Safety shoe label processing identification positioning method and system

Abstract

The application provides a kind of safety shoes label processing identification positioning method and system, method includes: obtaining historical safety shoes image;Extract the image feature vector of historical safety shoes image;Based on image feature vector, calculate the feature richness and feature uniqueness of each historical safety shoes image, and determine the target historical safety shoes image based on feature richness and feature uniqueness;Target historical safety shoes image is used to train convolutional neural network model, and the identification positioning parameter extraction model is obtained;The safety shoes image to be processed is input into identification positioning parameter extraction model, and the identification positioning parameter of the safety shoes image to be processed is obtained.The application calculates the aggregation coefficient, isomorphism coefficient, average euclidean distance and outlier of image feature vector, selects high-quality target image from two dimensions of representation and uniqueness for model training, outputs identification positioning parameter, and is beneficial to improve safety shoes production efficiency.

Description

Technical Field

[0001] This invention relates to the field of safety shoe technology, and more specifically, to a method and system for marking and positioning safety shoes in label processing. Background Technology

[0002] Safety shoes, as special labor protection equipment, typically require labels on their uppers to indicate brand, model, size, and other information. Labeling is a crucial final step in the production and processing of safety shoes. Currently, label placement in safety shoe labeling can be done manually, with operators determining the label's position based on the shoe type and size; however, this method is inefficient. Therefore, developing a more efficient label placement method that can adapt to different shoe types and sizes has become a pressing technical problem. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for marking and positioning safety shoes for label processing, so as to improve the above-mentioned problems.

[0004] To achieve the above objectives, this application provides the following technical solution: On one hand, embodiments of this application provide a method for identifying and positioning safety shoe labels, the method comprising: Obtain historical safety shoe images; Extract image feature vectors from historical safety shoe images; calculate the feature richness and feature uniqueness of each historical safety shoe image based on the image feature vectors, and determine the target historical safety shoe image based on the feature richness and feature uniqueness. A convolutional neural network model is trained using historical images of target safety shoes to obtain a marker localization parameter extraction model; the safety shoe image to be processed is then input into the marker localization parameter extraction model to obtain the marker localization parameters of the safety shoe image to be processed.

[0005] Secondly, this application provides an identification and positioning system for safety shoe labeling, the system comprising: The acquisition module is used to acquire historical safety shoe images; The extraction module is used to extract image feature vectors from historical safety shoe images; calculate the feature richness and feature uniqueness of each historical safety shoe image based on the image feature vectors, and determine the target historical safety shoe image based on the feature richness and feature uniqueness. The localization module is used to train a convolutional neural network model using historical images of the target safety shoes to obtain a marker localization parameter extraction model; the safety shoe image to be processed is input into the marker localization parameter extraction model to obtain the marker localization parameters of the safety shoe image to be processed.

[0006] Thirdly, this application provides a labeling and positioning device for safety shoe label processing, the device including a memory and a processor. The memory is used to store a computer program; the processor is used to execute the computer program to implement the steps of the above-described labeling and positioning method for safety shoe label processing.

[0007] Fourthly, this application provides a readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described identification and positioning method for safety shoe labeling.

[0008] The beneficial effects of this invention are as follows: This invention evaluates the representativeness of an image from two dimensions: local density and local structural consistency, by calculating the clustering coefficient and isomorphism coefficient of the image feature vector. Simultaneously, it evaluates the uniqueness of an image from two dimensions: global isolation and local density difference, by calculating the average Euclidean distance and outlier degree. A weighted fusion of representativeness and uniqueness yields a value score, allowing for the selection of high-quality target images for model training. Through diverse, high-quality training samples, the model can accurately learn the mapping relationship between safety shoe images and label locations, automatically and quickly outputting the corresponding label positioning parameters for new safety shoe images, thereby improving production efficiency.

[0009] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of the identification and positioning method for safety shoe label processing as described in this embodiment of the invention; Figure 2 This is a schematic diagram of the identification and positioning system for safety shoe labeling processing as described in this embodiment of the invention; Figure 3 This is a schematic diagram of the identification and positioning equipment for safety shoe labeling processing described in this embodiment of the invention. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0013] It should be noted that similar reference numerals or letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0014] Example 1 like Figure 1 As shown in the figure, this embodiment provides a method for marking and positioning safety shoes through label processing. The method includes steps S1, S2 and S3.

[0015] Step S1: Obtain historical safety shoe images; In this step, historical safety shoe images can be retrieved from the company's production database. These historical safety shoe images are two-dimensional images with labels already attached, which may include brand logos, model logos, size logos, etc. All historical safety shoe images are of the same size. The historical safety shoe images can cover various safety shoe samples of different shoe types and sizes, and the number of images can be determined according to actual needs.

[0016] Step S2: Extract the image feature vectors of historical safety shoe images; calculate the feature richness and feature uniqueness of each historical safety shoe image based on the image feature vectors, and determine the target historical safety shoe image based on the feature richness and feature uniqueness. The specific implementation steps of this step include step S21 and step S22; Step S21: For each historical safety shoe image: add T different levels of noise to the historical safety shoe image to obtain T noisy historical safety shoe images. Input the T noisy historical safety shoe images into the feature extraction model to obtain T feature vectors. Calculate the mean of the T feature vectors to obtain the mean vector. Perform L2 normalization on the mean vector to obtain the image feature vector corresponding to the historical safety shoe image. In this step, T different levels of noise can be added to each historical safety shoe image, where T can be 5. For example, Gaussian noise, salt-and-pepper noise, Poisson noise, multiplicative noise, and uniformly distributed noise can be added respectively, resulting in five noisy historical safety shoe images. These five noisy images are then input into a pre-trained feature extraction model to obtain five feature vectors. The feature extraction model uses a ResNet-50 network pre-trained on the ImageNet dataset. The last fully connected classification layer is removed, and the global average pooling layer is retained. The output of the global average pooling layer is used as the image's feature vector. The average of the five feature vectors is calculated to obtain the mean vector. Then, L2 normalization is performed on the mean vector, i.e., the mean vector is divided by its own magnitude, to obtain the final image feature vector corresponding to the historical safety shoe image.

[0017] Step S22: Calculate the Euclidean distance between any two image feature vectors, sort all Euclidean distances from smallest to largest, take the median, and multiply the preset value by the median to obtain the first value; calculate the feature richness and feature uniqueness of each historical safety shoe image based on the image feature vector of each historical safety shoe image and the first value; sum the feature richness and feature uniqueness of each image feature vector by weight to obtain the value of each image feature vector; sort the historical safety shoe images corresponding to the top N image feature vectors from largest to smallest value as the target historical safety shoe images.

[0018] In this step, the preset value can be 0.2, and the first value is used as the distance threshold for subsequent neighbor point judgment. The weight of feature richness can be set to 0.5, the weight of feature uniqueness can be set to 0.5, and N is a preset positive integer that can be customized according to actual needs.

[0019] In this step, feature richness reflects the representativeness of a sample in the feature space. Samples with high richness are located in dense regions of the data distribution, reflecting the common features of most samples. Using them for training allows the model to learn the main data distribution patterns. Feature uniqueness reflects the diversity of a sample in the feature space. Samples with high uniqueness are located in sparse regions or marginal positions of the data distribution, supplementing the lack of differential information in the dataset. Using them for training can enhance the model's generalization ability and avoid overfitting. Combining the two ensures that the training data covers mainstream features while also considering adaptability to marginal cases, thus selecting balanced, high-quality training samples that allow the model to maintain good generalization performance while learning the main patterns.

[0020] Meanwhile, in this step, the specific implementation steps for calculating the feature richness and feature uniqueness of each historical safety shoe image based on the image features and the first value include step S221; Step S221: Calculate the clustering coefficient of each image feature vector based on the image feature vector and the first numerical value, and calculate the isomorphism coefficient of each image feature vector based on the image feature vector and the first numerical value. Sum the clustering coefficient and the isomorphism coefficient by weight to obtain the feature richness of each image feature vector. Calculate the average Euclidean distance and outlier of each image feature vector based on the image feature vector and the first numerical value. Sum the average Euclidean distance and the outlier by weight to obtain the feature uniqueness of each image feature vector.

[0021] In this step, the specific implementation steps for calculating the aggregation coefficient of each image feature vector based on the image feature vector and the first numerical value include step S2211; Step S2211: For each image feature vector, count the number of other image feature vectors whose Euclidean distance to it is less than the first value, and record it as the second value; after calculating the second value corresponding to each image feature vector, record the largest second value as the third value; divide the second value corresponding to each image feature vector by the third value to obtain the clustering coefficient of each image feature vector.

[0022] This step can be understood as follows: the second value reflects the local density of the image feature vector in the feature space; the higher the density, the more representative the image is. After calculating the second value for all image feature vectors, find the maximum value and denote it as the third value. Divide the second value for each image feature vector by the third value to obtain the clustering coefficient of that vector. The clustering coefficient ranges from 0 to 1; a larger value indicates that the image is located in a denser region in the feature space, and is more representative.

[0023] In this step, the specific implementation steps for calculating the isomorphism coefficient of each image feature vector based on the image feature vector and the first numerical value include step S2212; Step S2212: For each image feature vector: Sort its Euclidean distances to all other image feature vectors in ascending order, select the first K Euclidean distances corresponding to the other image feature vectors and set them together to obtain the first set of each image feature vector; calculate the intersection of the first sets of each pair of image feature vectors, count the number of image feature vectors in the intersection, and record it as the fourth value; add up all the fourth values ​​corresponding to each image feature vector to obtain the fifth value; after calculating the fifth value corresponding to each image feature vector, record the largest fifth value as the sixth value; divide the fifth value corresponding to each image feature vector by the sixth value to obtain the isomorphism coefficient of each image feature vector.

[0024] This step can be understood as follows: For any two different image feature vectors, calculate the intersection of their respective first sets, and count the number of image feature vectors in the intersection, denoted as the fourth value. The fourth value reflects the similarity between the two vectors in their local neighborhood structure; the more shared nearest neighbors, the more similar their local structures. For each image feature vector, add it to the fourth value calculated from all other image feature vectors to obtain the fifth value. This value reflects the overall similarity of this vector with all other vectors in their local structure.

[0025] After calculating the fifth value corresponding to all image feature vectors, find the maximum value and denote it as the sixth value. Divide the fifth value of each image feature vector by the sixth value to obtain the isomorphism coefficient of that vector. The isomorphism coefficient ranges from 0 to 1; a larger value indicates a higher consistency in local structure between the image and its surrounding images. K is a preset positive integer that can be customized according to actual needs. The clustering coefficient and isomorphism coefficient were calculated through the above steps. The weights of the clustering coefficient and isomorphism coefficient can be set to 0.6 and 0.4, respectively. The feature richness is obtained by weighted summation of the two coefficients.

[0026] In addition, the specific implementation steps for calculating the average Euclidean distance and outlier degree of each image feature vector based on the image feature vector and the first numerical value include steps S2213-S2215; Step S2213: For each image feature vector: add its Euclidean distance to all other image feature vectors and calculate the average to obtain the average Euclidean distance; This step can be understood as follows: For each image feature vector, sum its Euclidean distances to all other image feature vectors, then divide by the total number of other image feature vectors to obtain the average Euclidean distance of that image feature vector. The average Euclidean distance reflects the degree of isolation of the image feature vector in the overall feature space; the larger the distance, the further the historical safety shoe image is from other historical safety shoe images, and the higher its uniqueness.

[0027] Step S2214: For each image feature vector: sort its Euclidean distances with other image feature vectors in ascending order to obtain an Euclidean distance sequence; calculate the difference between adjacent Euclidean distances in the Euclidean distance sequence to obtain a difference sequence; calculate the mean and standard deviation of all differences in the difference sequence; in the difference sequence, check each difference one by one starting from the first difference; when the first difference greater than the mean plus twice the standard deviation is found, increment the index corresponding to the difference by one as the neighborhood scale H of the image feature vector; if no difference greater than the mean plus twice the standard deviation is found after traversing all differences, then set the neighborhood scale H of the image feature vector to the total number of historical safety shoe images minus one. Step S2215: Calculate the local outlier factor of each image feature vector relative to its neighborhood scale; calculate the outlier degree of each image feature vector based on the local outlier factor.

[0028] In this step, the specific implementation steps for calculating the local outlier factor of each image feature vector relative to its neighborhood scale include steps S22151-S22153. Step S22151: For each image feature vector, find the Hth Euclidean distance in its Euclidean distance sequence, use it as the neighborhood distance threshold of each image feature vector, and use the image feature vectors corresponding to the first H Euclidean distances as the neighborhood of each image feature vector. Step S22152: For each image feature vector, calculate its reachability distance to each image feature vector in its neighborhood. The reachability distance is calculated as follows: For two different image feature vectors, denoted as the first image feature vector and the second image feature vector respectively, take the larger value between the Euclidean distance between the first image feature vector and the second image feature vector and the neighborhood distance threshold of the second image feature vector as the reachability distance from the first image feature vector to the second image feature vector; average all reachability distances of each image feature vector to obtain the average reachability distance; divide the neighborhood scale H of each image feature vector by the average reachability distance to obtain the local reachability density of each image feature vector. Step S22153: For each image feature vector, sum the local reachability densities of all image feature vectors in its neighborhood, and then divide by the number of image feature vectors in the neighborhood to obtain the neighborhood average density; divide the neighborhood average density by the local reachability density of each image feature vector to obtain the local outlier factor of each image feature vector relative to its neighborhood scale.

[0029] Following this step, a multi-scale fusion mechanism can be introduced, specifically including steps S22154-S22156; Step S22154: Calculate the smaller neighborhood scale for each image feature vector, where half of the neighborhood scale H is rounded down to obtain the smaller neighborhood scale; then calculate the local outlier factor of each image feature vector relative to the smaller neighborhood scale. In this step, the calculation of the local outlier factor of each image feature vector relative to a smaller neighborhood scale is the same as steps S22151-S22153; only the neighborhood scale is replaced with a smaller neighborhood scale. Step S22155: Calculate the larger neighborhood scale of each image feature vector, where the larger neighborhood scale is the minimum between twice the neighborhood scale and the total number of historical safety shoe images minus one; then calculate the local outlier factor of each image feature vector relative to the larger neighborhood scale. In this step, the calculation of the local outlier factor of each image feature vector relative to a larger neighborhood scale is the same as steps S22151-S22153; only the neighborhood scale is replaced with a larger neighborhood scale. Step S22156: Weighted summation of the local outlier factor of each image feature vector compared to the local outlier factor at a neighborhood scale, compared to the local outlier factor at a smaller neighborhood scale, and compared to the local outlier factor at a larger neighborhood scale, to obtain the multi-scale local outlier factor of each image feature vector.

[0030] The local outlier factors at the three scales are weighted and summed, where the weight of the neighborhood scale can be set to 0.5, the weight of the smaller neighborhood scale can be set to 0.25, and the weight of the larger neighborhood scale can be set to 0.25, to obtain the multi-scale local outlier factors of each image feature vector.

[0031] The multi-scale fusion mechanism effectively avoids the problem of a single scale being too sensitive to the size of the neighborhood by integrating local outlier factors at small, adaptive, and large scales, making the identification of outliers more stable and reliable. At the same time, different scales can take into account the outlier features of sparse and dense regions, making the evaluation of feature uniqueness more accurate.

[0032] In step S2215, the specific implementation steps for calculating the outlier degree of each image feature vector based on the local outlier factor include step S22157. Step S22157: Find the maximum and minimum values ​​of the local outliers among all image feature vectors; for each image feature vector, subtract the minimum value from its local outlier, and then divide by the difference between the maximum and minimum values ​​to obtain the outlier degree of each image feature vector.

[0033] In this step, the maximum and minimum values ​​of the local outlier factors among all image feature vectors are identified. For each image feature vector, its local outlier factor is subtracted from the minimum value, and then divided by the difference between the maximum and minimum values ​​to obtain the normalized outlier score, which ranges from 0 to 1. The outlier score reflects the degree of uniqueness of the image in the feature space; a larger value indicates that the image is more unique and has higher value for model training.

[0034] If the calculation is of the multi-scale local outlier factor, then step S22157 is as follows: find the maximum and minimum values ​​of the multi-scale local outlier factors of all image feature vectors; for each image feature vector, subtract the minimum value from its multi-scale local outlier factor, and then divide by the difference between the maximum and minimum values ​​to obtain the outlier degree of each image feature vector.

[0035] The above steps completed the calculation of the clustering coefficient, isomorphism coefficient, mean Euclidean distance, and outlier. The clustering coefficient counts the number of surrounding points, reflecting local density; higher density indicates more common feature vectors and stronger representativeness. The isomorphism coefficient counts the number of shared nearest neighbors, reflecting the consistency of local structure; higher consistency indicates similarity between the image and surrounding images, and better representation of typical features of the region. Combining these two factors comprehensively evaluates the representativeness of the image from both density and structure dimensions. The mean Euclidean distance measures the average distance between the image and the overall data distribution; a larger distance indicates a more isolated image and higher uniqueness. Outlier, based on the local outlier factor, reflects the density difference between the image and its neighborhood; a larger difference indicates a more anomalous image within a local area and higher uniqueness. Combining these two factors comprehensively evaluates the unique value of the image from both global and local dimensions.

[0036] Step S3: Train a convolutional neural network model using the target historical safety shoe image to obtain the identifier positioning parameter extraction model; input the safety shoe image to be processed into the identifier positioning parameter extraction model to obtain the identifier positioning parameters of the safety shoe image to be processed.

[0037] In this step, the size of the safety shoe image to be processed is consistent with that of the historical safety shoe image; in addition, in this step, the specific implementation steps of training a convolutional neural network model using the target historical safety shoe image to obtain the identifier positioning parameter extraction model include step S31. Step S31: Obtain the annotation information of the target historical safety shoe image. The annotation information includes the identification positioning parameters. Use the annotated target historical safety shoe image to train the convolutional neural network model to obtain the identification positioning parameter extraction model.

[0038] In this step, the identifier positioning parameters are the bounding box positioning parameters of the identifier. These parameters describe the position and size of the rectangle selected in the image, including the x and y coordinates of the rectangle's center point, its width, and its height. During training, the target historical safety shoe image is used as input, and the annotation information is used as output. After training, the identifier positioning parameter extraction model can automatically adapt to different shoe types and sizes, outputting identifier positioning parameters, which helps improve the overall efficiency of OEM processing. After outputting the identifier positioning parameters, they can also be used to complete OEM labeling, for example, by sending them to an industrial robot to attach the identifier within the bounding box.

[0039] Example 2 like Figure 2 As shown, this embodiment provides an identification and positioning system for safety shoe label processing. The system includes an acquisition module 1, an extraction module 2, and a positioning module 3.

[0040] Module 1 is used to acquire historical safety shoe images; Extraction module 2 is used to extract image feature vectors from historical safety shoe images; calculate the feature richness and feature uniqueness of each historical safety shoe image based on the image feature vectors, and determine the target historical safety shoe image based on the feature richness and feature uniqueness. The positioning module 3 is used to train a convolutional neural network model using historical safety shoe images of the target to obtain an identifier positioning parameter extraction model; the safety shoe image to be processed is input into the identifier positioning parameter extraction model to obtain the identifier positioning parameters of the safety shoe image to be processed.

[0041] In one specific embodiment of this disclosure, the extraction module 2 further includes an extraction unit 21 and a selection unit 22.

[0042] Extraction unit 21 is used for each historical safety shoe image: adding T different levels of noise to the historical safety shoe image to obtain T noisy historical safety shoe images, inputting the T noisy historical safety shoe images into the feature extraction model to obtain T feature vectors, calculating the mean of the T feature vectors to obtain the mean vector, and performing L2 normalization on the mean vector to obtain the image feature vector corresponding to the historical safety shoe image. Unit 22 is selected to calculate the Euclidean distance between pairwise image feature vectors. All Euclidean distances are sorted from smallest to largest, and the median is taken. A preset value is multiplied by the median to obtain a first value. Based on the image feature vector of each historical safety shoe image and the first value, the feature richness and feature uniqueness of each historical safety shoe image are calculated. The feature richness and feature uniqueness of each image feature vector are weighted and summed to obtain the value of each image feature vector. The historical safety shoe images corresponding to the top N image feature vectors are selected as target historical safety shoe images according to the value in descending order.

[0043] In one specific embodiment of this disclosure, the selection unit 22 further includes a summation unit 221.

[0044] The summation unit 221 is used to calculate the clustering coefficient of each image feature vector based on the image feature vector and the first numerical value, and to calculate the isomorphism coefficient of each image feature vector based on the image feature vector and the first numerical value. The clustering coefficient and the isomorphism coefficient are weighted and summed to obtain the feature richness of each image feature vector. The unit also calculates the average Euclidean distance and outlier of each image feature vector based on the image feature vector and the first numerical value, and the average Euclidean distance and outlier are weighted and summed to obtain the feature uniqueness of each image feature vector.

[0045] In one specific embodiment of this disclosure, the summing unit 221 further includes a statistics unit 2211.

[0046] The statistical unit 2211 is used to count the number of other image feature vectors whose Euclidean distance to each image feature vector is less than the first value, and record it as the second value; after calculating the second value corresponding to each image feature vector, the largest second value is recorded as the third value; and the second value corresponding to each image feature vector is divided by the third value to obtain the clustering coefficient of each image feature vector.

[0047] In one specific embodiment of this disclosure, the summing unit 221 further includes a calculation unit 2212.

[0048] The calculation unit 2212 is used for each image feature vector to: sort its Euclidean distances with all other image feature vectors in ascending order; select the first K Euclidean distances corresponding to other image feature vectors and set them together to obtain the first set of each image feature vector; calculate the intersection of the first sets of each pair of image feature vectors and count the number of image feature vectors in the intersection, which is recorded as the fourth value; add all the fourth values ​​corresponding to each image feature vector to obtain the fifth value; after calculating the fifth value corresponding to each image feature vector, record the largest fifth value as the sixth value; divide the fifth value corresponding to each image feature vector by the sixth value to obtain the isomorphism coefficient of each image feature vector.

[0049] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0050] Example 3 Corresponding to the above method embodiments, this disclosure also provides a marking and positioning device for safety shoe OEM processing. The marking and positioning device for safety shoe OEM processing described below can be referred to in correspondence with the marking and positioning method for safety shoe OEM processing described above.

[0051] Figure 3 This is a block diagram illustrating a labeling and positioning device 300 for safety shoe labeling according to an exemplary embodiment. Figure 3 As shown, the safety shoe labeling and positioning device 300 may include a processor 301 and a memory 302. The safety shoe labeling and positioning device 300 may also include one or more of a multimedia component 303, an I / O interface 304, and a communication component 305.

[0052] The processor 301 controls the overall operation of the safety shoe labeling and positioning device 300 to complete all or part of the steps in the aforementioned safety shoe labeling and positioning method. The memory 302 stores various types of data to support the operation of the safety shoe labeling and positioning device 300. This data may include, for example, instructions for any application or method operating on the safety shoe labeling and positioning device 300, as well as application-related data such as contact data, sent and received messages, images, audio, video, etc. The memory 302 can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 303 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 302 or transmitted via the communication component 305. The audio component also includes at least one speaker for outputting audio signals. I / O interface 304 provides an interface between processor 301 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 305 is used for wired or wireless communication between the safety shoe labeling and positioning device 300 and other devices. Wireless communication includes Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 305 may include a Wi-Fi module, a Bluetooth module, or an NFC module.

[0053] In an exemplary embodiment, the safety shoe labeling and positioning device 300 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the aforementioned safety shoe labeling and positioning method.

[0054] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the above-described safety shoe labeling and positioning method. For example, the computer-readable storage medium may be the memory 302 including the program instructions, which may be executed by the processor 301 of the safety shoe labeling and positioning device 300 to complete the above-described safety shoe labeling and positioning method.

[0055] Example 4 Corresponding to the above method embodiments, this disclosure also provides a readable storage medium. The readable storage medium described below can be referred to in conjunction with the identification and positioning method for safety shoe labeling processing described above.

[0056] A readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the identification and positioning method for safety shoe labeling processing described in the above method embodiments.

[0057] Specifically, the readable storage medium can be a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, or any other readable storage medium capable of storing program code.

[0058] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for marking and positioning safety shoe labels, characterized in that, include: Obtain historical safety shoe images; Extract image feature vectors from historical safety shoe images; calculate the feature richness and feature uniqueness of each historical safety shoe image based on the image feature vectors, and determine the target historical safety shoe image based on the feature richness and feature uniqueness. A convolutional neural network model is trained using historical images of target safety shoes to obtain a marker localization parameter extraction model; the safety shoe image to be processed is then input into the marker localization parameter extraction model to obtain the marker localization parameters of the safety shoe image to be processed.

2. The identification and positioning method for safety shoe labeling processing according to claim 1, characterized in that, Extract image feature vectors from historical safety shoe images; calculate the feature richness and feature uniqueness of each historical safety shoe image based on the image feature vectors, and determine the target historical safety shoe image based on the feature richness and feature uniqueness, including: For each historical safety shoe image: add T different levels of noise to the historical safety shoe image to obtain T noisy historical safety shoe images. Input the T noisy historical safety shoe images into the feature extraction model to obtain T feature vectors. Calculate the mean of the T feature vectors to obtain the mean vector. Perform L2 normalization on the mean vector to obtain the image feature vector corresponding to the historical safety shoe image. Calculate the Euclidean distance between any two image feature vectors, sort all Euclidean distances from smallest to largest, take the median, and multiply the preset value by the median to obtain the first value. Calculate the feature richness and feature uniqueness of each historical safety shoe image based on its image feature vector and the first value. Calculate the weighted sum of the feature richness and feature uniqueness of each image feature vector to obtain the value of each image feature vector. Sort the historical safety shoe images corresponding to the top N image feature vectors from largest to smallest value as the target historical safety shoe images.

3. The identification and positioning method for safety shoe labeling processing according to claim 2, characterized in that, Based on the image features of each historical safety shoe image and the first numerical value, the feature richness and feature uniqueness of each historical safety shoe image are calculated, including: The clustering coefficient and isomorphism coefficient of each image feature vector are calculated based on the image feature vector and the first numerical value. The clustering coefficient and isomorphism coefficient are then weighted and summed to obtain the feature richness of each image feature vector. The mean Euclidean distance and outlier of each image feature vector are calculated based on the image feature vector and the first numerical value. The mean Euclidean distance and outlier are then weighted and summed to obtain the feature uniqueness of each image feature vector.

4. The identification and positioning method for safety shoe labeling processing according to claim 3, characterized in that, The clustering coefficient of each image feature vector is calculated based on the image feature vector and the first numerical value, including: For each image feature vector, count the number of other image feature vectors whose Euclidean distance to it is less than the first value, and record it as the second value; after calculating the second value corresponding to each image feature vector, record the largest second value as the third value; divide the second value corresponding to each image feature vector by the third value to obtain the clustering coefficient of each image feature vector.

5. The identification and positioning method for safety shoe label processing according to claim 3, characterized in that, The isomorphism coefficients of each image feature vector are calculated based on the image feature vectors and the first numerical value, including: For each image feature vector: sort its Euclidean distances with all other image feature vectors in ascending order, select the first K Euclidean distances corresponding to the other image feature vectors and set them together to obtain the first set of each image feature vector; calculate the intersection of the first sets of each pair of image feature vectors, count the number of image feature vectors in the intersection, and record it as the fourth value; add up all the fourth values ​​corresponding to each image feature vector to obtain the fifth value; after calculating the fifth value corresponding to each image feature vector, record the largest fifth value as the sixth value; divide the fifth value corresponding to each image feature vector by the sixth value to obtain the isomorphism coefficient of each image feature vector.

6. A labeling and positioning system for safety shoe OEM manufacturing, characterized in that, include: The acquisition module is used to acquire historical safety shoe images; The extraction module is used to extract image feature vectors from historical safety shoe images; based on The image feature vector is used to calculate the feature richness and feature uniqueness of each historical safety shoe image, and the target historical safety shoe image is determined based on the feature richness and feature uniqueness. The localization module is used to train a convolutional neural network model using historical images of the target safety shoes to obtain a marker localization parameter extraction model; the safety shoe image to be processed is input into the marker localization parameter extraction model to obtain the marker localization parameters of the safety shoe image to be processed.

7. The identification and positioning system for safety shoe labeling processing according to claim 6, characterized in that, The extraction module includes: The extraction unit is used for each historical safety shoe image: adding T different levels of noise to the historical safety shoe image to obtain T noisy historical safety shoe images; inputting the T noisy historical safety shoe images into the feature extraction model to obtain T feature vectors; calculating the mean of the T feature vectors to obtain the mean vector; and performing L2 normalization on the mean vector to obtain the image feature vector corresponding to the historical safety shoe image. A unit is selected to calculate the Euclidean distance between pairwise image feature vectors. All Euclidean distances are sorted from smallest to largest, and the median is taken. A preset value is multiplied by the median to obtain a first value. Based on the image feature vector of each historical safety shoe image and the first value, the feature richness and feature uniqueness of each historical safety shoe image are calculated. The feature richness and feature uniqueness of each image feature vector are weighted and summed to obtain the value of each image feature vector. The historical safety shoe images corresponding to the top N image feature vectors are selected as target historical safety shoe images according to the value in descending order.

8. The identification and positioning system for safety shoe labeling processing according to claim 7, characterized in that, Selecting units, including: The summation unit is used to calculate the clustering coefficient of each image feature vector based on the image feature vector and the first numerical value, and to calculate the isomorphism coefficient of each image feature vector based on the image feature vector and the first numerical value. The clustering coefficient and the isomorphism coefficient are weighted and summed to obtain the feature richness of each image feature vector. The unit also calculates the average Euclidean distance and outlier of each image feature vector based on the image feature vector and the first numerical value, and the average Euclidean distance and outlier are weighted and summed to obtain the feature uniqueness of each image feature vector.

9. The identification and positioning system for safety shoe OEM processing according to claim 8, characterized in that, Summation unit, including: The statistical unit is used to count the number of other image feature vectors whose Euclidean distance to each image feature vector is less than the first value, and record it as the second value; after calculating the second value corresponding to each image feature vector, the largest second value is recorded as the third value; the second value corresponding to each image feature vector is divided by the third value to obtain the clustering coefficient of each image feature vector.

10. The identification and positioning system for safety shoe OEM processing according to claim 8, characterized in that, Summation unit, including: The calculation unit is used for each image feature vector to: sort its Euclidean distances with all other image feature vectors in ascending order; select the first K Euclidean distances corresponding to other image feature vectors and set them together to obtain the first set of each image feature vector; calculate the intersection of the first sets of each pair of image feature vectors and count the number of image feature vectors in the intersection, which is recorded as the fourth value; add all the fourth values ​​corresponding to each image feature vector to obtain the fifth value; after calculating the fifth value corresponding to each image feature vector, record the largest fifth value as the sixth value; divide the fifth value corresponding to each image feature vector by the sixth value to obtain the isomorphism coefficient of each image feature vector.

Images