System for providing delivery service based on QR code

Abstract

The present invention relates to a QR code-based delivery service provision system that utilizes a QR code attached to a delivery box to minimize the exposure of personal information, such as addresses, during the delivery process and reliably determine whether there is a misdelivery.

Description

Technology Field

[0001] The present invention relates to a QR code-based delivery service provision system, and more specifically, to a QR code-based delivery service provision system that minimizes the exposure of personal information, such as an address, during the delivery process by utilizing a QR code attached to a delivery box. Background Technology

[0002] With the rise of non-face-to-face transactions such as internet shopping and home shopping, the market size of courier services has expanded significantly, and it can now be said that courier services have established themselves as a standard basic convenience service.

[0003] In such a courier service, the entities constituting the service can be divided into three entities: a person who orders a product (hereinafter, customer), a person who packages the ordered product and submits it to the courier service for dispatch (hereinafter, shopping mall administrator), and a person who is assigned a courier and delivers it to the customer (hereinafter, delivery driver).

[0004] In this case, a shipping label containing the personal information of the sender and recipient (e.g., name, address, phone number, etc.) is attached to the dispatched parcel; the delivery driver makes the delivery based on the information on the shipping label, and the customer can check the delivery status or verify whether the received parcel is theirs based on the information on the shipping label.

[0005] Consequently, there have always been concerns regarding the leakage of personal information exposed on shipping labels in courier services, and there have actually been crimes committed using such personal information.

[0006] Based on this, one can consider QR shipping labels that utilize encrypted QR codes instead of shipping labels with exposed personal information.

[0007] However, while introducing QR shipping labels to provide courier services can be effective from a security perspective in preventing personal information leakage, focusing solely on security may lead to concerns about reduced delivery efficiency for drivers and decreased usability for customers.

[0008] Accordingly, there is a need to develop a new type of courier service technology that utilizes QR codes to prevent the leakage of personal information on shipping labels, while simultaneously increasing delivery efficiency from the driver's perspective and the usability of the service from the customer's perspective.

[0009] Meanwhile, the aforementioned background technology is technical information that the inventor possessed for the derivation of the present invention or acquired during the process of deriving the present invention, and it cannot necessarily be considered publicly known technology disclosed to the general public prior to the filing of the present invention. Prior art literature

[0011] Korean Registered Patent No. 10-2023-0124190 The problem to be solved

[0012] One aspect of the present invention provides a QR code-based delivery service provision system that utilizes a QR code attached to a delivery box to minimize the exposure of personal information, such as an address, during the delivery process and reliably determine whether there is a misdelivery.

[0013] The technical problems of the present invention are not limited to those mentioned above, and other unmentioned technical problems will be clearly understood by those skilled in the art from the description below. means of solving the problem

[0014] In a QR code-based delivery service provision system that minimizes the exposure of personal information during the delivery process and improves the efficiency and reliability of the delivery service by using a QR code attached to a delivery box according to one embodiment of the present invention,

[0015] The above system is,

[0016] A courier management server comprising a security setting unit that, for text data containing personal information, calculates a security score based on the characteristics of the text data and its association with user identification information, and if the calculated security score exceeds a preset threshold value, selectively encrypts the personal information to generate a QR code;

[0017] A delivery driver terminal that performs delivery tasks by scanning a parcel with the above-generated QR code attached, and optimizes the scan direction by calculating a search index using pattern and reference point information within the QR code when scanning the QR code;

[0018] Includes more,

[0019] The above parcel management server is,

[0020] It includes a delivery driver management unit that calculates a rating for each delivery driver using variables including the delivery driver's delivery volume, travel distance, customer evaluation score, and the number of negative words in reviews, based on delivery-related data and review data collected from the delivery driver terminal and customer terminal.

[0021] The above security setting unit is,

[0022] The security score (CS) may be calculated using the following mathematical formula 2, which includes as variables the total number of words (dl) of the text data, a first embedding vector (va) for the entire text data, a second embedding vector (vk) for keywords with high personal identifiability extracted from the text data, and an angle (θ) between the second embedding vector and the identification information of the user terminal.

[0023] [Mathematical Formula 2]

[0025] Here, ε is a first stabilization constant that ensures the argument of the logarithmic function has a valid value, and ε is a second stabilization constant that prevents the denominator from becoming zero.

[0026] The above delivery driver management department,

[0027] The method may be characterized by calculating the rating (QS) for each delivery driver using the following mathematical formula 1, which includes as variables the total number of delivery packages (s_m) of the delivery driver subject to evaluation during a certain period, the maximum number of delivery packages (s_a) of the delivery driver subject to evaluation on a daily basis, the average daily travel distance (d) of the delivery driver subject to evaluation, the number of words (n) included in a pre-registered negative word list among the words constituting the review data collected from the customer terminal, and the customer evaluation score (r) of the delivery driver subject to evaluation included in the review data collected from the customer terminal.

[0028] [Mathematical Formula 1]

[0029]

[0030] Here, max(1, n) is a function that treats n as 1 when n is 0, and ε is a third stabilization constant that ensures the argument of the logarithmic function has a valid value.

[0031] The above delivery driver terminal is,

[0032] When scanning a QR code, the search index (SE) is calculated using the following mathematical formula 3, which includes variables such as a barcode density (d1) in the horizontal right direction relative to the current search point, a barcode density (d2) in the vertical downward direction relative to the current search point, a distance (ri) between the current search point and a specific reference point i among a plurality of reference points pre-placed in the QR code, a first weight (a_i) that is differentially set for each reference point pre-placed in the QR code, and a distance (|p_j - p_(j+1)|) between two reference points j and j+1 among reference points pre-placed in the QR code, and the scan direction is determined in the direction in which the calculated search index (SE) exceeds a preset scan threshold.

[0033] [Mathematical Formula 3]

[0034]

[0035] Here, Σ represents the sum for all corresponding indices, m is the number of reference point pairs or the number of valid reference points, and ε is a fourth stabilization constant that prevents the denominator from becoming zero.

[0036] The above system is,

[0037] It further includes a customer terminal that performs the function of carrying a customer, scanning a QR code attached to a delivery box to request authentication from the delivery management server, displaying decrypted personal information if the scanned QR code is encrypted upon successful authentication, determining whether there is a misdelivery by comparing it with the customer's order information stored in the delivery management server, and transmitting a review and rating score for the delivery driver to the delivery management server after delivery is completed;

[0038] The 'identification information of the user terminal' in the above mathematical formula 2 uses an embedding vector generated by securely hashing the unique identifier of the customer terminal or the corresponding customer account ID, or converting it into a unique ID within the system, and the 'keyword' may be characterized as referring to information within the text data that has high personal identifiability or a high possibility of privacy infringement, such as a detailed address, a specific unique number pattern, or a sensitive product name.

[0039] The above parcel management server is,

[0040] It further includes a parameter optimization module that improves the accuracy and realism of each mathematical formula by periodically learning and updating the weights, sensitivity coefficients, influence coefficients, or stabilization constants used in the above mathematical formula 1, mathematical formula 2, or mathematical formula 3 through a machine learning model or statistical optimization technique based on actual delivery result data and delivery environment change data continuously collected through system operation;

[0041] The delivery driver rating (QS) calculated by the above delivery driver management department may be characterized by being mutually linked in such a way that it is reflected as a positive or negative weight in the calculation of variables related to the delivery driver's proficiency when calculating the delivery difficulty index (DDI) of other deliveries, or that the number of successful completions of deliveries with a high delivery difficulty index (DDI) is fed back as an evaluation factor when calculating the above delivery driver rating (QS). Effects of the invention

[0043] According to one aspect of the present invention described above, by using a QR code attached to a delivery box, not only is the exposure of personal information, such as an address, minimized during the delivery process, but it is also possible to reliably determine whether there is a misdelivery. Brief explanation of the drawing

[0045] FIG. 1 is a diagram showing the schematic configuration of a QR code-based courier service provision system according to one embodiment of the present invention. Specific details for implementing the invention

[0046] The following detailed description of the invention refers to the accompanying drawings, which illustrate specific embodiments in which the invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention. It should be understood that various embodiments of the invention are different but need not be mutually exclusive. For example, specific shapes, structures, and characteristics described herein may be implemented in other embodiments without departing from the spirit and scope of the invention in relation to one embodiment. It should also be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the invention. Accordingly, the following detailed description is not intended to be limiting, and the scope of the invention is limited only by the appended claims, including all equivalents to those claimed therein, provided appropriately described. Similar reference numerals in the drawings refer to the same or similar functions across various aspects.

[0047] Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the drawings.

[0048] In the present invention, the term 'personal information' subject to encryption comprehensively refers to all identifying information that, if exposed during the delivery process, may infringe upon an individual's privacy or be misused. This includes, at a minimum, the recipient's name, address (including detailed address), and contact information (telephone number, email address, etc.). Depending on the case, specific product names or purchase history that may be classified as sensitive information among the ordered product information may also be optionally included. The term 'text data' used in Equation 2 refers to the entirety of data containing such personal information.

[0049] FIG. 1 is a diagram showing the schematic configuration of a QR code-based courier service provision system according to one embodiment of the present invention.

[0050] The QR code-based delivery service provision system according to the present invention aims to minimize the exposure of personal information, such as addresses, during the delivery process by encrypting a QR code attached to a delivery box, and to prevent misdelivery of the delivery.

[0051] Specifically, a QR code-based courier service provision system according to one embodiment of the present invention includes a delivery driver terminal (100) and a customer terminal (200).

[0052] A delivery driver terminal (100) scans the QR code on a QR shipping label issued with the personal information of the sender and recipient encrypted, and is assigned a package with the QR shipping label attached.

[0053] When a customer terminal (200) scans a QR code on a specific courier with a QR shipping label attached, if the information of the specific courier exists in the delivery tracking list of the customer page accessed through authentication, the scanned QR code is decrypted and the personal information is displayed based on the information. Accordingly, in addition to the delivery driver terminal and the courier management server, the system of the present invention includes, as an essential component, a customer terminal that allows the customer to directly scan a QR code to verify whether they are the legitimate recipient, look up delivery-related information, and participate in evaluating the delivery driver through reviews. The customer terminal can also perform the function of determining whether there is a misdelivery by comparing the scanned QR code information with the customer's order information stored on the server.

[0054] To this end, the delivery driver terminal can construct a neural network that extracts contextual information from input data by training the data with the Word2Vec algorithm to understand or estimate the meaning of text-based data.

[0055] The Word2Vec algorithm may include a Neural Network Language Model (NNLM). A Neural Network Language Model is fundamentally a neural network composed of an Input Layer, a Projection Layer, a Hidden Layer, and an Output Layer. The Neural Network Language Model is used as a method for vectorizing words. Since the Neural Network Language Model is a well-known technology, a more detailed explanation will be omitted.

[0056] The Word2vec algorithm is designed for text mining and determines proximity based on the preceding and succeeding relationships between words. It is an unsupervised learning algorithm. As its name suggests, Word2vec is a quantitative technique that represents the meaning of words in vector form. The Word2vec algorithm can represent each word as a vector in a space of approximately 200 dimensions. By utilizing the Word2vec algorithm, a vector corresponding to each word can be obtained.

[0057] The Word2vec algorithm can enable a dramatic improvement in precision in the field of natural language processing compared to other conventional algorithms. Word2vec learns the meaning of words by utilizing the relationships between words and adjacent words within sentences of an input corpus. Based on artificial neural networks, the Word2vec algorithm starts from the premise that words with the same context carry similar meanings. The algorithm learns through text documents, training the neural network to identify related words by identifying other words that appear nearby (approximately 5 to 10 words before and after) a given word. Since words with related meanings are highly likely to appear close together within a document, the two words can gradually acquire closer vectors as the learning process is repeated.

[0058] There are two training methods for the Word2vec algorithm: CBOW (Continuous Bag Of Words) and skip-gram. The CBOW method predicts a target word by utilizing the context created by surrounding words. The skip-gram method predicts words that may follow a single word. The skip-gram method is known to be more accurate in large-scale datasets.

[0059] Accordingly, in the embodiments of the present invention, a Word2vec algorithm using the skip-gram method is used. For example, if training is successfully completed through the Word2vec algorithm, similar words can be located nearby in a high-dimensional space. According to the Word2vec algorithm described above, the closer the distribution of surrounding words within a training document, the more similar the calculated vector values ​​can be, and words with similar calculated vector values ​​can be considered similar. Since the Word2vec algorithm is a known technology, a more detailed explanation regarding the calculation of vector values ​​will be omitted.

[0060] The delivery driver terminal can input data collected into a neural network to extract an evaluation result vector value representing contextual information.

[0061] The delivery driver terminal calculates the similarity between the evaluation result vector value and each of the multiple reference vector values, and can extract the reference vector value among the multiple reference vector values ​​that has the highest similarity to the evaluation result vector value. In this case, Euclidean distance, cosine similarity, Tanimoto coefficient, etc., may be adopted as the similarity calculation method.

[0062] In the present invention, the above-described Word2Vec algorithm can be utilized specifically to analyze customer review text to extract the number of negative words 'n' in Equation 1 or to calculate a text-based customer evaluation score 'r'. Additionally, it can be applied to generate an embedding vector 'va' for the entire text data and an embedding vector 'vk' for keywords used in Equation 2.

[0063] In this case, the aforementioned keywords are not simple common nouns, but refer to words or phrases within text data that identify an individual's identity or represent sensitive information closely related to privacy. For example, this may include a full detailed address such as 'Room 607, Building 45, 123 Teheran-ro, Gangnam-gu, Seoul,' 'first digits of Resident Registration Number 900101,' or specific product names that allow one to infer health status, such as 'anticancer drug.' The system extracts these core personal information keywords through a predefined list of sensitive keywords or natural language processing technology and generates an embedding vector vk for them. Therefore, the relationship between vk and va and the θ value play a role in evaluating the qualitative sensitivity and identifiability of personal information beyond simple text similarity.

[0064] The delivery driver terminal can extract the word corresponding to the reference vector value with the highest similarity to the evaluation result vector value as the word corresponding to the recognized text.

[0065] Furthermore, the delivery driver terminal can train artificial neural networks and utilize completed artificial neural networks. The processor can train or execute artificial neural networks stored in memory, and memory can store completed artificial neural networks. The electronic device that trains the artificial neural network and the electronic device that utilizes it may be identical, but they may also be separate. Artificial intelligence is a computer system that partially implements the functions of the human brain, capable of learning, speculating, and making judgments on its own. As learning progresses, the probability of extracting the correct answer may increase. Artificial intelligence can be composed of learning and component technologies that utilize it. The learning of artificial intelligence is an algorithmic technology that classifies and learns features based on input data, and the component technologies may be technologies that partially implement the functions of the human brain by utilizing learning algorithms.

[0066] Artificial intelligence is a technology that facilitates the approach to problems where multiple probabilistic answers are possible, enabling it to logically and probabilistically infer optimal cycles, methods, and plans based on input data. AI inference techniques can include evaluating input data, optimization prediction, knowledge and probability-based reasoning, and preference-based planning.

[0067] Artificial neural networks are learning algorithms in the field of machine learning that programmatically implement the connections between neurons and synapses in the brain. By creating a neural network structure through programming and then training it, artificial neural networks can acquire desired functions. Although errors may exist, they can learn from massive datasets to produce appropriate output data from input data. They have the advantage of being able to obtain output data that has yielded statistically good results and are similar to human reasoning.

[0068] The delivery driver terminal can construct query / metric datasets required for learning using an artificial intelligence algorithm built on big data, and to this end, it may include multiple pre-trained artificial neural networks.

[0069] The system according to the present invention may include a plurality of pre-trained artificial neural networks for performing machine learning algorithms. Through machine learning, it can output output data based on input data and learn autonomously using the results, thereby improving its data processing capabilities. The artificial neural network can extract features and predict regularities based on input data to output result data, and as this process accumulates, the reliability of the result data increases.

[0070] In this embodiment, the artificial neural network may be an algorithm that outputs text data from at least one feature data among the shape, length, number, and height difference of an object recognized as text. The artificial neural network can infer the best output data by using big data as input data as is, or by using it as input data after undergoing a processing step to clean up unnecessary data.

[0071] Artificial intelligence machine learning models are classified according to the type of learning into Supervisory Learning, Unsupervisory Learning, Semi-supervised Learning, and Reinforcement Learning. In addition, machine learning algorithms include Decision Trees, K-Nearest Neighborhoods, Artificial Neural Networks, Support Vector Machines, Ensemble Learning, Gradient Descent, and Na Bayes Classifier, Hidden Markov Model, K-Means Clustering, etc. can be used.

[0072] Artificial neural networks can be pre-trained on various input values ​​that may be included in the input data. An artificial neural network can be trained using reinforcement learning, a learning method. Reinforcement learning is a method that gradually increases the probability of obtaining the correct result by setting rewards and constraints. Artificial neural networks can also be modeled based on Convectional Neural Networks (CNNs) or Recurrent Neural Networks (RNNs).

[0073] In this way, the delivery driver terminal can estimate the meaning of text data using big data and artificial neural networks.

[0074] In some other embodiments, the courier management server according to the present invention may further include a delivery driver management unit (not shown) that calculates a rating for each delivery driver based on review data collected from a terminal held by the courier recipient and written by the courier recipient.

[0075] The calculation of the delivery driver rating (QS) according to the above mathematical formula 1 is primarily performed by the delivery driver management unit (not shown) implemented within the courier management server. To this end, the delivery driver terminal (100) periodically or in real-time transmits the delivery driver's login information, the number of deliveries at each delivery completion, daily travel distance (e.g., based on GPS logs), delivery area information, etc., to the delivery driver management unit of the courier management server. At the same time, evaluation score data, such as review text and star ratings entered by a customer regarding a specific delivery case through the customer terminal (200), is safely transmitted and stored in the delivery driver management unit of the courier management server along with the unique identification information of the delivery case. The delivery driver management unit integrates these collected heterogeneous data, extracts variables n and r using a pre-configured negative word list and natural language processing (NLP) module, and finally calculates the rating QS for each delivery driver according to mathematical formula 1. The calculated QS is stored in the database within the delivery driver management department and can be used as an input value for future delivery assignments, incentive payments, or other system modules (e.g., when calculating DDI in Equation 4).

[0076] The delivery driver management department can calculate the rating for each delivery driver using the following mathematical formula 1.

[0077] [Mathematical Formula 1]

[0078]

[0079] Here, QS is the rating of the delivery driver being evaluated, s_m is the total number of delivery packages delivered by the delivery driver being evaluated in one month, s_a is the maximum number of delivery packages delivered by the delivery driver being evaluated in one day, d is the average daily travel distance of the delivery driver being evaluated, n is the number of words included in a pre-registered negative word list among the words constituting the review data, and r is the score of the delivery driver included in the review data collected from the terminal being possessed, characterized as being a natural number between 1 and 5.

[0080] In addition, s_m is in units of cases / month, which is calculated by aggregating the total number of parcels processed by a specific delivery driver during an evaluation period (e.g., the last month) based on delivery completion data transmitted from the delivery driver terminal (100) to the parcel management server (300) upon completion of each delivery case. s_a is in units of cases / day, which is calculated by determining the maximum value among the daily number of completed deliveries within the evaluation period from the delivery completion data used to calculate s_m. This reflects the instantaneous maximum processing capacity of the delivery driver.

[0081] d is in units of km / day, which is calculated by summing the total distance traveled on days when actual delivery work was performed, based on daily location log data collected from the GPS module of the delivery driver terminal (100) and transmitted to the parcel management server (300), and then dividing by the number of work days. This indirectly indicates the activity radius and effort level of the delivery driver.

[0082] In this case, the aforementioned list of negative words consists of terms that carry negative connotations regarding delivery service quality, such as 'damage,' 'unfriendliness,' 'delay,' 'loss,' 'misdelivery,' and 'dissatisfaction,' and this list can be periodically updated through big data analysis related to courier services and customer feedback. The value of 'n' is calculated by counting the frequency of these negative words appearing in review data using Natural Language Processing (NLP) techniques, such as morphological analysis and keyword extraction algorithms. Here, the aforementioned list of negative words is not merely selected subjectively; rather, it is a systematic list periodically updated after extracting core negative keywords through text mining of delivery service complaints from large-scale customer VOC data, online communities, and news articles, and undergoing review by a group of logistics experts and verification of statistical significance. Furthermore, when applying NLP techniques, a technique may be employed that goes beyond simple frequency calculation to differentiate the intensity of negative sentiment considering context (e.g., mild dissatisfaction vs. strong dissatisfaction) and reflect this differentially in the value of 'n.' If there are no negative words at all (n=0), exception handling is performed to prevent the denominator from becoming zero. This involves setting the value of n to the minimum value of 1, treating the contribution of the corresponding log term as 0, or using max(1, n) instead of n to ensure the denominator is always greater than or equal to 1, thereby ensuring the definability of the formula and computational stability. This is intended to prevent the rating from becoming excessively high or the system from falling into an uncalculable state even when there is no negative feedback.

[0083] In addition, in the log term `log( (d + r) / max(1, n) + ε)` of the above mathematical formula 1, `ε` is a very small positive stabilization constant (e.g., 1e-6) to ensure that the argument of the logarithm function always has a positive value, and to mitigate abrupt changes in the logarithm value that may occur, especially when the result of `(d+r) / max(1,n)` is very close to 0. This contributes to increasing the numerical stability of the rating calculation.

[0084] The above score r can directly reflect delivery satisfaction evaluations (e.g., star ratings) left by customers via their devices, or use a value obtained by performing sentiment analysis on the review text to quantify the degree of positivity or negativity. For example, reviews containing a large number of positive words can be converted to 4–5 points, neutral reviews to 3 points, and reviews containing negative words to 1–2 points. In the case of star ratings entered directly by customers, the reliability of the score can be enhanced by assigning weights based on the time of evaluation (e.g., higher weight for recent evaluations) or by including logic to filter out patterns suspected of being false or malicious evaluations (e.g., repeated extreme evaluations of a specific delivery driver). When calculating the r value through sentiment analysis, a multi-model ensemble technique based on various sentiment dictionaries is used to ensure the accuracy and robustness of the analysis.

[0085] The structure of mathematical formula 1 is designed to use objective performance indicators, such as a delivery driver's consistent delivery volume (s_m, s_a) and range of activity (d), as basic evaluation elements, while integratively considering multifaceted evaluation elements to positively reflect direct customer feedback (r) and allow negative feedback (n) to act as a penalty. This moves away from existing fragmentary evaluation methods that rely solely on the number of deliveries or travel distance, and enables a fairer and more realistic evaluation of delivery drivers by quantitatively reflecting actual customer experience and service quality. The use of square root and logarithmic functions is intended to prevent the influence of each variable from being excessively distorted by a specific variable and to adjust the rating distribution more stably, thereby increasing the reliability of the evaluation results. In particular, placing the number of negative words (n) in the denominator of the logarithmic term and applying the max(1,n) function and the stabilization constant (ε) ensures technical stability by clearly defining the influence of negative feedback without causing it to diverge to infinity or become uncalculable, which constitutes the advanced evaluation model configuration of the present invention.

[0086] For example, assuming that s_m is 100 cases, s_a is 20 cases, the average daily travel distance d is 50 km, the number of negative words extracted from reviews n is 2, and the customer rating score r is 4 points, QS = sqrt(100 + 20^2) + log((50+4) / 2) = sqrt(500) + log(27) ≈ 22.36 + 1.43 = approximately 23.79 points (value may vary depending on the base of the logarithm, assuming a common logarithm here).

[0087] In this way, the delivery driver management department enables objective evaluation of each delivery driver by calculating a rating for each driver.

[0088] In some other embodiments, the system according to the present invention further includes a delivery management server (300) that encrypts a QR code, and the delivery management server may include a security setting unit (not shown) that selectively encrypts according to the importance of the data to minimize the time required for encrypting the QR code.

[0089] Specifically, the security setting unit can calculate the security score using the following mathematical formula 2.

[0091] [Mathematical Formula 2]

[0092]

[0093] Here, CS is the security score, dl is the total number of words in the text data (address, name, zip code, etc.), va is the embedding vector for the entire text data, vk is the embedding vector for the keyword extracted from the text data, and θ is the angle formed by the embedding vector for the keyword and the embedding vector for the identification information of the user terminal.

[0094] The calculation of the security score (CS) according to the above mathematical formula 2 and the decision on selective QR code encryption based thereon are overseen by the security setting unit within the courier management server. Text data containing the recipient's personal information (including address, name, contact information, etc.) transmitted when a courier is sent from a shopping mall administrator or a dispatch system is input into the security setting unit. The security setting unit generates a full-text embedding vector (va) and an embedding vector (vk) for key personal information keywords using a language model (e.g., Word2Vec, FastText, etc.) for the input text data. At this time, user terminal identification information (e.g., encrypted customer ID) is retrieved from the customer database and converted into an embedding vector, thereby calculating the angle θ between the vectors. The security setting unit applies these calculated variables to mathematical formula 2 to calculate the CS, and determines whether to encrypt the corresponding personal information text data based on whether this CS exceeds a preset minimum security value. According to the determined encryption policy, the QR code generation module generates a standard QR code or an encrypted QR code and sends it to the delivery driver terminal (100) or the linked shipping label printing system. transmits.

[0095] At this time, the embedding vector va for the entire text data and the embedding vector vk for the extracted keywords can be generated using a word embedding model such as Word2Vec or FastText that has been trained in advance on a large amount of text data (e.g., general documents, news articles, or text data related to personal information). The generated vectors can take the form of high-density vectors of, for example, 100 to 300 dimensions.

[0096] Here, 'identification information of the user terminal' refers, for example, to the unique identifier of the customer terminal (e.g., encrypted Device ID or user account ID), and this identification information is also converted into a fixed-dimensional embedding vector. This can be achieved by placing a separate embedding layer for the identifier, or by treating the identifier as categorical data and projecting it onto a low-dimensional vector after one-hot encoding. Security is considered more sensitively when the angle θ between the keyword embedding vector and the user terminal identification information embedding vector is small, as it is determined that the keyword is more likely to be closely associated with a specific user, whereas it is considered general information when the angle is large. In the above Equation 2, ε is a very small positive first stabilization constant (e.g., 1e-6) to prevent the argument of the log term (va^2 + vk) / √ from becoming zero or negative, and to prevent the log value from diverging to negative infinity when the calculation result is a very small positive number. In addition, ε is a very small second stabilization constant (e.g., 1e-6) to prevent fractional calculation from becoming impossible when the denominator |va - vk| of the sine term becomes zero, and if va and vk are very similar and |va-vk| approaches zero, exception handling is performed by adding the above ε.

[0097] Here, ε represents the search index denominator stabilization constant and is a very small positive constant (e.g., 1e-6) introduced to prevent the search index (SE) from diverging or becoming unstable due to zero or very small values ​​of the denominator terms in Equation 3, particularly the terms related to the number of reference points `m` or the sum of distances between reference points. This ensures that the system can reliably attempt to determine the reading direction even if the quality of the QR code to be scanned is very low or it has an unexpected shape.

[0098] Mathematical Equation 2 calculates a security score by comprehensively considering the length of the text data (dl), the semantic similarity between the entire text and the core keywords (va, vk), and the angle (θ) indicating how closely the keywords relate to a specific user. The log term is designed to reflect the overall sensitivity of the data, while the sine term reflects the sensitivity within a specific user context. Through this, information with a higher potential for personal identification is assigned a higher security score, serving as a criterion for selective encryption.

[0099] That is, if the address is 10 characters long, the contact number is 10 digits long, and the name is 3 characters long, dl can be set to 23.

[0100] For example, if dl is 16, va is 2, vk is 1, and θ is 30, CS can be calculated as approximately 0.6.

[0101] The security setting unit encrypts and transmits the data if the calculated security score exceeds a preset minimum security value (e.g., 0.7), and transmits the data without encryption if the calculated security score is less than the preset minimum security value. The above minimum security value (e.g., 0.7) may be set empirically or experimentally at a point that balances false positives and false negatives, taking into account statistical analysis based on actual operational data and personal information protection regulations.

[0102] In this way, the security setting unit encrypts data requiring security, such as personal information, based on the calculated security score, and transmits text that is acceptable to be disclosed to third parties without encryption, thereby reducing the time required for encryption and decryption while maintaining security.

[0103] In some other embodiments, the delivery driver terminal may perform a series of processes to read a QR code in an encrypted state prior to decryption.

[0104] For example, a delivery driver can read an encrypted QR code during delivery to verify whether the address listed on the QR code matches the delivery address.

[0105] To this end, the delivery driver terminal extracts multiple feature points from an encrypted QR code, calculates a reading index for each scan point using the extracted feature points and the following mathematical formula 3, and determines the reading direction of the QR code based on the calculated reading index.

[0106] The calculation of the search index (SE) according to the above mathematical formula 3 and the determination of the QR code reading direction using it are performed in real time by the QR code scanning module and image processing engine installed in the delivery driver terminal (100). When the delivery driver attempts to scan the QR code on the delivery box, the image processing engine calculates the barcode density (d1, d2) based on the current scan search point in the image frame input from the camera of the delivery driver terminal (100), and extracts the distance (ri) to predefined reference points (e.g., finder pattern) within the QR code. The above information is transmitted to the QR code scanning module to calculate the SE value according to mathematical formula 3. By comparing the calculated SE value with a pre-set reference value, the scanning module determines the optimal scan direction (e.g., horizontal movement, vertical movement, or retry after maintaining the current position) to acquire the next image frame, and if necessary, automatically adjusts the camera's focus or exposure to increase the QR code recognition rate. The above process can be performed repeatedly until the QR code is successfully decoded or determined to be a reading failure.

[0107] [Mathematical Formula 3]

[0108]

[0109] Here, SE is the search index, d1 is the barcode density in the horizontal right direction relative to the current search point, d2 is the barcode density in the vertical downward direction relative to the current search point, ri is the distance between the current search point and a specific reference point i among the reference points pre-placed in the QR code, a is the first weight that is differentially set for each reference point pre-placed in the QR code, and pj-pj+1 is the distance between two reference points among the reference points pre-placed in the QR code.

[0110] At this time, barcode densities d1 and d2 can be quantified by setting a window of a specific size (e.g., a 5x5 pixel window) centered on the current scan search point and calculating the proportion of the black module (or pixel) of the QR code within that window. Alternatively, the density can be estimated by the number of edges detected per unit area using an edge detection algorithm.

[0111] The reference points pre-placed in the QR code may refer, for example, to the center point of the QR code's finder pattern (large squares located at the corners) or the center point of the alignment pattern (small squares). The first weight 'a' is set differentially according to the importance of each reference point; for example, a higher weight may be assigned to reference points closer to the center of the QR code or to reference points in areas expected to have high information density. This can be determined through empirical data or QR code structure analysis.

[0112] In mathematical formula 3, to handle extreme cases where there are no valid reference points (pj) within the QR code or the number of reference points m becomes zero when calculating the summation term Σ in the denominator, exception handling using a small positive constant (ε) such as max(εm) can be included so that the denominator does not become zero. This ensures that the system can reliably attempt to determine the reading direction even when the quality of the QR code to be scanned is very low or it has an unexpected shape.

[0113] For example, if d_1 is 3, d_2 is 9, r_i is 15.7, a_i is 2, and pj-pj+1 is 5, the search index is calculated to be approximately 6.4. If the search index calculated in this way is compared with a preset reference value (e.g., 5) and it is confirmed that the search index (6.4) is greater than the reference value (5), then the image can be scanned along the horizontal direction.

[0114] In this way, the delivery driver terminal of the present invention can reduce the time required for QR scanning by variably determining the next scanning direction based on a search index calculated for each scanning point, and by comparing the pattern and metadata (e.g., version information, error correction level, etc.) of the QR code obtained during the scanning process with a pre-stored pattern, it can be used to estimate the probability that the QR code contains a specific type of information (e.g., a pattern highly likely to contain address information) or to partially verify the validity of the QR code without decoding the entire code.

[0115] In another embodiment, the system according to the present invention may further include a function for calculating a Delivery Difficulty Index (DDI) to quantitatively evaluate the difficulty of each delivery. The Delivery Difficulty Index is calculated by comprehensively considering various factors such as delivery location information, product information, time constraints, weather conditions, and the proficiency of the delivery driver performing the delivery.

[0116] Specifically, the above Delivery Difficulty Index (DDI) can be calculated using the following mathematical formula 4.

[0117] The calculation of the Delivery Difficulty Index (DDI) according to the above mathematical formula 4 can be performed in a DDI calculation module that is separately provided within the courier management server (300) or linked to the delivery driver management department. Each variable required for DDI calculation is collected through various channels. For example, L_a is obtained by querying information about the delivery case or delivery driver from the GIS database and the delivery location characteristic information DB, T_wp from the order management system, V_l from the product information DB, C_ar from the CRM system, W_f from the external weather information API, and H_e and D_c from the delivery history DB of the delivery driver management department. The DDI calculation module calculates the DDI value for a specific delivery case by applying the collected data to mathematical formula 4. The calculated DDI value is stored in the database along with the delivery case information, and is used as important input data for decision-making in the delivery route optimization system, the delivery driver task assignment system, or the delivery fee calculation system. If necessary, the DDI value may be transmitted to the delivery driver terminal (100) to help the delivery driver recognize the difficulty of their next delivery in advance.

[0118] [Mathematical Formula 4]

[0119]

[0120] Here, each variable and coefficient is defined as follows:

[0121] DDI refers to the delivery difficulty index of the delivery being evaluated.

[0122] L_a is a delivery location accessibility score with a value ranging from 0 (very easy) to 1 (very difficult), and is a normalized value obtained by comprehensively evaluating factors such as parking availability, presence of an elevator and high floors, length of inaccessible vehicle zones, and security level of the common entrance.

[0123] T_wp is a delivery time constraint pressure score, ranging from 0 (no constraint) to 1 (very tight), and is a normalized value that takes into account the need to meet specific customer request times, the length of the allowed delivery time window, and the urgency of the deadline relative to the current time.

[0124] V_l is a product loading burden score, ranging from 0 (very light / small) to 1 (very heavy / large / multiple), and is a normalized value that comprehensively considers the total volume, total weight, number of products being shipped, and whether special handling such as refrigeration / freezing or fragility is required.

[0125] C_ar is a customer absence risk score with a value ranging from 0 (low probability of absence) to 1 (high probability of absence), and is a normalized value that takes into account the customer's past absence history, general absence rate statistics for delivery times, and whether non-face-to-face delivery was not requested.

[0126] W_f is a weather deterioration weight, having a value greater than or equal to 1 (good weather), and is a weight assigned differentially according to real-time weather information, such as 1.0 for clear / cloudy, 1.2 for light rain / snow, 1.5 for heavy rain / snow, and 2.0 for typhoon / heavy snow.

[0127] H_e is a delivery driver's rough terrain experience index, ranging from 0 (no rough terrain experience) to 1 (very rough terrain experience), and is a value calculated by normalizing the ratio of deliveries to areas previously classified as 'rough terrain', such as islands, mountainous areas, old city centers, and industrial complexes, or the cumulative number of deliveries among the areas where the delivery driver has delivered during a recent period.

[0128] D_c is a delivery driver experience and vehicle proficiency coefficient, ranging from 0 (newcomer or inexperienced with the vehicle) to 1 (veteran or proficient with the vehicle), and is a weighted and normalized value that comprehensively considers the delivery driver's total delivery experience, driving experience by type of vehicle currently in operation (e.g., 1-ton truck, motorcycle, bicycle), accident-free period, etc.

[0129] w_La, w_Twp, w_Vl, and w_Car are weights that reflect the relative importance of L_a, T_wp, V_l, and C_ar, respectively.

[0130] γ is the sensitivity coefficient of the weather factor W_f, and can have a value greater than 1.0 and less than or equal to 2.0, for example.

[0131] δ_He and δ are the influence coefficients of H_e and D_c on difficulty damping, respectively. (Changed from the existing δδ to be associated with variable names)

[0132] η is a coefficient that controls the nonlinear influence of D_c, and can have a value of, for example, between 0.5 and 1.5.

[0133] ε_lH is a small positive constant used to prevent the argument of the logarithm function from becoming less than 1 in the log(1 + δ + ε) term, or to control the default value when H_e is 0.

[0134] ε_pD is a small positive constant used to control the default value when D_c is 0 in the (1 + (δη + ε) term.

[0135] ε_div is a small positive constant to prevent the denominator from becoming 0 or a very small value during division operations.

[0136] K is a scaling constant used to adjust the scale of the final calculated shipping difficulty index.

[0137] C is a correction constant used to adjust the minimum value of the final calculated shipping difficulty index or to adjust the overall range of values.

[0138] The above variables can be obtained or estimated in the actual system in the following ways. For example, L_a can be calculated through the analysis of Geographic Information System (GIS) data regarding delivery addresses, feedback data from past delivery drivers, or integration with a building management database. T_wp can be calculated based on delivery request time information entered during customer orders and real-time delivery interval data managed by the Delivery Management System (TMS). V_l is calculated using product specifications (volume, weight) information stored in the product database and the number of products included in a single delivery order, while the need for special handling can be determined from product category information. C_ar can be estimated through the customer's past delivery receipt history accumulated in the Customer Relationship Management (CRM) system (e.g., return rate due to absence) or delivery options directly set by the customer (e.g., doorstep delivery request). W_f obtains weights based on the weather conditions of the current delivery area by integrating in real-time with a public weather information API. H_e and D_c are calculated by analyzing the delivery history database of delivery drivers recorded in the delivery driver management system (e.g., delivery completion area, delivery period, vehicle information).

[0139] According to one embodiment of the present invention, when each normalized variable (L_a, T_wp, V_l, C_ar, H_e, D_c) has a value between 0 and 1 and W_f has a value between 1 and 3, and each weight and coefficient (w_La et al., γ, δ_He et al., η) is set within a preset valid range (e.g., weights such that the sum becomes a specific value (e.g., 4), γ is 1.5, δ series is 0.5 to 1.0, η is 0.8), and the scaling constant K and correction constant C are appropriately adjusted (e.g., K=20, C=0), the final Delivery Difficulty Index (DDI) showed a distribution between approximately 5 and 95 points, and showed a logically valid change in DDI values ​​according to changes in each input variable. For example, it was confirmed that as the value of L_a or W_f increased, the DDI increased non-linearly, and as the value of H_e or D_c increased, the DDI tended to decrease. In particular, when stabilization constants such as ε_IH, ε_pD, and ε_div were each set to 1e-3, they effectively prevented cases where the denominator approached zero or the argument of the logarithm function had an inappropriate value. These weights and coefficients can initially be set based on the empirical knowledge of logistics experts, competitor benchmarking, or results from small-scale pilot tests; however, as mentioned earlier, it is desirable to continuously optimize them through machine learning or statistical modeling utilizing big data accumulated through system operations to improve realism and prediction accuracy. For example, the L_a weight (w_La) for a specific region can be adjusted to maximize the correlation between the actual delivery delay rate or accident rate of that region and the predicted DDI. The stabilization constants (ε series) are generally set to very small values ​​between 1e-6 and 1e-2, serving to ensure numerical stability while having almost no impact on normal calculations.

[0140] The calculation structure of the Delivery Difficulty Index (DDI) according to the above mathematical formula 4 has the following technical effects and rationality.

[0141] First, the sum-of-squares structure of objective difficulty elements, that is

[0142]

[0143] The part ensures that each objective difficulty element (L_a, T_wp, V_l, C_ar) acts as difficulty in independent dimensions, forming the overall base difficulty based on a concept similar to their vector sum. This structure has the effect of drastically increasing the overall base difficulty when the difficulty of a specific element is very high, as its influence becomes dominant due to the squaring operation. This mathematically reflects the reality that the perceived difficulty is much higher when a single element is extremely difficult to overcome, rather than when multiple elements are only slightly difficult. Each weight w_La, w_Twp, w_Vl, and w_Car allows the system designer to adjust the relative importance of each objective difficulty element.

[0144] Second, the exponential weighting of weather factors, namely (W_f^γ), is achieved by multiplying the basic objective difficulty by raising the sensitivity coefficient γ to the exponential level of the degree of weather deterioration (W_f). This ensures that when γ is greater than 1, the delivery difficulty increases steeply beyond a simple proportional relationship as weather deterioration intensifies. This indicates that bad weather has a complex negative impact on all aspects of delivery, and that the severity of this impact is not linear.

[0145] Third, the complex difficulty attenuation structure based on the delivery driver's proficiency, that is, the denominator's

[0146]

[0147] This part precisely reflects the effect of a delivery driver's experience and proficiency in lowering the actual perceived difficulty. Specifically, the rough terrain experience index (H_e) is reflected through a logarithmic function; while the accumulation of initial rough terrain experience has a relatively large impact on difficulty reduction, it simulates the characteristics of diminishing marginal utility, where the difficulty reduction effect resulting from additional experience gradually becomes milder once experience reaches a certain level. Additionally, career and vehicle proficiency (D_c) are reflected through exponential terms, allowing the pattern to be adjusted based on the value of the coefficient η. This ensures that the difficulty reduction effect due to accumulated experience becomes more pronounced once it exceeds a specific threshold, or that initial experience plays a more significant role. By combining the two proficiency-related factors in the form of a product, the system is designed to reduce difficulty more effectively through a synergistic effect when both types of proficiency are high.

[0148] Fourth, the final scaling and adjustment constants K and C play a role in adjusting the range of the calculated raw DDI values ​​so that they are expressed within a meaningful range (e.g., between 0 and 100 points) that is intuitively understandable and comparable in an actual operating environment.

[0149] As such, the delivery difficulty index calculated through Equation 4 can provide an objective indicator similar to the perceived difficulty in the field by more sophisticatedly modeling the non-linear relationships, critical effects, and interactions between factors that are difficult to capture through simple summation or multiplication of elements. This delivery difficulty index can effectively contribute to the optimization of various logistics operations, such as the efficient allocation of delivery resources, fair performance evaluation of delivery drivers, delivery route optimization, risk management, and prediction of estimated delivery times.

[0150] Furthermore, in another embodiment of the present invention, the Delivery Difficulty Index (DDI) calculated by Equation 4 can be organically linked with the results of other Equations. For example, the Delivery Driver Rating (QS) calculated by Equation 1 can be used to calculate or correct variables related to the delivery driver's proficiency, such as H_e or D_c in Equation 4. Specifically, a delivery driver with a high QS score is given a positive weight when calculating the H_e or D_c value, even if they have the same objective delivery history, which can consequently contribute to lowering the DDI value. This reflects the reality that excellent delivery drivers can handle difficult deliveries more efficiently, and can enhance the consistency of evaluation throughout the system and serve as a means of motivating delivery drivers. In addition, the DDI value itself is fed back as an evaluation factor (e.g., the number of times difficult deliveries were successfully completed) when calculating the delivery driver rating in Equation 1, so that the evaluation can reflect not only how many deliveries a delivery driver handles but also how well they perform difficult deliveries.

[0151] The technology according to the present invention, as described above, may be implemented in the form of program instructions that can be executed through various computer components or implemented as an application, and may be recorded on a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc., either individually or in combination.

[0152] The program instructions recorded on the above-mentioned computer-readable recording medium are those specifically designed and configured for the present invention, but may also be those known and available to those skilled in the art of computer software.

[0153] Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, and flash memory.

[0154] Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware device may be configured to operate as one or more software modules to perform processing according to the present invention, and vice versa.

[0155] Although the invention has been described above with reference to embodiments, those skilled in the art will understand that various modifications and changes can be made to the invention without departing from the spirit and scope of the invention as set forth in the following claims.

Claims

Claim 1 A QR code-based delivery service provision system that minimizes the exposure of personal information during the delivery process and improves the efficiency and reliability of the delivery service by utilizing a QR code attached to a delivery box, comprising: a delivery management server including a security setting unit that, for text data containing personal information, calculates a security score based on the characteristics of the text data and the association including an angle (θ) between a first embedding vector of a keyword with high personal identifiability extracted from the text data and a second embedding vector generated from identification information of a user terminal corresponding to the recipient of the personal information, and selectively encrypts the personal information to generate a QR code when the calculated security score exceeds a preset threshold value; and a delivery driver terminal that performs delivery tasks by scanning a delivery package with the generated QR code attached, and adjusts the scanning direction by calculating a search index (SE) using a pattern within the QR code and preset reference point information when scanning the QR code. A customer terminal possessed by the customer, which performs the function of scanning a QR code attached to a delivery box to request authentication from the delivery management server, and upon successful authentication, displaying decrypted personal information if the scanned QR code is encrypted, determining whether there is a misdelivery by comparing it with the customer's order information stored in the delivery management server, and transmitting a review and evaluation score for the delivery driver to the delivery management server after delivery is completed;[Equation 2]; wherein the reference point information refers to the center point of the QR code finder pattern or the center point of the alignment pattern; the courier management server includes a delivery driver management unit that calculates a delivery driver rating (QS) using variables including the delivery driver's delivery volume, travel distance, customer evaluation score, and the number of negative words in the review, based on delivery-related data and review data collected from the delivery driver terminal and the customer terminal; and the security setting unit calculates the security score (CS) using the following mathematical formula 2, which includes as variables the total number of words (dl) of the text data, a first embedding vector (va) for the entire text data, a second embedding vector (vk) for keywords with high personal identifiability extracted from the text data, and an angle (θ) between the second embedding vector and the identification information of the user terminal. Here, ε_cs_log is a first stabilization constant that prevents the argument of the logarithmic function from becoming zero or negative, and ε_denom is a second stabilization constant that prevents the denominator from becoming zero; the delivery driver management unit is characterized by calculating the delivery driver rating (QS) using the following mathematical formula 1, which includes as variables the total number of parcels delivered by the delivery driver subject to evaluation during a certain period (s_m), the maximum number of parcels delivered per day by the delivery driver subject to evaluation (s_a), the average daily travel distance of the delivery driver subject to evaluation (d), the number of words included in a pre-registered negative word list among the words constituting the review data collected from the customer terminal (n), and the customer evaluation score (r) of the delivery driver subject to evaluation included in the review data collected from the customer terminal. [Mathematical Formula 1] Here, max(1, n) is a function that treats n as 1 when n is 0, and ε_qs_log is a third stabilization constant that ensures the argument of the logarithmic function is always positive, and the delivery driver terminal calculates the search index (SE) using the following mathematical formula 3, which includes as variables a barcode density (d1) in the horizontal right direction relative to the current search point when scanning the QR code, a barcode density (d2) in the vertical downward direction relative to the current search point, a distance (ri) between the current search point and a specific reference point i which is the center point of the QR code's finder pattern or alignment pattern, a first weight (a_i) that is differentially set for each reference point pre-placed in the QR code, and a distance (|p_j - p_(j+1)|) between two reference points j and j+1 among the reference points pre-placed in the QR code, and determines the scan direction in the direction in which the calculated search index (SE) exceeds a preset scan threshold, [Mathematical Formula 3] (Here, Σ represents the sum for all corresponding indices, m is the number of reference point pairs, and ε_se_denom is a fourth stabilization constant that prevents the denominator from becoming zero; the 'identification information of the user terminal' in the above Equation 2 utilizes an embedding vector generated by securely hashing the unique identifier of the customer terminal or the corresponding customer account ID, or converting it into an internal system unique ID; the above 'keyword' is characterized by referring to a detailed address, a specific unique number pattern, or a sensitive product name within the text data; the above courier management server further includes a parameter optimization module that improves the accuracy of each equation by periodically learning and updating the weights, sensitivity coefficients, influence coefficients, or stabilization constants used in the above Equation 1, Equation 2, or Equation 3 through a machine learning model or statistical techniques based on actual delivery result data and delivery environment change data continuously collected through system operation; and using the following [Equation 4], the delivery location accessibility score (L_a), delivery time constraint pressure score (T_wp), and product loading burden It further includes a DDI calculation module that calculates a Delivery Difficulty Index (DDI) that comprehensively reflects objective difficulty factors including a score (V_l) and a customer absence risk score (C_ar), and driver proficiency including a weather deterioration weight (W_f), a driver's rough terrain experience index (H_e), and a career and vehicle proficiency coefficient (D_c), [Equation 4] The above DDI calculation module forms a base difficulty by taking the square root of the weighted sum of squares of the objective difficulty elements, multiplies it by the weather deterioration weight (W_f) raised to the power of exponential (γ), and calculates the delivery difficulty index (DDI) by dividing the result by a composite proficiency attenuation factor in which the delivery driver's rough terrain experience index (H_e) is reflected through a logarithmic function and the career and vehicle proficiency (D_c) is reflected through an exponential term (η); the delivery driver rating (QS) calculated by the delivery driver management unit is linked to be reflected as a positive weight when the DDI calculation module calculates the delivery driver's rough terrain experience index (H_e) or the career and vehicle proficiency coefficient (D_c) to lower the DDI value; and the delivery difficulty index (DDI) value calculated by the DDI calculation module is mutually linked to be fed back as an evaluation factor when the delivery driver management unit calculates the delivery driver rating (QS) as the 'number of successful completions of difficult deliveries'. A QR code-based delivery service provision system. Claim 2 delete Claim 3 delete Claim 4 delete Claim 5 delete Claim 6 delete

Images