Weighing cross-lattice dynamic track commodity identification system for automatic vending cabinet

Abstract

The invention discloses an automatic vending cabinet-oriented weighing cross-lattice dynamic track commodity identification system, which relates to the technical field of automatic vending article identification and specifically comprises the following modules: a weighing sensing module, a track updating module, an event identification module and a migration management module. A plurality of weighing sensor units are arranged at the bottom of a goods lattice, weight change data are collected in real time, zero point adjustment and temperature drift correction are carried out, a sensitivity matrix is established to calibrate sensors, a goods lattice coordinate system is established based on the physical position of an automatic vending cabinet, and a goods lattice state updating mechanism is established. Calculating the shortest distance between the dynamic three-dimensional trajectory of the trigger event and the reference trajectory, preliminarily judging the similarity through global and local threshold values, comprehensively evaluating whether the trigger event is the trigger event of the same commodity, selecting the nearest T most matched trajectories and mapping the most matched trajectories as the initial reference trajectory of a new commodity lattice, updating a migration log and a similarity score, and updating the migration log and the similarity score. And establishing a cross-lattice migration mechanism.

Description

Technical Field

[0001] This invention relates to the field of vending machine product recognition technology, specifically to a weighing cross-grid dynamic trajectory product recognition system for vending machines. Background Technology

[0002] Currently, most common automated vending machines rely on a single-point weighing module to determine when items are retrieved. This method is prone to false alarms or missed alarms when multiple people operate the machine simultaneously, when there is vibration in adjacent vending channels, or when the environment changes.

[0003] While visual perception systems can identify objects intuitively, they are limited by lighting, occlusion, and privacy, and are also costly.

[0004] The existing technology has the following shortcomings:

[0005] (1) Traditional methods often fail to maintain the continuity of the trajectory when switching between cargo compartments, which can easily lead to identification breakpoints, misidentification or omission, reducing the identification stability of high-density, multi-category goods.

[0006] (2) Using only a single point or a small amount of sensor information makes it difficult to depict rapid and localized changes in weight distribution at high resolution, which limits the ability to respond quickly to picking and placing events.

[0007] (3) The lack of systematic modeling of dynamic trajectories, constraint management of trajectory library capacity, and continuous training and version control of benchmark trajectories leads to insufficient consistency, maintainability and traceability in long-term operation.

[0008] (4) Using only simple distance or feature similarity to identify similar products lacks multi-level threshold evaluation that combines historical trajectory statistics, time window consistency and fault tolerance range of storage compartment, which can easily lead to misjudgment in high concurrency or abnormal pickup modes.

[0009] (5) In the scenario of goods migration across storage cells, existing solutions often lack selection based on the most recently matched trajectory, mapping of the initial baseline trajectory across storage cells, and rollback mechanism for conflicts after migration, which leads to increased uncertainty and insufficient traceability in the migration process.

[0010] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0011] The purpose of this invention is to provide a weighing cross-compartment dynamic trajectory merchandise recognition system for vending machines, in order to solve the problems mentioned in the background art.

[0012] To achieve the above objectives, the present invention provides the following technical solution: a weighing cross-compartment dynamic trajectory merchandise recognition system for vending machines, specifically comprising the following modules:

[0013] Weighing sensing module: By arranging multiple weighing sensor units at the bottom of the compartment, it collects weight change data in real time and performs zero-point calibration and temperature drift correction, and establishes a sensitivity matrix to calibrate the sensors;

[0014] Track update module: Establishes a grid coordinate system based on the physical location of the vending machine and establishes a grid status update mechanism;

[0015] Event recognition module: Calculates the shortest distance between the dynamic 3D trajectory of the triggering event and the baseline trajectory, makes a preliminary judgment on similarity through global and local thresholds, and comprehensively evaluates whether they are triggering events for the same product;

[0016] Migration Management Module: Select the T most recent matching trajectories and map them as the initial baseline trajectory for the new storage cell, update the migration log and similarity score, and establish a cross-storage cell migration mechanism.

[0017] As a preferred embodiment of the weighing cross-compartment dynamic trajectory merchandise recognition system for vending machines described in this invention, wherein:

[0018] M weighing sensor units are evenly arranged at the bottom of N layers of shelves in the vending machine to construct N horizontal two-dimensional weighing matrices. The shelves are divided into M sensing areas, that is, each sensing area corresponds to one weighing sensor unit. Each time the weight changes, the product is taken out or put in event. The weight change data in the corresponding sensing area is collected in parallel by each weighing sensor unit, and the sampling frequency is set to 50Hz.

[0019] Within each acquisition cycle, the sum of the output values ​​of all weighing sensor units is calculated as the total weight change data;

[0020] Zero-point calibration and temperature drift correction are performed on each weighing sensor unit. By recording the response of each weighing sensor unit under different loads, a sensitivity matrix is ​​established. Based on the response relationship of this matrix to different weight changes and distributions, the sensitivity of the weighing sensor unit is calibrated.

[0021] Save the unique barcode identifier and its corresponding weight for each product.

[0022] As a preferred embodiment of the weighing cross-compartment dynamic trajectory merchandise recognition system for vending machines described in this invention, wherein:

[0023] Establish a coordinate system for each vending machine based on its physical location within the vending machine;

[0024] The coordinates and status of each compartment in the coordinate system are stored as an object;

[0025] Bind weight change data, change timestamp, and product's storage compartment level information to the storage compartment;

[0026] Establish a storage compartment status update mechanism. Each time a product is taken out or put in, after finding the storage compartment corresponding to the product, the dynamic three-dimensional trajectory corresponding to the triggered event is mapped to the trajectory list, and the storage compartment status is updated.

[0027] N horizontal two-dimensional weighing matrices are mapped to the grid coordinate system. When there is a continuous change in weight, a dynamic three-dimensional trajectory of the goods is generated in the grid coordinate system.

[0028] The trajectory list is trained and stores the dynamic 3D trajectories of trigger events for all products in the vending machine as reference trajectories. The list capacity is set to an upper limit K, where K ≤ 50. The specific steps for establishing the list include...

[0029] The entrance and exit of the storage compartment are used as the initial endpoints of the storage compartment coordinate system. If there is a most recent endpoint of the storage compartment's trajectory, it is directly used as the new starting point.

[0030] Concurrent events are processed one by one based on their timestamps;

[0031] Based on the agreed-upon level and structural orientation of the vending machine, the weight change data is mapped to the spatial displacement component of the vending machine. The endpoint of the product's trigger event is calculated by adding the spatial displacement component to the starting coordinates.

[0032] The new track fragment is appended to the track list. The track list version number is incremented, and the last update time is updated to the current time.

[0033] As a preferred embodiment of the weighing cross-compartment dynamic trajectory merchandise recognition system for vending machines described in this invention, wherein:

[0034] The shortest distance between the dynamic three-dimensional trajectory of the product's triggering event and the baseline trajectory is calculated using a three-dimensional line segment distance measurement method as the initial confidence score;

[0035] The dynamic 3D trajectory is used to initially determine whether the triggering events are for the same product by setting global and local thresholds. Specifically, this includes...

[0036] When the initial confidence score is greater than or equal to the global threshold, further judgment is initiated, which specifically includes...

[0037] Compare the historical trajectory statistics of known similar products in the product database and calculate the similarity score;

[0038] Consistency assessment between timestamps and time windows of known shipping patterns;

[0039] Whether the hierarchical combination of the product's storage compartment information is within the same fault tolerance range.

[0040] As a preferred embodiment of the weighing cross-compartment dynamic trajectory merchandise recognition system for vending machines described in this invention, wherein:

[0041] Define cross-cargo cell migration rules, specifically including:

[0042] When the same product appears in a new storage cell, identify the most recent T trajectory list and select the dynamic 3D trajectory that best matches the current storage cell status;

[0043] Map the dynamic 3D trajectory that best matches the current state of the cargo cell to the cargo cell coordinate system of the new cargo cell, and use it as the initial reference trajectory of the new cargo cell.

[0044] The migration log between the old and new storage compartments is updated based on the initial baseline trajectory, and the migration time, product identifier, and similarity score are recorded.

[0045] Within a certain time window after migration, the dynamic three-dimensional trajectory of the new storage cell is monitored. If an overlap or conflict deviating from the expected behavior is detected, a rollback mechanism is triggered to restore the state before migration.

[0046] When the same product has more than two different dynamic 3D trajectories in the same storage compartment, the conflict is decomposed based on the most recent trigger time point, and the latest dynamic 3D trajectory is retained.

[0047] On the other hand, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements the steps of a weighing cross-compartment dynamic trajectory merchandise recognition system for vending machines as described above.

[0048] On the other hand, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements the steps of a weighing cross-compartment dynamic trajectory merchandise recognition system for vending machines as described above.

[0049] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0050] (1) By establishing an independent coordinate system for each compartment, the weight change data, timestamps and compartment hierarchy information are bound together to form a dynamic three-dimensional trajectory and realize continuous tracking and mapping across compartments. This solves the problem of the break in the traditional identification based on single-point sensing or static weight distribution in the cross-compartment scenario and improves the consistency and continuity of cross-compartment commodity identification.

[0051] (2) By using a preliminary confidence score plus a global / local threshold for hierarchical judgment, and combining historical trajectory statistics, time window consistency and fault tolerance range of the cargo compartment level, the robustness of triggering events for similar commodities is enhanced and the risk of misidentification is reduced.

[0052] (3) The cross-cargo cell migration rules, combined with the rollback mechanism, make the cargo cell transformation process controllable and traceable. Especially in high-concurrency scenarios where multiple trajectories coexist in the same cargo cell, it can effectively decompose conflicts, retain the most recent valid trajectory, and reduce the uncertainty brought about by migration. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0054] Figure 1 This is a flowchart of the method for a weighing cross-compartment dynamic trajectory product recognition system for vending machines according to the present invention.

[0055] Figure 2 This is a schematic diagram of the module of the weighing cross-compartment dynamic trajectory product recognition system for vending machines according to the present invention. Detailed Implementation

[0056] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0057] Example 1, referring to Figure 1 and Figure 2 This is the first embodiment of the present invention, which provides a weighing cross-compartment dynamic trajectory product recognition system for vending machines, specifically including the following modules:

[0058] Weighing sensing module: By arranging multiple weighing sensor units at the bottom of the compartment, it collects weight change data in real time and performs zero-point calibration and temperature drift correction, and establishes a sensitivity matrix to calibrate the sensors;

[0059] M weighing sensor units are evenly arranged at the bottom of N layers of shelves in the vending machine to construct N horizontal two-dimensional weighing matrices. The shelves are divided into M sensing areas, that is, each sensing area corresponds to one weighing sensor unit. Each time the weight changes, the product is taken out or put in event. The weight change data in the corresponding sensing area is collected in parallel by each weighing sensor unit, and the sampling frequency is set to 50Hz.

[0060] Within each acquisition cycle, the sum of the output values ​​of all weighing sensor units is calculated as the total weight change data;

[0061] Zero-point calibration and temperature drift correction are performed on each weighing sensor unit. By recording the response of each weighing sensor unit under different loads, a sensitivity matrix is ​​established. Based on the response relationship of this matrix to different weight changes and distributions, the sensitivity of the weighing sensor unit is calibrated.

[0062] It should also be noted that the relationship between the output of each weighing sensor unit and the actual load is characterized by the sensitivity matrix. Under each load condition, the response data of each sensor unit is recorded one by one. The specific correction steps are to fit the response data of the sensor by the least squares method and construct the sensitivity matrix of each sensor unit based on the fitting results.

[0063] Save the unique barcode identifier and its corresponding weight for each product.

[0064] Track update module: Establishes a grid coordinate system based on the physical location of the vending machine and establishes a grid status update mechanism;

[0065] Establish a coordinate system for each vending machine based on its physical location within the vending machine;

[0066] The coordinates and status of each compartment in the coordinate system are stored as an object;

[0067] Bind weight change data, change timestamp, and product's storage compartment level information to the storage compartment;

[0068] Establish a storage compartment status update mechanism. Each time a product is taken out or put in, after finding the storage compartment corresponding to the product, the dynamic three-dimensional trajectory corresponding to the triggered event is mapped to the trajectory list, and the storage compartment status is updated.

[0069] It should also be noted that the status definition of the storage compartment includes the quantity, weight, whether it has been taken out, and remaining inventory of the goods. Whenever a goods are taken out or put in, the weight sensor detects a change in weight and triggers a status update event.

[0070] It should also be noted that the trajectory list uses a linked list structure to store the dynamic 3D trajectory of each product. Each trajectory includes a timestamp, weight change, and coordinate change; the trajectory list can store a maximum of 50 trajectories. When a new trajectory is added, if the list is full, the oldest trajectory is deleted.

[0071] N horizontal two-dimensional weighing matrices are mapped to the grid coordinate system. When there is a continuous change in weight, a dynamic three-dimensional trajectory of the goods is generated in the grid coordinate system.

[0072] The trajectory list is trained and stores the dynamic 3D trajectories of trigger events for all products in the vending machine as reference trajectories. The list capacity is set to an upper limit K, where K ≤ 50. The specific steps for establishing the list include...

[0073] The entrance and exit of the storage compartment are used as the initial endpoints of the storage compartment coordinate system. If there is a most recent endpoint of the storage compartment's trajectory, it is directly used as the new starting point.

[0074] Concurrent events are processed one by one based on their timestamps;

[0075] Based on the agreed-upon level and structural orientation of the vending machine, the weight change data is mapped to the spatial displacement component of the vending machine. The endpoint of the product's trigger event is calculated by adding the spatial displacement component to the starting coordinates.

[0076] The new track fragment is appended to the track list. The track list version number is incremented, and the last update time is updated to the current time.

[0077] Event recognition module: Calculates the shortest distance between the dynamic 3D trajectory of the triggering event and the baseline trajectory, makes a preliminary judgment on similarity through global and local thresholds, and comprehensively evaluates whether they are triggering events for the same product;

[0078] The shortest distance between the dynamic three-dimensional trajectory of the product's triggering event and the baseline trajectory is calculated using a three-dimensional line segment distance measurement method as the initial confidence score;

[0079] It should also be noted that the three-dimensional line segment distance metric divides the trajectory into multiple fine points and directly calculates the straight-line distance between corresponding two points;

[0080] The dynamic 3D trajectory is used to initially determine whether the triggering events are for the same product by setting global and local thresholds. Specifically, this includes...

[0081] When the initial confidence score is greater than or equal to the global threshold, further judgment is initiated, which specifically includes...

[0082] Compare the historical trajectory statistics of known similar products in the product database and calculate the similarity score;

[0083] Consistency assessment between timestamps and time windows of known shipping patterns;

[0084] Whether the hierarchical combination of the product's storage compartment information is within the same fault tolerance range;

[0085] It should be further explained that the global threshold is used to determine whether an event is likely to be a trigger event for the same product. Based on the preliminary confidence score calculated from the shortest distance, the global threshold is set to 0.8 using historical data and product type. If the shortest distance calculation result between the trigger event of product A and the baseline trajectory is 0.75, then this event will be considered to be a trigger event for the same product, but further judgment is still required. If the shortest distance calculation result is 0.9, then the event obviously does not belong to the same product, and the system will filter out this event.

[0086] It should be further noted that the local threshold is used for fine-grained matching of specific products or scenarios, and is set to 0.85 based on the changes in the grid. If the shortest distance between the triggering event of product B and the baseline trajectory is calculated to be 0.72, and product B is a relatively simple product, then a local threshold of 0.7 can classify it as the same product. If product C is a more special product, and its trajectory distance is calculated to be 0.82, then setting the local threshold to 0.85 may exclude this event from being considered as the triggering event of the same product.

[0087] Migration Management Module: Select the T most recent matching trajectories and map them as the initial baseline trajectory for the new cargo cell, update the migration log and similarity score, and establish a cross-cargo cell migration mechanism;

[0088] Define cross-cargo cell migration rules, specifically including:

[0089] When the same product appears in a new storage cell, identify the most recent T trajectory list and select the dynamic 3D trajectory that best matches the current storage cell status;

[0090] Map the dynamic 3D trajectory that best matches the current state of the cargo cell to the cargo cell coordinate system of the new cargo cell, and use it as the initial reference trajectory of the new cargo cell.

[0091] The migration log between the old and new storage compartments is updated based on the initial baseline trajectory, and the migration time, product identifier, and similarity score are recorded.

[0092] Within a certain time window after migration, the dynamic three-dimensional trajectory of the new storage cell is monitored. If an overlap or conflict deviating from the expected behavior is detected, a rollback mechanism is triggered to restore the state before migration.

[0093] When the same product has more than two different dynamic 3D trajectories in the same storage compartment, the conflict is decomposed based on the most recent trigger time point, and the latest dynamic 3D trajectory is retained.

[0094] Example 2

[0095] The following is another embodiment of the present invention, which provides a weighing cross-grid dynamic trajectory merchandise recognition system for vending machines. In order to verify the beneficial effects of the present invention, a simulation experiment is conducted for scientific demonstration.

[0096] The experimental objective is defined as follows:

[0097] Verify the effectiveness of dynamic trajectory recognition, cross-grid migration, and rollback mechanism based on multi-point weighing in high-density commodity distribution scenarios.

[0098] Evaluate the impact of trajectory database capacity on recognition stability (comparing two groups with K values ​​of 50 and 40).

[0099] The tracking accuracy, false recognition rate, and migration time delay were tested at 50Hz sampling.

[0100] The container structure is defined as two layers of compartments, with four columns per layer, for a total of eight usable compartments. Four weighing sensor units are arranged at the bottom of each compartment. High-precision strain gauge weighing sensors are used as weighing sensor units, with a range of ±5 kg, a resolution of 1 g, and a sampling frequency of 50 Hz.

[0101] Zero-point calibration is performed on each sensing unit, and the initial weight zero point is recorded.

[0102] The responses were collected under different loads (no load, light load, medium load, full load) to construct a sensitivity matrix S (4×4 matrix).

[0103] Each of the four products has its barcode linked to its nominal weight, and a product weight database is established.

[0104] Simulate two concurrent pickup and drop-off scenarios, generating 1–2 events per second for 30 minutes.

[0105] Record the event timestamp, triggering compartment, compartment level information, and actual weight change distribution.

[0106] Weight change data is mapped to the grid coordinate system to generate a dynamic three-dimensional trajectory.

[0107] When a continuous weight change is detected, a trajectory is generated in the corresponding grid coordinate system and added to the trajectory list.

[0108] Calculate the shortest distance between the dynamic trajectory and the nearest set of reference trajectories to obtain a preliminary confidence score.

[0109] A preliminary judgment is made using global and local thresholds; if the criteria are met, the process proceeds to the second stage.

[0110] The final judgment result is output by comprehensively considering the historical trajectory statistics scores of similar products, time window consistency, and hierarchical fault tolerance range L.

[0111] When the same product appears in a new storage cell, the nearest T trajectories are selected for matching, mapped to the initial baseline trajectory of the new storage cell, and the migration log is recorded.

[0112] After migration, a 1000-second time period is set to monitor the trajectory. If a conflict occurs, a rollback to the state before migration is executed.

[0113] Experiment A (K=50, 30 minutes, Experimental Environment A)

[0114] Total number of events: 1800

[0115] Cross-compartment recognition success rate: 92.0%

[0116] Average migration latency: 248ms

[0117] Rollback trigger count: 7 times

[0118] Track library hit rate (percentage of tracks hitting the baseline): 74.5%

[0119] False alarm rate: 1.8%

[0120] Missed alarm rate: 3.2%

[0121] Improved robustness during high-concurrency periods: the accuracy of identification decreases by no more than 1.5 percentage points during peak periods of concurrent events.

[0122] Experiment B (K=40, 30 minutes, experimental environment B)

[0123] Total number of events: 1720

[0124] Cross-compartment recognition success rate: 89.4%

[0125] Average migration latency: 266ms

[0126] Rollback trigger count: 9 times

[0127] Tracking database hit rate: 69.2%

[0128] False alarm rate: 2.1%

[0129] Missed alarm rate: 3.8%

[0130] The observed slight decrease is due to the reduced capacity of the trajectory library, resulting in some boundary trajectories not being effectively covered.

[0131] Compared to the baseline, this solution improves the cross-grid recognition rate by approximately 8–12 percentage points, significantly reduces the average latency related to migration, and the rollback mechanism effectively reduces misordering caused by misidentification.

[0132] The size of the trajectory database significantly affects robustness and hit rate. A larger K value results in better long-term stability but also increases storage and retrieval overhead. This experimental comparison shows that K=50 provides the best overall performance, achieving high cross-cell continuity and a low false positive rate.

[0133] This embodiment verifies the effectiveness of dynamic trajectory recognition, cross-cell migration, and rollback mechanisms based on multi-point weighing in a high-density commodity distribution environment, specifically including:

[0134] Stability of cross-cell recognition: Under the condition of K=50, the cross-cell recognition rate reaches more than 90%, which is significantly higher than the baseline method of single-point weight threshold determination, indicating that trajectory-level temporal modeling effectively reduces cross-cell breakpoints.

[0135] Migration performance: The average migration latency is approximately 248ms (Experiment A), which is far superior to the millisecond to second-level response of previous mechanisms, supporting the real-time requirements of high-concurrency scenarios.

[0136] Rollback effect: The number of rollbacks was low (7 times), and the system was able to quickly restore the correct trajectory after the rollback, indicating that the migration log and conflict decomposition strategy have good traceability and stability.

[0137] The role of the trajectory library: The trajectory library hit rate was approximately 74.5% (Experiment A) and 69.2% (Experiment B), indicating that the size of the trajectory library has a direct impact on the robustness of recognition. 50 trajectories can cover most common triggering patterns in most scenarios.

[0138] Compared with the baseline: Compared with the method that only uses single-point weight and threshold judgment, the cross-grid dynamic trajectory recognition scheme has significantly improved the overall performance in terms of cross-grid continuity, false recognition rate and migration time, with an improvement of 8-12 percentage points. It also has a significant improvement in robustness during high-concurrency periods.

[0139] By establishing an independent coordinate system for each storage compartment, binding weight change data, timestamps, and storage compartment hierarchy information, a dynamic three-dimensional trajectory is formed, enabling continuous tracking and mapping across storage compartments. This solves the problem of fragmentation in traditional single-point sensing or static weight distribution-based identification in cross-storage compartment scenarios, and improves the consistency and continuity of cross-storage compartment commodity identification.

[0140] By using a tiered approach that combines preliminary confidence scores with global / local thresholds, and by integrating historical trajectory statistics, time window consistency, and the fault tolerance range of the storage compartment level, the robustness to events triggered by similar products is enhanced, and the risk of misidentification is reduced.

[0141] Cross-cargo cell migration rules, combined with a rollback mechanism, make the cargo cell transformation process controllable and traceable. Especially in high-concurrency scenarios where multiple trajectories coexist in the same cargo cell, it can effectively decompose conflicts, retain the most recent valid trajectory, and reduce the uncertainty brought about by migration.

[0142] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A weighing-based, cross-compartment dynamic trajectory product recognition system for automated vending machines, characterized in that, Specifically, it includes the following modules: Weighing sensing module: By arranging multiple weighing sensor units at the bottom of the compartment, it collects weight change data in real time and performs zero-point calibration and temperature drift correction, and establishes a sensitivity matrix to calibrate the sensors; Track update module: Establishes a grid coordinate system based on the physical location of the vending machine and establishes a grid status update mechanism; Event recognition module: Calculates the shortest distance between the dynamic 3D trajectory of the triggering event and the baseline trajectory, makes a preliminary judgment on similarity through global and local thresholds, and comprehensively evaluates whether they are triggering events for the same product; Migration Management Module: Select the T most recent matching trajectories and map them as the initial baseline trajectory for the new storage cell, update the migration log and similarity score, and establish a cross-storage cell migration mechanism.

2. The weighing cross-compartment dynamic trajectory merchandise recognition system for vending machines according to claim 1, characterized in that: The weighing sensing module evenly arranges M weighing sensor units at the bottom of N layers of shelves in the vending machine to construct N horizontal two-dimensional weighing matrices, dividing the shelves into M sensing areas. Each time the weight changes, an event of taking out or putting in the product is triggered, and the weight change data in the corresponding sensing area is collected in parallel by each weighing sensor unit. Within each acquisition cycle, the sum of the output values ​​of all weighing sensor units is calculated as the total weight change data; Zero-point calibration and temperature drift correction are performed on each weighing sensor unit. By recording the response of each weighing sensor unit under different loads, a sensitivity matrix is ​​established. Based on the response relationship of this matrix to different weight changes and distributions, the sensitivity of the weighing sensor units is calibrated.

3. The weighing cross-compartment dynamic trajectory merchandise recognition system for vending machines according to claim 1, characterized in that: The trajectory update module establishes a grid coordinate system based on the physical location of each grid in the vending machine; The coordinates and status of each compartment in the coordinate system are stored as an object; Bind weight change data, change timestamp, and product's storage compartment level information to the storage compartment; Establish a storage compartment status update mechanism. Each time a product is taken out or put in, after finding the storage compartment corresponding to the product, the dynamic three-dimensional trajectory corresponding to the triggered event is mapped to the trajectory list, and the storage compartment status is updated. N horizontal two-dimensional weighing matrices are mapped to the grid coordinate system. When there is a continuous change in weight, a dynamic three-dimensional trajectory of the goods is generated in the grid coordinate system. The trajectory list is trained and stored as a reference trajectory using the dynamic 3D trajectories of trigger events for all products in the vending machine. The specific steps for establishing this trajectory list include... The entrance and exit of the storage compartment are used as the initial endpoints of the storage compartment coordinate system. If there is a most recent endpoint of the storage compartment's trajectory, it is directly used as the new starting point. Based on the agreed-upon level and structural orientation of the vending machine, the weight change data is mapped to the spatial displacement component of the vending machine. The endpoint of the product's trigger event is calculated by adding the spatial displacement component to the starting coordinates. The new track fragment is appended to the track list. The track list version number is incremented, and the last update time is updated to the current time.

4. The weighing cross-compartment dynamic trajectory merchandise recognition system for vending machines according to claim 1, characterized in that: The event recognition module calculates the shortest distance between the dynamic three-dimensional trajectory of the product's triggering event and the baseline trajectory using a three-dimensional line segment distance measurement method as a preliminary confidence score. The dynamic 3D trajectory is used to initially determine whether the triggering events are for the same product by setting global and local thresholds. Specifically, this includes... When the initial confidence score is greater than or equal to the global threshold, further judgment is initiated, which specifically includes... Compare the historical trajectory statistics of known similar products in the product database and calculate the similarity score; Consistency assessment between timestamps and time windows of known shipping patterns; Whether the hierarchical combination of the product's storage compartment information is within the same fault tolerance range.

5. The weighing cross-compartment dynamic trajectory merchandise recognition system for vending machines according to claim 1, characterized in that: The migration management module defines cross-cargo cell migration rules, specifically including... When the same product appears in a new storage cell, identify the most recent T trajectory list and select the dynamic 3D trajectory that best matches the current storage cell status; Map the dynamic 3D trajectory that best matches the current state of the cargo cell to the cargo cell coordinate system of the new cargo cell, and use it as the initial reference trajectory of the new cargo cell. The migration log between the old and new storage compartments is updated based on the initial baseline trajectory, and the migration time, product identifier, and similarity score are recorded. Within a certain time window after migration, the dynamic three-dimensional trajectory of the new storage cell is monitored. If an overlap or conflict deviating from the expected behavior is detected, a rollback mechanism is triggered to restore the state before migration. When the same product has more than two different dynamic 3D trajectories in the same storage compartment, the conflict is decomposed based on the most recent trigger time point, and the latest dynamic 3D trajectory is retained.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the module of the weighing cross-compartment dynamic trajectory merchandise recognition system for vending machines as described in any one of claims 1 to 5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the module of the weighing cross-compartment dynamic trajectory merchandise recognition system for vending machines as described in any one of claims 1 to 5.

Images