Method and system for calculating moving line trajectory of client field

Abstract

The invention discloses a method and a system for calculating a moving line trajectory of a customer field, which belong to the field of calculation of the moving line trajectory of the customer field, and comprise a data acquisition module, a data processing module, a rule engine module and a trajectory generation module, the data acquisition module is used for capturing facial features and first behavior data of a customer and recording a timestamp and a position coordinate as basic data; the data processing module is used for processing the collected data, confirming the identity of the client and converting the processed data into second behavior data; the rule engine module is used for generating a customer behavior state and regional information; and the track generation module is used for calculating the moving track of the customer according to the customer behavior state and the regional information output by the rule engine. According to the method, the tracks generated in various actual scenes are ensured to have the advantages of high accuracy, high applicability and high efficiency from multiple dimensions.

Description

Technical Field

[0001] The present invention relates to the field of customer site moving line trajectory calculation, and more specifically, to a method and system for calculating customer site moving line trajectory. Background Art

[0002] The realization of intelligence relies on the support of digital technology. Through visual recognition technology, customer behavior and related visual data are digitized to build accurate basic data. This data is not only used to analyze customer portraits, but also provides a basis for generating effective transaction recommendations.

[0003] As an important part of the digitalization process, customer trajectory calculation can provide accurate customer movement lines. However, due to the cost limitation of camera deployment, it is impossible to achieve full coverage in the case site, which affects the accuracy of data collection and customer trajectory. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a method and system for calculating the trajectory of customer site movement, which ensures from multiple dimensions that the trajectory generated in various actual scenarios has high accuracy, high applicability and high efficiency.

[0005] The object of the present invention is achieved through the following solutions:

[0006] A system for calculating a customer's case site moving line trajectory, comprising: a data acquisition module, a data processing module, a rule engine module and a trajectory generation module; the data acquisition module is used to capture the customer's facial features and first behavior data, and record the timestamp and position coordinates as basic data; the data processing module is used to process the collected data, confirm the customer's identity, and convert the processed data into second behavior data; the rule engine module is used to generate customer behavior status and area information; the trajectory generation module calculates the customer's moving trajectory according to the customer behavior status and area information output by the rule engine.

[0007] Furthermore, in the data collection module, the first behavior data includes the residence time, moving direction, walking distance and step frequency; the second behavior data includes the residence time, moving direction, walking distance and step frequency of a customer with confirmed identity at each point.

[0008] Furthermore, the data processing module specifically includes: an image recognition submodule, a feature extraction submodule and a face recognition submodule; the image recognition submodule is used to use a deep learning algorithm to process the image captured by the camera, perform face recognition and customer identity confirmation; the feature extraction submodule is used to use a CNN deep learning model to extract features from the image, and use the model to extract key features in the image; the face recognition submodule is used to use a pre-trained deep learning model ResNet for face recognition.

[0009] Furthermore, the use of the pre-trained deep learning model ResNet for face recognition specifically includes the following processes: inputting the processed image into the deep neural network, and performing face detection using the trained model; after identifying the face area in the image, further performing face feature extraction to obtain a unique "face feature vector" for each customer; and comparing the feature vector with the customer information stored in the database to confirm the customer's identity.

[0010] Furthermore, the rule engine module specifically includes: a custom area data submodule, an initialization modeling data submodule, an area division configuration submodule and a rule engine implementation submodule; the custom area data submodule is used to define area functions and area characteristics; the initialization modeling data submodule is used to enter modeling data, data calibration and measurement; the area division configuration submodule is used to analyze the actual environment, define area boundaries and configure area division rules; the rule engine implementation submodule is used to load custom areas and modeling data, real-time behavior data processing and output behavior status and trajectory data.

[0011] Furthermore, in the area division configuration submodule, the actual environment analysis, definition of area boundaries and configuration of area division rules are specifically performed in the following process: Step 1: Actual environment analysis. The system first analyzes the actual environment within the case site to ensure that the area division conforms to the actual situation and truly reflects the customer's activity trajectory; Step 2: Define area boundaries. According to the environmental analysis results, the user defines the spatial boundaries of each area. The boundaries will determine the flow and behavior status of the customer between different areas; Step 3: Configure area division rules. The user configures the area division rules in the system, combines the actual environmental factors, and performs dynamic division. The rule configuration is used to ensure that the flow and interaction between different areas are logical.

[0012] Furthermore, in the rule engine implementation submodule, the loading of custom regions and modeling data, real-time behavior data processing, and output of behavior status and trajectory data are specifically performed as follows:

[0013] Step 1: Load custom areas and modeling data. The rule engine loads user-defined area data, initialized modeling data, and area division rules to form a complete area configuration and behavior model. Step 2: Real-time behavior data processing. When the camera captures the customer's behavior data, the rule engine processes the data in real time according to preset rules, determines the customer's area affiliation, residence time, flow path, and generates the customer's behavior status. Step 3: Output behavior status and trajectory data. The rule engine outputs the customer's behavior status data and its corresponding trajectory information based on the real-time processing results.

[0014] Furthermore, in the trajectory generation module, the customer's movement trajectory is calculated based on the customer behavior status and area information output by the rule engine, which specifically includes the following processes: using a Markov chain model to simulate the customer's behavior pattern in the case site, and combining a dynamic path generation algorithm to realize the calculation and optimization of the customer's trajectory.

[0015] A method for calculating the trajectory of a customer's case site, based on the system for calculating the trajectory of a customer's case site as described above, using a Markov chain model to simulate the behavior pattern of the customer in the case site, and combining a dynamic path generation algorithm to realize the calculation and optimization of the customer's trajectory, specifically includes the following steps:

[0016] Step 1, constructing a Markov chain model, using the Markov chain model to describe the probability of a customer's transfer between different regions, and the transfer probability between each region is calculated based on the flow of customers between each region;

[0017] Step 2: Design a dynamic path generation algorithm. During the customer trajectory generation process, use the dynamic path generation algorithm to optimize the customer's path in real time.

[0018] Step 3, output of trajectory generation: output the calculated customer trajectory and store it in the database.

[0019] Furthermore, in step 1, the Markov chain model is constructed, and the Markov chain model is used to describe the transfer probability of customers between different regions. The transfer probability between each region is calculated based on the flow of customers between each region, which specifically includes the following process:

[0020] State space definition: Each key area in the case is set as a "state", and the movement of customers between different areas is the transfer from one state to another;

[0021] State transfer matrix: If there are n regions, then the state transfer matrix PPP is an n×n matrix, where Pij represents the probability of a customer transferring from region i to region j;

[0022]

[0023] Among them, P ij It is obtained through statistical analysis of historical data and is calculated using the following formula:

[0024]

[0025] Among them, N ij is the frequency of customer movement from region i to region j, is the total customer turnover frequency in region i;

[0026] Formula for state transfer: Assume that the customer is in area i at the initial moment. After a period of time, the customer may move to other areas. The probability distribution of the area where the customer is located at a certain moment is calculated through the state transfer matrix by using the properties of the Markov chain. Assume that the initial state is vector v0, then the state of the customer at time t is:

[0027] v t =P t v0;

[0028] Among them, Vt is the state distribution of customers at time t, P t It is the t-th power of the state transfer matrix, which represents the probability distribution after t state transfers;

[0029] In step 2, the dynamic path generation algorithm is designed. During the customer trajectory generation process, the dynamic path generation algorithm is used to optimize the customer's path in real time, which specifically includes the following process:

[0030] Step 2.1, trajectory initialization: At the beginning of trajectory generation, the customer's trajectory is initialized according to the customer's stay time, movement direction, and transfer probability information in each area. The customer's trajectory is dynamically updated based on historical data and real-time data;

[0031] Step 2.2, trajectory calculation: At each moment, the customer's current area and its possible future transfer area are calculated according to the Markov chain model constructed in step 1; according to the transfer probability P ij Predict the customer's trajectory; at time t, the customer is in area i, then the next possible area j is determined by the transition probability P ij To determine; the calculation formula of the trajectory is:

[0032] x t+1 =x t +Δx;

[0033] Among them, x t represents the location of the customer at time t, and Δx represents the displacement of the customer from the current area to the next area;

[0034] Step 2.3, path smoothing and optimization: The real-time calculated trajectory is smoothed to eliminate the discontinuity or unevenness of the trajectory caused by the error of the camera or data acquisition; the trajectory smoothing process is achieved by the following formula:

[0035]

[0036] in, is the smoothed trajectory point, α is the smoothing coefficient, which controls the weight of the current point and the previous point;

[0037] Step 2.4, dynamically optimize the path: As the customer moves in the case, update the path in real time and adjust it dynamically based on historical data; if the customer's trajectory at a certain moment deviates from the expected path, use dynamic adjustment to re-optimize the path to make it more consistent with the actual customer flow pattern; the dynamic optimization process is achieved through the following optimization algorithms:

[0038]

[0039] Among them, v opt is the optimized trajectory, x i is the actual location of the customer, is the position predicted by the model;

[0040] In step 3, each trajectory contains the customer's behavior data, and the customer's complete trajectory is represented by the following format:

[0041] trajectory = (x1, x2, ..., x t );

[0042] Among them, x t Represents the location of the customer at time t. The trajectory contains all time points and corresponding locations.

[0043] The beneficial effects of the present invention include:

[0044] (1) The present invention provides a method for generating customer trajectory lines based on camera face recognition. The method identifies the appearance of customers at key points, combines time series links, and relies on custom rules to calculate regional affiliation, thereby generating customer trajectory lines based on comprehensive data, aiming to improve the accuracy of customer behavior analysis.

[0045] (2) The present invention integrates face recognition technology and uses advanced face recognition technology to capture the customer's behavior data in real time in the case site to ensure accurate identification of the customer's identity; the flexible application of customized regional data and rule engine allows users to customize regional data according to actual needs, and process customer behavior information in real time through the rule engine to realize dynamic calculation of customer trajectory; the initialization modeling data can be accurately entered to ensure that the modeling data between points accurately reflects the customer's behavior characteristics and provides a reliable basis for trajectory calculation; using a specially designed rule engine, the customer behavior data is integrated to calculate the trajectory, and the path is optimized in real time to adapt to different customer flow patterns.

[0046] (3) The present invention combines face recognition technology, the configuration and application of custom region association, the input and use of modeling data, and the configuration and application of region division to calculate trajectories. Through this overall process and method, it is ensured from multiple dimensions that the trajectories generated in various practical scenarios have high accuracy, applicability, and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] 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.

[0048] Figure 1 A schematic diagram of camera deployment of a data acquisition module in an embodiment of the present invention;

[0049] Figure 2 It is a flowchart of the overall process of an embodiment of the present invention. DETAILED DESCRIPTION

[0050] 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.

[0051] In an embodiment, the present invention provides a method and system for calculating a customer's trajectory based on a rule engine. The data returned by the camera, the area data between the points, the initialization modeling data between the points, and the area division between the points are input into the rule engine to calculate the customer's trajectory, which mainly includes the following key components and corresponding processing flows:

[0052] like Figure 2 As shown, in the system embodiment of the present invention, it includes:

[0053] The data collection module rationally arranges multiple cameras in the case site to cover key points, such as entrances, exits, and negotiation areas, to ensure that the customer's movement can be captured. Figure 1 As shown in the figure, multiple high-resolution cameras are arranged at key points of the case site to ensure comprehensive monitoring of customer activities. The cameras should have the ability to transmit real-time video streams and be connected to the server. The cameras are used to capture the customer's facial features and behavior data in real time. The behavior data includes the stay time, movement direction, walking distance, step frequency and other data, record the timestamp and location coordinates, and use face recognition algorithms to identify the customer's identity in real time and analyze the customer's basic information (such as gender, age, etc.) to provide basic data for subsequent analysis.

[0054] The data processing module specifically includes an image recognition submodule, a feature extraction submodule and a face recognition submodule.

[0055] In the image recognition submodule, deep learning algorithms are used to process images captured by the camera for face recognition and customer identity confirmation. The specific steps are as follows: Preprocess the captured images, including image denoising, cropping, brightness adjustment, and standardization, to improve image quality and reduce interference factors. This stage mainly adjusts the contrast, brightness, and size of the image to make the image content clearer and adapt to the subsequent deep learning model.

[0056] In the feature extraction submodule, the convolutional neural network (CNN) deep learning model is used to extract features from the image. The model is used to extract key features in the image, such as face contours, eyes, nose, mouth and other important information. These features will become the key basis for the subsequent recognition process.

[0057] In the face recognition submodule, the pre-trained deep learning model ResNet is used for face recognition. The specific process is as follows: the processed image is input into the deep neural network, and the trained model is used for face detection; after the face area in the image is identified, the face feature is further extracted to obtain the unique "face feature vector" for each customer; the feature vector is compared with the customer information stored in the database to confirm the customer's identity (such as gender, age, identity identification, etc.).

[0058] The rule engine module specifically includes a custom region data submodule, an initialization modeling data submodule, a region division configuration submodule and a rule engine implementation submodule.

[0059] In the custom area data submodule, the following sub-steps are specifically performed: Step 1: Define area functions. According to business needs, users first define the area functions between key points, such as display areas, negotiation areas, rest areas, etc. The functions and characteristics of each area will directly affect the results of customer behavior analysis. Step 2: Define area characteristics. Users define relevant feature data for each area, such as the area of ​​the area, channel width, customer flow path, etc. These characteristics describe the physical environment of the area and help the system understand the customer's behavior patterns in the case site.

[0060] In the initialization modeling data submodule, the following sub-steps are specifically performed: Step 1: Enter modeling data. The user manually enters the behavioral characteristic data in each area according to the actual situation. These data include the average stay time, flow frequency, and transfer probability of customers in each area. These data provide basic information for subsequent rule engine processing; Step 2: Data calibration and measurement. The system calibrates and measures the entered data to ensure the accuracy and consistency of the modeling data. The calibrated data is used to describe the flow patterns and behavioral laws of customers between different areas.

[0061] In the area division configuration submodule, the following sub-steps are specifically performed: Step 1: Actual environment analysis. The system first analyzes the actual environment in the case site, including obstacles, layout design, channels, and customer flow paths. This step is to ensure that the area division conforms to the actual situation and can truly reflect the customer's activity trajectory. Step 2: Define area boundaries. Based on the results of the environmental analysis, the user defines the spatial boundaries of each area. These boundaries will determine the flow and behavior status of customers between different areas. Step 3: Configure area division rules. The user configures the area division rules in the system and performs dynamic division based on actual environmental factors, such as channel width, customer density, etc. The rule configuration ensures that the flow and interaction between different areas are logical.

[0062] In the rule engine implementation submodule, the following substeps are specifically performed: Step 1: Load custom regions and modeling data. The rule engine loads user-defined region data, initialized modeling data, and region division rules to form a complete region configuration and behavior model. Step 2: Real-time behavior data processing. When the camera captures the customer's behavior data, the rule engine processes the data in real time according to the preset rules, determines the customer's region affiliation, residence time, flow path, etc., and generates the customer's behavior status. Step 3: Output behavior status and trajectory data. The rule engine outputs the customer's behavior status data and its corresponding trajectory information based on the real-time processing results. These data will provide necessary information for the subsequent trajectory generation module.

[0063] The trajectory generation module specifically includes a trajectory generation process starting submodule and a region attribution analysis submodule.

[0064] In the trajectory generation process startup submodule, the following substeps are specifically performed: Step 1: Receive the behavior status and regional information output by the rule engine. After the rule engine module completes the customer behavior data processing, the trajectory generation module receives the customer behavior status and regional attribution information output by the rule engine. These data include the customer's stay time, flow path, regional attribution, etc. Step 2: Data integration and initialization of trajectory data. According to the regional information provided by the rule engine, the trajectory generation module first integrates the behavior data and generates a preliminary trajectory data structure. This structure includes basic information such as the starting position, target area, and movement path of each customer.

[0065] In the regional attribution analysis submodule, the following substeps are specifically performed: Step 1: Analyze the customer's activity area based on the regional attribution information. Through the regional attribution information provided by the rule engine, the system will analyze the customer's activity area in the case. Each behavior state of the customer (such as staying, moving) will be mapped to a specific area to help the system identify the customer's activities in different areas. Step 2: Calculate the main activity area. Based on the customer's stay time and visit frequency in each area, the system will identify the customer's main activity area and provide reference for subsequent trajectory generation based on this information.

[0066] Specifically, based on the above embodiment system, in the method embodiment of the present invention, the Markov chain model is used to simulate the behavior pattern of customers in the case field, and combined with the dynamic path generation algorithm to realize the calculation and optimization of customer trajectories. The whole process includes the following key steps:

[0067] Step 1. Build a Markov chain model: The Markov chain model is mainly used to describe the probability of customers transferring between different regions. The transfer probability between each region is calculated based on historical data (customer flow between regions), which includes the following process:

[0068] State space definition: Set each key area in the site as a "state", such as entrance (S1), negotiation area (S2), display area (S3), etc. The movement of customers between different areas is the transfer from one state to another.

[0069] State transfer matrix: Assuming there are n regions, the state transfer matrix PPP is an n×n matrix, where Pij represents the probability of a customer transferring from region i to region j.

[0070]

[0071] Among them, Pij can be obtained through statistical analysis of historical data, and the specific formula is:

[0072]

[0073] Among them, N ij is the frequency of customer movement from region i to region j, is the total customer turnover frequency in region i.

[0074] State transfer formula: Assume that the customer is in region i at the initial moment. After a period of time, the customer may move to other regions. Based on the properties of the Markov chain, the probability distribution of the region where the customer is located at a certain moment can be calculated through the state transfer matrix P. Assuming that the initial state is vector v0, the state of the customer at time t is:

[0075] v t =P t v0;

[0076] Among them, V t is the state distribution of customers at time t, P t It is the t-th power of the state transfer matrix, which represents the probability distribution after t state transfers.

[0077] Step 2. Design the following dynamic path generation algorithm: During the customer trajectory generation process, the dynamic path generation algorithm can optimize the customer's path in real time to ensure that the path has high accuracy and readability. This algorithm mainly includes the following steps:

[0078] Step 2.1 Trajectory initialization: At the beginning of trajectory generation, the system will initialize the customer's trajectory based on the customer's stay time in each area, movement direction, transfer probability, etc. The customer's trajectory will be dynamically updated based on historical data and real-time data.

[0079] Step 2.2 Trajectory calculation: At each moment, the customer's current region and possible future transfer regions are calculated based on the Markov chain model constructed in step 1. The system predicts the customer's trajectory based on the transfer probability. At time t, the customer is in region i, then the next possible region j is determined by the transfer probability P ij The trajectory is calculated as:

[0080] x t+1 =x t +△x;

[0081] Among them, Xt represents the location of the customer at time t, and Δx represents the displacement of the customer from the current area to the next area.

[0082] Step 2.3 Path smoothing and optimization: In order to improve the accuracy and readability of the trajectory, the dynamic path generation algorithm will smooth the trajectory calculated in real time. Commonly used smoothing algorithms include Kalman filtering or Bayesian filtering. This process is mainly used to eliminate trajectory discontinuities or unevenness caused by errors in the camera or data collection.

[0083] The trajectory smoothing process can be achieved by the following formula:

[0084]

[0085] in, is the smoothed trajectory point, α is the smoothing coefficient, which controls the weight of the current point and the previous point.

[0086] Step 2.4 Dynamically optimize the path: As the customer moves through the case, the system will update its path in real time and dynamically adjust it based on historical data. If the customer's trajectory at a certain moment deviates from the expected path, the algorithm will use a dynamic adjustment mechanism to optimize the path to make it more consistent with the actual customer flow pattern.

[0087] The dynamic optimization process can be achieved through the following optimization algorithms:

[0088]

[0089] Among them, v opt is the optimized trajectory, x i is the actual location of the customer, is the position predicted by the model.

[0090] Step 3. Output of trajectory generation: Finally, the system will output the calculated customer trajectory and store it in the database. Each trajectory contains the customer's behavior data, including residence time, flow path, transition probability and other information. The customer's complete trajectory can be represented in the following format:

[0091] trajectory = (x1, x2, ..., x t )

[0092] Among them, x i Represents the location of the customer at time i. The trajectory contains all time points and corresponding locations.

[0093] 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.

[0094] 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.

[0095] 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 system for calculating the trajectory of a customer's case site, characterized in that: include: Data acquisition module, data processing module, rule engine module and trajectory generation module; The data acquisition module is used to capture the customer's facial features and first behavior data, and record the timestamp and location coordinates as basic data; the data processing module is used to process the collected data, confirm the customer's identity, and convert the processed data into second behavior data; the rule engine module is used to generate customer behavior status and area information; the trajectory generation module calculates the customer's movement trajectory based on the customer behavior status and area information output by the rule engine.

2. The system for calculating the customer's case site movement trajectory according to claim 1, characterized in that: In the data collection module, the first behavior data includes the residence time, moving direction, walking distance and step frequency; the second behavior data includes the residence time, moving direction, walking distance and step frequency of a customer with confirmed identity at each point.

3. The system for calculating the customer's case site movement trajectory according to claim 1, characterized in that: The data processing module specifically includes: an image recognition submodule, a feature extraction submodule and a face recognition submodule; the image recognition submodule is used to use a deep learning algorithm to process the image captured by the camera, perform face recognition and customer identity confirmation; the feature extraction submodule is used to use a CNN deep learning model to extract features from the image, and use the model to extract key features in the image; the face recognition submodule is used to use a pre-trained deep learning model ResNet for face recognition.

4. The system for calculating the customer's case site movement trajectory according to claim 3, characterized in that: The face recognition using the pre-trained deep learning model ResNet specifically includes the following process: inputting the processed image into the deep neural network, and performing face detection using the trained model; after identifying the face area in the image, further performing face feature extraction to obtain a unique "face feature vector" for each customer; and comparing the feature vector with the customer information stored in the database to confirm the customer's identity.

5. The system for calculating the customer's case site movement trajectory according to claim 1, characterized in that: The rule engine module specifically includes: a custom area data submodule, an initialization modeling data submodule, an area division configuration submodule and a rule engine implementation submodule; the custom area data submodule is used to define area functions and area characteristics; the initialization modeling data submodule is used to enter modeling data, data calibration and measurement; the area division configuration submodule is used to analyze the actual environment, define area boundaries and configure area division rules; the rule engine implementation submodule is used to load custom areas and modeling data, real-time behavior data processing and output behavior status and trajectory data.

6. The system for calculating the customer's case site movement trajectory according to claim 1, characterized in that: In the area division configuration submodule, the actual environment analysis, definition of area boundaries and configuration of area division rules are specifically performed in the following process: Step 1: Actual environment analysis. The system first analyzes the actual environment in the case site to ensure that the area division conforms to the actual situation and truly reflects the customer's activity trajectory; Step 2: Define area boundaries. According to the results of the environmental analysis, the user defines the spatial boundaries of each area. The boundaries will determine the flow and behavior status of customers between different areas; Step 3: Configure area division rules. The user configures the area division rules in the system, combines the actual environmental factors, and performs dynamic division. The rule configuration is used to ensure that the flow and interaction between different areas are logical.

7. The system for calculating the customer's case site movement trajectory according to claim 1, characterized in that: In the rule engine implementation submodule, the loading of custom areas and modeling data, real-time behavior data processing and output of behavior status and trajectory data are specifically performed as follows: Step 1: Loading custom areas and modeling data, the rule engine loads user-defined area data, initialized modeling data and area division rules to form a complete area configuration and behavior model; Step 2: Real-time behavior data processing, when the camera captures the customer's behavior data, the rule engine processes the data in real time according to preset rules, determines the customer's area affiliation, residence time, flow path, and generates the customer's behavior status; Step 3: Output behavior status and trajectory data, the rule engine outputs the customer's behavior status data and its corresponding trajectory information based on the real-time processing results.

8. The system for calculating the customer's case site movement trajectory according to claim 1, characterized in that: In the trajectory generation module, the customer's movement trajectory is calculated based on the customer behavior status and area information output by the rule engine, which specifically includes the following processes: using the Markov chain model to simulate the customer's behavior pattern in the case site, and combining the dynamic path generation algorithm to realize the calculation and optimization of the customer's trajectory.

9. A method for calculating the trajectory of a customer's case site, characterized in that: Based on the system for calculating the customer's case site moving line trajectory according to claim 8, the Markov chain model is used to simulate the customer's behavior pattern in the case site, and the dynamic path generation algorithm is combined to realize the calculation and optimization of the customer's trajectory, which specifically includes the following steps: Step 1, constructing a Markov chain model, using the Markov chain model to describe the probability of a customer's transfer between different regions, and the transfer probability between each region is calculated based on the flow of customers between each region; Step 2: Design a dynamic path generation algorithm. During the customer trajectory generation process, use the dynamic path generation algorithm to optimize the customer's path in real time. Step 3, output of trajectory generation: output the calculated customer trajectory and store it in the database.

10. The method for calculating the trajectory of a customer's case site according to claim 9, characterized in that: In step 1, the Markov chain model is constructed, and the Markov chain model is used to describe the transfer probability of customers between different regions. The transfer probability between each region is calculated based on the flow of customers between each region, which specifically includes the following process: State space definition: Each key area in the case is set as a "state", and the movement of customers between different areas is the transfer from one state to another; State transfer matrix: If there are n regions, then the state transfer matrix PPP is an n×n matrix, where Pij represents the probability of a customer transferring from region i to region j; Among them, Pij is obtained through statistical analysis of historical data and is calculated using the following formula: Among them, N ij is the frequency of customer movement from region i to region j, is the total customer turnover frequency in region i; State transfer formula: Assume that the customer is in area i at the initial moment. After a period of time, the customer may move to other areas. The probability distribution of the area where the customer is located at a certain moment is calculated through the state transfer matrix by using the properties of the Markov chain. Assume that the initial state is vector v0, then the state of the customer at time t is: v t =P t ·v0; Among them, Vt is the state distribution of customers at time t, P t It is the t-th power of the state transfer matrix, which represents the probability distribution after t state transfers; In step 2, the dynamic path generation algorithm is designed. During the customer trajectory generation process, the dynamic path generation algorithm is used to optimize the customer's path in real time, which specifically includes the following process: Step 2.1, trajectory initialization: At the beginning of trajectory generation, the customer's trajectory is initialized according to the customer's stay time, movement direction, and transfer probability information in each area. The customer's trajectory is dynamically updated based on historical data and real-time data; Step 2.2, trajectory calculation: At each moment, calculate the customer's current area and its possible future transfer area according to the Markov chain model constructed in step 1; predict the customer's trajectory according to the transfer probability; at time t, the customer is in area i, then the next possible area j is determined by the transfer probability P ij To determine; the calculation formula of the trajectory is: x t+1 =x t +△x; Among them, x t represents the location of the customer at time t, and Δx represents the displacement of the customer from the current area to the next area; Step 2.3, path smoothing and optimization: The real-time calculated trajectory is smoothed to eliminate the discontinuity or unevenness of the trajectory caused by the error of the camera or data acquisition; the trajectory smoothing process is achieved by the following formula: in, is the smoothed trajectory point, α is the smoothing coefficient, which controls the weight of the current point and the previous point; Step 2.4, dynamically optimize the path: As the customer moves in the case, update the path in real time and adjust it dynamically based on historical data; if the customer's trajectory at a certain moment deviates from the expected path, use dynamic adjustment to re-optimize the path to make it more consistent with the actual customer flow pattern; the dynamic optimization process is achieved through the following optimization algorithms: Among them, v opt is the optimized trajectory, x i is the actual location of the customer, is the position predicted by the model; In step 3, each trajectory contains the customer's behavior data, and the customer's complete trajectory is represented by the following format: trajectory = (x1, x2, ..., x t ); Among them, x t Represents the location of the customer at time t. The trajectory contains all time points and corresponding locations.

Images