An unmanned self-service cash collection system
By combining mobile phone locking with deep learning-based full-process verification and door lock linkage control logic, the problem of unmanned vending equipment being unable to adapt to non-standard products and having shortcomings in loss prevention has been solved. This has enabled the equipment to adapt to all categories of products and achieve zero-loss vulnerability management, thereby improving its stability and commercialization capabilities.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- WUHAN LANTAI CHUANGDA TECH CO LTD
- Filing Date
- 2026-04-24
- Publication Date
- 2026-07-10
AI Technical Summary
Unmanned vending machines cannot be adapted to non-standard goods such as fresh fruits and vegetables, irregularly shaped packaging, and bulk weighing, which leads to problems such as taking too much, taking the wrong items, and malicious evasion of payment. Existing loss prevention technologies have shortcomings and cannot protect the interests of merchants.
The system employs a control logic that combines mobile phone locking with deep learning-based end-to-end verification and door lock linkage. Through multimodal image acquisition, edge computing, and federated learning frameworks, it achieves end-to-end control of all product categories and zero-loss vulnerabilities. It uses deep learning models for real-time analysis and user risk prediction, and combines edge computing and federated learning frameworks for incremental model iteration.
Eliminate over-delivery, misdelivery, and malicious evasion of payment; achieve full-process control with zero loss prevention loopholes and adaptability to all product categories; reduce system development costs and edge deployment difficulty; and improve equipment operation stability and commercial promotion capabilities.
Smart Images
Figure CN122369155A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the technical field of self-service payment collection, and in particular to an unmanned self-service payment collection system. Background Technology
[0002] With the rapid development of the unmanned retail industry, vending machines have been widely used in public places such as shopping malls, office buildings, communities, and subway stations. Currently, the mainstream vending equipment is mainly divided into two categories: The first type is the traditional spring / track vending machine, which has a simple structure and mature technology; The second type is the open-door unmanned vending machine, which allows users to retrieve goods independently by unlocking the cabinet door, thus solving the problem of poor product compatibility of traditional equipment.
[0003] With the increasing prevalence of autonomous driving, driverless vending vehicles are becoming more common. However, during use, it has been found that the first type of vending equipment can only accommodate standardized goods of fixed size and regular shape (such as bottled beverages and boxed snacks), and cannot accommodate non-standard goods such as fresh fruits and vegetables, irregularly shaped packaging, and bulk weighing. The applicable range of goods is extremely narrow, failing to meet diversified retail needs. While the second type of vending equipment can handle a wide variety of retail goods, it has serious operational pain points. Existing loss prevention technologies have significant shortcomings, making it easy for customers to pay for more or take the wrong items, or even maliciously skip out on their payments, causing continuous economic losses to merchants. Generally, unmanned stores using this type of vending equipment can prevent customers from maliciously skipping out on their payments or taking the wrong items by locking the store doors. However, the limited space of the vending vehicle makes it difficult to perform these operations, thus failing to provide protection for merchants and resulting in poor practicality. Therefore, there is an urgent need for a driverless self-service payment system to improve these problems. Summary of the Invention
[0004] To address the aforementioned technical challenges, this invention provides a core control logic that combines mobile phone locking with deep learning-based end-to-end verification and door lock linkage. This logic fundamentally prevents over-the-top, incorrect, and malicious evasion of payment, achieving end-to-end control with zero loss vulnerabilities and adaptability to all product categories. The entire process utilizes mature, standardized deep learning models, eliminating the need for customized modifications, significantly reducing system development costs and edge deployment difficulty, and enhancing device stability and commercialization capabilities. Based on edge computing and federated learning frameworks, it supports in-vehicle offline real-time inference and incremental model iteration under privacy protection, making it an autonomous self-service payment system adapted for mobile scenarios of driverless vending vehicles.
[0005] The present invention provides an unmanned self-service payment system, comprising: Data acquisition module: Collects basic data from the entire transaction process, including user identity data, order data, device status data, and image data of the pickup process. After preprocessing the collected basic data, it transmits it to the central control module and storage module in real time. Data analysis module: It is equipped with a deep learning model adapted to vehicle edge computing, which is used to receive detection tasks issued by the central control module. Based on the deep learning model, it compares and analyzes real-time collected data with benchmark data to complete user identity detection, order product consistency detection, and user risk level detection, and sends the analysis results to the central control module in real time. Storage module: Stores basic product information, user information data, historical order data, risky usernames, equipment operating parameters, raw data throughout the entire process, as well as pre-trained data and model update parameters for deep learning models; Alarm module: Receives instructions from the central control module and performs operations such as voice broadcast reminders, on-site sound and light alarms, and remote operation terminal alarm push; Storage facility: Stores and locks customers' mobile phones; Central control module: Receives data uploaded by each module, issues control commands, verification tasks and deep learning model inference scheduling commands, coordinates the collaborative operation of each module, and executes the linkage control logic of the equipment door lock.
[0006] Preferably, the data acquisition module includes: Multimodal image acquisition unit: including visible light high-definition camera, infrared supplementary light camera and depth camera, which are deployed inside the vending machine and in the pick-up area, respectively, to collect visible light and infrared dual-modal images of the user's face during the pick-up process, video of the goods being picked up and placed and depth images of the goods; Image preprocessing unit: performs normalization, noise reduction, cropping, and scaling operations on the acquired image data, and outputs standardized data that conforms to the input format of the deep learning model; Storage mechanism status detection unit: includes infrared beam sensor, pressure sensor and electronic lock status sensor, used for dual verification of whether the mobile phone is locked, and real-time detection of the status of the storage mechanism.
[0007] Preferred deep learning models include: Face liveness detection model: Equipped with a dual-modal liveness detection algorithm of visible light and infrared, it is used to verify the identity of the user picking up the goods and the user placing the order, and at the same time identify forgery attacks such as photos, videos and 3D masks. The detection results are sent to the central control module in real time. Product recognition model: used to identify the type, quantity, and volume of products picked up and placed by users, and to calculate the weight of bulk products by combining product density data; Pickup behavior recognition model: Models the timing of user actions of picking up and putting down goods, distinguishes between picking up and putting back actions, corrects the product counting error in occluded scenarios, and generates an accurate actual pickup list; User risk dynamic prediction model: It takes the user's historical transaction behavior, frequency of abnormal behavior, detection pass rate and operation characteristics as input, and outputs the user's risk level. At the same time, it adopts a federated learning framework to realize joint training of multi-device models, and local data is not transmitted back to the central server, so as to complete the incremental iteration of the model while protecting user privacy.
[0008] Preferably, the central control module has a bidirectional communication connection with the vehicle control system of the unmanned vending vehicle. When the vehicle control system reports that the vehicle is stationary and the parking brake is engaged, the central control module starts the transaction service and real-time inference scheduling of the deep learning model. At the same time, the central control module can dynamically adjust the inference frame rate and accuracy threshold of the deep learning model according to the on-board computing power load.
[0009] Preferably, the alarm module has a built-in hierarchical alarm strategy, which executes alarm operations at the corresponding level based on the user risk level and anomaly type output by the data analysis module.
[0010] Preferred alarm actions corresponding to the following levels: minor anomalies trigger voice guidance reminders; moderate anomalies trigger on-site audio-visual alarms; and severe malicious anomalies trigger real-time alarms and video stream pushes from the remote operation terminal.
[0011] Preferably, the linkage control logic includes: Once a user's order payment is completed and the phone is locked, an unlock command can be sent to open the vending machine door. Once the locker door is closed, the product consistency check begins. When the verification result shows that the product is completely consistent with the order, an unlock command can be issued to open the mobile phone storage locker. If the test results are inconsistent, keep the phone locked and trigger an error alert.
[0012] Preferably, the storage mechanism includes a housing, an inspection door, two sets of limiting plates, multiple sets of electric cylinders, and a bottom-laying mechanism. The housing is installed on the autonomous vehicle, the inspection door is installed on the side of the housing, and a storage cavity is provided on the top of the housing. An equipment cavity is provided inside the housing. The two sets of limiting plates are slidably installed on the top of the storage cavity of the housing. Multiple sets of electric cylinders are installed in the storage cavity of the housing, and one end of each set of electric cylinders is connected to the two sets of limiting plates. A weight sensor is also provided in the storage cavity of the housing. The bottom-laying mechanism is installed in the equipment cavity of the housing. The bottom-laying mechanism keeps the storage cavity of the housing clean. After the customer puts the mobile phone into the storage cavity of the housing, the weight sensor in the inspection door detects the mobile phone. After a preset time, the multiple sets of electric cylinders extend, causing the two sets of limiting plates to move towards each other, thus securing the mobile phone in the storage cavity of the housing.
[0013] Preferably, the bottom-laying mechanism includes a dual-output reducer, a raw material shaft, a waste material shaft, and a drive motor. The dual-output reducer is installed in the equipment cavity of the housing. The raw material shaft and the waste material shaft are respectively installed on the two sets of output ends of the dual-output reducer. The drive motor is fixedly installed on the dual-output reducer, and the output shaft of the drive motor is connected to the input end of the dual-output reducer. The non-woven fabric is rolled onto the raw material shaft, and one end of the non-woven fabric is passed from the top of the equipment cavity of the housing into the storage cavity of the housing, and then passed back into the equipment cavity of the housing from the bottom of the storage cavity. Then, one end of the non-woven fabric is fixed onto the waste material shaft. After the customer takes the mobile phone out of the storage cavity of the housing, the drive motor runs, and through the transmission of the dual-output reducer, the raw material shaft releases the non-woven fabric on it, and the waste material shaft rolls up the non-woven fabric in the storage cavity of the housing, thereby replacing the non-woven fabric inside the housing.
[0014] Preferably, when using an unmanned self-service payment system for payment collection, the following steps are included: S1. The central control module confirms that the driverless vending vehicle is parked and stationary, starts the transaction service, the user scans the code to select the goods, generates an order and completes the payment. The system simultaneously completes the pre-verification of the user's risk level. High-risk users are directly intercepted and refunded via the original payment method. S2. After receiving the payment completion signal, the central control module guides the user to place the phone used for order payment into the storage mechanism, ensuring the phone's touchscreen faces upwards. After double verification confirms the phone is locked,... S3: After the control module confirms the order payment is completed and the mobile phone storage cabinet is locked, the vending cabinet door is unlocked. At the same time, the deep learning model runs and starts real-time reasoning verification. S4. During the pickup process, the data acquisition module and the data analysis module work together to confirm that the pickup user and the ordering user are the same entity, and record the category and quantity of goods picked up and put down by the user in real time, and generate a dynamic pickup list. S5. After the user closes the vending machine door, the product consistency check begins, and the deep learning model completes the comparison and verification of the final pick-up list and the payment order. S6, Detection passed: The central control module unlocks the storage mechanism, the user retrieves the mobile phone to complete the transaction, and the transaction data is synchronized to the storage module for later training of the deep learning model; If the test fails: Keep the phone locked and guide the user to check and adjust the product via voice. After the adjustment is completed, re-verify until the verification is passed and the storage mechanism is unlocked. If the test fails multiple times, guide the customer via voice to use the phone to call the service hotline through the gap between the two sets of limit plates so that a person can remotely confirm and operate. S7. For users with malicious or abnormal behavior, update their risk level and include them in the list of risky users. Regularly update the federated learning of the deep learning model based on newly added local transaction data, in conjunction with the central server.
[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By combining mobile phone locking with deep learning-based full-process verification and door lock linkage core control logic, the system can fundamentally prevent the taking of too many items, the wrong items, and malicious evasion of payment, achieving full-process control with zero loss prevention loopholes and adaptability to all categories of goods. 2. The entire process adopts mature and standardized deep learning models in the industry, without the need for customized improvements, which greatly reduces system development costs and edge deployment difficulty, and improves the stability of device operation and commercial promotion capabilities; 3. Based on edge computing and federated learning framework, it supports in-vehicle offline real-time inference and incremental model iteration under privacy protection, adapting to the mobile scenario of driverless vending vehicles. Attached Figure Description
[0016] Figure 1 This is a schematic diagram of the unmanned self-service payment system of the present invention; Figure 2 This is a schematic diagram of the data acquisition module of the present invention; Figure 3 This is a schematic diagram of the data analysis module of the present invention; Figure 4 This is a first isometric structural schematic diagram of the storage mechanism of the present invention; Figure 5 This is a second isometric structural schematic diagram of the storage mechanism of the present invention; Figure 6 This is a schematic diagram of the right-side cross-sectional structure of the present invention.
[0017] The following are labeled in the attached diagram: 1. Outer casing; 2. Inspection door; 3. Limiting plate; 4. Electric cylinder; 5. Dual-output reducer; 6. Raw material shaft; 7. Waste shaft; 8. Drive motor. Detailed Implementation
[0018] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. The present invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.
[0019] Example: Figures 1 to 6 As shown, an unmanned self-service payment system includes: Data acquisition module: Collects basic data from the entire transaction process, including user identity data, order data, device status data, and image data of the pickup process. After preprocessing the collected basic data, it transmits it to the central control module and storage module in real time. Data analysis module: It is equipped with a deep learning model adapted to vehicle edge computing, which is used to receive detection tasks issued by the central control module. Based on the deep learning model, it compares and analyzes real-time collected data with benchmark data to complete user identity detection, order product consistency detection, and user risk level detection, and sends the analysis results to the central control module in real time. Storage module: Stores basic product information, user information data, historical order data, risky usernames, equipment operating parameters, raw data throughout the entire process, as well as pre-trained data and model update parameters for deep learning models; Alarm module: Receives instructions from the central control module and performs operations such as voice broadcast reminders, on-site sound and light alarms, and remote operation terminal alarm push; Storage facility: Stores and locks customers' mobile phones; Central control module: Receives data uploaded by each module, issues control commands, verification tasks and deep learning model inference scheduling commands, coordinates the collaborative operation of each module, and executes the linkage control logic of the equipment door lock; The data acquisition module includes: Multimodal image acquisition unit: including visible light high-definition camera, infrared supplementary light camera and depth camera, which are deployed inside the vending machine and in the pick-up area, respectively, to collect visible light and infrared dual-modal images of the user's face during the pick-up process, video of the goods being picked up and placed and depth images of the goods; Image preprocessing unit: performs normalization, noise reduction, cropping, and scaling operations on the acquired image data, and outputs standardized data that conforms to the input format of the deep learning model; Storage mechanism status detection unit: includes infrared beam sensor, pressure sensor and electronic lock status sensor, used for dual verification of whether the mobile phone is locked, and real-time detection of the status of the storage mechanism; Deep learning models include: Face liveness detection model: Equipped with a dual-modal liveness detection algorithm of visible light and infrared, it is used to verify the identity of the user picking up the goods and the user placing the order, and at the same time identify forgery attacks such as photos, videos and 3D masks. The detection results are sent to the central control module in real time. Product recognition model: used to identify the type, quantity, and volume of products picked up and placed by users, and to calculate the weight of bulk products by combining product density data; Pickup behavior recognition model: Models the timing of user actions of picking up and putting down goods, distinguishes between picking up and putting back actions, corrects the product counting error in occluded scenarios, and generates an accurate actual pickup list; User risk dynamic prediction model: It takes the user's historical transaction behavior, frequency of abnormal behavior, detection pass rate and operation characteristics as input, and outputs the user's risk level. At the same time, it adopts a federated learning framework to realize joint training of multi-device models, and local data is not transmitted back to the central server, so as to complete the incremental iteration of the model while protecting user privacy. The central control module has a two-way communication connection with the vehicle control system of the unmanned vending vehicle. When the vehicle control system reports that the vehicle is stationary and the parking brake is engaged, the central control module starts the transaction service and real-time inference scheduling of the deep learning model. At the same time, the central control module can dynamically adjust the inference frame rate and accuracy threshold of the deep learning model according to the on-board computing power load. The alarm module has a built-in hierarchical alarm strategy, which executes alarm operations at the corresponding level based on the user risk level and anomaly type output by the data analysis module. Corresponding alarm actions: minor anomalies trigger voice guidance reminders, moderate anomalies trigger on-site audio and visual alarms, and severe malicious anomalies trigger real-time alarms and video stream pushes from the remote operation terminal; The linkage control logic includes: Once a user's order payment is completed and the phone is locked, an unlock command can be sent to open the vending machine door. Once the locker door is closed, the product consistency check begins. When the verification result shows that the product is completely consistent with the order, an unlock command can be issued to open the mobile phone storage locker. If the test results are inconsistent, keep the phone locked and trigger an error alert. The storage mechanism includes an outer shell 1, an inspection door 2, two sets of limiting plates 3, multiple sets of electric cylinders 4, and a bottom-laying mechanism. The outer shell 1 is installed on the unmanned vehicle, the inspection door 2 is installed on the side of the outer shell 1, and a storage cavity is provided on the top of the outer shell 1. An equipment cavity is provided inside the outer shell 1. The two sets of limiting plates 3 are slidably installed on the top of the storage cavity of the outer shell 1. The multiple sets of electric cylinders 4 are installed in the storage cavity of the outer shell 1, and one end of each set of electric cylinders 4 is connected to the two sets of limiting plates 3. A weight sensor is also provided in the storage cavity of the outer shell 1, and the bottom-laying mechanism is installed in the equipment cavity of the outer shell 1. The bottom-laying mechanism includes a dual-output reducer 5, a raw material shaft 6, a waste shaft 7, and a drive motor 8. The dual-output reducer 5 is installed in the equipment cavity of the housing 1. The raw material shaft 6 and the waste shaft 7 are respectively installed on the two sets of output ends of the dual-output reducer 5. The drive motor 8 is fixedly installed on the dual-output reducer 5, and the output shaft of the drive motor 8 is connected to the input end of the dual-output reducer 5.
[0020] When using an unmanned self-service payment system, the following steps are included: S1. The central control module confirms that the driverless vending vehicle is parked and stationary, starts the transaction service, the user scans the code to select the goods, generates an order and completes the payment. The system simultaneously completes the pre-verification of the user's risk level. High-risk users are directly intercepted and refunded via the original payment method. S2. After receiving the payment completion signal, the central control module guides the user to place the phone used for order payment into the storage mechanism, ensuring the phone's touchscreen faces upwards. After double verification confirms the phone is locked,... S3: After the control module confirms the order payment is completed and the mobile phone storage cabinet is locked, the vending cabinet door is unlocked. At the same time, the deep learning model runs and starts real-time reasoning verification. S4. During the pickup process, the data acquisition module and the data analysis module work together to confirm that the pickup user and the ordering user are the same entity, and record the category and quantity of goods picked up and put down by the user in real time, and generate a dynamic pickup list. S5. After the user closes the vending machine door, the product consistency check begins, and the deep learning model completes the comparison and verification of the final pick-up list and the payment order. S6, Detection passed: The central control module unlocks the storage mechanism, the user retrieves the mobile phone to complete the transaction, and the transaction data is synchronized to the storage module for later training of the deep learning model; If the test fails: Keep the phone locked and guide the user to check and adjust the product via voice. After the adjustment is completed, re-verify until the verification is passed and the storage mechanism is unlocked. If the test fails multiple times, guide the customer via voice to use the phone to make a service call through the gap between the two sets of limit plates 3 so that a person can remotely confirm and operate. S7. For users with malicious or abnormal behavior, update their risk level and include them in the list of risky users. Regularly update the federated learning of the deep learning model based on newly added local transaction data, in conjunction with the central server.
[0021] The unmanned self-service payment system of this invention uses common mechanical methods for installation, connection, and setup, and any method that achieves the desired beneficial effect can be implemented. The face liveness detection model adopts the MobileFaceNet face liveness verification model, the product recognition model adopts the YOLOv8n instance segmentation recognition model, and the goods retrieval behavior recognition model adopts the SlowFast temporal behavior recognition model. The electric cylinder 4, dual-output reducer 5, and drive motor 8 of the unmanned self-service payment system of this invention are commercially available. Technical personnel in this industry only need to install and operate them according to the accompanying instruction manual, without requiring any creative effort from those skilled in the art.
[0022] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. An unmanned self-service payment system, characterized in that, include: Data acquisition module: Collects basic data from the entire transaction process, including user identity data, order data, device status data, and image data of the pickup process. After preprocessing the collected basic data, it transmits it to the central control module and storage module in real time. Data analysis module: It is equipped with a deep learning model adapted to vehicle edge computing, which is used to receive detection tasks issued by the central control module. Based on the deep learning model, it compares and analyzes real-time collected data with benchmark data to complete user identity detection, order product consistency detection, and user risk level detection, and sends the analysis results to the central control module in real time. Storage module: Stores basic product information, user information data, historical order data, risky usernames, equipment operating parameters, raw data throughout the entire process, as well as pre-trained data and model update parameters for deep learning models; Alarm module: Receives instructions from the central control module and performs operations such as voice broadcast reminders, on-site sound and light alarms, and remote operation terminal alarm push; Storage facility: Stores and locks customers' mobile phones; Central control module: Receives data uploaded by each module, issues control commands, verification tasks and deep learning model inference scheduling commands, coordinates the collaborative operation of each module, and executes the linkage control logic of the equipment door lock.
2. The driverless self-service payment system as described in claim 1, characterized in that, The data acquisition module includes: Multimodal image acquisition unit: including visible light high-definition camera, infrared supplementary light camera and depth camera, which are deployed inside the vending machine and in the pick-up area, respectively, to collect visible light and infrared dual-modal images of the user's face during the pick-up process, video of the goods being picked up and placed and depth images of the goods; Image preprocessing unit: performs normalization, noise reduction, cropping, and scaling operations on the acquired image data, and outputs standardized data that conforms to the input format of the deep learning model; Storage mechanism status detection unit: includes infrared beam sensor, pressure sensor and electronic lock status sensor, used for dual verification of whether the mobile phone is locked, and real-time detection of the status of the storage mechanism.
3. The driverless self-service payment system as described in claim 1, characterized in that, Deep learning models include: Face liveness detection model: Equipped with a dual-modal liveness detection algorithm of visible light and infrared, it is used to verify the identity of the user picking up the goods and the user placing the order, and at the same time identify forgery attacks such as photos, videos and 3D masks. The detection results are sent to the central control module in real time. Product recognition model: used to identify the type, quantity, and volume of products picked up and placed by users, and to calculate the weight of bulk products by combining product density data; Pickup behavior recognition model: Models the timing of user actions of picking up and putting down goods, distinguishes between picking up and putting back actions, corrects the product counting error in occluded scenarios, and generates an accurate actual pickup list; User risk dynamic prediction model: It takes the user's historical transaction behavior, frequency of abnormal behavior, detection pass rate and operation characteristics as input, and outputs the user's risk level. At the same time, it adopts a federated learning framework to realize joint training of multi-device models, and local data is not transmitted back to the central server, so as to complete the incremental iteration of the model while protecting user privacy.
4. The driverless self-service payment system as described in claim 1, characterized in that, The central control module has a two-way communication connection with the vehicle control system of the unmanned vending vehicle. When the vehicle control system reports that the vehicle is stationary and the parking brake is engaged, the central control module starts the transaction service and real-time inference scheduling of the deep learning model. At the same time, the central control module can dynamically adjust the inference frame rate and accuracy threshold of the deep learning model according to the on-board computing power load.
5. The driverless self-service payment system as described in claim 1, characterized in that, The alarm module has a built-in hierarchical alarm strategy, which executes the corresponding alarm operation based on the user risk level and anomaly type output by the data analysis module.
6. The driverless self-service payment system as described in claim 5, characterized in that, Corresponding alarm actions: minor anomalies trigger voice guidance reminders, moderate anomalies trigger on-site audio and visual alarms, and severe malicious anomalies trigger real-time alarms and video stream pushes from the remote operation terminal.
7. The driverless self-service payment system as described in claim 1, characterized in that, The linkage control logic includes: Once a user's order payment is completed and the phone is locked, an unlock command can be sent to open the vending machine door. Once the locker door is closed, the product consistency check begins. When the verification result shows that the product is completely consistent with the order, an unlock command can be issued to open the mobile phone storage locker. If the test results are inconsistent, keep the phone locked and trigger an error alert.
8. The driverless self-service payment system as described in claim 1, characterized in that, The storage mechanism includes an outer shell (1), an inspection door (2), two sets of limit plates (3), multiple sets of electric cylinders (4), and a bottom-laying mechanism. The outer shell (1) is installed on the unmanned vehicle, the inspection door (2) is installed on the side of the outer shell (1), and a storage cavity is provided on the top of the outer shell (1). An equipment cavity is provided inside the outer shell (1). The two sets of limit plates (3) are slidably installed on the top of the storage cavity of the outer shell (1). The multiple sets of electric cylinders (4) are installed in the storage cavity of the outer shell (1), and one end of the multiple sets of electric cylinders (4) is connected to the two sets of limit plates (3) respectively. A weight sensor is also provided in the storage cavity of the outer shell (1). The bottom-laying mechanism is installed in the equipment cavity of the outer shell (1).
9. The driverless self-service payment system as described in claim 8, characterized in that, The bottom-laying mechanism includes a dual-output reducer (5), a raw material shaft (6), a waste shaft (7), and a drive motor (8). The dual-output reducer (5) is installed in the equipment cavity of the housing (1). The raw material shaft (6) and the waste shaft (7) are respectively installed on the two sets of output ends of the dual-output reducer (5). The drive motor (8) is fixedly installed on the dual-output reducer (5), and the output shaft of the drive motor (8) is connected to the input end of the dual-output reducer (5).
10. An unmanned self-service payment system as described in any one of claims 1-9, characterized in that, When using an unmanned self-service payment system, the following steps are included: S1. The central control module confirms that the driverless vending vehicle is parked and stationary, starts the transaction service, the user scans the code to select the goods, generates an order and completes the payment. The system simultaneously completes the pre-verification of the user's risk level. High-risk users are directly intercepted and refunded via the original payment method. S2. After receiving the payment completion signal, the central control module guides the user to place the phone used for order payment into the storage mechanism, ensuring the phone's touchscreen faces upwards. After double verification confirms the phone is locked,... S3: After the control module confirms the order payment is completed and the mobile phone storage cabinet is locked, the vending cabinet door is unlocked. At the same time, the deep learning model runs and starts real-time reasoning verification. S4. During the pickup process, the data acquisition module and the data analysis module work together to confirm that the pickup user and the ordering user are the same entity, and record the category and quantity of goods picked up and put down by the user in real time, and generate a dynamic pickup list. S5. After the user closes the vending machine door, the product consistency check begins, and the deep learning model completes the comparison and verification of the final pick-up list and the payment order. S6, Detection passed: The central control module unlocks the storage mechanism, the user retrieves the mobile phone to complete the transaction, and the transaction data is synchronized to the storage module for later training of the deep learning model; If the test fails: keep the phone locked and guide the user to check and adjust the product. After the adjustment is completed, re-verify until the verification is passed and the storage mechanism is unlocked. If the test fails multiple times, guide the customer to make a service call through the gap between the two sets of limit plates (3) so that the human can remotely confirm and operate. S7. For users with malicious or abnormal behavior, update their risk level and include them in the list of risky users. Regularly update the federated learning of the deep learning model based on newly added local transaction data, in conjunction with the central server.