selected actual reference areas to construct a sample coordinate system;

[0012] In the actual coordinate system, form the actual contour function of the label, and in the sample coordinate system, form the sample contour function of the label;

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

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

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

[0016] Enlarge the label with the center of the label to obtain a label enlarged area;

[0017] When the label enlarged area first covers more than three objects around the label, the covered object is taken as the reference object.

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

[0019] Determine the sample area of the label in the sample attachment image, and enlarge the sample area with the center of the sample area to obtain a sample enlarged area;

[0020] When the sample enlarged area first covers more than three objects around the label, identify the area where the covered object is located as the sample reference area.

[0021] Preferably, the steps for identifying the reference object in the actual attachment image to obtain the actual reference area include the following:

[0022] Determine the actual area of the label in the actual attachment image, and enlarge the actual area with the center of the actual area to obtain an actual enlarged area;

[0023] When the actual enlarged area first covers more than three objects around the label, identify the area where the covered object is located as the actual reference area.

[0024] Preferably, the steps for pairing the sample reference area and the actual reference area corresponding to the same reference object include the following:

[0025] Perform coordinate modeling on both the sample reference area and the actual reference area using a random coordinate system;

[0026] Uniformly take at least one sample point on the contour line of the sample reference area, and fit the sample points to obtain a sample fitting function;

[0027] Uniformly take at least one actual point on the contour line of the actual reference area, and fit the actual points to obtain an actual fitting function;

[0028] Use the sample fitting function to calculate the curvature at the sample points and take the average value to obtain the sample curvature;

[0029] Use the actual fitting function to calculate the curvature at the actual points and take the average value to obtain the actual curvature;

[0030] Take the sample reference region with the maximum sample curvature as the target sample reference region, and take the actual reference region with the maximum actual curvature as the target actual reference region;

[0031] Number the sample reference regions in a clockwise direction, with the starting point being the target sample reference region, and number the actual reference regions in a clockwise direction, with the starting point being the target actual reference region;

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

[0033] Preferably, the steps for constructing the actual coordinate system by selecting three non-collinear actual reference regions include the following:

[0034] Take the selected actual reference regions as the first actual reference region, the second actual reference region, and the third actual reference region respectively;

[0035] Use the center of the first actual reference region as the actual origin, use the line connecting the centers of the first actual reference region and the second actual reference region as the actual x-axis, and use the line connecting the centers of the first actual reference region and the third actual reference region as the actual y-axis, and summarize to form 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 the 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 the unit distance on the actual y-axis.

[0037] Preferably, the steps for constructing the sample coordinate system by using the sample reference regions corresponding to the selected actual reference regions include the following:

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

[0039] Use the center of the first sample reference region as the origin, use the line connecting the centers of the first sample reference region and the second sample reference region as the x-axis, and use the line connecting the centers of the first sample reference region and the third sample reference region as the y-axis, and summarize to form the sample coordinate system;

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

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

[0042] In the actual coordinate system, obtain the coordinates of the actual points, and fit to obtain the actual contour function. In the sample coordinate system, obtain the coordinates of the sample points, and fit to obtain the sample contour function;

[0043] Obtaining the coordinates of the actual points includes the following steps:

[0044] Draw lines parallel to the actual x-axis and the actual y-axis through the actual points respectively, and intersect the actual x-axis and the actual y-axis at the actual x-point and the actual y-point. The coordinates of the actual point 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 the sample points includes the following steps:

[0046] Draw lines parallel to the sample x-axis and the sample y-axis through the sample points respectively, and intersect the sample x-axis and the sample y-axis at the sample x-point and the sample y-point. The coordinates of the sample point are (e / f, g / h), where e is the value of the sample x-point on the sample x-axis, g is the value of the 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, calculating the position gap between the sample contour function and the actual contour function, and calculating the critical value of the position gap includes the following steps:

[0048] Subtract the sample contour function from the actual contour function to obtain the gap function;

[0049] Take the union of the value ranges of the independent variables of the sample contour function and the actual contour function to obtain the overall range;

[0050] Integrate the gap function over the overall range to obtain the position gap;

[0051] Randomly combine at least one sample attachment image to obtain at least one sample attachment image group;

[0052] Subtract the sample contour functions in the sample attachment image group to obtain the sample function;

[0053] Take the union of the value ranges of the independent variables of two sample contour functions in the sample attachment image group to obtain the comprehensive range;

[0054] Integrate the sample function over the comprehensive range to obtain the gap sample value;

[0055] Take the maximum value among at least one gap sample value as the critical value of the position gap.

[0056] A label position anomaly recognition system based on image feature recognition is used to implement the above-mentioned label position anomaly recognition method based on image feature recognition, including:

[0057] A reference setting module, which obtains at least one sample attachment image of the label attachment, determines the reference objects on the object to which the label is attached, and the number of reference objects is at least three, and the reference objects are arranged around the label;

[0058] An area acquisition module, which determines the area where the reference object is located in the sample attachment image to obtain the sample reference area, obtains the actual attachment image of the label, and identifies the reference object in the actual attachment image to obtain the actual reference area;

[0059] An area pairing module, which pairs the sample reference area and the actual reference area corresponding to the same reference object;

[0060] A coordinate establishment module, which selects three non-collinear actual reference areas to construct an actual coordinate system, and uses the sample reference areas corresponding to the selected actual reference areas to construct a sample coordinate system;

[0061] A contour acquisition module, which forms an actual contour function of the label in the actual coordinate system and forms a sample contour function of the label in the sample coordinate system;

[0062] A gap acquisition module, which calculates the position gap between the sample contour function and the actual contour function and calculates the critical value of the position gap;

[0063] An anomaly recognition module, which identifies the label position as abnormal when the position gap exceeds the critical value, otherwise, 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 an area acquisition module, an area pairing module, a coordinate establishment module, and a contour acquisition module, during recognition, through the recognition of reference objects, the coordinates are established in the same way in the sample attachment image and the actual attachment image. Thus, the coordinate situations in the sample attachment image and the actual attachment image are consistent. Furthermore, the label position anomaly can be recognized based on these coordinates. According to the processing process of this method, whether the size of the label in the image is inconsistent or the shooting angle is inconsistent, it will not interfere with the recognition. Thus, the anomaly can be accurately recognized. Description of the Drawings

[0066] Figure 1 Schematic flow diagram of the method for identifying abnormal label positions based on image feature recognition according to the present invention;

[0067] Figure 2 Schematic flow diagram of determining a reference object on an object to which a label is attached according to the present invention;

[0068] Figure 3 Schematic flow diagram of determining the area where the reference object is located in the sample attachment image to obtain the sample reference area according to the present invention;

[0069] Figure 4 Schematic flow diagram of identifying the reference object in the actual attachment image to obtain the actual reference area according to the present invention;

[0070] Figure 5 Schematic flow diagram of pairing the sample reference area and the actual reference area corresponding to the same reference object according to the present invention;

[0071] Figure 6 Schematic flow diagram of selecting three actual reference areas with non - collinear centers to construct an actual coordinate system according to the present invention;

[0072] Figure 7 Schematic flow diagram of using the sample reference area corresponding to the selected actual reference area to construct a sample coordinate system according to the present invention;

[0073] Figure 8 Schematic flow diagram of calculating the position gap between the sample contour function and the actual contour function and calculating the critical value of the position gap according to the present invention. Detailed implementation manners

[0074] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and those skilled in the art can think of other obvious variations.

[0075] Referring to Figure 1 As shown, a method for identifying abnormal label positions based on image feature recognition includes:

[0076] Obtaining at least one sample attachment image of the attached label, determining a reference object on the object to which the label is attached, where the number of reference objects is at least three and the reference objects are arranged around the label;

[0077] Determining the area where the reference object is located in the sample attachment image to obtain the sample reference area;

[0078] Obtaining the actual attachment image of the label, and identifying the reference object in the actual attachment image to obtain the actual reference area;

[0079] Pair the sample reference region and the actual reference region corresponding to the same reference object;

[0080] Select three actual reference regions whose centers are not collinear to construct an actual coordinate system, and use the sample reference regions corresponding to the selected actual reference regions to construct a sample coordinate system;

[0081] In the actual coordinate system, form the actual contour function of the label, and in the sample coordinate system, form the sample contour function of the label;

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

[0083] When the position gap exceeds the critical value, the label position is identified as abnormal, otherwise, the label position is identified as normal.

[0084] During actual image acquisition, since the position of the object with the label attached may be unfixed, the label may not be directly facing the camera during shooting, and the distances are also different. At the same time, there may be a situation of angular rotation. The images of the markers used for judging the label position will all change and are difficult to be used as the basis for identification. All these will lead to the inability to use the standard comparison method to identify the abnormal label position. Therefore, in this solution, corresponding steps are set to overcome these problems.

[0085] Refer to Figure 2 As shown, the steps to determine the reference object on the object to which the label is attached include the following:

[0086] Enlarge the label with the center of the label to obtain a label enlarged region;

[0087] When the label enlarged region 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. Its determination is different from the usual method because, during image acquisition, due to the different positions of the objects with the label attached, it is impossible to obtain the reference object based on contour recognition. It is necessary to determine the reference object according to the relative relationship between the objects around the label and the label. Since the label attachment has a small error, it can be considered that the relative positions of the objects around it and the label are consistent. Therefore, no matter in what shooting situation, when the label is gradually enlarged, the coverage situation of first covering more than three objects around the label is the same. Thus, the same method can be used to determine the sample reference region and the actual reference region.

[0089] Refer to Figure 3 As shown, the steps to determine the region where the reference object is located in the sample attachment image to obtain the sample reference region include the following:

[0090] Determine the sample area of the label in the sample attachment image, magnify the sample area with the center of the sample area to obtain a sample magnification area;

[0091] When the sample magnification area first covers objects around more than three labels, identify the area where the covered object is located as the sample reference area.

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

[0093] Determine the actual area of the label in the actual attachment image, magnify the actual area with the center of the actual area to obtain an actual magnification area;

[0094] When the actual magnification area first covers objects around more than three labels, identify the area where the covered object is located as the actual reference area.

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

[0096] Perform coordinate modeling on both the sample reference area and the actual reference area using a random coordinate system;

[0097] Uniformly take at least one sample point on the contour line of the sample reference area, fit the sample points to obtain a sample fitting function;

[0098] Uniformly take at least one actual point on the contour line of the actual reference area, fit the actual points to obtain an actual fitting function;

[0099] Use the sample fitting function to calculate the curvature at the sample points and take the average value to obtain the sample curvature;

[0100] Use the actual fitting function to calculate the curvature at the actual points and take the average value to obtain the actual curvature;

[0101] Take the sample reference area with the maximum sample curvature as the target sample reference area, and take the actual reference area with the maximum actual curvature as the target actual reference area;

[0102] Number the sample reference areas in a clockwise direction, starting from the target sample reference area, and number the actual reference areas in a clockwise direction, starting from the target actual reference area;

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

[0104] After the sample reference region and the actual reference region are determined, they still cannot be paired because the object with the label attached may be rotated. Therefore, calculate the average curvature of the sample reference region and the actual reference region. When the image is transformed, its curvature will change, but the relative magnitude between the curvatures will not change. That is, under the same shooting conditions, the object with a greater curvature will definitely have a greater curvature in the image. Therefore, determine the initial numbered positions in the sample reference region and the actual reference region in this way, and thus correspond the two to each other.

[0105] Refer to Figure 6 As shown, the steps to construct the actual coordinate system by selecting three actual reference regions whose centers are not collinear include the following:

[0106] Respectively use the selected actual reference regions as the first actual reference region, the second actual reference region, and the third actual reference region;

[0107] Use the center of the first actual reference region as the actual origin, use the line connecting the centers of the first actual reference region and the second actual reference region as the actual x-axis, and use the line connecting the centers of the first actual reference region and the third actual reference region as the actual y-axis, and summarize to form 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 the 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 the unit distance on the actual y-axis.

[0109] Due to possible differences in the acquisition methods of the sample attachment image and the actual attachment image, directly performing coordinate modeling, the positions with the same coordinates are not corresponding points. Therefore, it is impossible to identify abnormalities. It is necessary to perform the same coordinate modeling on the two so that in the established coordinate system, the positions with the same coordinates correspond to the same points, so as to be able to determine the coordinates according to;

[0110] The basis for establishing its coordinate system is that under each shooting condition, if the label position attachment is completely error-free, the relative position relationship between the same points in the image remains unchanged. Therefore, taking the actual attachment image as an example:

[0111] Select the centers of the first actual reference region, the second actual reference region, and the third actual reference region whose centers are not collinear to establish a coordinate system. At the same time, determine the unit distances on the actual x-axis and the actual y-axis. Thus, the relative positional relationships between each point in the actual attached image and the centers of the first actual reference region, the second actual reference region, and the third actual reference region can be determined. In the sample attached image, the centers of the regions corresponding to the first actual reference region, the second actual reference region, and the third actual reference region are also used to establish the coordinate system. Therefore, what is obtained is the relative positional relationship of each point with respect to the centers of the first sample reference region, the second sample reference region, and the third sample reference region. Since the first actual reference region, the second actual reference region, and the third actual reference region correspond to the first sample reference region, the second sample reference region, and the third sample reference region, and the unit distances used for measurement are also corresponding, in this coordinate system, the same points correspond to the same positions, so abnormal identification can be carried out.

[0112] Refer to Figure 7 As shown, the steps for constructing a sample coordinate system using the sample reference regions corresponding to the selected actual reference regions are as follows:

[0113] Respectively use the sample reference regions corresponding to the first actual reference region, the second actual reference region, and the third actual reference region as the first sample reference region, the second sample reference region, and the third sample reference region;

[0114] Use the center of the first sample reference region as the origin, use the line connecting the centers of the first sample reference region and the second sample reference region as the x-axis, and use the line connecting the centers of the first sample reference region and the third sample reference region as the y-axis to summarize and form a sample coordinate system;

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

[0116] In the actual coordinate system, form the actual contour function of the label. In the sample coordinate system, the steps for forming the sample contour function of the label are as follows:

[0117] In the actual coordinate system, obtain the coordinates of the actual points and fit to obtain the actual contour function. In the sample coordinate system, obtain the coordinates of the sample points and fit to obtain the sample contour function;

[0118] The steps for obtaining the coordinates of the actual points are as follows:

[0119] Draw lines parallel to the actual x-axis and the actual y-axis through the actual point respectively, which intersect the actual x-axis and the actual y-axis at the actual x-point and the actual y-point. The coordinates of the actual point 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 the sample points includes the following steps:

[0121] Draw lines parallel to the sample x-axis and the sample y-axis through the sample point respectively, which intersect the sample x-axis and the sample y-axis at the sample x-point and the sample y-point. The coordinates of the sample point are (e / f, g / h), where e is the value of the sample x-point on the sample x-axis, g is the value of the 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] Refer to Figure 8 As shown, calculating the position gap between the sample profile function and the actual profile function, and calculating the critical value of the position gap includes the following steps:

[0123] Subtract the actual profile function from the sample profile function to obtain the gap function;

[0124] Take the union of the value ranges of the independent variables of the sample profile function and the actual profile function to obtain the overall range;

[0125] Integrate the gap function over the overall range to obtain the position gap;

[0126] Randomly combine at least one sample attachment image to obtain at least one sample attachment image group;

[0127] Subtract the sample profile functions in the sample attachment image group to obtain the sample function;

[0128] Take the union of the value ranges of the independent variables of two sample profile functions in the sample attachment image group to obtain the comprehensive range;

[0129] Integrate the sample function over the comprehensive range to obtain the gap sample value;

[0130] Take the maximum value among at least one gap sample value as the critical value of the position gap.

[0131] Here, a critical value needs to be obtained for the position gap. Due to the existence of certain fluctuations in the error, therefore, select the maximum value of the gap sample value as the critical value because the fluctuations caused by the sample attachment image are all acceptable.

[0132] A label position anomaly recognition system based on image feature recognition for implementing the above-mentioned label position anomaly recognition method based on image feature recognition, including:

[0133] A reference setting module, which obtains at least one sample attachment image of the label attachment, determines a reference object on the object to which the label is attached, and there are at least three reference objects, and the reference objects are arranged around the label;

[0134] An area acquisition module, which determines the area where the reference object is located in the sample attachment image to obtain a sample reference area, obtains the actual attachment image of the label, and identifies the reference object in the actual attachment image to obtain an actual reference area;

[0135] An area pairing module, which pairs the sample reference area and the actual reference area corresponding to the same reference object;

[0136] A coordinate establishment module, which selects three non-collinear actual reference areas to construct an actual coordinate system, and uses the sample reference areas corresponding to the selected actual reference areas to construct a sample coordinate system;

[0137] A contour acquisition module, which forms an actual contour function of the label in the actual coordinate system and forms a sample contour function of the label in the sample coordinate system;

[0138] A gap acquisition module, which calculates the position gap between the sample contour function and the actual contour function and calculates the critical value of the position gap;

[0139] An anomaly recognition module, which, when the position gap exceeds the critical value, recognizes the label position as abnormal, otherwise, recognizes the label position as normal.

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

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

[0142] In summary, the advantages of the present invention are as follows: By setting up a region acquisition module, a region pairing module, a coordinate establishment module, and a contour acquisition module, during recognition, through the recognition of fiducial objects, coordinates are established in the same way in the sample attachment image and the actual attachment image. Thus, the coordinate situations in the sample attachment image and the actual attachment image are consistent. Furthermore, the recognition of label position anomalies can be carried out based on these coordinates. According to the processing process of this method, regardless of whether the sizes of the labels in the image are inconsistent or the shooting angles are inconsistent, it will not interfere with the recognition. Therefore, accurate recognition of anomalies can be achieved.

[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 by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, various changes and improvements will occur to the present invention, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for identifying abnormal label positions based on image feature recognition, characterized in that, Including: Obtain at least one sample attachment image with the label attached, determine fiducial markers on the object to which the label is attached, where the number of fiducial markers is at least three and the fiducial markers are arranged around the label; Determine the regions where the fiducial markers are located in the sample attachment image to obtain sample fiducial regions; Obtain the actual attachment image of the label, identify the fiducial markers in the actual attachment image to obtain actual fiducial regions; Pair the sample fiducial regions and the actual fiducial regions corresponding to the same fiducial markers; Select three actual fiducial regions with non - collinear centers to construct an actual coordinate system, and use the sample fiducial regions corresponding to the selected actual fiducial regions to construct a sample coordinate system; In the actual coordinate system, form the actual contour function of the label, and in the sample coordinate system, form the sample contour function of the label; Calculate the position gap between the sample contour function and the actual contour function, and calculate the critical value of the position gap; When the position gap exceeds the critical value, identify the label position as abnormal, otherwise, identify the label position as normal.

2. The method for identifying abnormal label positions based on image feature recognition according to claim 1, characterized in that, The determination of the fiducial markers on the object to which the label is attached includes the following steps: Enlarge the label centered on the center of the label to obtain a label enlarged region; When the label enlarged region first covers more than three objects around the label, take the covered objects as fiducial markers.

3. The method for identifying abnormal label positions based on image feature recognition according to claim 2, wherein, The determination of the regions where the fiducial markers are located in the sample attachment image to obtain sample fiducial regions includes the following steps: Determine the sample region of the label in the sample attachment image, and enlarge the sample region centered on the center of the sample region to obtain a sample enlarged region; When the sample enlarged region first covers more than three objects around the label, identify the region where the covered objects are located as the sample fiducial region.

4. The method for identifying abnormal label positions based on image feature recognition according to claim 3, wherein The identification of the fiducial markers in the actual attachment image to obtain actual fiducial regions includes the following steps: Determine the actual region of the label in the actual attachment image, and enlarge the actual region centered on the center of the actual region to obtain an actual enlarged region; When the actual enlarged region first covers more than three objects around the label, identify the region where the covered objects are located as the actual fiducial region.

5. The method for identifying abnormal label positions based on image feature recognition according to claim 4, characterized in that The pairing of the sample fiducial regions and the actual fiducial regions corresponding to the same fiducial markers includes the following steps: Perform coordinate modeling on both the sample fiducial regions and the actual fiducial regions using a random coordinate system; Uniformly take at least one sample point on the contour line of the sample fiducial region, and fit the sample points to obtain a sample fitting function; Uniformly take at least one actual point on the contour line of the actual fiducial region, and fit the actual points to obtain an actual fitting function; Use the sample fitting function to calculate the curvature at the sample points and take the average value to obtain the sample curvature; Use the actual fitting function to calculate the curvature at the actual points and take the average value to obtain the actual curvature; Take the sample fiducial region with the maximum sample curvature as the target sample fiducial region, and take the actual fiducial region with the maximum actual curvature as the target actual fiducial region; Number the sample fiducial regions clockwise, starting from the target sample fiducial region, and number the actual fiducial regions clockwise, starting from the target actual fiducial region; Pair the sample fiducial regions and the actual fiducial regions with the same number.

6. The method for identifying abnormal label positions based on image feature recognition according to claim 5, characterized in that The steps for constructing an actual coordinate system by selecting three non - collinear actual reference regions are as follows: Respectively take the selected actual reference regions as the first actual reference region, the second actual reference region, and the third actual reference region; Use the center of the first actual reference region as the actual origin, use the line connecting the centers of the first actual reference region and the second actual reference region as the actual x - axis, and use the line connecting the centers of the first actual reference region and the third actual reference region as the actual y - axis, and summarize to form 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 the 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 the unit distance on the actual y - axis.

7. The method for identifying abnormal label positions based on image feature recognition according to claim 6, characterized in that, The steps for constructing a sample coordinate system by using the corresponding sample reference regions of the selected actual reference regions are as follows: Respectively take the sample reference regions corresponding to the first actual reference region, the second actual reference region, and the third actual reference region as the first sample reference region, the second sample reference region, and the third sample reference region; Use the center of the first sample reference region as the origin, use the line connecting the centers of the first sample reference region and the second sample reference region as the x - axis, and use the line connecting the centers of the first sample reference region and the third sample reference region as the y - axis, and summarize to form the sample coordinate system; In the sample coordinate system, the distance between the centers of the first sample reference region and the second sample reference region is the unit distance on the sample x - axis, and the distance between the centers of the first sample reference region and the third sample reference region is the unit distance on the sample y - axis.

8. The method for identifying abnormal label positions based on image feature recognition according to claim 7, characterized in that, The steps for 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 are as follows: In the actual coordinate system, obtain the coordinates of the actual points and fit to obtain the actual contour function. In the sample coordinate system, obtain the coordinates of the sample points and fit to obtain the sample contour function; The steps for obtaining the coordinates of the actual points include the following: Draw lines parallel to the actual x - axis and the actual y - axis through the actual point, intersecting the actual x - axis and the actual y - axis at the actual x - point and the actual y - point. The coordinates of the actual point 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; The steps for obtaining the coordinates of the sample points include the following: Draw lines parallel to the sample x - axis and the sample y - axis through the sample point, intersecting the sample x - axis and the sample y - axis at the sample x - point and the sample y - point. The coordinates of the sample point are (e / f, g / h), where e is the value of the sample x - point on the sample x - axis, g is the value of the 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.

9. A method for identifying abnormal label positions based on image feature recognition according to claim 8, characterized in that The steps for calculating the position gap between the sample contour function and the actual contour function and calculating the critical value of the position gap are as follows: Subtract the sample contour function from the actual contour function to obtain the gap function; Take the union of the value range of the independent variable of the sample contour function and the value range of the independent variable of the actual contour function to obtain the overall range; Integrate the gap function over the overall range to obtain the position gap; Randomly combine at least one sample attachment image to obtain at least one sample attachment image group; Subtract the sample contour functions in the sample attachment image group to obtain the sample function; Take the union of the value ranges of the independent variables of two sample contour functions in the sample attachment image group to obtain the comprehensive range; Integrate the sample function over the comprehensive range to obtain the gap sample value; Take the maximum value among at least one gap sample value as the critical value of the position gap.

10. A label position anomaly recognition system based on image feature recognition, which is used to implement the label position anomaly recognition method based on image feature recognition according to any one of claims 1-9, characterized in that, Including: A reference setting module, which obtains at least one sample attachment image of the label attachment, determines the reference objects on the object to which the label is attached, and the number of reference objects is at least three and the reference objects are arranged around the label; A region acquisition module, which determines the region where the reference object is located in the sample attachment image to obtain the sample reference region, obtains 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 the sample reference region and the actual reference region corresponding to the same reference object; A coordinate establishment module, which selects three non-collinear actual reference regions to construct the actual coordinate system, and uses the sample reference regions corresponding to the selected actual reference regions to construct the sample coordinate system; A contour acquisition module, which forms the actual contour function of the label in the actual coordinate system and forms the sample contour function of the label in the sample coordinate system; A gap acquisition module, which calculates the position gap between the sample contour function and the actual contour function and calculates the critical value of the position gap; An abnormality identification module, which identifies the label position as abnormal when the position gap exceeds the critical value, otherwise, identifies the label position as normal.

Images