Visiting customer matching method and system

Abstract

The invention discloses a visiting customer matching method and system, and belongs to the field of passenger flow recognition, and the system comprises a data collection module which is used for collecting basic information and passenger flow recognition data registered when a customer visits; the data preprocessing module is used for carrying out cleaning and feature extraction on the data; the feature matching module is used for comprehensively utilizing time, space and feature extraction information, matching the visit registration data and the passenger flow identification data, and determining whether the visit registration data and the passenger flow identification data correspond to the same customer or not; the matching result judgment and data fusion module is used for fusing the visit registration data and the passenger flow identification data according to the matching structure of the feature matching module; the data storage and management module is used for storing the fused customer data to a database; and the system security and privacy protection module is used for establishing an authority control and data encryption mechanism and preventing data leakage and unauthorized access. According to the invention, the accuracy, reliability and real-time performance of the matching of the visiting clients can be improved, and the method also has the advantage of high expandability.

Description

Technical Field

[0001] The present invention relates to the technical field of passenger flow identification, and more specifically, to a visiting customer matching method and system. Background Art

[0002] In the real estate industry, customer flow recognition technology is widely used to analyze customer visit behavior, helping sales managers understand customer needs and optimize sales strategies. However, the customer groups identified by the current customer flow recognition system through the algorithm model cannot directly establish an effective association with the customer groups registered offline. This situation leads to deviations in data display, affecting the sales manager's judgment of the actual visiting customer group, thereby affecting subsequent sales decisions and customer service.

[0003] The existing customer behavior data collection and analysis has the following deficiencies:

[0004] 1) Data is isolated and cannot be integrated: The visitor registration system can only obtain basic customer information (such as name, contact information, visit time, etc.), and cannot capture the customer's behavior trajectory in the venue; although the passenger flow identification system can monitor the customer's behavior trajectory, it cannot associate with specific customer identity information, resulting in difficulty in data integration and inability to form a complete customer portrait.

[0005] 2) Inaccurate matching affects strategy formulation: Due to the lack of an effective matching mechanism, visitor registration data and customer flow identification data cannot be accurately matched, resulting in incomplete analysis of customer behavior, affecting marketing strategies and service quality, and further affecting customer satisfaction and sales conversion rates.

[0006] 3) Lack of real-time performance and missed service opportunities: The existing system is unable to monitor and analyze customer behavior in real time. Sales staff are unable to obtain customers’ interests and needs in a timely manner, thus missing out on the best service and marketing opportunities.

[0007] In order to improve the accuracy of sales and services, it is necessary to develop a solution that can effectively link passenger flow data with offline registration data. Summary of the invention

[0008] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a visiting customer matching method and system, which realizes the effective integration of visitor registration data and passenger flow identification data, and overcomes the problem of data isolation; improves the accuracy of customer behavior identification through multi-dimensional feature matching; ensures real-time acquisition and processing of data, and timely captures customer dynamics; the algorithm and system architecture have good scalability and can adapt to different business scenarios and needs.

[0009] The object of the present invention is achieved through the following solutions:

[0010] A visiting customer matching system includes a data collection module, a data preprocessing module, a feature matching module, a matching result determination and data fusion module, a data storage and management module, and a system security and privacy protection module; the data collection module is used to collect basic information and passenger flow identification data registered by customers when they visit; the data preprocessing module is used to clean and extract features from the data; the feature matching module is used to comprehensively utilize time, space and feature extraction information to match visiting registration data and passenger flow identification data to determine whether the two correspond to the same customer; the matching result determination and data fusion module is used to fuse visiting registration data and passenger flow identification data according to the matching structure of the feature matching module to form a complete customer behavior record; the data storage and management module is used to save the fused customer data to a database, establish a data management system, and support data addition, deletion, modification and query; the system security and privacy protection module is used to establish an authority control and data encryption mechanism to prevent data leakage and unauthorized access.

[0011] Furthermore, the data collection module specifically includes: a visitor registration data collection submodule and a passenger flow identification data collection submodule; the visitor registration data collection submodule is used to collect the basic information of the customer, and with the customer's consent, collect the customer's facial photo or biometric information, and enter it in real time and save it to the database; the passenger flow identification data collection submodule installs camera equipment at key nodes to collect the customer's image data in real time, and the key nodes include the entrances and important areas within the venue; the collection content includes the customer's facial image, behavior trajectory, stay area and stay time, and the image data is transmitted in real time to the background server for processing using the transmission network.

[0012] Furthermore, the data preprocessing module specifically includes a data cleaning submodule and a feature extraction submodule; the data cleaning submodule is used to clean the visitor registration data and passenger flow identification data, remove duplicate, abnormal, ambiguous or invalid data, and ensure data quality; the feature extraction submodule includes facial feature extraction, attribute feature extraction and behavioral feature extraction; the attribute feature extraction is based on an image processing algorithm to extract the customer's attribute information, and the attribute information includes gender, age group and emotion; the behavioral feature extraction tracks the customer's behavior trajectory through video analysis, and records the customer's stay time and frequency in each area.

[0013] Furthermore, the feature matching module specifically includes a time window screening submodule, a spatial area screening submodule, a face feature matching submodule, an attribute feature matching submodule, a behavior feature matching submodule and a comprehensive scoring model module; the time window screening submodule is used to set a time window, screen possible passenger flow identification data within the time window, and narrow the matching range; the spatial area screening submodule is used to screen customers appearing in the corresponding area according to the location when the visitor registered, and further narrow the matching objects; the face feature matching submodule is used to calculate the similarity of the face feature vectors of the visitor registered customer and the passenger flow identification customer, set a similarity threshold, and if it is higher than the threshold, it is considered to be the same person, and the successfully matched object is recorded for subsequent comprehensive scoring; otherwise, it is not the same person; the attribute feature matching submodule is used to compare the customer's attribute information, set a consistency score, and accumulate the consistency score of each attribute as the attribute matching score; the behavior feature matching submodule is used to compare the customer's behavior characteristics and associate them with the demand type in the visitor registration; the comprehensive scoring model module is used to set weights for different feature matching results and calculate the comprehensive matching score.

[0014] Furthermore, in the matching result determination and data fusion module, the matching result of the feature matching module is used to fuse the visitor registration data and the passenger flow identification data to form a complete customer behavior record, which specifically includes: judging whether the matching result is successful based on the comprehensive score; for successfully matched customers, fusing the visitor registration data and the passenger flow identification data to form a complete customer behavior record.

[0015] Furthermore, it also includes a matching model optimization module for optimizing the accuracy and robustness of the matching model to adapt to changes in different scenarios and data.

[0016] Furthermore, the data storage and management module specifically adopts a relational database, which is used to store structured customer information and visitor registration data; wherein, a NoSQL database is used to store semi-structured feature data and behavior data; and a time series database is used to store time series data.

[0017] Furthermore, in the attribute feature matching submodule, the attribute information includes gender and age group; the gender and age group information of the visiting registered customer and the customer flow identified customer are compared, and if the gender is consistent, the set value is one point, the set value is two points if the age group is consistent, and the set value is three points if the emotion is consistent.

[0018] Furthermore, in the behavior feature matching submodule, the behavior features include entry time, stay area and stay time; if the customer is interested in product A during visit registration and stays in the product A exhibition area in the venue for a long time, the behavior feature matching score is improved; when calculating the score, the corresponding scoring rules are set according to the stay time and visit frequency.

[0019] A visiting customer matching method is based on the visiting customer matching system as described in any one of the above items, and the system includes a computer device, a computer program is stored in the memory of the computer device, and when the computer program is loaded by the processor of the computer device, the following steps are performed: in the comprehensive scoring model module, the program is run to execute the following comprehensive scoring formula to calculate the comprehensive score:

[0020] Comprehensive score = set weight value 1 × face matching score + set weight value 2 × attribute matching score + set weight value 3 × behavior matching score comprehensive score;

[0021] Among them, the set weight value one, the set weight value two and the set weight value three are all fixed values ​​that can be adjusted, and a threshold is set for the comprehensive score. If it is higher than the threshold, the match is considered successful, otherwise the match is unsuccessful.

[0022] The beneficial effects of the present invention include:

[0023] (1) The present invention forms a complete customer portrait based on data fusion: the visitor registration data and passenger flow identification data are effectively integrated to overcome the problem of data isolation and form a complete and three-dimensional customer portrait, providing a solid data foundation for the company's precise marketing and services.

[0024] (2) The present invention can improve matching accuracy and reliability: through multi-dimensional feature matching and comprehensive scoring model, the accuracy and reliability of customer matching are effectively improved, and the matching accuracy rate can reach more than 90%.

[0025] (3) The present invention can realize real-time monitoring and feedback of customer behavior: the system has real-time data processing capabilities, and sales personnel can promptly understand customer behavior dynamics, quickly respond to customer needs, and improve service quality and customer satisfaction.

[0026] (4) The present invention supports marketing strategies and management decisions: Based on in-depth analysis of customer behavior data, it provides strong data support for enterprises to formulate marketing strategies, optimize resource allocation, and improve operational efficiency.

[0027] (5) The present invention can enhance the versatility and scalability of the system: the system architecture design has good versatility and can be applied to different industries and application scenarios; the modular design facilitates functional expansion and system upgrades. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0029] Figure 1 is an overall flow chart of the system of the embodiment of the present invention;

[0030] Figure 2 The visitor registration data collection process of the system according to the embodiment of the present invention;

[0031] Figure 3 The matching result determination and data fusion process of the system according to the embodiment of the present invention;

[0032] Figure 4 The passenger flow identification data collection process of the system according to the embodiment of the present invention;

[0033] Figure 5 The data preprocessing process of the system according to the embodiment of the present invention;

[0034] Figure 6 This is an example of the matching process specifically for Mr. Zhang in the system of the embodiment of the present invention. DETAILED DESCRIPTION

[0035] All features disclosed in all embodiments in this specification, or steps in all methods or processes implicitly disclosed, except for mutually exclusive features and / or steps, can be combined and / or expanded or replaced in any manner.

[0036] The present invention aims to solve the problems of isolated data, inaccurate customer matching and poor real-time performance of the visitor registration system and passenger flow identification system in the prior art, and proposes a customer behavior matching method and system that integrates visitor registration data and passenger flow identification data. Through multi-dimensional feature matching and comprehensive scoring models, accurate identification and real-time tracking of customer behavior can be achieved, providing enterprises with comprehensive customer behavior analysis tools to support marketing and management decisions.

[0037] In a preferred embodiment, if Figures 1 to 6 As shown, a visiting customer matching system is provided, including a data collection module, a data preprocessing module, a feature matching module, a matching result determination and data fusion module, a data storage and management module, and a system security and privacy protection module.

[0038] The data collection module specifically includes a visitor registration data collection submodule and a passenger flow identification data collection submodule. The visitor registration data collection submodule is used to collect the basic information of customers, including name, contact information, visit time, demand type, etc. With the customer's consent, collect the customer's facial photo or other biometric information. The data is entered in real time and saved to the database. Furthermore, in terms of hardware equipment settings, computer terminals, high-definition cameras, touch screens and other equipment are configured at the front desk or reception desk to facilitate the entry and collection of information. In terms of software system design, a friendly visitor registration software is developed to support the rapid entry of information, photo collection, encrypted transmission and storage of data. In terms of data security and privacy protection, a data encryption function is set up to encrypt and store sensitive information; when collecting biometric information, customers must be prompted and their explicit consent must be obtained.

[0039] In the passenger flow identification data collection submodule, video cameras are installed at the entrances and important areas of the venue to collect customer image data in real time. The collected content includes the customer's facial image, behavior trajectory, stay area, stay time, etc. The image data is transmitted to the background server in real time for processing. Furthermore, through the video cameras installed in the venue, the customer's image data is collected in real time to monitor the customer's behavior trajectory, stay area, stay time and other information. The video cameras are arranged at the entrance of the venue, various functional areas, main channels, important exhibition areas and other locations, and high-definition cameras are installed. At the same time, the camera should have the following characteristics:

[0040] High-definition resolution: at least 1080P, to ensure image clarity and facilitate subsequent image processing and recognition;

[0041] Wide Dynamic Range (WDR): Adapts to different lighting conditions and avoids overexposure or underexposure;

[0042] Infrared night vision function: clear images can still be obtained in low light conditions;

[0043] Waterproof and dustproof: For outdoor or special environments, the camera needs to have a certain level of protection.

[0044] In terms of data transmission design, wired or wireless networks are used to transmit real-time image data to the backend server. The network bandwidth must meet the transmission requirements of high concurrency and large data volumes. In terms of data storage and management, image data can be cached or stored as needed, and a suitable storage period can be set to comply with relevant laws and regulations.

[0045] The data preprocessing module specifically includes a data cleaning submodule and a feature extraction submodule. The data cleaning submodule cleans the visitor registration data and passenger flow identification data, removes duplicate, abnormal, ambiguous or invalid data, and ensures data quality. Furthermore, in the visitor registration data cleaning, the registration data is formatted and standardized to remove duplicate, erroneous and abnormal data. Among them, duplicate data detection specifically detects duplicate registrations based on unique identifiers such as contact information (mobile phone number, email address), ID card number, etc. Abnormal data identification marks and processes obviously erroneous data, such as blank fields, format errors, and ages beyond a reasonable range. Error correction and completion: For missing or erroneous data, it can be supplemented and corrected manually in combination with historical records or prompts. In the passenger flow identification data cleaning, the quality of the collected image data is evaluated, and image frames that do not meet the requirements are filtered out, such as blur, incomplete face capture, poor lighting conditions, etc. In image quality detection, the image clarity, brightness, contrast and other indicators are calculated, and thresholds are set to filter low-quality images. In face integrity detection, the face detection algorithm is used to determine whether the face is complete, blocked, and whether it meets the requirements of feature extraction. In abnormal scene filtering, camera shaking, lens blocking, sudden changes in ambient light, etc. are identified and processed accordingly.

[0046] The feature extraction submodule includes facial feature extraction, attribute feature extraction, and behavior feature extraction. Facial feature extraction uses deep learning face recognition algorithms (FaceNet, ArcFace, etc.) to extract the customer's facial feature vector. Further, facial feature extraction detects faces from cleaned image data and extracts high-dimensional facial feature vectors. Among them, the face detection algorithm uses advanced face detection algorithms such as FaceNet and ArcFace to quickly and accurately locate the face area. In multi-scale detection, multi-scale detection is performed for faces of different sizes and angles in the image to improve the detection rate. In real-time design, the algorithm can be optimized to improve execution efficiency and meet the needs of real-time processing. In face alignment design, the detected faces are aligned and normalized to eliminate the influence of posture and angle. Among them, face key point detection (such as eyes, nose, mouth, etc.) is used to perform affine transformation or perspective transformation. When extracting specific features, deep learning models such as FaceNet, ArcFace, SphereFace, etc. are used to extract 128-dimensional or 512-dimensional facial feature vectors. Among them, the model training uses large-scale face datasets (such as CASIA-WebFace, MS-Celeb-1M) for pre-training, and fine-tunes the model to adapt to specific scenes and groups. In feature storage, the extracted feature vectors are stored together with the corresponding timestamps, camera positions and other information to facilitate subsequent matching.

[0047] Attribute feature extraction is based on image processing algorithms to extract attribute information such as gender, age group, and emotions of customers. In particular, attribute feature extraction is used to analyze facial images and extract attribute information such as gender, age group, emotion, and race. In specific implementation, in gender recognition, a convolutional neural network (CNN) model is used to classify facial images and output gender probabilities. In age estimation, a multi-task learning model is used to regress and predict age, or segmented prediction of age groups. In emotion recognition, based on the expression recognition model, the customer's emotional state, such as happiness, anger, sadness, and surprise, is identified. When labeling data and training models, the labeling data set uses public labeling data sets (such as IMDB-WIKI, FER2013) or self-labeled data for training and verification. In model optimization, the model structure and parameters are adjusted according to the data distribution of the actual scenario to improve the prediction accuracy.

[0048] Behavior feature extraction uses video analysis technology to track the customer's behavior trajectory and record the customer's stay time and frequency in each area. Further, behavior feature extraction is used to track the customer's movement trajectory in the venue and analyze their behavior characteristics, such as entry and exit time, stay area, path preference, etc. In specific implementation, target detection algorithm: Use target detection algorithms such as YOLOv11, SSD, and Faster R-CNN to identify and locate the customer's position in the video. The target tracking algorithm uses online multi-target tracking algorithms such as DeepSORT and ByteTrack (the fine-tuned model has an accuracy of no less than 80% on the international standard complex data sets MOT16, MOT17, and MOT20), assigns a unique ID, and keeps continuous tracking of customers. In trajectory generation and analysis, the customer's location coordinates at different times are recorded to generate complete trajectory data. In area mapping, the video coordinate system is mapped to the plane coordinate system of the venue to facilitate identification of the functional area where the customer is located. For example, a customer visited area 3, area 2, and area 4 in 5 minutes, and stayed in area 2 for more than 3 minutes. From this, it is concluded that the customer is particularly interested in the products in area 2. In behavioral analysis, statistics are collected on the time customers enter and leave each area, length of stay, visit frequency, path preference, etc., to extract key behavioral features.

[0049] In the feature matching module, a variety of information such as time, space, facial features, attribute features, and behavioral features are used to match visitor registration data and passenger flow identification data to determine whether the two correspond to the same customer. In the specific implementation, it also includes time window screening and spatial area screening. The time window screening sets a reasonable time window (such as 30 minutes before and after the visitor registration time) to screen possible passenger flow identification data within this time range to narrow the matching range. The spatial area screening screens customers who appear in the corresponding area according to the location of the visitor registration (such as the front desk, reception room), further narrowing the matching objects.

[0050] In addition, the feature matching module specifically includes a facial feature matching submodule, an attribute feature matching submodule, a behavior feature matching submodule and a comprehensive scoring model. The facial feature matching submodule calculates the similarity (such as cosine similarity or Euclidean distance) between the facial feature vectors of the visiting registered customer and the customer identified by the passenger flow. A similarity threshold is set, and if it is higher than the threshold, it is considered to be the same person. Further, when comparing facial features, the similarity between the facial feature vector of the visiting registered customer and the facial feature vector of the customer identified by the passenger flow is calculated. The similarity index specifically uses cosine similarity or Euclidean distance as a measurement indicator. In terms of threshold setting, a similarity threshold (such as cosine similarity greater than 0.8) is set according to experimental results and actual needs, and if it is higher than the threshold, it is considered to be a match. Finally, the successfully matched objects are recorded for subsequent comprehensive scoring. The attribute feature matching submodule compares the customer's gender, age group, clothing type and other attribute information, and the higher the consistency, the higher the matching credibility. Specifically, the gender, age group and other attribute information of the visiting registered customer and the customer identified by the passenger flow are compared. In the consistency calculation design, 1 point is awarded for gender consistency, 1 point is awarded for age group consistency, 0.8 points is awarded for clothing type consistency, and 0.5 points is awarded for emotion consistency (depending on specific needs). In the attribute matching score design, the consistency scores of each attribute are accumulated as the attribute matching score. The behavioral feature matching submodule compares the customer's behavioral characteristics, such as entry time, stay area, stay time, etc., and associates them with the demand type in the visit registration. Further, the customer's behavioral characteristics are matched with the needs expressed in the visit registration. An example is as follows: If the customer is interested in product A in the visit registration and stays in the product A exhibition area in the venue for a long time, the behavioral feature matching score will increase. Score calculation: Set corresponding scoring rules based on stay time, visit frequency, etc. The comprehensive scoring model sets weights for different feature matching results (such as 0.5 for facial features, 0.3 for attribute features, and 0.2 for behavioral features) to calculate the comprehensive matching score. The comprehensive scoring formula is as follows:

[0051] Comprehensive score = 0.5 × face matching score + 0.3 × attribute matching score + 0.2 × behavior matching score Comprehensive score = 0.5 × face matching score + 0.3 × attribute matching score + 0.2 × behavior matching score.

[0052] And set a successful matching score threshold (such as 0.7), and if it is higher than the threshold, the match is considered successful.

[0053] In the matching result judgment and data fusion module, the matching result is judged to be successful based on the comprehensive score. For successfully matched customers, the visitor registration data and customer flow identification data are integrated to form a complete customer behavior record.

[0054] Furthermore, the system of the embodiment of the present invention also includes a matching model optimization module, which is used to continuously optimize the accuracy and robustness of the matching model to adapt to changes in different scenarios and data. Specifically, a large amount of matching sample data is collected during the model training process, the matching results are manually annotated, and the training set and test set are constructed. The machine learning algorithm uses machine learning algorithms such as logistic regression, support vector machine (SVM), random forest, XGBoost, etc. to establish a matching model. Feature inputs include face similarity, attribute consistency, behavioral feature scores, etc. Model evaluation uses cross-validation, ROC curve, precision, recall and other indicators to evaluate model performance. Parameter adjustment adjusts model parameters and feature weights according to the evaluation results to improve matching accuracy.

[0055] In the data storage and management module, the integrated customer data is saved to the database, and a complete data management system is established to support the addition, deletion, modification and query of data to ensure data consistency and security. Furthermore, the cleaned and extracted feature data is effectively stored and managed to support efficient query and access. In terms of specific implementation, the database selects a relational database, which is used to store structured customer information, visitor registration data, etc. NoSQL databases such as MongoDB and Cassandra are used to store semi-structured feature data and behavioral data. Time series databases such as InfluxDB are used to store time series data, such as customer behavior trajectories. In terms of data indexing, appropriate indexes are established, such as customer ID, timestamp, and feature vector indexes, to improve query efficiency. The data backup and recovery mechanism uses regular database backup and disaster recovery mechanism to ensure data security and integrity.

[0056] In the system security and privacy protection module, in terms of data security mechanism, a complete permission control and data encryption mechanism is established to prevent data leakage and unauthorized access. In terms of privacy protection, relevant laws and regulations are complied with to ensure that data collection and analysis are carried out with the explicit consent of the customer. Privacy setting options are provided so that customers can choose not to share personal sensitive information.

[0057] The technical solution of the present invention has the following advantages:

[0058] 1) A multi-dimensional feature fusion matching algorithm is proposed: a matching algorithm that combines facial features, attribute features, and behavioral features is proposed, which greatly improves the accuracy and robustness of matching.

[0059] 2) Real-time data processing and comprehensive scoring model was proposed: A real-time data processing mechanism and a comprehensive scoring model were designed to achieve real-time monitoring of customer behavior and quantitative evaluation of matching results.

[0060] 3) Realized visual customer behavior analysis: Through data visualization technology, complex customer behavior data is converted into intuitive graphics and reports, which are convenient for sales staff and managers to understand and apply.

[0061] 4) System security and privacy protection mechanism: Taking data security and customer privacy protection into consideration, a complete security mechanism has been established to ensure the legality and compliance of the system.

[0062] It should be further noted that the present invention can be widely applied to the following fields:

[0063] Real estate sales center: improve customer service quality, accurately follow up on potential customers, and increase sales conversion rate.

[0064] Large shopping malls and retail industry: analyze customer flow dynamics, optimize store layout, and formulate precise marketing strategies.

[0065] Exhibition halls, museums and other public places: understand visitors’ interests and behavior patterns, and optimize exhibition content and services.

[0066] Chain stores and catering industry: monitor customer visits to stores, improve customer experience, and conduct effective membership management.

[0067] The units involved in the embodiments of the present invention may be implemented by software or hardware, and the units described may also be arranged in a processor. The names of these units do not, in some cases, limit the units themselves.

[0068] According to one aspect of an embodiment of the present invention, a computer program product or a computer program is provided, the computer program product or the computer program includes a computer instruction, and the computer instruction is stored in a computer-readable storage medium. A processor of a computer device reads the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the method provided in the above various optional implementations.

[0069] As another aspect, an embodiment of the present invention further provides a computer-readable medium, which may be included in the electronic device described in the above embodiment; or may exist independently without being assembled into the electronic device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by an electronic device, the electronic device implements the method described in the above embodiment.

Claims

1. A visiting customer matching system, characterized in that: It includes data acquisition module, data preprocessing module, feature matching module, matching result determination and data fusion module, data storage and management module and system security and privacy protection module; The data collection module is used to collect basic information and customer flow identification data registered by customers when they visit; The data preprocessing module is used to clean and extract features from data; The feature matching module is used to comprehensively utilize time, space and feature extraction information to match visitor registration data and passenger flow identification data to determine whether the two correspond to the same customer; The matching result determination and data fusion module is used to fuse the visitor registration data and the passenger flow identification data according to the matching structure of the feature matching module to form a complete customer behavior record; The data storage and management module is used to save the integrated customer data to the database, establish a data management system, and support the addition, deletion, modification and query of data; The system security and privacy protection module is used to establish permission control and data encryption mechanisms to prevent data leakage and unauthorized access.

2. The visiting customer matching system according to claim 1, characterized in that: The data collection module specifically includes: a visitor registration data collection submodule and a passenger flow identification data collection submodule; the visitor registration data collection submodule is used to collect the basic information of the customer, and with the customer's consent, collect the customer's facial photo or biometric information, and enter it in real time and save it to the database; the passenger flow identification data collection submodule installs camera equipment at key nodes to collect the customer's image data in real time, and the key nodes include the entrances and important areas in the venue; the collection content includes the customer's facial image, behavior trajectory, stay area and stay time, and the image data is transmitted in real time to the background server for processing using the transmission network.

3. The visiting customer matching system according to claim 1, characterized in that: The data preprocessing module specifically includes a data cleaning submodule and a feature extraction submodule; the data cleaning submodule is used to clean the visitor registration data and passenger flow identification data, remove duplicate, abnormal, ambiguous or invalid data, and ensure data quality; the feature extraction submodule includes facial feature extraction, attribute feature extraction and behavior feature extraction; the attribute feature extraction is based on an image processing algorithm to extract the customer's attribute information, and the attribute information includes gender, age group and emotion; the behavior feature extraction tracks the customer's behavior trajectory through video analysis and records the customer's stay time and frequency in each area.

4. The visiting customer matching system according to claim 3, characterized in that: The feature matching module specifically includes a time window screening submodule, a space area screening submodule, a face feature matching submodule, an attribute feature matching submodule, a behavior feature matching submodule and a comprehensive scoring model module; the time window screening submodule is used to set a time window, screen possible passenger flow identification data within the time window, and narrow the matching range; the space area screening submodule is used to screen customers appearing in the corresponding area according to the location when the visitor registered, and further narrow the matching objects; the face feature matching submodule is used to calculate the similarity of the face feature vectors of the visitor registered customer and the passenger flow identification customer, set a similarity threshold, and if it is higher than the threshold, it is considered to be the same person, and the successfully matched object is recorded for subsequent comprehensive scoring; otherwise, it is not the same person; the attribute feature matching submodule is used to compare the attribute information of the customer, set a consistency score, and accumulate the consistency score of each attribute as the attribute matching score; the behavior feature matching submodule is used to compare the behavior characteristics of the customer and associate them with the demand type in the visitor registration; the comprehensive scoring model module is used to set weights for different feature matching results and calculate the comprehensive matching score.

5. The visiting customer matching system according to claim 4, characterized in that: In the matching result determination and data fusion module, the matching result of the feature matching module is used to fuse the visitor registration data and the passenger flow identification data to form a complete customer behavior record, which specifically includes: judging whether the matching result is successful based on the comprehensive score; for successfully matched customers, fusing the visitor registration data and the passenger flow identification data to form a complete customer behavior record.

6. The visiting customer matching system according to claim 4, characterized in that: It also includes a matching model optimization module, which is used to optimize the accuracy and robustness of the matching model to adapt to changes in different scenarios and data.

7. The visiting customer matching system according to claim 4, characterized in that: The data storage and management module specifically adopts a relational database, which is used to store structured customer information and visitor registration data; wherein, a NoSQL database is used to store semi-structured feature data and behavior data; and a time series database is used to store time series data.

8. The visiting customer matching system according to claim 4, characterized in that: In the attribute feature matching submodule, the attribute information includes gender and age group; the gender and age group information of the visiting registered customers and the customer flow identification customers are compared, and the set value of one point is obtained if the gender is consistent, the set value of two points is obtained if the age group is consistent, and the set value of three points is obtained if the emotion is consistent.

9. The visiting customer matching system according to claim 4, characterized in that: In the behavior feature matching submodule, the behavior features include entry time, stay area and stay time; if the customer is interested in product A during visit registration and stays in the product A exhibition area in the venue for a long time, the behavior feature matching score will be improved; when calculating the score, the corresponding scoring rules are set according to the stay time and visit frequency.

10. A visiting customer matching method, characterized in that: A visiting customer matching system based on any one of claims 4 to 9, and the system includes a computer device, a computer program is stored in the memory of the computer device, and when the computer program is loaded by the processor of the computer device, the following steps are performed: in the comprehensive scoring model module, the program is run to execute the following comprehensive scoring formula to calculate the comprehensive score: Comprehensive score = set weight value 1 × face matching score + set weight value 2 × attribute matching score + set weight value 3 × behavior matching score comprehensive score; Among them, the set weight value one, the set weight value two and the set weight value three are all fixed values ​​that can be adjusted, and a threshold is set for the comprehensive score. If it is higher than the threshold, the match is considered successful, otherwise the match is unsuccessful.

Images