Warehouse inventory intelligent management method, device and terminal based on AI visual monitoring

Abstract

The embodiments of the present application relate to the fields of data processing and AI vision technology, and provide a warehouse inventory intelligent management method, device and terminal based on AI vision monitoring, the method comprising: obtaining inventory management basic data and external dynamic impact data of the warehouse to be monitored; based on AI vision technology, dynamically optimizing the current inventory data in the inventory management basic data according to a sample image data set and a label image data set to obtain dynamically optimized inventory data; performing inventory structure evaluation based on the inventory management basic data and the dynamically optimized inventory data to obtain inventory structure evaluation data; performing real-time dynamic adjustment processing on the inventory structure evaluation data according to the external dynamic impact data to obtain inventory dynamic adjustment data; generating an inventory health evaluation report based on the dynamically optimized inventory data, the inventory structure evaluation data and the inventory dynamic adjustment data, which is conducive to achieving the purpose of more accurate and comprehensive warehouse inventory intelligent management.

Description

Technical Field

[0001] The present application relates to the fields of data processing technology and AI vision technology, and specifically to a warehouse inventory intelligent management method, device and terminal based on AI vision monitoring. Background Art

[0002] Traditional warehouse management relies heavily on manual inventory counting, especially in large warehouses. This process often requires significant manpower and takes days or even weeks, disrupting warehouse operations. It can also lead to omissions or misrecording of inventory quantities due to human negligence, such as fatigue or carelessness. Furthermore, the misplaced storage of goods can lead to duplicate or missed counts, leading to discrepancies between inventory and actuals, impacting the accuracy of corporate cost accounting and profits. Therefore, achieving more accurate and comprehensive intelligent warehouse inventory management has become a pressing issue. Summary of the Invention

[0003] The embodiments of the present application provide a warehouse inventory intelligent management method, device and terminal based on AI visual monitoring, which can generate a more accurate and comprehensive inventory health assessment report based on inventory management basic data and external dynamic impact data, which is conducive to achieving the purpose of more accurate and comprehensive warehouse inventory intelligent management.

[0004] A first aspect of an embodiment of the present application provides a method for intelligent warehouse inventory management based on AI visual monitoring, the method comprising:

[0005] Obtain basic inventory management data and external dynamic impact data of the warehouse to be monitored;

[0006] Based on AI vision technology, the current inventory data in the inventory management basic data is dynamically optimized according to the sample image data set and the label image data set to obtain dynamically optimized inventory data;

[0007] Based on the basic inventory management data and the dynamically optimized inventory data, the inventory structure is evaluated to obtain the inventory structure evaluation data;

[0008] Perform real-time dynamic adjustment processing on inventory structure assessment data based on external dynamic impact data to obtain inventory dynamic adjustment data;

[0009] Generate an inventory health assessment report based on dynamically optimized inventory data, inventory structure assessment data, and inventory dynamic adjustment data.

[0010] In this example, by obtaining the inventory management basic data and external dynamic impact data of the warehouse to be monitored, the current inventory data in the inventory management basic data can be dynamically optimized based on the sample image data set and the label image data set based on AI vision technology to obtain dynamically optimized inventory data, and the inventory structure evaluation can be performed based on the inventory management basic data and the dynamically optimized inventory data to obtain inventory structure evaluation data, so that the inventory structure evaluation data can be dynamically adjusted in real time based on the external dynamic impact data to obtain inventory dynamic adjustment data, and then an inventory health evaluation report can be generated based on the dynamically optimized inventory data, the inventory structure evaluation data and the inventory dynamic adjustment data, which is conducive to achieving the purpose of more accurate and comprehensive warehouse inventory intelligent management.

[0011] A second aspect of an embodiment of the present application provides a warehouse inventory intelligent management device based on AI visual monitoring, the device comprising:

[0012] An acquisition unit, used to obtain basic inventory management data and external dynamic impact data of the warehouse to be monitored;

[0013] a first processing unit, configured to dynamically optimize the current inventory data in the inventory management basic data based on the sample image data set and the label image data set based on AI vision technology to obtain dynamically optimized inventory data;

[0014] a second processing unit, configured to perform inventory structure evaluation based on the inventory management basic data and the dynamically optimized inventory data to obtain inventory structure evaluation data;

[0015] a third processing unit, configured to perform real-time dynamic adjustment processing on the inventory structure evaluation data according to the external dynamic impact data to obtain inventory dynamic adjustment data;

[0016] The fourth processing unit is configured to generate an inventory health assessment report based on the dynamically optimized inventory data, the inventory structure assessment data, and the inventory dynamic adjustment data.

[0017] A third aspect of an embodiment of the present application provides a terminal, comprising a processor, an input device, an output device, and a memory, wherein the processor, input device, output device, and memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions and execute the step instructions in the first aspect of the embodiment of the present application.

[0018] The fourth aspect of the embodiments of the present application provides a computer-readable storage medium, wherein the above-mentioned computer-readable storage medium stores a computer program for electronic data exchange, wherein the above-mentioned computer program enables a computer to execute some or all of the steps described in the first aspect of the embodiments of the present application.

[0019] A fifth aspect of the embodiments of the present application provides a computer program product, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program, wherein the computer program is operable to cause a computer to perform some or all of the steps described in the first aspect of the embodiments of the present application. The computer program product may be a software installation package. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0021] Figure 1 A schematic diagram of the structure of a warehouse inventory intelligent management system is provided for the embodiment of the present application;

[0022] Figure 2 A flowchart of a warehouse inventory intelligent management method based on AI visual monitoring is provided for an embodiment of the present application;

[0023] Figure 3 A schematic diagram of the structure of a terminal provided in an embodiment of the present application;

[0024] Figure 4 A structural schematic diagram of a warehouse inventory intelligent management device based on AI visual monitoring is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0025] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0026] The terms "first," "second," and the like in the specification and claims of this application and the accompanying drawings are used to distinguish between different objects, not to describe a particular order. Furthermore, the terms "including," "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements but may optionally include steps or elements not listed, or may optionally include other steps or elements inherent to the process, method, product, or apparatus.

[0027] References to "embodiments" in this application mean that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described in this application may be combined with other embodiments.

[0028] In order to better understand the warehouse inventory intelligent management method based on AI visual monitoring provided in the embodiment of the present application, the following first briefly introduces the scenario of applying the warehouse inventory intelligent management method based on AI visual monitoring.

[0029] When managing warehouse inventory, the most common method is to count and record items. Traditional, manually operated inventory counting is extremely tedious. Large warehouses often require significant manpower and take days or even weeks to complete each count. This not only disrupts warehouse operations but is also prone to human negligence, such as fatigue and carelessness, leading to omissions or misrecorded quantities of goods. Furthermore, due to irregular storage locations, inventory counts can be repeated or missed, resulting in discrepancies between accounts and actual inventory, impacting the accuracy of corporate cost accounting and profits. Furthermore, procurement often lacks rational planning and foresight, often disconnected from actual production practices. This often results in purchased materials of incorrect models, excessive quantities, or materials becoming unusable due to process improvements, resulting in significant capital waste. Furthermore, material reserves are not always rational, with both overstocking and shortages. This not only occupies significant inventory capital but also leads to an increase in urgently needed materials due to a lack of timely inventory management, driving up logistics and procurement costs. Ultimately, traditional warehouse management systems are outdated and lack real-time data updates, making it difficult to respond to changes in market demand in a timely manner. This hinders companies from capturing market opportunities and seriously restricts the improvement of their operational efficiency and market competitiveness. Therefore, it is urgent to introduce innovative management methods and technical means to solve this problem.

[0030] In response to the above-mentioned technical problems, the present application provides a warehouse inventory intelligent management method based on AI visual monitoring. By obtaining the inventory management basic data and external dynamic influence data of the warehouse to be monitored, the current inventory data in the inventory management basic data can be dynamically optimized based on the sample image data set and the label image data set based on artificial intelligence (AI) visual technology to obtain dynamically optimized inventory data. The inventory structure evaluation can be performed based on the inventory management basic data and the dynamically optimized inventory data to obtain inventory structure evaluation data. The inventory structure evaluation data can be dynamically adjusted in real time based on the external dynamic influence data to obtain inventory dynamic adjustment data. The inventory health evaluation report can be generated based on the dynamically optimized inventory data, the inventory structure evaluation data and the inventory dynamic adjustment data, which is conducive to achieving the purpose of more accurate and comprehensive warehouse inventory intelligent management.

[0031] The warehouse inventory intelligent management method based on AI visual monitoring is applied to the warehouse inventory intelligent management system. Figure 1 Figure 2 shows a schematic diagram of the structure of a warehouse inventory intelligent management system. Figure 1 As shown in Figure 1, the intelligent warehouse inventory management system includes a data acquisition module, a dynamic inventory optimization processing module (including a target damage identification module), an inventory structure assessment module, a real-time dynamic adjustment module, and an inventory health assessment module. The data acquisition module collects basic inventory management data (such as historical sales, current inventory, supply chain, and cost data) from the monitored warehouse, as well as external dynamic impact data, providing comprehensive data support for subsequent modules. The dynamic inventory optimization processing module leverages AI vision technology to dynamically optimize current inventory data. For example, the target damage identification module can accurately identify anomalies such as damaged items and correct inventory data to reflect actual inventory conditions. The inventory structure assessment module evaluates inventory structure from multiple dimensions (such as category, value, time, and region) based on basic inventory management data, generating inventory structure assessment data to determine the rationality of the inventory structure. The real-time dynamic adjustment module combines external dynamic impact data to make real-time adjustments to the inventory structure assessment data, enabling the inventory structure to promptly adapt to external changes in the market, supply chain, and other factors. The inventory health assessment module can comprehensively and dynamically optimize inventory data, inventory structure assessment data and inventory dynamic adjustment data, comprehensively assess the health status of warehouse inventory, generate inventory health assessment reports and provide improvement suggestions, etc. This application does not impose any restrictions on this.

[0032] See also Figure 2 , Figure 2 The present invention provides a flowchart of a method for intelligent warehouse inventory management based on AI visual monitoring, which includes:

[0033] S10: Obtain the basic inventory management data and external dynamic impact data of the warehouse to be monitored.

[0034] The warehouse to be monitored can be understood as any warehouse that has a need for intelligent inventory management. The warehouse inventory intelligent management method based on AI visual monitoring provided in this application can be used to perform intelligent inventory management on the warehouse to be monitored, and this application does not impose any restrictions on this.

[0035] The basic data for inventory management can be understood as a set of basic information that warehouse inventory management relies on. The basic data for inventory management may include but is not limited to historical in-and-out data, current inventory data, supply chain inventory data, and inventory cost data, which is not limited in this application. Historical in-and-out data may be data including the in-and-out quantity, in-and-out time, and in-and-out reasons of various items in the warehouse over a period of time in the past, which may reflect the demand for historical items in the warehouse. Current inventory data may be data that records in real time the types, quantities, storage locations, and item status (such as whether they are intact) of existing items in the warehouse, which may show the current inventory status of the warehouse. Supply chain inventory data may be inventory data in the procurement process, as well as inventory data in logistics distribution, etc. Inventory cost data may be cost data related to warehouse inventory, such as inventory procurement cost, inventory storage cost, inventory transportation cost, inventory loss cost, etc., which may be used to calculate the economic cost of the inventory management process.

[0036] External dynamic impact data can be used to indicate real-time changing data from outside the warehouse that can affect inventory management, such as weather factors, such as humid weather may affect the placement of inventory, high temperature weather may affect the quality of inventory, etc., as well as changes in market demand, such as changes in consumer preferences may lead to inventory hoarding and thus affect the quality of inventory, etc. This application does not impose any restrictions on this.

[0037] S20: Based on AI vision technology, according to the sample image data set and the label image data set, the current inventory data in the inventory management basic data is dynamically optimized to obtain dynamically optimized inventory data.

[0038] Among them, AI vision technology can be understood as the use of computer vision algorithms, deep learning models and other technical means to enable computers to understand and analyze the content in images or videos, to simulate human visual perception and comprehension capabilities, and to achieve related technologies for tasks such as object recognition, detection, tracking and analysis.

[0039] The sample image data set may include one or more sample image data, which can be understood as image data of various types of monitoring items in the inventory used in the model training process. The sample image data in the sample image data set can be image data of inventory monitoring items at different angles, different lighting conditions, and different scenes, obtained by image acquisition devices in AI vision technology, such as cameras, etc., which can provide the model with rich visual image information, allowing the model to learn the appearance characteristics, placement, quantity, and other information of inventory monitoring items in various situations, so that the current inventory data can be better analyzed and processed in the subsequent process. This application does not impose any restrictions on this.

[0040] The label image data set may include one or more label image data, which can be understood as data after the relevant information of the inventory monitoring items in the sample image data is annotated. Among them, the label may include the category, quantity, status (such as intact or damaged, etc.), storage location, etc. of the monitoring items, which is not limited in this application. The label image data can provide a supervision signal for the model, allowing the model to understand the true situation of the inventory monitoring items contained in the sample image data, so that the model can learn how to more accurately predict or identify the real information based on the sample image data, and then realize the dynamic optimization of the current inventory data in subsequent steps, which is not limited in this application.

[0041] The current inventory data can be used to indicate the actual inventory status data recorded in the current warehouse inventory intelligent management system, and can reflect the actual quantity, location, status, etc. of various items in the current inventory. Dynamically optimized inventory data can be understood as optimized inventory data obtained after dynamic optimization processing of the current inventory data based on AI vision technology. The dynamically optimized inventory data can be the data after optimizing and verifying the current inventory data by comprehensively considering factors such as the abnormal location, damage, and expiration of the monitored items. It can reflect the optimal status of the inventory in real time and provide more scientific and reasonable decision-making support for intelligent inventory management, such as procurement plan formulation, inventory allocation, etc. This application does not impose any restrictions on this.

[0042] It should be understood that dynamically optimizing the current inventory data in the basic inventory management data based on the sample image data set and the labeled image data set based on AI vision technology to obtain dynamically optimized inventory data refers to the process of dynamically optimizing the current inventory data using AI vision technology to obtain more accurate dynamically optimized inventory data. Step S20, i.e., dynamically optimizing the current inventory data in the basic inventory management data based on the sample image data set and the labeled image data set based on AI vision technology to obtain dynamically optimized inventory data, may include the following steps:

[0043] S21: Based on the AI ​​vision technology, perform image acquisition and preprocessing on the items in the warehouse to be monitored to obtain a set of monitored item image data;

[0044] S22: Based on the AI ​​vision technology, perform position anomaly recognition processing on each monitored object image data in the monitored object image data set to obtain out-of-position optimized data;

[0045] S23: Based on the AI ​​vision technology, perform abnormal identification processing of damaged items according to the sample image data set, the label image data set, and the monitored item image data set to obtain optimized damaged item data;

[0046] S24: Based on the AI ​​vision technology, the current inventory data in the inventory management basic data is reviewed and verified according to the out-of-place optimization data and the damaged item optimization data to obtain verified inventory data;

[0047] S25: Based on the AI ​​vision technology, the verified inventory data is optimized for outbound delivery to obtain dynamically optimized inventory data.

[0048] The monitored item image data set may include one or more monitored item image data. This monitored item image data can be understood as image data corresponding to the monitored items in the monitored warehouse captured by an image acquisition device (such as a video camera or a still camera) used in AI vision technology. This monitored item image data may include the monitored item's appearance, shape, spatial location, placement, packaging text, etc., and can be understood as the raw data for subsequent analysis and processing.

[0049] The off-site optimization data can be understood as the optimization data obtained after the position anomaly identification processing is performed on the image data of each monitored item. Specifically, it can be determined by using AI visual technology, such as identifying the position information of the monitored item in the monitored item image data based on AI visual technology, and further comparing it with the preset normal position to determine whether the monitored item is in the correct position, such as whether the height is inconsistent, whether the height exceeds the limit, whether the area is abnormal, whether the stacking stability is unbalanced, and whether there is abnormal displacement and other position anomalies, so as to further generate corresponding off-site optimization data, which is not limited by this application.

[0050] It should be understood that performing position anomaly identification processing on each monitored item image data set to obtain out-of-position optimization data refers to the process of performing position anomaly identification processing on the monitored item image data and providing position optimization suggestions for the items with abnormal positions to obtain out-of-position optimization data. Among them, step S22, that is, performing position anomaly identification processing on each monitored item image data set to obtain out-of-position optimization data, may include the following steps:

[0051] S221: constructing a three-dimensional model of the object based on each monitored object image data in the monitored object image data set to obtain a three-dimensional model of the monitored object;

[0052] S222: Performing height anomaly monitoring processing based on the three-dimensional model of the monitored object to obtain height anomaly data;

[0053] S223: Performing regional anomaly monitoring processing based on the three-dimensional model of the monitored object to obtain regional anomaly data;

[0054] S224: performing stacking stability detection processing based on the three-dimensional model of the monitored object to obtain stability imbalance data;

[0055] S225: Performing real-time dynamic position tracking processing based on the three-dimensional model of the monitored object to obtain abnormal displacement data;

[0056] S226: Perform an abnormal fusion process based on the height abnormal data, the regional abnormal data, the stability imbalance data, and the abnormal shift data to obtain out-of-position optimized data.

[0057] The 3D model of a monitored object can be used to represent the 3D model of the object after constructing the 3D model based on the image data of each monitored object. This 3D model of the monitored object can be understood as a digital representation of the object's shape, size, position, and other characteristics in 3D space, which can more intuitively and accurately reflect the spatial information of the monitored object.

[0058] It should be understood that performing a three-dimensional object model construction process based on each monitored object image data in the monitored object image data set to obtain a three-dimensional object model refers to the process of performing a three-dimensional object model construction process based on the monitored object image data to obtain a three-dimensional object model. Specifically, step S221, i.e., performing a three-dimensional object model construction process based on each monitored object image data in the monitored object image data set to obtain a three-dimensional object model, may include the following steps:

[0059] S2211: Acquire a first monitored object image data subset taken by a first photographing device and a second monitored object image data subset taken by a second photographing device from the monitored object image data set;

[0060] S2212: Acquire first plane coordinate data of a first imaging point from the first monitored object image data subset, and acquire second plane coordinate data of the first imaging point from the second monitored object image data subset;

[0061] S2213: Determine first three-dimensional coordinates of the first imaging point based on the first plane coordinate data, the second plane coordinate data, the first optical center coordinates of the first photographing device, the first focal length of the first photographing device, the second optical center coordinates of the second photographing device, and the second focal length of the second photographing device;

[0062] S2214: Acquire third plane coordinate data of a second imaging point from the first monitored object image data subset, and acquire fourth plane coordinate data of the second imaging point from the second monitored object image data subset;

[0063] S2215: Determine second three-dimensional coordinates of the second imaging point based on the third plane coordinate data, the fourth plane coordinate data, the first optical center coordinates of the first photographing device, the first focal length of the first photographing device, the second optical center coordinates of the second photographing device, and the second focal length of the second photographing device;

[0064] S2216: Obtain a three-dimensional model of the monitored object according to the first three-dimensional coordinates and the second three-dimensional coordinates.

[0065] The first photographic device may be any photographic device that captures and pre-processes images of items in the warehouse to be monitored. The second photographic device may be any photographic device other than the first photographic device that captures and pre-processes images of items in the warehouse to be monitored. In other words, the first photographic device and the second photographic device are different photographic devices.

[0066] The first monitoring item image data subset may include one or more first monitoring item image data, which can be understood as monitoring item image data obtained after the first photographic device performs image acquisition and preprocessing on the items in the warehouse to be monitored. The second monitoring item image data subset may include one or more second monitoring item image data, which can be understood as monitoring item image data obtained after the second photographic device performs image acquisition and preprocessing on the items in the warehouse to be monitored.

[0067] The first imaging point can be any imaging point that has been photographed by both the first photographic device and the second photographic device. The first plane coordinate data can be understood as the plane coordinate data obtained after the first photographic device photographs the first imaging point. The second plane coordinate data can be understood as the plane coordinate data obtained after the second photographic device photographs the first imaging point. It should be noted that a coordinate system can be established with a preset point on the first photographic device (such as the upper left corner endpoint or the lower left corner endpoint) as the origin (such as called the first origin), so that the first plane coordinate data is determined based on the first origin. Optionally, a coordinate system can be established with a preset point on the second photographic device (such as the upper left corner endpoint or the lower left corner endpoint) as the origin (such as called the second origin), so that the second plane coordinate data is determined based on the second origin, and this application does not impose any restrictions on this.

[0068] The optical center is a specific point in a photographic lens, which can be understood as the central reference point where light rays converge into an image after passing through the lens. The first optical center coordinates can be understood as the coordinates corresponding to the optical center of the first photographic lens. The second optical center coordinates can be understood as the coordinates corresponding to the optical center of the second photographic lens. The focal length is the focal length of a photographic lens and reflects its optical properties. The first focal length is the focal length of the first photographic lens, and the second focal length is the focal length of the second photographic lens.

[0069] Based on the first plane coordinate data obtained by photographing the first imaging point using the first photographic device, the second plane coordinate data obtained by photographing the first imaging point using the second photographic device, the first optical center coordinates of the first photographic device, and the second optical center coordinates of the second photographic device, the first three-dimensional coordinates of the first imaging point can be further determined using the principle of triangulation. In other words, the first three-dimensional coordinates can be understood as the three-dimensional coordinates corresponding to the first imaging point in the space captured by both the first and second photographic devices.

[0070] Optionally, the process of determining the first three-dimensional coordinates of the first imaging point based on the first plane coordinate data, the second plane coordinate data, the first optical center coordinates of the first photographic device, the first focal length of the first photographic device, the second optical center coordinates of the second photographic device, and the second focal length of the second photographic device may refer to the following formula:

[0071]

[0072] in, It can represent the abscissa in the first plane coordinate data; It can represent a first focal length of the first photographic device; A horizontal coordinate that can represent a first three-dimensional coordinate of a first imaging point; The abscissa may represent the coordinates of the first optical center of the first photographic device; A depth coordinate that may represent a first three-dimensional coordinate of the first imaging point; A depth coordinate that can represent the coordinates of the first optical center of the first photographic device; It can represent the vertical coordinate in the first plane coordinate data; A ordinate that may represent a first three-dimensional coordinate of a first imaging point; A ordinate that may represent the coordinates of the first optical center of the first photographic device; It can represent the abscissa in the second plane coordinate data; It can represent a second focal length of the second photographic device; The horizontal coordinate may represent the coordinates of the second optical center of the second photographic device; A depth coordinate that can represent the coordinates of the second optical center of the second photographic device; It can represent the vertical coordinate in the second plane coordinate data; The ordinate of the second optical center coordinate of the second photographic device can be represented by , the above formula can be used to construct a system of equations to solve The specific value of is not limited in this application.

[0073] Furthermore, the second imaging point may be any imaging point other than the first imaging point that has been photographed by both the first and second imaging devices. The third plane coordinate data may be understood as the plane coordinate data obtained after the first imaging device photographs the second imaging point. The fourth plane coordinate data may be understood as the plane coordinate data obtained after the second imaging device photographs the second imaging point.

[0074] Optionally, the second three-dimensional coordinates of the second imaging point are determined based on the third plane coordinate data, the fourth plane coordinate data, the first optical center coordinates of the first photographic device, the first focal length of the first photographic device, the second optical center coordinates of the second photographic device and the second focal length of the second photographic device. For relevant content, please refer to the detailed description of determining the first three-dimensional coordinates mentioned above, and this application will not repeat it here.

[0075] Optionally, this application uses the example of constructing a three-dimensional model of a monitored object by determining the three-dimensional coordinates of a first imaging point and a second imaging point, which does not constitute a limitation on this application. Optionally, the three-dimensional coordinates of other imaging points (such as a third imaging point, a fourth imaging point, etc.) can be further determined, thereby further constructing a three-dimensional model of the monitored object using the three-dimensional coordinates of the first imaging point, the second imaging point, the third imaging point, and the fourth imaging point, etc., which is not a limitation on this application.

[0076] By acquiring the three-dimensional coordinates of multiple imaging points (e.g., the first imaging point and the second imaging point), these discrete three-dimensional coordinate points can be further used to perform interpolation, fitting, and other operations to gradually construct a three-dimensional model of the monitored object. It is understood that these three-dimensional coordinate points can constitute point cloud data of the monitored object, and based on this point cloud data, surface data of the monitored object can be further generated, thereby obtaining a three-dimensional model of the monitored object. This application does not impose any restrictions on this.

[0077] It can be understood that the above-mentioned monitoring item image data set may include subsets of monitoring item image data taken by various photographic devices at different times. For the monitoring item image data subset at each moment, the three-dimensional coordinate points of the monitoring items corresponding to each moment can be generated. The three-dimensional coordinate points of the monitoring items corresponding to each moment are merged according to the time sequence to obtain a three-dimensional model of the monitoring items that changes in time sequence. This application does not impose any restrictions on this.

[0078] Height anomaly data can be understood as height anomaly data obtained after height anomaly monitoring based on the 3D model of the monitored item. Specifically, the height of each monitored item can be monitored based on the 3D model. For example, a conventional height interval threshold can be set to determine whether there are significant height inconsistencies among items of the same type, or a height threshold can be set to determine whether the height of the items exceeds the limit. This can generate height anomaly data indicating significant height inconsistencies among items of the same type or that the height of the items exceeds the limit.

[0079] Regional anomaly data can be understood as regional anomaly data obtained after regional anomaly monitoring processing is performed based on the three-dimensional model of the monitored item. Specifically, regional monitoring can be performed on each monitored item based on the three-dimensional model of the monitored item. Specifically, the regular placement position of the monitored item can be preset, and the position of the monitored item in the three-dimensional model can be located, so as to further compare whether the monitored position of the monitored item is consistent with the regular placement position to determine whether the monitored item is in the area where it should be located. For example, by obtaining the regular placement position of the monitored item (that is, the area where it should be located), such as monitoring item A is usually placed in the first area of ​​shelf C, when item A is detected at a position outside the first area of ​​shelf C, it can be determined that the monitored item A is a regional anomaly. This application does not impose any restrictions on this.

[0080] Stability imbalance data can be understood as stability anomaly data obtained after stacking stability testing based on the three-dimensional model of the monitored items. Specifically, stacking stability testing can be performed on each monitored item based on the three-dimensional model of the monitored items. Specifically, the center of gravity of the stacked items can be calculated using the three-dimensional model of the monitored items to further determine whether each monitored item is prone to collapse or falling, thereby further evaluating the stability of the stacked items. This is not limited to this application.

[0081] Abnormal displacement data can be understood as abnormal displacement data obtained after real-time dynamic position tracking based on the three-dimensional model of the monitored object. Specifically, real-time dynamic position tracking can be performed on each monitored object based on the three-dimensional model of the monitored object, that is, the real-time changes of the monitored object can be obtained through the three-dimensional model of the monitored object to continuously track the position changes of the monitored object in space. For example, the displacement of the monitored object B can be discovered in a timely manner by monitoring the spatial data of the detected object B corresponding to time t1 in the three-dimensional model of the monitored object and the spatial data of the detected object B corresponding to time t2. If unauthorized movement is discovered, it can be determined that the monitored object B has abnormal displacement data between time t1 and time t2. This application does not impose any restrictions on this.

[0082] Furthermore, the highly abnormal data, regional abnormal data, stability imbalance data and abnormal displacement data obtained through monitoring are comprehensively processed, such as through list integration to clearly reflect the abnormalities of each position in the table. Optionally, different levels of priority warnings can be given to different position abnormal data, such as stability imbalance data can be set as a warning with a higher priority, and highly abnormal data can be set as a warning with a lower priority, so as to finally obtain ectopic optimization data. This application does not impose any restrictions on this.

[0083] Optionally, the obtained highly abnormal data, regional abnormal data, stability imbalance data and abnormal displacement data can be converted accordingly (such as being expressed in the form of scoring values), and the converted data can be further weighted and summed to obtain a numerical value that can be used to evaluate the overall degree of dislocation, and by setting a dislocation threshold, a dislocation warning can be issued when the dislocation threshold is exceeded, thereby better avoiding the dislocation of the monitored items. This application does not impose any restrictions on this.

[0084] The damaged item optimization data can be understood as the optimized data obtained after performing damage anomaly identification processing on each monitored item image data. Specifically, the initial damage identification module can be trained to obtain a more accurate target damage identification module, which can then be used to perform damage identification processing on each monitored item image data, thereby further obtaining the above-mentioned damaged item optimization data. This application does not impose any restrictions on this.

[0085] It should be understood that performing abnormal item damage identification processing based on the sample image data set, the labeled image data set, and the monitored item image data set to obtain optimized damaged item data refers to the process of performing abnormal item damage identification processing on the monitored item image data and providing optimization suggestions for abnormally damaged items to obtain optimized damaged item data. Step S23, i.e., performing abnormal item damage identification processing based on the sample image data set, the labeled image data set, and the monitored item image data set to obtain optimized damaged item data, may include the following steps:

[0086] S231: using an initial damage recognition module to perform appearance damage recognition processing on each sample image data in the sample image data set to obtain an appearance damaged object image data set;

[0087] S232: using an initial damage recognition module to perform quality damage recognition processing on each sample image data in the sample image data set to obtain a quality damaged item image data set;

[0088] S233: using an initial damage recognition module to perform expiration recognition processing on each sample image data in the sample image data set to obtain an expired item image data set;

[0089] S234: Constructing a damage recognition loss function based on the image data set of the appearance-damaged item, the image data set of the quality-damaged item, the image data set of the expired item, and the label image data set;

[0090] S235: Training the initial damage recognition module based on the damage recognition loss function to obtain a target damage recognition module;

[0091] S236: Using a target damage recognition module, performing damage recognition processing on each monitored object image data in the monitored object image data set to obtain a target damaged object image data set;

[0092] S231: Performing anomaly fusion processing on each target damaged item image data in the target damaged item image data set to obtain damaged item optimized data.

[0093] The initial damage identification module can be a pre-defined model or program based on artificial intelligence or machine learning algorithms, used to initially identify and analyze item image data to identify various possible damage conditions. It is understood that the parameters of this initial damage identification module can be randomly set and can be further optimized and adjusted during subsequent training to obtain parameters that better match the damage identification task, thereby further obtaining a more accurate target damage identification module. This application does not impose any restrictions on this.

[0094] The image data set of visually damaged items may include one or more images of visually damaged items. These images may be image data that indicates visual damage to the monitored items in the sample image data set, filtered out by the initial damage identification module after performing visual damage identification processing on each sample image data set. Visual damage may refer to damage to the physical appearance of the monitored items. The scope of identification may include, but is not limited to, deformation (e.g., change in shape), openings (e.g., cracks in the surface), fissures (e.g., cracks in the surface), or bloating (e.g., bloating of vacuumed items), etc. This application does not impose any restrictions on this.

[0095] The image data set of quality-damaged items may include one or more image data of quality-damaged items, and the image data of quality-damaged items may be image data that is screened out after the initial damage recognition module performs quality damage recognition processing on each sample image data in the sample image data set, showing that the monitored items in the sample image data have quality damage. Quality damage may refer to the phenomenon of damage to the quality of the monitored items, and the scope of recognition may include but is not limited to leakage (such as internal substances leaking out), packaging soaking (such as packaging being soaked due to moisture, etc.), etc. Optionally, AI vision technology can be further combined with Internet of Things technology, such as identifying weight anomalies through weight sensors (such as actual weight not matching the standard weight), or identifying temperature anomalies through temperature sensors, or identifying humidity anomalies through humidity sensors, etc., to more comprehensively analyze and judge the quality damage of each monitored item image data, and this application does not impose any restrictions on this.

[0096] The expired item image data set may include one or more expired item image data. The expired item image data may be image data indicating expiration of the monitored item in the sample image data set, filtered out by the initial damage identification module after performing expiration identification processing on each sample image data in the sample image data set. Specifically, information related to the expiration date of the monitored item, such as identification information such as the production date and shelf life, may be identified based on AI visual recognition to determine whether the monitored item is expired. This application does not impose any restrictions on this.

[0097] The damage identification loss function can be constructed based on the visually damaged item image dataset, the quality-damaged item image dataset, the expired item image dataset, and the labeled image dataset, and used for identifying abnormal damaged items. This damage identification loss function can be used to measure the difference between the prediction results of the initial damage identification module and the true labels in the labeled image dataset. By minimizing this difference, the parameters of the damage identification module are optimized, thereby improving the damage identification accuracy of the damage identification module.

[0098] Optionally, the damage identification loss function involved in this application can be constructed based on a multi-classification cross entropy loss function. Optionally, the process of constructing the damage identification loss function based on the image dataset of the items with damaged appearance, the image dataset of the items with damaged quality, the image dataset of the items with expired quality, and the label image dataset can be seen in the following formula:

[0099]

[0100] in, It can be expressed as a damage identification loss function; It can represent the number of monitored object image data in the monitored object image data set, that is, the number of samples; Can represent the index of the sample; Indicates the damage category. This application uses three damage categories (i.e., appearance damage, quality damage, and expiration) as an example for illustration and does not limit this application. An index that can represent the damage category; It can be said that the i-th sample data (i.e., the monitored item image data) is the true label of the j-th type of damage; Indicates the operation of performing logarithmic operations; It can represent the predicted probability that the i-th sample data (i.e., the monitored item image data) is damaged by the j-th category.

[0101] It should be noted that in the process of training the initial damage identification module, the parameters of the initial damage identification module can be calculated first, and then the damage identification loss function can be further calculated based on the parameters or data of the initial damage identification module (such as the parameters of the initial damage identification module itself, or the result data obtained by calculating using the initial damage identification module). That is, the parameters or data of the above-mentioned initial damage identification module are substituted into the damage identification loss function to calculate and obtain the damage identification loss value. When the damage identification loss value is less than the preset loss threshold, that is, at this time, the damage identification loss function converges, the training of the initial damage identification module is completed, and the target damage identification module is obtained. This application does not impose any restrictions on this.

[0102] The target damage identification module can be understood as a module that is trained and optimized using a damage identification loss function on the initial damage identification module, resulting in a more accurate identification of item damage. The target damaged item image data set may include one or more target damaged item image data. This target damaged item image data can be understood as image data containing item damage information obtained after damage identification processing is performed on each monitored item image data using the target damage identification module. It is understood that this target damaged item image data is more accurate and reliable than the identification results obtained using the initial damage identification module.

[0103] Furthermore, each target damaged item image data set in the target damaged item image data set undergoes anomaly fusion processing, such as by tabulating and consolidating the data to clearly reflect each damage anomaly and damage degree in a table, thereby obtaining optimized damaged item data. Optionally, different damage anomalies can be assigned different levels of priority warnings, such as a higher priority warning for expired items and a lower priority warning for cosmetic damage, ultimately obtaining optimized damaged item data. This is not a limitation of this application.

[0104] After obtaining the out-of-position optimization data and damaged item optimization data, the current inventory data in the inventory management basic data can be optimized and compared and verified based on the out-of-position optimization data and damaged item optimization data. On the one hand, the data with anomalies (such as out-of-position anomalies and damage anomalies) can be optimized, and on the other hand, the quantity information of the monitored items in the current inventory data can be checked and verified for accuracy, thereby correcting inventory data deviations caused by abnormal position or damage of items. The quantity in the current inventory data can also be further verified to obtain verified inventory data. Verified inventory data can be understood as inventory data that has been reviewed and verified, with position anomalies and damage anomalies corrected and quantity verified, which can more accurately reflect the actual quantity, location and status of the monitored items in the inventory.

[0105] Outbound optimization processing can be understood as the use of AI visual technology to perform relevant identification on monitored items, such as expiration date identification, fragility identification, etc., to further determine the related operations of outbound priority of monitored items. Specifically, taking expiration date identification as an example, it is possible to identify information such as the shelf life and production date on the monitored items, and further combine it with the current time to analyze the remaining validity period of the monitored items, so as to determine the outbound priority for the monitored items. For example, for monitored items with a shorter remaining validity period, it can be indicated to be prioritized for outbound delivery, and for monitored items with a longer remaining validity period, it can be indicated not to be temporarily outbound. This application does not impose any restrictions on this.

[0106] Based on AI vision technology, the current inventory data is dynamically optimized and processed. Image acquisition and analysis can be used to accurately identify abnormal location, damage status, expiration date and other information of items, so that inventory data can be updated in real time. For example, inventory deviations caused by misplacement, damage, expiration, etc. of items can be corrected, and the order of item delivery can be reasonably planned. This can significantly improve the accuracy and timeliness of inventory data, which is conducive to providing a reliable basis for subsequent procurement, allocation, sales and other decisions, and can effectively reduce inventory costs, improve inventory turnover, and enhance the competitiveness of enterprises in the market.

[0107] S30: performing inventory structure evaluation based on the inventory management basic data and the dynamically optimized inventory data to obtain inventory structure evaluation data.

[0108] Inventory structure assessment data can be understood as an assessment of the inventory structure of various products in inventory based on basic inventory management data and dynamically optimized inventory data. This data is derived from analyzing and evaluating characteristics such as category share, value distribution, inventory time distribution, and regional distribution. This inventory structure assessment data can be used to determine the rationality of the inventory structure.

[0109] Specifically, this application uses the example of inventory structure evaluation, including the evaluation of category structure, value structure, time structure, and regional structure, which does not limit this application. Optionally, the basic inventory management data and dynamic optimization inventory data can be classified and summarized according to the category of goods. For example, the inventory goods can be divided into categories such as food, daily necessities, and electronic products, and the proportion of each category in the total inventory can be further calculated to obtain category structure evaluation data (that is, category proportion).

[0110] Optionally, inventory can be categorized based on the value of the goods. An ABC classification method can be used to calculate the inventory share and related indicators (i.e., value distribution) of the three categories of goods, A, B, and C. The ABC classification method can generally be divided according to the following rules: Category A goods: cumulative sales account for approximately 70%-80%, and the number of varieties accounts for approximately 10%-20%; Category B goods: cumulative sales account for approximately 15%-25%, and the number of varieties accounts for approximately 20%-30%; Category C goods: cumulative sales account for approximately 5%-15%, and the number of varieties accounts for approximately 60%-70%. This application does not impose any restrictions on this.

[0111] Optionally, inventory can be classified according to time factors such as the time of entry or shelf life of the goods. For example, it can be divided into recently entered goods, medium-term entered goods, and long-term unsold goods; or it can be divided into goods nearing the expiration date, medium-term expiration date goods, and long-term expiration date goods according to the shelf life; and the proportion of goods in the inventory and related indicators in different time intervals (that is, inventory time distribution) can be further calculated.

[0112] Optionally, you can categorize inventory based on the storage area of ​​the goods, such as different warehouses, different shelf areas, etc., and further calculate the inventory share and related indicators of each area (that is, regional distribution).

[0113] Specifically, certain key indicators can be further calculated to better assess inventory structure. For example, the cost of sales and average inventory value over a certain period can be obtained from the basic inventory management data to calculate inventory turnover; actual sales volume and available inventory quantity over a certain period can be counted to calculate inventory fill rate; standards for slow-moving inventory can be pre-defined, such as long-term unsold goods (for example, goods that have not been sold for more than 90 days), to further calculate the amount or quantity of slow-moving inventory and calculate its proportion of total inventory, etc. This application does not impose any restrictions on this.

[0114] Furthermore, the proportion data and key indicator data for each of the aforementioned dimensions can be summarized and organized. Specifically, this can be summarized in a table and analyzed and interpreted (e.g., observing the proportion of each dimension, the changing trends of key indicators, and the differences with historical data or industry standards; summarizing the overall characteristics of the inventory structure, identifying existing problems and potential optimization points), thereby obtaining inventory structure assessment data. It is understood that this inventory structure assessment data may include an overall overview of the inventory structure, a detailed analysis of each dimension, the calculation results of key indicators and comparative analysis with historical data or industry standards, existing problems and improvement suggestions, etc., and this application does not impose any restrictions on this.

[0115] S40: Performing real-time dynamic adjustment processing on the inventory structure evaluation data according to the external dynamic impact data to obtain inventory dynamic adjustment data.

[0116] Inventory dynamic adjustment data can be understood as the dynamic adjustment data obtained by real-time dynamic adjustment of inventory structure assessment data based on external dynamic influence data. It should be noted that the data obtained by real-time adjustment of inventory structure assessment data in combination with external dynamic influence data can further enable the inventory structure to better adapt to changes in external factors such as market changes and supply chain fluctuations, thereby maintaining an optimal inventory status.

[0117] Specifically, by sorting out external dynamic influencing data, such as analyzing market demand fluctuation data, supplier supply capacity data, competitor strategy data and / or macroeconomic and policy data, and further analyzing the impact of each data on the inventory structure, such as the impact on category structure (for example, according to market demand fluctuations, when the demand for a certain category increases, the inventory of that category can be increased), the impact on the value structure (for example, when the economy is down, consumers tend to buy cost-effective goods, so the inventory of Class B medium-value goods can be appropriately increased, and the proportion of Class A high-end goods inventory can be reduced), the impact on the time structure (for example, in autumn, the inventory of down jackets that have been recently warehoused needs to be increased in advance, and at the same time, the turnover of summer clothing inventory needs to be accelerated to reduce long-term inventory backlogs), and the impact on the regional structure (taking into account regional differences in policies and regulations, such as some regions have strong subsidy policies for Class C products, and the inventory of Class C products in this region can be appropriately increased), etc., so as to obtain inventory dynamic adjustment data, and this application does not impose any restrictions on this.

[0118] Optionally, an inventory dynamic adjustment strategy can be further formulated based on the inventory dynamic adjustment data to more comprehensively optimize the inventory quantity, inventory layout and inventory turnover, and real-time monitoring and feedback can be performed during the adjustment process to further adjust the strategy to obtain more appropriate and more in line with the current situation inventory dynamic adjustment data. This application does not impose any restrictions on this.

[0119] S50: Generate an inventory health assessment report based on the dynamically optimized inventory data, the inventory structure assessment data, and the inventory dynamic adjustment data.

[0120] The inventory health assessment report can be understood as an assessment report generated based on the dynamically optimized inventory data, the inventory structure assessment data, and the inventory dynamic adjustment data, indicating the overall health of the warehouse inventory. Optionally, the inventory health assessment report may include, but is not limited to, an evaluation of inventory accuracy, the rationality of the inventory structure, the adaptability of the inventory to external changes, and suggestions for improvements to existing issues, etc., which are not limited in this application.

[0121] Specifically, an overview of the current inventory status can be achieved based on the dynamic optimization inventory data and inventory structure assessment data, and the inventory dynamic adjustment data can be further interpreted to obtain response measures and adjusted data changes, so as to further demonstrate the effectiveness of the dynamic adjustment and obtain the above-mentioned inventory health assessment report. Optionally, a scoring system can be established to score the inventory health status based on dimensions such as inventory accuracy, structural rationality, and dynamic adjustment timeliness, and to give a conclusion on the health status. Optionally, potential risks in inventory management can be pointed out, and specific improvement measures can be proposed for the problems and risks found. This application does not limit this.

[0122] It can be seen that in the above scheme, by obtaining the inventory management basic data and external dynamic impact data of the warehouse to be monitored, the current inventory data in the inventory management basic data can be dynamically optimized based on AI vision technology to obtain dynamically optimized inventory data, and the inventory structure evaluation can be performed based on the inventory management basic data and the dynamically optimized inventory data to obtain inventory structure evaluation data, so that the inventory structure evaluation data can be dynamically adjusted in real time according to the external dynamic impact data to obtain inventory dynamic adjustment data, and then an inventory health evaluation report can be generated based on the dynamically optimized inventory data, the inventory structure evaluation data and the inventory dynamic adjustment data, which is conducive to achieving the purpose of more accurate and comprehensive warehouse inventory intelligent management.

[0123] For the same example as above, please refer to Figure 3 , Figure 3 A schematic diagram of the structure of a terminal provided in an embodiment of the present application is shown in FIG. Figure 3 As shown, the system comprises a processor, an input device, an output device and a memory, which are connected to each other, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions, and the program includes instructions for executing the following steps;

[0124] Obtain basic inventory management data and external dynamic impact data of the warehouse to be monitored;

[0125] Based on AI vision technology, according to the sample image data set and the label image data set, the current inventory data in the inventory management basic data is dynamically optimized to obtain dynamically optimized inventory data;

[0126] Performing inventory structure evaluation based on the inventory management basic data and the dynamically optimized inventory data to obtain inventory structure evaluation data;

[0127] Performing real-time dynamic adjustment processing on the inventory structure assessment data according to the external dynamic impact data to obtain inventory dynamic adjustment data;

[0128] An inventory health assessment report is generated based on the dynamically optimized inventory data, the inventory structure assessment data, and the inventory dynamic adjustment data.

[0129] In this example, by obtaining the inventory management basic data and external dynamic impact data of the warehouse to be monitored, the current inventory data in the inventory management basic data can be dynamically optimized based on AI vision technology to obtain dynamically optimized inventory data, and inventory structure evaluation can be performed based on the inventory management basic data and the dynamically optimized inventory data to obtain inventory structure evaluation data, so that the inventory structure evaluation data can be dynamically adjusted in real time based on the external dynamic impact data to obtain inventory dynamic adjustment data, and then an inventory health evaluation report can be generated based on the dynamically optimized inventory data, the inventory structure evaluation data and the inventory dynamic adjustment data, which is conducive to achieving the purpose of more accurate and comprehensive warehouse inventory intelligent management.

[0130] The above mainly introduces the scheme of the embodiment of the present application from the perspective of the execution process on the method side. It is understandable that, in order to implement the above functions, the terminal includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of the various examples described in the embodiments provided herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in a hardware or computer software driven hardware manner depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0131] The embodiment of the present application can divide the terminal into functional units according to the above method example. For example, each functional unit can be divided according to each function, or two or more functions can be integrated into one processing unit. The above integrated unit can be implemented in the form of hardware or in the form of software functional units. It should be noted that the division of units in the embodiment of the present application is schematic and is only a logical functional division. In actual implementation, there may be other division methods.

[0132] In line with the above, please see Figure 4 , Figure 4 The present invention provides a schematic diagram of a warehouse inventory intelligent management device based on AI visual monitoring. Figure 4 As shown, the device includes:

[0133] An acquisition unit 101 is used to acquire basic inventory management data and external dynamic impact data of the warehouse to be monitored;

[0134] A first processing unit 102 is configured to dynamically optimize the current inventory data in the inventory management basic data based on AI vision technology to obtain dynamically optimized inventory data;

[0135] A second processing unit 103 is configured to perform inventory structure evaluation based on the inventory management basic data and the dynamically optimized inventory data to obtain inventory structure evaluation data;

[0136] The third processing unit 104 is configured to perform real-time dynamic adjustment processing on the inventory structure evaluation data according to the external dynamic impact data to obtain inventory dynamic adjustment data;

[0137] The fourth processing unit 105 is configured to generate an inventory health assessment report based on the dynamically optimized inventory data, the inventory structure assessment data, and the inventory dynamic adjustment data.

[0138] In one possible implementation, the first processing unit 102 is configured to dynamically optimize the current inventory data in the inventory management basic data based on the sample image data set and the label image data set based on AI vision technology to obtain dynamically optimized inventory data, specifically for:

[0139] Based on the AI ​​vision technology, images of the items in the warehouse to be monitored are collected and preprocessed to obtain a set of monitored item image data;

[0140] Based on the AI ​​vision technology, position anomaly recognition processing is performed according to the sample image data set, the label image data set and the monitored object image data set to obtain out-of-position optimization data;

[0141] Based on the AI ​​vision technology, each monitored item image data in the monitored item image data set is subjected to an abnormal item damage identification process to obtain optimized damaged item data;

[0142] Based on the AI ​​vision technology, the current inventory data in the inventory management basic data is reviewed and verified according to the off-site optimization data and the damaged item optimization data to obtain verified inventory data;

[0143] Based on the AI ​​vision technology, the verified inventory data is subjected to outbound optimization processing to obtain dynamically optimized inventory data.

[0144] In one possible implementation, the first processing unit 102 is configured to perform position anomaly identification processing on each monitored object image data in the monitored object image data set to obtain out-of-position optimized data, specifically for:

[0145] Performing object three-dimensional model construction processing based on each monitored object image data in the monitored object image data set to obtain a monitored object three-dimensional model;

[0146] Performing height anomaly monitoring processing based on the three-dimensional model of the monitored object to obtain height anomaly data;

[0147] Performing regional anomaly monitoring processing based on the three-dimensional model of the monitored object to obtain regional anomaly data;

[0148] Performing stacking stability detection processing based on the three-dimensional model of the monitored object to obtain stability imbalance data;

[0149] Performing real-time dynamic position tracking processing based on the three-dimensional model of the monitored object to obtain abnormal displacement data;

[0150] Anomaly fusion processing is performed according to the height abnormality data, the regional abnormality data, the stability imbalance data and the abnormal shift data to obtain ex-situ optimized data.

[0151] In one possible implementation, the first processing unit 102 is configured to perform object 3D model construction processing based on each monitored object image data in the monitored object image data set to obtain the monitored object 3D model, specifically for:

[0152] From the monitored object image data set, obtaining a first monitored object image data subset taken by a first photographing device and a second monitored object image data subset taken by a second photographing device;

[0153] Acquire first plane coordinate data of a first imaging point from the first monitored object image data subset, and acquire second plane coordinate data of the first imaging point from the second monitored object image data subset;

[0154] determining first three-dimensional coordinates of the first imaging point based on the first plane coordinate data, the second plane coordinate data, the first optical center coordinates of the first photographing device, the first focal length of the first photographing device, the second optical center coordinates of the second photographing device, and the second focal length of the second photographing device;

[0155] Acquire third plane coordinate data of a second imaging point from the first monitored object image data subset, and acquire fourth plane coordinate data of the second imaging point from the second monitored object image data subset;

[0156] determining second three-dimensional coordinates of the second imaging point based on the third plane coordinate data, the fourth plane coordinate data, the first optical center coordinates of the first photographing device, the first focal length of the first photographing device, the second optical center coordinates of the second photographing device, and the second focal length of the second photographing device;

[0157] A three-dimensional model of the monitored object is obtained according to the first three-dimensional coordinates and the second three-dimensional coordinates.

[0158] In one possible implementation, the first processing unit 102 is configured to perform abnormal item damage identification processing based on the sample image data set, the label image data set, and the monitored item image data set to obtain optimized damaged item data, specifically for:

[0159] Using the initial damage recognition module, each sample image data in the sample image data set is processed for appearance damage recognition to obtain an appearance damaged object image data set;

[0160] Using an initial damage recognition module, performing quality damage recognition processing on each sample image data in the sample image data set to obtain a quality damaged item image data set;

[0161] Using an initial damage recognition module, performing expiration recognition processing on each sample image data in the sample image data set to obtain an expired item image data set;

[0162] Constructing a damage recognition loss function based on the image dataset of the appearance-damaged items, the image dataset of the quality-damaged items, the image dataset of the expired items, and the label image dataset;

[0163] Based on the damage identification loss function, the initial damage identification module is trained to obtain a target damage identification module;

[0164] Using a target damage recognition module, performing damage recognition processing on each monitored object image data in the monitored object image data set to obtain a target damaged object image data set;

[0165] Anomaly fusion processing is performed on each target damaged item image data in the target damaged item image data set to obtain damaged item optimized data.

[0166] An embodiment of the present application also provides a computer storage medium, wherein the computer storage medium stores a computer program for electronic data exchange, and the computer program enables a computer to execute part or all of the steps of any one of the warehouse inventory intelligent management methods based on AI visual monitoring as recorded in the above method embodiments.

[0167] An embodiment of the present application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program enables a computer to execute part or all of the steps of any one of the warehouse inventory intelligent management methods based on AI visual monitoring as recorded in the above method embodiments.

[0168] It should be noted that for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by this application.

[0169] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0170] In the several embodiments provided in this application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0171] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0172] In addition, the functional units in the various embodiments of the application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software program modules.

[0173] If the integrated unit is implemented in the form of a software program module and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a memory and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned memory includes: U disk, read-only memory (ROM), random access memory (RAM), mobile hard disk, magnetic disk, or optical disk, etc., various media that can store program code.

[0174] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable memory, which may include: a flash drive, a read-only memory, a random access memory, a magnetic disk or an optical disk, etc.

[0175] The above is a detailed introduction to the embodiments of the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for those skilled in the art, according to the idea of ​​the present application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A warehouse inventory intelligent management method based on AI visual monitoring, characterized in that: The method comprises: Obtain basic inventory management data and external dynamic impact data of the warehouse to be monitored; Based on AI vision technology, according to the sample image data set and the label image data set, the current inventory data in the inventory management basic data is dynamically optimized to obtain dynamically optimized inventory data; Performing inventory structure evaluation based on the inventory management basic data and the dynamically optimized inventory data to obtain inventory structure evaluation data; Performing real-time dynamic adjustment processing on the inventory structure assessment data according to the external dynamic impact data to obtain inventory dynamic adjustment data; generating an inventory health assessment report based on the dynamically optimized inventory data, the inventory structure assessment data, and the inventory dynamic adjustment data; The method of dynamically optimizing the current inventory data in the inventory management basic data based on the sample image data set and the label image data set based on the AI ​​vision technology to obtain the dynamically optimized inventory data includes: Based on the AI ​​vision technology, images of the items in the warehouse to be monitored are collected and preprocessed to obtain a set of monitored item image data; Based on the AI ​​vision technology, each monitored object image data in the monitored object image data set is subjected to position anomaly recognition processing to obtain out-of-position optimized data; Based on the AI ​​vision technology, abnormal identification processing of damaged items is performed according to the sample image data set, the label image data set and the monitored item image data set to obtain optimized data of damaged items; Based on the AI ​​vision technology, the current inventory data in the inventory management basic data is reviewed and verified according to the off-site optimization data and the damaged item optimization data to obtain verified inventory data; Based on the AI ​​vision technology, the verified inventory data is optimized for outbound delivery to obtain dynamically optimized inventory data; The performing of abnormal identification processing of damaged items based on the sample image data set, the label image data set, and the monitored item image data set to obtain optimized damaged item data includes: Using the initial damage recognition module, each sample image data in the sample image data set is processed for appearance damage recognition to obtain an appearance damaged object image data set; Using an initial damage recognition module, performing quality damage recognition processing on each sample image data in the sample image data set to obtain a quality damaged item image data set; Using an initial damage recognition module, performing expiration recognition processing on each sample image data in the sample image data set to obtain an expired item image data set; Constructing a damage recognition loss function based on the image dataset of the appearance-damaged items, the image dataset of the quality-damaged items, the image dataset of the expired items, and the label image dataset; Based on the damage identification loss function, the initial damage identification module is trained to obtain a target damage identification module; Using a target damage recognition module, performing damage recognition processing on each monitored object image data in the monitored object image data set to obtain a target damaged object image data set; performing anomaly fusion processing on each target damaged item image data in the target damaged item image data set to obtain damaged item optimized data; The damage identification loss function is represented by the following formula: in, represents the damage identification loss function; Indicates the number of monitored object image data in the monitored object image data set, that is, the number of samples; Indicates the index of the sample; Indicates the damage category; An index representing the damage category; Indicates that the i-th sample data is the true label of the j-th type of damage; Indicates the operation of performing logarithmic operations; It represents the predicted probability that the i-th sample data is damaged by the j-th type.

2. The method for intelligent warehouse inventory management based on AI visual monitoring according to claim 1 is characterized in that: The performing position anomaly recognition processing on each monitored object image data in the monitored object image data set to obtain out-of-position optimized data includes: Performing object three-dimensional model construction processing based on each monitored object image data in the monitored object image data set to obtain a monitored object three-dimensional model; Performing height anomaly monitoring processing based on the three-dimensional model of the monitored object to obtain height anomaly data; Performing regional anomaly monitoring processing based on the three-dimensional model of the monitored object to obtain regional anomaly data; Performing stacking stability detection processing based on the three-dimensional model of the monitored object to obtain stability imbalance data; Performing real-time dynamic position tracking processing based on the three-dimensional model of the monitored object to obtain abnormal displacement data; Anomaly fusion processing is performed according to the height abnormality data, the regional abnormality data, the stability imbalance data and the abnormal shift data to obtain ex-situ optimized data.

3. The intelligent warehouse inventory management method based on AI visual monitoring according to claim 2 is characterized in that: The step of constructing a three-dimensional model of an object based on each monitored object image data in the monitored object image data set to obtain a three-dimensional model of the monitored object includes: From the monitored object image data set, obtaining a first monitored object image data subset taken by a first photographing device and a second monitored object image data subset taken by a second photographing device; Acquire first plane coordinate data of a first imaging point from the first monitored object image data subset, and acquire second plane coordinate data of the first imaging point from the second monitored object image data subset; determining first three-dimensional coordinates of the first imaging point based on the first plane coordinate data, the second plane coordinate data, the first optical center coordinates of the first photographing device, the first focal length of the first photographing device, the second optical center coordinates of the second photographing device, and the second focal length of the second photographing device; Acquire third plane coordinate data of a second imaging point from the first monitored object image data subset, and acquire fourth plane coordinate data of the second imaging point from the second monitored object image data subset; determining second three-dimensional coordinates of the second imaging point based on the third plane coordinate data, the fourth plane coordinate data, the first optical center coordinates of the first photographing device, the first focal length of the first photographing device, the second optical center coordinates of the second photographing device, and the second focal length of the second photographing device; A three-dimensional model of the monitored object is obtained according to the first three-dimensional coordinates and the second three-dimensional coordinates.

4. An intelligent warehouse inventory management device based on AI visual monitoring, characterized in that: The device comprises: An acquisition unit, used to obtain basic inventory management data and external dynamic impact data of the warehouse to be monitored; a first processing unit, configured to dynamically optimize the current inventory data in the inventory management basic data based on the sample image data set and the label image data set based on AI vision technology to obtain dynamically optimized inventory data; a second processing unit, configured to perform inventory structure evaluation based on the inventory management basic data and the dynamically optimized inventory data to obtain inventory structure evaluation data; a third processing unit, configured to perform real-time dynamic adjustment processing on the inventory structure evaluation data according to the external dynamic impact data to obtain inventory dynamic adjustment data; a fourth processing unit, configured to generate an inventory health assessment report based on the dynamically optimized inventory data, the inventory structure assessment data, and the inventory dynamic adjustment data; The first processing unit is configured to dynamically optimize the current inventory data in the inventory management basic data based on the sample image data set and the label image data set based on AI vision technology to obtain dynamically optimized inventory data, specifically for: Based on the AI ​​vision technology, images of the items in the warehouse to be monitored are collected and preprocessed to obtain a set of monitored item image data; Based on the AI ​​vision technology, each monitored object image data in the monitored object image data set is subjected to position anomaly recognition processing to obtain out-of-position optimized data; Based on the AI ​​vision technology, abnormal identification processing of damaged items is performed according to the sample image data set, the label image data set and the monitored item image data set to obtain optimized data of damaged items; Based on the AI ​​vision technology, the current inventory data in the inventory management basic data is reviewed and verified according to the off-site optimization data and the damaged item optimization data to obtain verified inventory data; Based on the AI ​​vision technology, the verified inventory data is optimized for outbound delivery to obtain dynamically optimized inventory data; In the aspect of performing abnormal identification processing of damaged items based on the sample image data set, the label image data set, and the monitored item image data set to obtain optimized damaged item data, the first processing unit is specifically configured to: Using the initial damage recognition module, each sample image data in the sample image data set is processed for appearance damage recognition to obtain an appearance damaged object image data set; Using an initial damage recognition module, performing quality damage recognition processing on each sample image data in the sample image data set to obtain a quality damaged item image data set; Using an initial damage recognition module, performing expiration recognition processing on each sample image data in the sample image data set to obtain an expired item image data set; Constructing a damage recognition loss function based on the image dataset of the appearance-damaged items, the image dataset of the quality-damaged items, the image dataset of the expired items, and the label image dataset; Based on the damage identification loss function, the initial damage identification module is trained to obtain a target damage identification module; Using a target damage recognition module, performing damage recognition processing on each monitored object image data in the monitored object image data set to obtain a target damaged object image data set; performing anomaly fusion processing on each target damaged item image data in the target damaged item image data set to obtain damaged item optimized data; The damage identification loss function is represented by the following formula: in, represents the damage identification loss function; Indicates the number of monitored object image data in the monitored object image data set, that is, the number of samples; Indicates the index of the sample; Indicates the damage category; An index representing the damage category; Indicates that the i-th sample data is the true label of the j-th type of damage; Indicates the operation of performing logarithmic operations; It represents the predicted probability that the i-th sample data is damaged by the j-th type.

5. The intelligent warehouse inventory management device based on AI visual monitoring according to claim 4 is characterized in that: The first processing unit is configured to perform position anomaly recognition processing on each monitored object image data in the monitored object image data set to obtain out-of-position optimized data, specifically for: Performing object three-dimensional model construction processing based on each monitored object image data in the monitored object image data set to obtain a monitored object three-dimensional model; Performing height anomaly monitoring processing based on the three-dimensional model of the monitored object to obtain height anomaly data; Performing regional anomaly monitoring processing based on the three-dimensional model of the monitored object to obtain regional anomaly data; Performing stacking stability detection processing based on the three-dimensional model of the monitored object to obtain stability imbalance data; Performing real-time dynamic position tracking processing based on the three-dimensional model of the monitored object to obtain abnormal displacement data; Anomaly fusion processing is performed according to the height abnormality data, the regional abnormality data, the stability imbalance data and the abnormal shift data to obtain ex-situ optimized data.

6. A terminal, characterized in that: The method comprises a processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute the method according to any one of claims 1 to 3.

7. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor is caused to perform the method according to any one of claims 1 to 3.

Images