An IoT label scale with AI recognition module

Abstract

This invention provides an IoT tag scale with an AI recognition module. By having the user press a category set change button, the classification matching object is limited to all weighed objects that have been matched and bound to that category set change button. This narrows the matching range of classification features for subsequent weighed objects and improves classification speed. The provided classification algorithm uses the boundary shape image between the first contour of the weighed object and the second contour within the set field of view of the AI ​​recognition module in the tag scale as the classification feature. This eliminates the need to capture subtle feature differences between the weighed object and weighed objects with similar but different appearances. In other words, it improves the classification accuracy of weighed objects with similar but different appearances without improving the classification algorithm itself. The provided IoT batch parameter update method for the tag scale achieves synchronous batch parameter updates for all weighed objects under the same category by simply pressing the category set change button.

Description

Technical Field

[0001] This invention relates to the field of smart label scale technology, and more specifically to an IoT label scale with an AI recognition module. Background Technology

[0002] Smart label scales are currently widely used in the market. For example, after placing fruit trees such as pineapples and tomatoes on the label scale, the scale identifies the objects through image recognition, outputs a classification result, obtains the unit price of the goods, and then weighs and prices the successfully classified goods. However, existing smart label scales mainly have the following three problems:

[0003] 1. When the objects being weighed have a high degree of similarity in appearance, such as size and color, like oranges and tangerines that are very similar in size and color, conventional image recognition classification methods may be inaccurate. Classification algorithms capable of capturing more detailed features are needed for identification and classification. However, higher algorithm accuracy usually means longer computation time, which is generally unacceptable in scenarios like commodity weighing where minimizing waiting time is crucial.

[0004] 2. Image recognition classification methods are more prone to errors when dealing with weighed items of the same type but different sizes. For example, existing image classification algorithms are prone to pricing errors when dealing with melons of the same type that are highly similar in appearance features, including shape and color, but differ only in size. For instance, melons of the same type may have essentially the same appearance features except for size, and their sizes may be similar, but their prices differ due to size differences. How to identify the size difference between two frames of weighed items to reduce pricing classification errors is a pressing technical problem in this field.

[0005] 3. Multiple smart label scales are typically used in the same shopping mall. When setting up the label scales in batches, the method of setting each item individually is generally adopted. For example, a pineapple is placed on the label scale, and after the scale recognizes the pineapple through image recognition, it redirects to a parameter input page. The user enters the pineapple's information, such as the unit price, on the redirected page. After the setting is completed, other label scales connected to this scale and in setting mode simultaneously modify the pineapple's parameter information. This method of setting each item individually is inefficient and takes a long time when the number of items to be weighed is large. Summary of the Invention

[0006] This invention aims to improve the classification accuracy of weighed objects with similar but different appearance features, as well as those of the same type but whose size features are captured by conventional classification algorithms, and to enable batch setting and updating of parameter information of weighed objects for various smart tag scales connected to the Internet of Things. It provides an IoT tag scale with an AI recognition module.

[0007] To achieve this object, the present invention adopts the following technical solutions:

[0008] An IoT tag scale with an AI recognition module is provided. The object to be weighed belongs to a classification set. When image recognition fails to classify the object, the following steps are taken to identify, classify, and price the object:

[0009] S1, press the category set change button corresponding to the object to be weighed, and the label scale adjusts the set field of view and set field of view depth of the AI ​​recognition module by pressing the button as an instruction, and obtains the standard classification features of each historical weighed object in the category set to which the object to be weighed belongs.

[0010] S2, select the object to be weighed using the minimum bounding rectangle, and then calculate the center point of the minimum bounding rectangle;

[0011] S3, under the set field of view depth, adjust the position of the AI ​​recognition module so that the center point is the center point of the field of view generated by the AI ​​recognition module within the set field of view range;

[0012] S4, extract the boundary shape image between the first contour of the object to be weighed selected by the minimum bounding rectangle and the second contour of the set field of view as the classification feature of the object to be weighed, and then perform maximum similarity matching with each of the standard classification features of the associated historical weighed objects obtained in step S1. The matching result is used as the classification result and priced.

[0013] Preferably, the weighing objects belonging to different classification sets have differences in size characteristics; the weighing objects in the same classification set have similar appearance characteristics but are not of the same type; the similarity of appearance characteristics of the weighing objects includes any one or more of size characteristics, color characteristics, and shape characteristics; wherein, size characteristics are similar when the difference in the coverage area of ​​the smallest bounding rectangle of the weighing object is less than a preset first difference threshold; color characteristics are similar when the difference in the average color value of the weighing objects is less than a preset second difference threshold; and shape characteristics are similar when the difference in the outline shape of the weighing objects is less than a preset third difference threshold.

[0014] Preferably, step S2 specifically includes the following steps:

[0015] S21, using the standard classification features as an instruction, activate and illuminate the light-projecting module located under the transparent base of the label scale, and assign the initial projection aperture range of the light-projecting module to the set field of view range;

[0016] S22, the projection module detects whether the object to be weighed exists within the initial projection aperture defined by the set field of view, projected upwards from its initial position.

[0017] If so, proceed to step S23;

[0018] If not, adjust the position until the object to be weighed is detected, then proceed to step S23;

[0019] S23, the light projection module reduces the projection aperture range of the initial projection aperture projected upward according to the set aperture reduction step size, and the light receiving module set on the top of the label scale identifies and draws the disappearance point of the projection light at the moment of disappearance;

[0020] S24, the light receiving module selects the shape of the object to be weighed represented by each of the vanishing points using a minimum bounding rectangle selection method, and then calculates the center point of the minimum bounding rectangle.

[0021] Preferably, lines are drawn connecting each pair of opposite corners of the minimum bounding rectangle, and the intersection of the two connecting lines is taken as the center point of the minimum bounding rectangle.

[0022] Preferably, the set field of view is a perfect circle, and the classification features of the object to be weighed include the boundary shape features in the boundary shape image, and / or the gap width features between the first contour and the second contour of the boundary shape image.

[0023] Preferably, the first contour of the object to be weighed is the smallest bounding rectangle that frames the object; the gap width feature is the average distance between each vertex of the smallest bounding rectangle and the boundary of the second contour, and the connecting line of the distance points to the center of the set field of view.

[0024] Preferably, the method for forming the standard classification features of the historical weighed objects obtained in step S1 includes the following steps:

[0025] A1, for each vertex of the minimum bounding rectangle corresponding to each historical weighed object of the same type and with the same pricing and similar boundary shape features, calculate the distance of the second contour within the set field of view bound to the historical weighed object.

[0026] A2, calculate the average distance of each vertex in the minimum bounding rectangle corresponding to each historical weighing object that has a point-position correspondence in step A1;

[0027] A3. For any of the historical weighing objects obtained in step A1, expand or shrink the boundary shape using the average value corresponding to each vertex obtained in step A2 to obtain a standard boundary shape. The classification features of the standard boundary shape are used as the standard classification features.

[0028] Preferably, step S4, the method for performing maximum similarity matching, includes the following steps:

[0029] S41, perform boundary shape feature similarity matching between the object to be weighed and the standard classification features corresponding to each of the historical weighed objects obtained in step S1;

[0030] S42, determine whether the number of similarity matching results in step S41 is greater than or equal to "2".

[0031] If so, proceed to step S43;

[0032] If not, when the number of similarity matching results output in step S41 is "1", the matching result is taken as the classification result of the object to be classified; when the number of similarity matching results output in step S41 is "0", the classification is determined to be a failure.

[0033] S43, perform gap width feature similarity matching between the object to be weighed and each similarity matching result output in step S41;

[0034] S44, Determine whether the similarity matching result from step S43 is unique.

[0035] If so, the matching result shall be taken as the classification result for the object to be classified;

[0036] If not, the classification is deemed a failure.

[0037] This application also provides a method for batch parameter updating of IoT tag scales, which updates the parameters of weighed objects in batches for an IoT tag scale with an AI recognition module, including the following steps:

[0038] L1, press the same category set change button on each smart tag scale of the Internet of Things. Each smart tag scale obtains the category set bound to the button by pressing the button. The appearance characteristics of the weighed objects under the category set are similar but not the same type.

[0039] L2, each weighing object in the classification set whose parameters are to be updated is placed on a corresponding smart label scale for classification and identification;

[0040] L3, for the first weighed item successfully classified on the first smart tag scale, the parameters of the first weighed item are synchronously updated and marked in the classification set associated with the currently pressed classification set change button on each of the other second smart tag scales connected to the first smart tag scale;

[0041] L4, each of the second smart label scales takes the user's completion of parameter update for the first weighed object on the first smart label scale as an instruction, and performs parameter synchronization update for the first weighed object that has been marked with a synchronization update mark in the category set associated with the currently pressed category set change button in step L3.

[0042] The present invention has the following beneficial effects:

[0043] 1. The label scale uses the user's press of the category set change button as an instruction to limit the category matching object to each weighed object that has been bound to the category set change button, thus narrowing the matching range of the classification characteristics of the objects to be weighed and greatly improving the classification speed.

[0044] 2. The classification algorithm uses the boundary shape image between the first contour of the object to be weighed and the second contour within the set field of view of the AI ​​recognition module in the label scale as the classification feature. It no longer needs to capture the subtle feature differences between the object to be weighed and weighing objects that are similar in appearance but not of the same type. That is, it improves the classification accuracy of objects that are similar in appearance but not of the same type without improving the classification accuracy from the perspective of improving the classification algorithm.

[0045] 3. For weighing objects that are similar in size and have different prices but the same type, by setting the same field of view depth for the AI ​​recognition module, the different size characteristics between the weighing objects of the same type are reflected in the gap width between the first contour and the second contour mentioned above. The gap width is used as one of the classification features of the boundary shape image. Combined with the use of the classification set change button, the accurate classification of weighing objects that are similar in size and have different prices is achieved.

[0046] 4. When updating the parameters of weighed items in batches on various smart tag scales connected to the Internet of Things (IoT), the following steps are first performed: Classify and identify the weighed items whose parameters are to be updated within the category set bound to the currently pressed category set change button on each smart tag scale. Then, for the first weighed item successfully classified on the first smart tag scale, mark it for parameter synchronization update in the category set associated with the currently pressed category set change button on each of the other second smart tag scales connected to the first smart tag scale. Finally, after the user completes the parameter update for the first weighed item on the first smart tag scale, each second smart tag scale uses this as an instruction to synchronize the parameter update of the first weighed item marked for synchronization update. This achieves batch parameter synchronization update of weighed items belonging to the same category set on multiple smart tag scales connected to the IoT.

[0047] 5. The shape features of weighing objects of the same type are smoothed by using the minimum bounding rectangle to reduce the difference in boundary shape between the first contour (preferably the minimum bounding rectangle) and the second contour within the set field of view of weighing objects of the same type with different shapes. This helps to reduce the algorithm complexity of subsequent classification based on boundary shape images as classification features and improves the classification accuracy of weighing objects with similar but different appearance features.

[0048] 6. Using the center point of the smallest bounding rectangle as the reference point for finding the center point of the set field of view, it becomes possible for the AI ​​recognition module to cover the first outline of the weighed object with its set field of view by adjusting its position. This further makes it possible to find the subtle differences in the boundary shape image features between weighed objects that are similar in appearance but not of the same type.

[0049] 7. Set the field of view to a perfect circle. For weighing objects that are not perfectly circular, no matter how the weighing object is rotated or placed within a 360° range, the boundary shape formed by the first outline obtained by projecting light onto the weighing object and the second outline of the set field of view that is perfect circle is always the same. This ensures the consistency of the classification characteristics of the weighing object and helps to improve the classification accuracy of weighing objects that are similar in appearance but not of the same type. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0051] Figure 1 This is a step diagram illustrating a method for identifying and classifying objects to be weighed using an IoT tag scale with an AI recognition module, as provided in an embodiment of the present invention.

[0052] Figure 2 It is a schematic diagram of the boundary shape image between the first contour of the weighing object and the second contour of the corresponding set field of view, which is collected under the set field of view range and set field of view depth bound to the weighing object.

[0053] Figure 3 This is a schematic diagram showing the light projected from the transparent bottom of the label scale by the light projection module and received by the light receiving module;

[0054] Figure 4 This is an example image of the boundary shape between the first contour of the object to be weighed, which is selected by the smallest bounding rectangle, and the second contour, which sets the field of view. Detailed Implementation

[0055] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0056] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual images. They should not be construed as limiting the scope of this patent. To better illustrate the embodiments of the present invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions of the product. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0057] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "inner," and "outer" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present patent. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0058] In the description of this invention, unless otherwise explicitly specified and limited, the term "connection" or similar designation indicating a connection between components should be interpreted broadly. For example, it can refer to a fixed connection, a detachable connection, or an integral part; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can refer to the internal communication between two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0059] First, it should be noted that the items to be weighed in this embodiment refer to various types of goods to be weighed and priced, and these items belong to a specific category set. In this embodiment, the classification set is based on differences in size characteristics between the items to be weighed, and items within the same category set have similar appearance characteristics but different types. For example, oranges and melons have sizes d1 and d2, respectively. Assuming that there is a size difference between d1 and d2 according to the size classification rules, oranges are classified into the first category set, and melons are classified into the second category set.

[0060] Similarity in appearance characteristics among weighed objects within the same classification set includes any one or more of the following: similarity in size, color, and shape. Size similarity refers to the difference in the coverage area of ​​the smallest bounding rectangle of the symmetrical weighed objects being less than a preset first difference threshold. For example, assuming oranges and tangerines belong to the same first classification set, let a1 be the coverage area of ​​the smallest bounding rectangle for tangerines and a2 be the coverage area for tangerines, and let a1 - a2 be less than the preset area difference threshold (first difference threshold). Then, oranges and tangerines are considered to have similar size characteristics. The method of selecting weighed objects using the smallest bounding rectangle will be explained in detail later. Color similarity refers to the difference in the average color value of the weighed objects being less than a preset second difference threshold. For example, for oranges and tangerines, the color features of each pixel in the image of the weighed object selected by the smallest bounding rectangle are identified, and then the average color value of each pixel is taken (i.e., the color mean). The difference in the color mean is also the absolute value of the difference between the color mean values ​​of the two weighed objects. Shape similarity refers to the difference in the outline shape of the weighed objects being less than a preset third difference threshold. The outline shape of the object being weighed referred to here is Figure 2 The image shows the boundary shape between the first contour of the weighing object and the second contour within a defined field of view. The difference in the contour shape of the weighing object refers to the dissimilarity value between the boundary shape images of the two weighing objects, where the dissimilarity value is 1 minus the similarity value. How the similarity value between the boundary shape images of the two weighing objects is calculated will be explained in detail later.

[0061] This invention provides an IoT tag scale with an AI recognition module. The object to be weighed belongs to a classification set. When the built-in conventional image recognition method fails to classify the object, such as... Figure 1 As shown, the following steps are used to identify, classify, and price the items to be weighed:

[0062] S1, press the category set change button corresponding to the item to be weighed, and the label scale will adjust the set field of view and set field of view depth of the AI ​​recognition module by pressing the button;

[0063] For example, if the item to be weighed is an orange, its classification set is the first classification set, which corresponds to the classification set change button b1. Therefore, the classification set change button for oranges is b1. When the label scale fails to classify the orange, it will sound an alarm and prompt the user to press the classification set change button b1. After the user presses button b1, the label scale will adjust the field of view of its AI recognition module to the set field of view corresponding to button b1, and also adjust the depth of view to the set depth of view corresponding to button b1. Figure 2In the diagram, circle 100 represents the set field of view range, and height H represents the set field of view depth. The size of the set field of view range and the set field of view depth are related to the size of the object being weighed. Since how to set the set field of view range and set field of view depth for each object with different size characteristics is not within the scope of protection claimed in this application, and this technical point does not affect the solution to the technical problem, it will not be elaborated in detail.

[0064] The label scale automatically adjusts the set field of view and set depth of view of the AI ​​recognition module based on the user's button press command. Simultaneously, it acquires the standard classification features of historical weighed items within the classification set to which the item to be weighed belongs. These standard classification features are related to the distances between the vertices of the minimum bounding rectangle of the selected historical weighed items and the second contour of the set field of view bound to the historical weighed items. The method for forming the standard classification features will be explained in detail after the technical aspects such as the generation method of the minimum bounding rectangle are elaborated upon later.

[0065] In step S1, the field of view and depth of view of the AI ​​recognition module are adjusted to match the set field of view and depth of view corresponding to the object to be weighed. Figure 1 As shown, the method for identifying, classifying, and pricing the items to be weighed using the label scale provided in this embodiment proceeds to the following steps:

[0066] S2, Select the object to be weighed using the minimum bounding rectangle, and then calculate the center point of the minimum bounding rectangle. Specific steps include:

[0067] S21, using the standard classification characteristics of each historical weighed object in the classification set to which the object to be weighed belongs as an instruction, activate and light up the projection module set under the transparent base of the label scale, and assign the initial projection aperture range of the projection module to the set field of view range adjusted by the AI ​​recognition module in step S1.

[0068] It should be noted that the light projection module is located under the transparent base of the label scale, and the light projection direction is vertically upward, projecting light upward in the form of an aperture. After the light projection module is activated, the initial projection aperture range of its upward light projection is the set field of view range.

[0069] S22, the projection module detects whether there is an object to be weighed within the initial projection aperture defined by the set field of view, projected upwards from its initial position.

[0070] If so, proceed to step S23;

[0071] If not, adjust the position according to the preset trajectory until the object to be weighed is detected, then proceed to step S23;

[0072] There are many existing methods for detecting objects using a projection module. For example, a distance sensor can be used to detect the object. The projection module emits a distance detection signal upwards within the initial projection aperture range. When it encounters the object, the detection signal is blocked, and the distance between the projection module and the object can be calculated based on the transmission speed and duration of the detection signal. This technique can also be used to determine whether the object is within the initial projection aperture range. For instance, if the detection signals at all boundary points of the aperture are blocked at similar distances, it is determined that an object exists within the initial projection aperture range.

[0073] It should be noted that the number of objects detected within the initial projection aperture may not be just one. For example, if a consumer buys five oranges to be weighed and priced together, multiple oranges may be detected within the initial projection aperture. The method by which the projection module adjusts its position according to a preset trajectory is, for example, to adjust its position from the initial position by moving from left to right and from top to bottom according to a set detection step size.

[0074] S23, the projection module projects the initial projection aperture range upward by reducing the projection aperture step size according to the set aperture, and the light receiving module set on the top of the label scale identifies and draws the disappearance point when the projected light disappears;

[0075] For example, Figure 3 In this process, the light-projecting module 200 projects light upwards from the transparent bottom of the label scale, which is then received by the light-receiving module 300. When the light-projecting module projects light within the initial projection aperture range, all the light projected upwards from the boundaries of the initial projection aperture range is received by the light-receiving module 300. As the light-projecting module gradually reduces the projection aperture range according to the set aperture reduction step size, if the aperture boundary encounters an obstruction from the object to be weighed, the light-receiving module 300 will be unable to receive the light projected upwards from that boundary point. At this time, the light-receiving module 300 will record the disappearance time of the light projected at that boundary point and the location of that boundary point, until all the upward-projected light from all boundary points disappears. Then, the boundary points at each disappearance time of the light generated by the aperture boundary are plotted to characterize the shape features of the object to be weighed.

[0076] It should be noted that in conventional solutions, the shape features of the object to be weighed are identified through image recognition. However, in weighing applications, due to varying lighting conditions, identifying the shape features of the object from top to bottom using image recognition may result in inaccurate shape features due to factors such as lighting and shadows. This is unacceptable in the scenario where this application aims to improve the classification accuracy of objects with similar but different appearance features, especially when conventional image classification algorithms have failed to classify objects with similar but different appearance features.

[0077] S24, the light receiving module selects the shape of the object to be weighed by each vanishing point using the minimum bounding rectangle selection method, and then calculates the center point of the minimum bounding rectangle;

[0078] The shape of the object to be symmetrical is selected by using the minimum bounding rectangle method, which is an existing method and will not be explained in detail.

[0079] In this embodiment, the method for calculating the center point of the minimum bounding rectangle is as follows:

[0080] like Figure 2 As shown, lines are drawn connecting each pair of opposite corners of the minimum circumscribed rectangle 500, and the intersection point 400 of the two connecting lines serves as the center point of the minimum circumscribed rectangle 500. It should be noted that the minimum circumscribed rectangle is formed on the light receiving module, i.e., on the top of the label scale.

[0081] It's important to note that using the center point of the minimum bounding rectangle as the center point for finding the defined field of view makes it possible for the AI ​​recognition module to adjust its position to cover the first contour of the object to be weighed within its field of view. This further makes it possible to find subtle differences in the boundary shape image features between weighing objects that are similar in appearance but not of the same type. Furthermore, using the minimum bounding rectangle smooths the shape features of weighing objects of the same type, reducing the difference in boundary shape between the first contour of different weighing objects of the same type and the second contour of the defined field of view. This helps reduce the algorithmic complexity of subsequent classification using boundary shape images as classification features (i.e., maximum similarity matching in step S4), because only the difference in boundary shape between the shape contour of the minimum bounding rectangle (the first contour) and the second contour formed by the defined field of view needs to be considered, without needing to consider the details of the shape contour of the weighing objects selected within the minimum bounding rectangle. This improves the classification accuracy for weighing objects with similar but different appearance features.

[0082] After selecting the object to be weighed using the minimum bounding rectangle in step S2 and calculating the center point of the minimum bounding rectangle, the method provided in this embodiment for identifying, classifying, and pricing the object using an IoT tag scale with an AI recognition module proceeds to the next step:

[0083] S3, under the set field of view depth, adjust the position of the AI ​​recognition module so that the center point is the center point of the field of view generated by the AI ​​recognition module within the set field of view range;

[0084] For example, such as Figure 2As shown, under a set field of view depth H, the position of the AI ​​recognition module is adjusted so that the center point 400 becomes the center point of the field of view generated by the AI ​​recognition module within the set field of view, i.e., the center point 400 overlaps with the center point of the field of view. In this embodiment, the set field of view is preferably a perfect circle. For weighing objects that are not perfectly circular, no matter how the weighing object is rotated or placed within a 360° range, the boundary shape formed between the first contour obtained by projecting light onto the weighing object and the second contour of the set field of view, which is a perfect circle, is always the same. This ensures the consistency of the classification features of the weighing object and helps to improve the classification accuracy of weighing objects with similar but different appearance features.

[0085] S4. Extract the boundary shape image between the first contour of the object to be weighed selected by the minimum bounding rectangle and the second contour within the set field of view as the classification feature of the object to be weighed. Then, perform maximum similarity matching with each standard classification feature of each historical weighed object obtained in step S1. The matching result is used as the classification result and priced accordingly.

[0086] Figure 4 In the image, the gray area represents an example of a boundary shape image. In this embodiment, the classification features include the boundary shape features in the boundary shape image (i.e.,...) Figure 4 The gap width feature between the first contour (preferably the smallest bounding rectangle of the object to be weighed) and the second contour (preferably a perfect circle) of the shape of the gray portion shown), and / or the boundary shape image. The gap width feature is the average distance between each vertex of the smallest bounding rectangle and the boundary of the second contour, and the connecting line forming this distance points to the center of the circle within the defined field of view.

[0087] For example, Figure 4 In the attached diagram, the reference numeral "600" marks the center of the defined field of view. The distances between the vertices p1, p2, p3, and p4 of the smallest bounding rectangle and the boundary of the second contour are l1, l2, l3, and l4, respectively. Therefore, the gap width characteristic of the object to be weighed is (l1+l2+l3+l4) / 4.

[0088] The following explains the method for forming the standard classification characteristics of historically weighed objects:

[0089] In this embodiment, the method for forming standard classification features includes the following steps:

[0090] A1, for each vertex of the minimum bounding rectangle corresponding to each historical weighing object of the same type and with the same price and similar boundary shape features, calculate the distance of the second contour bound to the set field of view of the historical weighing object.

[0091] The method for calculating distance is described above. Figure 4The example provided will not be repeated here. It's important to clarify that "same type and same price" means, for instance, that to establish a standard classification feature for "oranges," all historically weighed items obtained in step A1 are "oranges." "Same price" means that the selling price of these "oranges" classified through steps S1-S4 is the same. "Similarity in boundary shape features between weighed items" means, for example, that the boundary shape features of a certain orange are... Figure 4 As shown in the gray area, if the overlap between the boundary shape features of another orange and this gray area is greater than a preset overlap threshold, then the two are determined to have similar boundary shape features. The calculation of the overlap between the gray areas uses existing methods and will not be specifically explained.

[0092] A2, calculate the average distance of each vertex with a point-to-point correspondence in the smallest bounding rectangle corresponding to each historical weighing object for which the distance calculation was completed in step A1;

[0093] For example, if vertex px1 in the historical weighed object x1 obtained in step A1 overlaps with vertex px2 in the historical weighed object x2, then it is determined that px1 and px2 have a point correspondence relationship, and then the average value of the distances calculated by step A1 corresponding to px1 and px2 is calculated respectively.

[0094] A3. For any historical weighing object boundary shape obtained in step A1, expand or shrink the boundary shape by using the average value of each vertex obtained in step A2 to obtain the standard boundary shape. The classification features of the standard boundary shape are used as the standard classification features.

[0095] For example, in step A3, the historical weighed object that underwent the expansion or contraction operation is... Figure 4 The minimum bounding rectangle 700 is shown. When it is necessary to expand or shrink the vertex p1 in the minimum bounding rectangle 700, the average distance between each vertex of other historical weighed objects obtained in step A1 that has a point correspondence with vertex p1 and the second contour is extracted, and then the distance between vertex p1 and the second contour is expanded or shrunk to this average value.

[0096] In step S4, the method for performing maximum similarity matching includes the following steps:

[0097] S41, perform boundary shape feature similarity matching between the object to be weighed and the standard classification features corresponding to each historical weighed object obtained in step S1 (such as the overlap matching in the example above).

[0098] S42, determine whether the number of similarity matching results in step S41 is greater than or equal to "2".

[0099] If so, proceed to step S43;

[0100] If not, when the number of similarity matching results output in step S41 is "1", the matching result is taken as the classification result of the object to be compared; when the number of similarity matching results output in step S41 is "0", the classification is determined to be unsuccessful and an alarm is triggered.

[0101] S43, perform further gap width feature similarity matching between the object to be compared with each similarity matching result output in step S41;

[0102] S44, Determine whether the similarity matching result from step S43 is unique.

[0103] If so, the matching result shall be used as the classification result for the object to be classified;

[0104] If not, the classification will be deemed a failure and an alarm will be triggered.

[0105] It should be noted that for weighing items that are similar in size but different in price but the same type, such as two types of textured melons priced differently due to their different sizes, by setting the same field of view depth for the AI ​​recognition module, the different size characteristics between these similar weighing items are reflected in the gap width between the first and second contours mentioned above. The gap width is one of the classification features of the boundary shape image. Through steps S42-S44, combined with the use of the classification set change button, the accurate classification of weighing items that are similar in size but different in price is achieved.

[0106] This embodiment also provides a method for batch parameter updating of IoT tag scales, specifically for updating the parameters of weighed objects in batches for an IoT tag scale with an AI recognition module, including the following steps:

[0107] L1, press the same category set change button on each smart tag scale of the Internet of Things. Each smart tag scale obtains the category set bound to the button by pressing the button. The appearance characteristics of the weighed objects under the category set are similar but not the same type. The similar appearance characteristics are explained in detail above and will not be repeated. The method of connecting the tag scales through the Internet of Things is an existing method and will not be explained here either.

[0108] L2, each weighed object in the classification set whose parameters need to be updated is placed on a corresponding smart tag scale for classification and identification. The classification and identification method is the above steps S1-S4, which will not be repeated here.

[0109] L3, for the first weighed item successfully classified on the first smart tag scale, in the classification set associated with the currently pressed classification set change button on each of the other second smart tag scales connected to the first smart tag scale, the parameters of the first weighed item are synchronously updated and marked.

[0110] For example, in step L2, the first smart tag scale successfully classifies "oranges" as the first weighed item. In the classification sets associated with the currently pressed classification set change buttons on other second smart tag scales connected to the first smart tag scale, the parameters of the first weighed item, i.e., "oranges," are synchronously updated and marked.

[0111] L4, each of the second smart label scales takes the user's completion of parameter update for the first weighed object on the first smart label scale as an instruction, and performs parameter synchronization update for the first weighed object that has been marked with a synchronization update mark in the category set associated with the currently pressed category set change button in step L3.

[0112] When the category change button is pressed, it remains pressed and will only pop up when pressed again.

[0113] In summary, the label scale provided in this application uses the user pressing the category set change button as an instruction to limit the classification matching objects to all weighed objects that have been bound to the category set change button, thus narrowing the matching range of classification features for subsequent weighed objects and significantly improving classification speed. The provided classification algorithm uses the boundary shape image between the first contour of the weighed object and the second contour within the set field of view of the AI ​​recognition module in the label scale as the classification feature. This eliminates the need to capture subtle feature differences between the weighed object and weighed objects with similar but different appearances, improving the classification accuracy for weighed objects with similar but different appearances without needing to improve the classification algorithm itself. The provided label scale IoT batch parameter update method achieves synchronous batch parameter updates for all weighed objects under the same category simply by pressing the category set change button.

[0114] It should be stated that the above-described specific embodiments are merely preferred embodiments of the present invention and the technical principles employed. Those skilled in the art should understand that various modifications, equivalent substitutions, and variations can be made to the present invention. However, such variations, as long as they do not depart from the spirit of the present invention, should be within the scope of protection of the present invention. Furthermore, some terminology used in this specification and claims is not limiting, but merely for ease of description.

Claims

1. An IoT tag scale with an AI recognition module, characterized in that, The item to be weighed belongs to a classification set. When image recognition fails to classify the item, the following steps are taken to identify, classify, and price the item: S1, press the category set change button corresponding to the object to be weighed, and the label scale adjusts the set field of view and set field of view depth of the AI ​​recognition module by pressing the button as an instruction, and obtains the standard classification features of each historical weighed object in the category set to which the object to be weighed belongs. S2, select the object to be weighed using the minimum bounding rectangle, and then calculate the center point of the minimum bounding rectangle; S3, under the set field of view depth, adjust the position of the AI ​​recognition module so that the center point is the center point of the field of view generated by the AI ​​recognition module within the set field of view range; S4, extract the boundary shape image between the first contour of the object to be weighed selected by the minimum bounding rectangle and the second contour of the set field of view as the classification feature of the object to be weighed, and then perform maximum similarity matching with each of the standard classification features of the associated historical weighed objects obtained in step S1. The matching result is used as the classification result and priced. Step S2 specifically includes the following steps: S21, using the standard classification features as an instruction, activate and illuminate the light-projecting module located under the transparent base of the label scale, and assign the initial projection aperture range of the light-projecting module to the set field of view range; S22, the projection module detects whether the object to be weighed exists within the initial projection aperture defined by the set field of view, projected upwards from its initial position. If so, proceed to step S23; If not, adjust the position until the object to be weighed is detected, then proceed to step S23; S23, the light projection module reduces the projection aperture range of the initial projection aperture projected upward according to the set aperture reduction step size, and the light receiving module set on the top of the label scale identifies and draws the disappearance point of the projection light at the moment of disappearance; S24, the light receiving module selects the shape of the object to be weighed represented by each of the vanishing points using a minimum bounding rectangle selection method, and then calculates the center point of the minimum bounding rectangle.

2. The IoT tag scale with AI recognition module according to claim 1, characterized in that, The weighing objects belonging to different classification sets have differences in size characteristics; the weighing objects in the same classification set have similar appearance characteristics but are not of the same type; the similarity of appearance characteristics of the weighing objects includes any one or more of size characteristics, color characteristics, and shape characteristics; wherein, size characteristics are similar when the difference in the coverage area of ​​the smallest bounding rectangle of the weighing object is less than a preset first difference threshold; color characteristics are similar when the difference in the average color value of the weighing objects is less than a preset second difference threshold; and shape characteristics are similar when the difference in the outline shape of the weighing objects is less than a preset third difference threshold.

3. The IoT tag scale with AI recognition module according to claim 1, characterized in that, Draw lines connecting each pair of opposite corners of the minimum bounding rectangle, and the intersection of the two connecting lines is taken as the center point of the minimum bounding rectangle.

4. The IoT tag scale with AI recognition module according to claim 1, characterized in that, The set field of view is a perfect circle, and the classification features of the object to be weighed include the boundary shape features in the boundary shape image and / or the gap width features between the first contour and the second contour of the boundary shape image.

5. An IoT tag scale with an AI recognition module according to claim 4, characterized in that, The first contour of the object to be weighed is the smallest bounding rectangle that frames the object to be weighed; the gap width feature is the average distance between each vertex of the smallest bounding rectangle and the boundary of the second contour, and the connecting line of the distance points to the center of the set field of view.

6. An IoT tag scale with an AI recognition module according to claim 1, characterized in that, The method for forming the standard classification features of the historical weighed objects obtained in step S1 includes the following steps: A1, for each vertex of the minimum bounding rectangle corresponding to each historical weighed object of the same type and with the same pricing and similar boundary shape features, calculate the distance of the second contour within the set field of view bound to the historical weighed object. A2, calculate the average distance of each vertex in the minimum bounding rectangle corresponding to each historical weighing object that has a point-position correspondence in step A1; A3. For any of the historical weighing objects obtained in step A1, expand or shrink the boundary shape using the average value corresponding to each vertex obtained in step A2 to obtain a standard boundary shape. The classification features of the standard boundary shape are used as the standard classification features.

7. An IoT tag scale with an AI recognition module according to claim 4, characterized in that, In step S4, the method for performing maximum similarity matching includes the following steps: S41, perform boundary shape feature similarity matching between the object to be weighed and the standard classification features corresponding to each of the historical weighed objects obtained in step S1; S42, determine whether the number of similarity matching results in step S41 is greater than or equal to "2". If so, proceed to step S43; If not, when the number of similarity matching results output in step S41 is "1", the matching result is taken as the classification result of the object to be classified; when the number of similarity matching results output in step S41 is "0", the classification is determined to be a failure. S43, perform gap width feature similarity matching between the object to be weighed and each similarity matching result output in step S41; S44, Determine whether the similarity matching result from step S43 is unique. If so, the matching result shall be taken as the classification result for the object to be classified; If not, the classification is deemed a failure.

8. A method for batch parameter updating of IoT tag scales, for batch updating of weighing parameters of an IoT tag scale with an AI recognition module as described in any one of claims 1-7, characterized in that, Including the following steps: L1, press the same category set change button on each smart tag scale of the Internet of Things. Each smart tag scale obtains the category set bound to the button by pressing the button. The appearance characteristics of the weighed objects under the category set are similar but not the same type. L2, each weighing object in the classification set whose parameters are to be updated is placed on a corresponding smart label scale for classification and identification; L3, for the first weighed item successfully classified on the first smart tag scale, the parameters of the first weighed item are synchronously updated and marked in the classification set associated with the currently pressed classification set change button on each of the other second smart tag scales connected to the first smart tag scale; L4, each of the second smart label scales takes the user's completion of parameter update for the first weighed object on the first smart label scale as an instruction, and performs parameter synchronization update for the first weighed object that has been marked with a synchronization update mark in the category set associated with the currently pressed category set change button in step L3.

Images