Warehouse logistics cargo identification management system

Abstract

The application discloses a warehouse logistics goods identification management system, and relates to the technical field of internet, cloud computing and big data services, comprising a multi-modal identification layer, a cloud big data computing engine layer and a front-end graphical user interface rendering and interaction layer. The cloud computing engine fuses the importance weight of goods business flow, the multi-modal comprehensive identification confidence and the state abnormality identification variable to calculate the comprehensive priority score, thereby driving the front-end hierarchical rendering scheduling. The front-end generates three progressive confirmation interactions, namely automatic confirmation, sliding confirmation and multi-step click review, according to the confidence difference. The system also estimates the cognitive load according to the operation personnel touch reaction time, frequency and accuracy. When the load is high, the system adaptively combines and enlarges the control and generates an edge adsorption touch area. The data of each link is returned to the cloud to form a closed loop, thereby solving the problems of interface rendering lag and low confidence data confirmation efficiency under the condition of massive goods concurrent changes, and being suitable for sorting screens, central control large screens and handheld terminals and other warehouse business scenarios.

Description

Technical Field

[0001] This invention relates to the fields of Internet and cloud computing and big data service technology, specifically to a warehouse logistics cargo identification and management system. Background Technology

[0002] With the deep application of big data and cloud computing technologies in the warehousing and logistics industry, full-warehouse cargo tracking has gradually evolved from the early single barcode scanning method to a multimodal heterogeneous identification method that integrates visual image recognition, radio frequency identification, and IoT sensors. In the existing warehouse management system, cargo identification data is usually collected by visual cameras, radio frequency identification readers and IoT sensors distributed at various operational nodes in the warehouse. After the data is aggregated, deduplicated and formatted by the back-end server, it is pushed to the front-end graphical user interface terminal for display in the form of incremental data streams, so that sorting, inventory, verification and other business personnel can view and operate in real time.

[0003] An existing implementation scheme that is similar to this invention is a warehouse visualization system based on incremental rendering and threshold switching. In this scheme, a multimodal recognition device collects cargo identity and status data at each operation node in the warehouse.

[0004] The backend server aggregates, deduplicates, and formats the collected data to form a structured data change record;

[0005] The front-end graphical user interface terminal compares the old and new data and performs partial redrawing only on the elements in the interface that have changed, thereby reducing rendering overhead.

[0006] When the amount of data change exceeds the preset threshold within a unit of time, the front-end graphical user interface abandons the partial redraw strategy and instead performs a full re-render of the entire interface component tree.

[0007] For cases with low confidence or no label in the identification results, the solution will uniformly prompt the abnormality in the form of a pop-up window, or force the corresponding goods to be transferred to the manual review station for secondary confirmation.

[0008] The above solution can reduce rendering overhead to some extent in scenarios with small data changes, but in scenarios with tens of thousands of inventory units in the entire warehouse undergoing concurrent state changes, the business terminal still faces obvious performance bottlenecks: on the one hand, the front-end rendering strategy only uses the total amount of data changes per unit time as the sole criterion for whether to switch to full rendering, without taking into account the value level of the goods in the business chain, the current business node, whether there is an abnormal state, and the confidence level of the recognition result for differentiated processing.

[0009] On the other hand, for low-confidence or unlabeled identification results, existing solutions generally adopt a uniform pop-up prompt or forced manual review method, without designing differentiated confirmation interactions that match the confidence level, and without perceiving and responding to the cognitive load of operators under high-load operation scenarios. The above shortcomings will be further amplified in warehouse scenarios with high data scale and high concurrency intensity, leading to main thread blocking, interface lag, or even system crashes on business terminals. At the same time, it causes problems such as low operator confirmation efficiency, a surge in cognitive load, and an increase in the error rate. There is an urgent need for a warehouse logistics cargo identification and management system that can be collaboratively optimized at the cloud semantic classification and front-end interaction levels. Summary of the Invention

[0010] The technical problem to be solved by the present invention lies in the analysis of the defects of the prior art.

[0011] The existing solution uses the total amount of data change per unit time as the sole criterion for switching rendering strategies, without incorporating semantic dimensions such as the business value of goods and identification confidence for differentiated processing. Therefore, when tens of thousands of inventory units in the entire warehouse undergo concurrent status changes, even if the vast majority of them are low-value normal circulation data, as long as the total amount of data change exceeds the preset threshold, the system will still uniformly trigger full rendering, causing the main thread of the graphical user interface to be blocked. Consequently, high-value abnormal goods or low-confidence identification results that truly require focus cannot be presented in a timely manner, delaying business decisions. This is the first technical problem.

[0012] Existing solutions use a uniform pop-up prompt or forced manual review method for low-confidence identification results, without designing differentiated confirmation interactions based on confidence levels. This forces even high-confidence data to undergo unnecessary interaction interruptions for confirmation, while data in the critical confidence range is difficult to obtain rigorous review commensurate with its risk level. Operators frequently switch their focus between the touch screen and the physical shelf, making it difficult to balance confirmation efficiency and accuracy. This is the second technical problem.

[0013] The existing solution does not perceive or model the real-time cognitive load of operators. The size and layout of the interface controls remain consistent between peak and off-peak sorting periods. During high-load periods, operators' visual attention is distracted, and the small and dense layout of controls further exacerbates the accidental touch rate. This creates a cycle of increased accidental touches under high load conditions, which in turn triggers secondary verification, which further increases the cognitive load. This is the third technical problem.

[0014] To address the aforementioned three technical problems, the present invention aims to provide a warehouse logistics cargo identification and management system. This system addresses the rendering lag issues and low confirmation efficiency of low-confidence data by employing a cloud-based big data semantic hierarchical front-end graphical user interface hierarchical rendering and scheduling mechanism, a progressive touch-based human-machine collaborative confirmation mechanism for low-confidence recognition results, and a touch interaction adaptive reconstruction mechanism based on operator cognitive load. These mechanisms work collaboratively to solve the rendering lag issues under massive heterogeneous recognition data and the low confirmation efficiency of low-confidence data from three levels: data semantic hierarchical classification, interaction process design, and interface adaptive adjustment.

[0015] To achieve the above objectives, the warehousing and logistics cargo identification and management system provided by the present invention includes a multimodal identification layer, a cloud big data computing engine layer, and a front-end graphical user interface rendering and interaction layer.

[0016] The multimodal recognition layer is used to collect the identity, location and status data of goods at each operational node of warehousing and logistics, specifically including collection devices such as visual recognition cameras, radio frequency identification readers and IoT sensors.

[0017] The cloud-based big data computing engine layer is communicatively connected to the multimodal recognition layer. It is used to receive the identity, location, and status data, calculate the business flow importance weight for each item, fuse the confidence of multiple recognition methods to obtain the multimodal comprehensive recognition confidence, and perform weighted fusion calculation based on the business flow importance weight, the multimodal comprehensive recognition confidence, and the status anomaly identifier variable to obtain a comprehensive priority score. Then, based on the comprehensive priority score, a hierarchical rendering scheduling instruction is generated and sent to the front-end terminal.

[0018] The front-end graphical user interface rendering and interaction layer is deployed on touch business terminals such as sorting screens, central control screens, and handheld terminals. It is used to receive the hierarchical rendering scheduling instructions, divide the goods into three levels of high priority, medium priority and low priority according to the comprehensive priority score, and render and present them using different rendering methods.

[0019] Simultaneously, based on the multimodal comprehensive recognition confidence level, the cargo recognition results are divided into three levels: high confidence, medium confidence, and low confidence. Three differentiated interaction methods are generated to complete cargo confirmation: automatic confirmation, sliding confirmation, and multi-step click verification confirmation.

[0020] In addition, it continuously monitors the touch events of operators and calculates the estimated cognitive load. When the cognitive load is high, it merges and enlarges the interface controls and generates edge-adhesive touch areas to reduce the false touch rate.

[0021] The front-end graphical user interface rendering and interaction layer also sends the rendering time and user interaction response data, the goods confirmation result and confirmation operation time, and touch behavior data back to the cloud big data computing engine layer. This is used to continuously correct the calculation parameters of the importance weight of business flow and the confidence of multimodal comprehensive recognition, as well as the judgment threshold corresponding to the cognitive load estimate, forming a closed loop of data collection, cloud computing, terminal presentation, and behavior feedback.

[0022] Compared with the prior art, the beneficial effects of the present invention by adopting the above technical solution are as follows:

[0023] 1. This invention uses a front-end hierarchical rendering scheduling mechanism driven by cloud semantic hierarchy to significantly reduce the rendering overhead of massive low-priority data while ensuring real-time visualization of high-value, abnormal, and low-confidence goods information. It avoids full rendering triggered by the total amount of data changes exceeding the threshold, and achieves smooth response of the business terminal interface in scenarios with tens of thousands of inventory units.

[0024] 2. This invention achieves dynamic matching between the complexity of the confirmation interaction and the confidence level of recognition through a progressive touch confirmation mechanism for low-confidence data, reduces unnecessary manual intervention for high-confidence data, ensures the rigor of confirmation results in low-confidence scenarios, and improves the overall throughput efficiency of recognition and confirmation.

[0025] 3. This invention uses a cognitive load-based adaptive reconstruction mechanism for touch interaction to automatically merge and enlarge high-frequency operation controls and generate edge-adhesive touch areas when the operator is under high load. This reduces the rate of accidental touches caused by the operator's visual attention being distracted and mitigates the adverse effects of accidental touches and secondary verification under high load conditions. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the closed-loop process of the overall system architecture in an embodiment of the present invention;

[0027] Figure 2 This is a flowchart of the cloud-based hierarchical + front-end hierarchical rendering scheduling process in an embodiment of the present invention;

[0028] Figure 3 This is a flowchart illustrating the interaction between confidence level confirmation and cognitive load adaptation in an embodiment of the present invention. Detailed Implementation

[0029] The following is in conjunction with the appendix Figure 1-3 The specific embodiments of the present invention will be further described below. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0030] The technical solution of the present invention will be further described in detail below with reference to specific embodiments, so that those skilled in the art can implement the present invention accordingly. It should be noted that the following embodiments are only used to explain the present invention and are not intended to limit the scope of protection of the present invention.

[0031] The warehouse logistics cargo identification and management system provided by this invention consists of three main parts: a multimodal recognition layer, a cloud-based big data computing engine layer, and a front-end graphical user interface rendering and interaction layer. These parts form a closed-loop collaborative relationship through low-latency network channels, encompassing data acquisition, cloud computing, terminal presentation, and behavioral feedback. The multimodal recognition layer, deployed at various warehouse operation nodes, includes visual recognition cameras, RFID readers, and IoT sensors, responsible for collecting cargo identity, location, and status data. The cloud-based big data computing engine layer is deployed on a cloud server cluster and internally includes a data access module based on a stream processing framework and a business importance weight calculation module. The system comprises a module, a multimodal confidence fusion calculation module, and a comprehensive priority scoring calculation module. These modules are responsible for semantically classifying the collected data and generating rendering and interaction scheduling instructions. The front-end graphical user interface rendering and interaction layer is deployed on touch terminals such as sorting screens, central control screens, and handheld terminals. Internally, it includes a hierarchical rendering engine, a progressive confirmation gesture interface generation module, and a cognitive load monitoring and interface reconstruction module. In addition, the system is equipped with a feedback loop to send rendering time, confirmation operation results, and touch behavior data back to the cloud computing engine layer for continuous correction of weight coefficients, confidence thresholds, and cognitive load model parameters, forming an end-to-end closed loop.

[0032] I. Cloud-based Big Data Semantic Classification and Front-end Adaptive Rendering Scheduling Mechanism

[0033] The cloud-based big data computing engine receives real-time cargo identification data streams from the multimodal recognition layer based on a stream processing framework. Each data record includes a unique cargo identifier, initial identification confidence level, information on the current business node of the cargo, and a data collection timestamp. For each cargo record, the cloud computing engine calculates the business flow importance weight of the cargo based on a pre-defined business rule base, denoted as... The preset business rule base covers four dimensions: cargo value level, timeliness requirements, current business node of the cargo, and whether there are any abnormal status alarms. Simultaneously, the cloud computing engine integrates visual recognition confidence, radio frequency identification confidence, and IoT sensor cross-validation results, calculating the multimodal comprehensive recognition confidence of the cargo through a weighted fusion method, denoted as... .

[0034] After obtaining the importance weights of business flows and the confidence level of multimodal comprehensive identification, the cloud computing engine combines the identifier variables of whether the goods are in an abnormal state. The overall priority score of the cargo in the graphical user interface rendering schedule is calculated according to the following formula. :

[0035] ;

[0036] In the formula, Indicates the first Rendering priority score for each item of goods;

[0037] Indicates the first The importance weight of each item in the business flow, with a value range of [value range missing]. The higher the value, the greater the importance of the goods in the business chain;

[0038] Indicates the first The confidence level for multimodal comprehensive identification of a shipment ranges from [value range missing]. , This indicates the degree of lack of confidence in the identification result of the goods; the larger the value, the less reliable the identification result.

[0039] This is a status anomaly indicator variable; it takes a value of 1 when the goods are in an abnormal state, and a value of 0 otherwise.

[0040] , , These are the first weighted coefficient, second weighted coefficient, and third weighted coefficient corresponding to the business importance item, the confidence level missing item, and the state anomaly item, respectively, and the three satisfy the following: The specific values ​​can be set based on experience in warehousing business scenarios, or obtained through regression training on historical rendering feedback data. It can be understood that the comprehensive priority score integrates three factors: the business value of the goods, the reliability of identification, and whether there are any anomalies. This avoids the problem of missing semantic information caused by using only the total amount of data change as the sole criterion.

[0041] The cloud computing engine scores based on overall priority. Combining the preset first priority threshold and second priority threshold (the first priority threshold is greater than the second priority threshold), goods are divided into three levels: high priority, medium priority, and low priority. Goods with a comprehensive priority score not lower than the first priority threshold are classified as high priority.

[0042] Goods with a comprehensive priority score that is lower than the first priority threshold but not lower than the second priority threshold are classified as medium priority.

[0043] Goods with a comprehensive priority score below the second priority threshold are classified as low priority. The cloud computing engine generates corresponding rendering scheduling instructions and sends them to the front-end graphical user interface terminal through a low-latency message channel.

[0044] After receiving the rendering scheduling instruction, the front-end graphical user interface terminal renders the full attributes of high-priority goods at a higher refresh rate and places them in a visually prominent position on the interface.

[0045] Medium-priority goods are rendered using standard frame rate;

[0046] For low-priority goods, rendering degradation is triggered. Low-priority goods in the same batch are dynamically aggregated into a heatmap and a lazy loading method is adopted. Rendering is only performed on demand when the user actively performs an expansion operation on the corresponding interface area. Through the above hierarchical rendering method, the front-end graphical user interface terminal can significantly reduce the rendering overhead of massive low-priority data while ensuring real-time visualization of information on high-value, abnormal, and low-confidence goods. This avoids the main thread being blocked by full rendering due to the total amount of data changes exceeding the threshold.

[0047] After the front-end graphical user interface is rendered, the actual rendering time and user interaction response data are sent back to the cloud big data computing engine. This is used to continuously optimize the parameter configuration of the business flow importance weight and multimodal comprehensive identification confidence fusion model, forming a closed-loop adaptive optimization of the rendering scheduling strategy.

[0048] II. Progressive Touch Confirmation Interaction Mechanism for Low-Confidence Data

[0049] The front-end graphical user interface terminal receives the confidence level of the multimodal comprehensive identification of goods from the cloud. The information includes the physical location coordinates of the goods in the warehouse space. Based on the preset first confidence threshold and second confidence threshold (the first confidence threshold is greater than the second confidence threshold), the goods identification results are divided into three levels: high confidence, medium confidence, and low confidence. The multimodal comprehensive identification confidence is not lower than the first confidence threshold and is classified as high confidence.

[0050] Multimodal integrated recognition with a confidence level lower than the first confidence threshold but not lower than the second confidence threshold is classified as medium confidence.

[0051] Multimodal integrated recognition with a confidence level below the second confidence threshold is classified as low confidence.

[0052] For high-confidence recognition results, the front-end graphical user interface automatically completes the confirmation process and presents the confirmed status of the goods with a lightweight status icon. No confirmation control that requires active interaction from the operator is generated, thereby avoiding unnecessary human-computer interaction interruptions. For medium-confidence recognition results, the front-end graphical user interface dynamically generates a sliding confirmation control at the interface coordinates that map to the physical location of the goods. The operator can complete the confirmation by performing a single sliding gesture.

[0053] If the operator does not perform the sliding operation within the preset time, the system will re-enter the record into the pending queue and increase its comprehensive priority score in subsequent rendering scheduling, so that it will receive attention earlier. For low confidence recognition results, the front-end graphical user interface will forcibly generate a multimodal verification and confirmation interface, which will present the original visual recognition image, the original radio frequency identification data, and the auxiliary information of the Internet of Things sensor in parallel. The operator must complete the three sub-steps of image verification, data comparison, and confirmation and approval in sequence to confirm the recognition result of the goods, so as to ensure the rigor of the confirmation result in low confidence scenarios.

[0054] After the operator completes or abandons the confirmation, the front-end graphical user interface sends the confirmation result, operation time, and operation path back to the cloud big data computing engine. This is used to dynamically correct the first confidence threshold and the second confidence threshold, and to feed back into the confidence output calibration of the multimodal recognition model, forming a two-way closed-loop optimization between the confirmation interaction strategy and the accuracy of the recognition model.

[0055] III. Adaptive Reconstruction Mechanism for Touch Interaction Based on Cognitive Load

[0056] The front-end graphical user interface terminal continuously monitors touch events generated by the operator on the interface and records the reaction time of each touch event relative to the corresponding data presentation time. Touch operation frequency within a unit time window And the accuracy of the gesture trajectory offset relative to the center of the target control. ,in This indicates the current statistical time window. The front-end graphical user interface terminal calculates the operator's time window based on the above parameters using the following formula. Cognitive load estimates within :

[0057] ;

[0058] In the formula, Indicates time window The estimated cognitive load of internal operators; the higher the value, the higher the cognitive load.

[0059] This indicates the average reaction time for an operator from data presentation to completing a click or swipe response within that time window. This indicates that the shorter the reaction time, the greater the contribution of the load estimation.

[0060] This indicates the frequency of touch operations per unit time within this time window;

[0061] This indicates the average offset accuracy of the gesture trajectory relative to the center of the target control within the given time window, with a value range of [value range missing]. The smaller the value, the lower the accuracy of the gesture. Indicates the degree of inadequacy in accuracy;

[0062] , , These are the weighting coefficients for the reciprocal of response time, touch frequency, and inaccuracy, respectively, which can be obtained by training based on historical calibration data.

[0063] The front-end graphical user interface terminal will calculate the estimated cognitive load. Compared with the preset cognitive load threshold Comparison:

[0064] like Not less than If so, it is determined that the operator is currently under high load, triggering the adaptive reconstruction process of the interactive interface;

[0065] like Below If the current interface layout remains unchanged, monitoring continues. After triggering the interactive interface adaptive reconstruction process, the front-end graphical user interface merges and groups the parameter setting controls and manipulation object controls that were originally scattered in the interface according to the touch event listening interface, and uniformly enlarges their visible and touchable areas according to the preset magnification ratio.

[0066] At the same time, edge-adhesive touch areas are generated around each merged and enlarged control. That is, when the operator's touch point falls within a certain tolerance range outside the geometric boundary of the control, the system still judges it as a valid touch of the control, thereby reducing the accidental touch rate caused by visual distraction.

[0067] After the interface reconstruction is completed, the front-end graphical user interface continues to monitor the operator's subsequent touch behavior in the above manner and dynamically updates the cognitive load estimate. ,when Falling back to the threshold After a preset time period, the interface will be gradually and smoothly restored to its normal layout, forming a dynamic closed loop of cognitive load monitoring and interaction reconstruction.

[0068] IV. Examples

[0069] Taking a batch of inventory units experiencing concurrent status changes during peak sorting hours at a certain warehouse as an example, for one item in an abnormal state whose current business flow node is an urgent shipment node, the cloud computing engine calculates its business flow importance weight based on a preset business rule base. The value is relatively high;

[0070] The confidence level of the cargo was obtained through multimodal comprehensive identification after cross-verification using visual recognition and radio frequency identification. The value is relatively high;

[0071] Because the goods are in an abnormal state, their abnormal status identifier variable... The value is set to 1. Substituting the above values ​​into the aforementioned comprehensive priority scoring formula for weighted fusion calculation, the comprehensive priority score of the goods can be obtained. Goods with high values, reaching the first priority threshold, are classified as high priority and rendered in a prominent position on the interface at a high refresh rate. Simultaneously, due to their high confidence level, the front-end graphical user interface automatically confirms the identification result of the goods without requiring manual interaction. Meanwhile, for normally circulating goods in the same batch that are numerous and have low business flow importance weights and abnormal status indicators, their comprehensive priority score is in a lower value range, below the second priority threshold. These are classified as low priority and aggregated into a heatmap for delayed loading, thus avoiding main thread blocking due to the large number of goods in the batch.

[0072] During peak sorting periods, operators experience shorter reaction times, increased touch operation frequency, and decreased gesture accuracy due to continuous high-intensity work. The cognitive load estimate calculated by the front-end graphical user interface terminal based on the above three parameters is... Not lower than the preset cognitive load threshold The system then triggers an adaptive interface reconstruction process, merging and enlarging the originally scattered confirmation controls and generating edge-attached touch areas, thus reducing the rate of accidental touches by operators. When the peak period ends and operators' touch behavior returns to normal, the estimated cognitive load gradually falls below the threshold and continues for a preset duration. The front-end graphical user interface automatically and smoothly restores the layout to the normal state.

[0073] In one alternative implementation, semantic hierarchical computation does not necessarily have to be performed by a centralized big data stream processing engine in the cloud. A lightweight hierarchical model can also be deployed locally at the edge gateway or the front end. For example, based on a rule engine or a lightweight decision tree model, the edge node performs the fusion calculation of the importance weight of business flow and the confidence of multimodal comprehensive identification and generates rendering scheduling instructions locally. This reduces the dependence on cloud network latency and is suitable for warehouse scenarios with unstable network conditions or high real-time requirements. It can serve as an effective supplement to the cloud computing architecture of this invention.

[0074] In another alternative implementation, in addition to implicitly estimating the cognitive load estimate based on the reaction time, touch frequency, and gesture offset accuracy of the touch event, the cognitive load estimate can also be selectively connected to a camera visual analysis module to collect visual features such as the operator's facial orientation and blink frequency to help correct the cognitive load estimate. This further improves the accuracy of the cognitive load estimate in warehouse scenarios with camera deployment capabilities, serving as an effective supplement to pure touch behavior estimation methods.

[0075] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of the present invention. Therefore, any modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope defined by the claims of the present invention.

Claims

1. A warehouse logistics cargo identification and management system, characterized in that, include: A multimodal recognition layer is used to collect data on the identity, location, and status of goods at various operational nodes in warehousing and logistics. The cloud-based big data computing engine layer communicates with the multimodal recognition layer to receive the identity, location, and status data. It calculates the importance weight of the business flow for each item, integrates the confidence of multiple recognition methods to obtain the multimodal comprehensive recognition confidence, performs weighted fusion calculation based on the importance weight of the business flow, the multimodal comprehensive recognition confidence, and the status anomaly identifier variable to obtain a comprehensive priority score, and generates and issues hierarchical rendering scheduling instructions based on the comprehensive priority score. The front-end graphical user interface rendering and interaction layer, deployed on the touch terminal, communicates with the cloud-based big data computing engine layer. It receives the hierarchical rendering scheduling instructions, classifies goods into three levels—high priority, medium priority, and low priority—based on the comprehensive priority score, and renders each level using different rendering methods. The front-end graphical user interface rendering and interaction layer also classifies the goods recognition results into three levels—high confidence, medium confidence, and low confidence—based on the multimodal comprehensive recognition confidence level, and generates three differentiated interaction methods—automatic confirmation, swipe confirmation, and multi-step click verification confirmation—to complete goods confirmation. Furthermore, the front-end graphical user interface rendering and interaction layer continuously monitors the operator's touch events and calculates the cognitive load estimate. When the cognitive load estimate is not lower than a preset cognitive load threshold, it merges and enlarges the interface controls, generating edge-attached touch areas around the merged and enlarged controls. The front-end graphical user interface rendering and interaction layer transmits rendering time and user interaction response data, goods confirmation results and confirmation operation time, and touch behavior data back to the cloud big data computing engine layer. The cloud big data computing engine layer corrects the calculation parameters of the business flow importance weight and the multimodal comprehensive recognition confidence based on the goods confirmation results and confirmation operation time, and corrects the judgment threshold corresponding to the cognitive load estimate based on the touch behavior data. This enables the parameters of rendering scheduling, confirmation interaction, and interface reconstruction to achieve linkage correction based on the same closed-loop feedback data, forming a closed loop of data collection, cloud computing, terminal presentation, and behavior feedback.

2. The warehouse logistics cargo identification and management system according to claim 1, characterized in that, The importance weight of the business flow corresponds to the first weighting coefficient, the confidence level missing corresponding to the multimodal comprehensive identification confidence level corresponds to the second weighting coefficient, and the state anomaly identifier variable corresponds to the third weighting coefficient. The sum of the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient is 1. The comprehensive priority score is obtained by summing the product of the importance weight of the business flow and the first weighting coefficient, the product of the confidence level missing and the second weighting coefficient, and the product of the state anomaly identifier variable and the third weighting coefficient.

3. The warehouse logistics cargo identification and management system according to claim 1, characterized in that, The cloud-based big data computing engine layer is set with a first priority threshold and a second priority threshold, wherein the first priority threshold is greater than the second priority threshold. Goods with a comprehensive priority score not lower than the first priority threshold are classified as high priority; goods with a comprehensive priority score lower than the first priority threshold but not lower than the second priority threshold are classified as medium priority; and goods with a comprehensive priority score lower than the second priority threshold are classified as low priority. The front-end graphical user interface rendering and interaction layer renders the full attributes of high-priority goods at a refresh rate no lower than a preset first frame rate threshold and places them in a visually prominent position on the interface. It performs regular rendering of medium-priority goods at a refresh rate lower than the preset first frame rate threshold. Low-priority goods are aggregated into a heat map and rendered using a delayed loading method, only when the user performs an expand operation on the corresponding interface area.

4. The warehouse logistics cargo identification and management system according to claim 1, characterized in that, The cloud-based big data computing engine layer is configured with a first confidence threshold and a second confidence threshold, wherein the first confidence threshold is greater than the second confidence threshold. The cargo identification results with a multimodal comprehensive recognition confidence level not lower than the first confidence level threshold are classified as high confidence; the cargo identification results with a multimodal comprehensive recognition confidence level lower than the first confidence level threshold but not lower than the second confidence level threshold are classified as medium confidence; and the cargo identification results with a multimodal comprehensive recognition confidence level lower than the second confidence level threshold are classified as low confidence.

5. The warehouse logistics cargo identification and management system according to claim 4, characterized in that, The front-end graphical user interface rendering and interaction layer automatically completes the confirmation process for high-confidence cargo identification results, and presents the confirmed status with status icons, without generating confirmation controls that require active interaction from the operator. For cargo identification results with medium confidence, a sliding confirmation control is generated at the interface coordinates that map to the physical location of the cargo. The operator performs a sliding gesture to complete the confirmation. If the sliding operation is not performed within a preset time, the cargo identification result is re-entered into the processing queue and its subsequent comprehensive priority score is increased. For cargo identification results with low confidence, a multimodal verification and confirmation interface is generated, which presents the original visual recognition image, the original radio frequency identification data, and the auxiliary information from the Internet of Things sensors in parallel. After the operator completes the three sub-steps of image verification, data comparison, and confirmation and approval, the cargo identification result is confirmed.

6. The warehouse logistics cargo identification and management system according to claim 1, characterized in that, The front-end graphical user interface rendering and interaction layer records the reaction time of each touch event relative to the corresponding data presentation time, the touch operation frequency within a unit time window, and the accuracy of the gesture trajectory offset relative to the center of the target control. The cognitive load estimate is obtained by weighting and summing the reciprocal of the reaction time, the touch operation frequency, and the degree of inadequacy of the offset accuracy.

7. The warehouse logistics cargo identification and management system according to claim 6, characterized in that, After the interface reconstruction is triggered, the front-end graphical user interface rendering and interaction layer continuously monitors the operator's subsequent touch behavior and dynamically updates the cognitive load estimate. After the cognitive load estimate falls below the preset cognitive load threshold and remains below it for a preset duration, the interface is gradually restored to the normal layout state.

8. The warehouse logistics cargo identification and management system according to claim 1, characterized in that, The cloud-based big data computing engine layer calculates the importance weight of the business flow based on a preset business rule base. The preset business rule base includes the value level of the goods, timeliness requirements, the current business node of the goods, and whether there are any abnormal status alarms.

9. A method for identifying and managing goods in warehousing and logistics, characterized in that, include: The multimodal recognition layer collects data on the identity, location, and status of goods at various operational nodes in warehousing and logistics. The cloud-based big data computing engine layer receives the identity, location, and status data, calculates the importance weight of the business flow for each item, integrates the confidence of multiple identification methods to obtain a multimodal comprehensive identification confidence score, performs weighted fusion calculation based on the importance weight of the business flow, the multimodal comprehensive identification confidence score, and the status anomaly identifier variable to obtain a comprehensive priority score, and divides the items into three levels: high priority, medium priority, and low priority based on the comprehensive priority score, generates a hierarchical rendering scheduling instruction, and sends it to the front-end graphical user interface terminal; After receiving the hierarchical rendering scheduling instruction, the front-end graphical user interface terminal renders the high-priority goods with full attributes at a high refresh rate, renders the medium-priority goods with normal attributes at a standard frame rate, and aggregates the low-priority goods into a heatmap and loads them with a delay. The front-end graphical user interface terminal divides the cargo recognition results into three levels: high confidence, medium confidence, and low confidence, based on the multimodal comprehensive recognition confidence level, and generates three differentiated interaction methods for cargo confirmation: automatic confirmation, swipe confirmation, and multi-step click verification confirmation. The front-end graphical user interface terminal continuously monitors the operator's touch events, calculates the cognitive load estimate based on the reaction time, touch operation frequency and gesture offset accuracy, and when the cognitive load estimate is not lower than the preset cognitive load threshold, merges and enlarges the interface controls and generates an edge-adhesive touch area. The rendering time and user interaction response data, goods confirmation results and confirmation operation time, and touch behavior data are respectively sent back to the cloud big data computing engine layer for the purpose of correcting calculation parameters and judgment thresholds.

10. The warehousing and logistics cargo identification and management method according to claim 9, characterized in that, The importance weight of the business flow corresponds to the first weighting coefficient, the confidence level of the multimodal comprehensive identification confidence corresponds to the second weighting coefficient, and the state anomaly identifier variable corresponds to the third weighting coefficient. The sum of the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient is 1.

Images