Goods in-warehouse checking method, system and device and computer equipment

Abstract

The invention relates to the technical field of warehouse management, and discloses an in-warehouse goods checking method, system and device and computer equipment, and the method comprises the steps: obtaining a tray code on a target checking goods location and the information of a material box on a corresponding tray; identifying the material box information by adopting a deep learning identification algorithm, and determining material specifications; based on the material box information, a preset stacking rule corresponding to the material specification and an AI algorithm, the number of the material boxes on the tray is calculated; and binding the material specifications, the number of the material boxes and the tray codes, and comparing with the upper inventory data based on the tray codes to obtain a cargo inventory result. According to the invention, the tray code and the material box information on the target inventory goods allocation can be directly obtained without continuously moving out the goods at the target inventory goods allocation, the difference between the inventory data and the actual inventory data can be found in time, and the problems of low inventory efficiency and increased inventory time caused by continuously moving out the goods at the target inventory goods allocation are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of warehouse management, and in particular to a method, system, device and computer equipment for counting goods in a warehouse. Background Art

[0002] A high-bay warehouse is a storage method that utilizes height to store goods, typically requiring the use of equipment such as stackers for handling, accessing, and storing goods. Inventory in a high-bay warehouse involves verifying the quantity and status of goods stored in the high-bay warehouse to ensure the accuracy and timeliness of storage information.

[0003] In the past, when there was only a single camera, for double-deep target inventory locations, when there were goods in both the deep and shallow locations, the shallow location pallets had to be moved to other locations first, and then the deep location pallets were taken out, and then the goods in the deep and shallow locations were counted separately. This process not only significantly increased the inventory time, but also resulted in reduced inventory efficiency and accuracy. Summary of the Invention

[0004] In view of this, the present invention provides a method, system, device and computer equipment for inventory counting of goods in warehouse to solve the problem that the target inventory location needs to continuously remove goods, resulting in reduced inventory efficiency and increased inventory time.

[0005] In a first aspect, the present invention provides a method for inventorying goods in a warehouse, the method comprising:

[0006] Obtain the pallet code on the target inventory location and the material box information on the corresponding pallet;

[0007] Use deep learning recognition algorithms to identify material box information and determine material specifications;

[0008] Calculate the number of material boxes on the pallet based on the material box information, preset stacking rules corresponding to the material specifications, and AI algorithms;

[0009] After binding the material specifications, material box quantity and pallet code, the pallet code is compared with the upper-level inventory data to obtain the inventory results.

[0010] The present invention provides a method for in-warehouse inventory counting of goods. This method directly obtains the pallet code and material box information on the target inventory location without the need to continuously remove goods from the target inventory location. This method provides comprehensive and fundamental data support for the subsequent inventory process, reduces the time cost of data collection, and ensures the timeliness and integrity of the data. A deep learning recognition algorithm is used to identify material box information and determine material specifications. Compared to traditional manual recognition or simple image recognition methods, deep learning algorithms have stronger learning capabilities and adaptability, can handle complex and changing material box information, improve the accuracy of material specification identification, and reduce inventory errors caused by specification identification errors. The method calculates the number of material boxes on a pallet based on material box information, preset stacking rules, and an AI algorithm. This method fully utilizes existing rules and intelligent algorithms, not only accurately calculating the number, but also, for some material boxes with irregular stacking or occlusion, intelligent analysis using the AI ​​algorithm can determine a relatively accurate number, improving counting efficiency and accuracy. By binding the material specification and material box quantity to the pallet code and comparing them with higher-level inventory data, the inventory results can be quickly and intuitively obtained. This process realizes automatic matching and difference analysis of data, reduces the workload and error probability of manual data verification, and can promptly discover the differences between inventory data and actual inventory data, providing accurate data basis for inventory management decisions, helping to timely adjust inventory strategies and optimize inventory layout, etc., and solves the problem of reduced inventory efficiency and increased inventory time caused by the need to continuously move goods out of the target inventory location.

[0011] In an optional embodiment, the target inventory storage locations include single-depth locations and double-depth locations; the double-depth locations include deep storage locations and shallow storage locations;

[0012] Obtain the pallet code on the target inventory location and the material box information on the corresponding pallet, including:

[0013] When the target inventory location is a single-deep location, a close-range barcode scanner is used to scan the pallet code on the single-deep location, and an image of the goods on the pallet is captured to obtain the single-deep location material box information;

[0014] When the target inventory storage location is a double-deep location and there are goods placed in both the deep and shallow locations, a stacker crane is used to move the goods in the shallow location to the loading platform. The long-distance scanning device installed on the stacker crane is used to scan and obtain the pallet code on the deep location, and the goods on the deep location pallet are imaged to obtain the deep location material box information. At the same time, the scanning device of the loading platform detection mechanism is used to scan and obtain the pallet code on the shallow location, and the goods on the shallow location pallet are imaged to obtain the shallow location material box information.

[0015] When the target inventory location is double-deep and there are goods in the deep location but no goods in the shallow location, a long-distance scanning device is directly used to scan the pallet code on the deep location, and the image of the goods on the deep location pallet is captured to obtain the deep location material box information.

[0016] The present invention provides a method for inventorying goods in a warehouse, which adopts a special information collection method for different types of target inventory locations. For single-depth locations, short-range code scanning equipment can accurately and quickly obtain pallet codes, and simultaneously complete cargo image collection to obtain material box information. There is no need for complex equipment scheduling, which greatly shortens the data collection time, adapts to the simple layout characteristics of single-depth locations, and ensures that the data is timely and complete. In the double-depth location scenario, when there are goods in both deep and shallow locations, the stacker, long-distance scanning equipment, and cargo platform detection mechanism work together. The stacker accurately moves the shallow location goods to create conditions for scanning and image collection at different locations, comprehensively covering the information acquisition of goods at different locations in the double-depth location, and ensuring that no information is missed. When only the deep location in the double-depth location has goods, the long-distance scanning equipment is directly used to flexibly simplify the operating process and improve inventory efficiency. This information acquisition method tailored to the location greatly improves the efficiency and adaptability of the entire inventory process, laying a solid data foundation for subsequent inventory links.

[0017] In an optional implementation, a deep learning recognition algorithm is used to identify the material box information and determine the material specifications, including:

[0018] Use deep learning recognition algorithms to train material logo information and / or material text information to identify material-specific pattern features and / or text identification features;

[0019] The material specifications are obtained by comparing the material's specific pattern features and / or text identification features with the preset material specification feature library.

[0020] The present invention provides a method for inventorying goods in warehouses, which adopts a deep learning recognition algorithm to identify material box information and determine material specifications. Compared with traditional manual recognition or simple image recognition methods, the deep learning algorithm has stronger learning ability and adaptability, and can handle complex and changeable material box information, greatly improving the accuracy of material specification identification and reducing inventory errors caused by specification identification errors.

[0021] In an optional embodiment, the material box information also includes point cloud data of material stacking; the AI ​​algorithm includes a target detection algorithm and an OCR algorithm;

[0022] Calculate the number of material boxes on the pallet based on the material box information, preset stacking rules corresponding to the material specifications, and AI algorithms, including:

[0023] Use target detection algorithms to identify the location and outline of material boxes in the point cloud data of material palletizing, and calculate the actual arrangement and stacking order of materials based on the location and outline of the material boxes using the stacker crane full pallet counting method or the OCR algorithm-based loose pallet counting method;

[0024] The number of material boxes is calculated based on the actual material arrangement, stacking order and preset stacking rules corresponding to the material specifications.

[0025] The present invention provides a method for inventorying goods in warehouses. It incorporates point cloud data of material pallets. An object detection algorithm accurately identifies the position and outline of material boxes within the point cloud data. Point cloud data comprehensively and intuitively reflects the three-dimensional shape of the material pallets. Even with complex situations such as irregular material placement and partial occlusion, the object detection algorithm, with its powerful feature extraction capabilities, accurately delineates the position and outline of each box, providing a solid foundation for subsequent calculations. This is difficult to achieve with traditional two-dimensional image analysis. Based on the identified box positions and outlines, a flexible approach can be used: stacker pallet counting or loose pallet counting using an optical character recognition (OCR) algorithm. For relatively regular, nearly full-pallet materials, the stacker pallet counting method efficiently utilizes the characteristics of the equipment to quickly determine the material arrangement and stacking status. For more scattered loose pallet materials, the OCR algorithm can accurately determine the material arrangement and stacking order by identifying information such as characters and patterns on the box surfaces and combining it with position and outline data. This flexible switching of inventory strategies based on actual cargo conditions significantly improves the efficiency and accuracy of quantity calculations.

[0026] In an optional embodiment, the method for counting goods in inventory further includes:

[0027] In the online warehouse management system, a global matching retrieval algorithm is used to perform a global search of virtual storage locations. If the virtual storage location at the time the goods first enter the system does not change from the actual physical storage location at the time of the current inventory count, and the storage location has not been used for order allocation, the goods in the virtual storage location in the warehouse management system are considered to be consistent with the goods in the actual physical storage location. The inventory result is directly based on the data stored in the warehouse management system.

[0028] When the virtual storage location at the time of first online warehousing changes from the actual physical storage location at the current inventory time and / or the storage location is involved in order allocation, the goods at the actual physical storage location will be counted using the stacker crane full pallet method or the OCR algorithm-based loose pallet method. The inventory result will be based on the actual inventory result.

[0029] The present invention provides a method for inventorying goods in a warehouse. When the virtual cargo location and the actual physical cargo location have not changed and are not involved in order allocation, the warehouse management system is directly used to store data as the inventory result, avoiding the tedious on-site inventory process. This operation greatly shortens the inventory time. Especially in large-scale storage systems, if many cargo locations need to be inventoried on-site, it will consume a lot of time and manpower. This method can quickly complete the inventory of most stable cargo locations, greatly improving the overall inventory efficiency. When the virtual cargo location and the actual physical cargo location change and / or the cargo location participates in order allocation, it is timely switched to the stacker full pallet counting method or the loose pallet counting method based on the OCR algorithm. This flexible inventory strategy adjustment can closely fit the actual situation of warehousing and accurately reflect the current inventory status. Whether it is the efficient inventory of whole pallets of goods or the precise sorting of loose pallets of goods, it can ensure that the inventory results are consistent with the actual inventory, avoiding distortion of inventory data due to changes in warehousing.

[0030] In an optional embodiment, after binding the material specifications, material box quantity and pallet code, the pallet code is compared with the upper-level inventory data to obtain the inventory results, including:

[0031] Bind the material specifications, material box quantity and pallet code to obtain complete pallet material data;

[0032] Get the upper-level inventory data corresponding to the pallet code;

[0033] Compare the complete data of the pallet material with the upper-level inventory data to obtain the inventory results of the goods that are consistent and the inventory results of the goods that are different.

[0034] The present invention provides a method for inventorying goods in warehouses, which binds material specifications, the number of material boxes, and pallet codes to form complete data for pallet materials, establishing a closely related and comprehensive information set. This ensures that the material information of each pallet has a unique identifier and detailed attribute records, facilitating subsequent query, traceability, and management. By obtaining the upper-level inventory data corresponding to the pallet code and comparing it with the complete data, inventory discrepancies can be quickly and intuitively discovered. Clearly distinguishing between inventory results that are consistent with the inventory of goods and those that are discrepant provides clear inventory status information to enterprise decision-makers.

[0035] In an optional embodiment, the method for counting goods in inventory further includes:

[0036] When the inventory results of the goods inventory differences are obtained, the current inventory pallet and the goods on the pallet are frozen, and the corresponding handling tasks are retrieved for the goods with inventory differences in the online warehouse management system log. According to the handling tasks, the key information matching algorithm is used to find the equipment scheduling operation trajectory and the corresponding timestamp in the online equipment scheduling system. Combined with the images taken by the camera or camera in the current inventory stereo library, the image capture technology is combined with the deep learning algorithm to capture the image features, and the reasons for the differences in the goods inventory are found based on the image features. The reasons for the differences in the goods inventory include manual handling, box drop caused by equipment failure in the middle, or other reasons other than manual handling and box drop caused by equipment failure in the middle.

[0037] The present invention provides a method for inventorying goods in a warehouse. Once a discrepancy in the goods inventory is found, the currently counted pallets and goods are immediately frozen, which can effectively prevent the problematic goods from continuing to participate in the warehouse circulation and prevent the discrepancy problem from further expanding. The handling tasks corresponding to the differences in the goods inventory are retrieved through the online warehouse management system log, and the equipment scheduling operation track and timestamp are obtained from the online equipment scheduling system with the help of the key information matching algorithm. This tracing method can accurately restore the entire handling process of the goods from the time they enter the warehouse to before the inventory. The time and path of each operation link are clearly traceable. The images taken by the camera or camera in the stereoscopic warehouse are combined with image capture technology and deep learning algorithms to extract image features, so as to accurately locate the cause of the difference in the goods inventory.

[0038] In a second aspect, the present invention provides a system for counting goods in a warehouse, the system comprising:

[0039] Data collection terminal, used to obtain the pallet code on the target inventory location and the material box information on the corresponding pallet;

[0040] The data processing industrial computer is connected to the data acquisition terminal. The industrial computer is used to use a deep learning recognition algorithm to identify the material box information and determine the material specifications; calculate the number of material boxes on the pallet based on the preset stacking rules corresponding to the material box information and material specifications and the AI ​​algorithm; bind the material specifications and the number of material boxes with the pallet code, and then compare the pallet code with the upper inventory data to obtain the goods inventory result.

[0041] In a third aspect, the present invention provides a device for counting goods in a warehouse, the device comprising:

[0042] Data collection module, used to obtain the pallet code on the target inventory location and the material box information on the corresponding pallet;

[0043] The material box information recognition module is used to identify the material box information using a deep learning recognition algorithm to determine the material specifications;

[0044] The material box quantity calculation module is used to calculate the number of material boxes on the pallet based on the material box information, preset stacking rules corresponding to the material specifications, and AI algorithms;

[0045] The cargo comparison and inventory module is used to bind material specifications, material box quantities and pallet codes, and then compare the pallet codes with upper-level inventory data to obtain cargo inventory results.

[0046] In a fourth aspect, the present invention provides a computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the method for inventorying goods in a warehouse according to the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0047] In a fifth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method for inventorying goods in a warehouse according to the first aspect or any corresponding embodiment thereof.

[0048] In a sixth aspect, the present invention provides a computer program product comprising computer instructions for enabling a computer to execute the method for inventorying goods in a warehouse according to the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0050] Figure 1 1 is a flow chart of a method for counting goods in a warehouse according to an embodiment of the present invention;

[0051] Figure 2 is a flow chart of another method for counting goods in warehouse according to an embodiment of the present invention;

[0052] Figure 3 1 is a flow chart of another method for inventory counting of goods in warehouse according to an embodiment of the present invention;

[0053] Figure 4 is a structural diagram of a data acquisition terminal installed on a detection mechanism according to an embodiment of the present invention;

[0054] Figure 5 1 is a flow chart of another method for inventory counting of goods in warehouse according to an embodiment of the present invention;

[0055] Figure 6 This is a schematic diagram showing the inventory result according to an embodiment of the present invention;

[0056] Figure 7 2. It is a schematic diagram showing an inventory review interface according to an embodiment of the present invention;

[0057] Figure 8 This is a structural block diagram of a device for counting goods in a warehouse according to an embodiment of the present invention;

[0058] Figure 9 Schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0059] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0060] According to an embodiment of the present invention, an embodiment of a method for inventorying goods in a warehouse is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0061] In this embodiment, a method for counting goods in warehouse is provided, which can be used in the above-mentioned computer device. The computer device is installed with a system for counting goods in warehouse, which includes a data acquisition terminal, Figure 1 FIG. 1 is a flow chart of a method for counting goods in a warehouse according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0062] Step S101: Obtain the pallet code on the target inventory location and the material box information on the corresponding pallet.

[0063] Specifically, the data collection terminal includes a high-definition camera and a scanning device. The scanning device is used to scan the QR code or barcode on the pallet of the target inventory location to obtain the pallet code. The high-definition camera is used to capture the material box information on the pallet, including the top surface image of the material box and the side images of the four surfaces of the front, back, left and right.

[0064] Step S102: Use a deep learning recognition algorithm to identify the material box information and determine the material specifications.

[0065] Specifically, in an actual inventory scenario, after obtaining the image information of the material box, the feature information on the material box is identified. The feature information includes but is not limited to the product logo or obvious text information. It is processed in the same way as the preprocessing method when training the model to ensure that the image format, size, normalization and other parameters are consistent, and then the processed image is input into the trained deep learning recognition model. The convolutional layer and pooling layer in the model extract features from the input image layer by layer, converting the original image into a series of abstract feature maps. These feature maps contain key information such as the shape, texture, and color of the material box. The fully connected layer integrates the extracted features and calculates the probability distribution of each material specification category through the softmax function, outputting the prediction result, that is, the probability that the model believes that the material box belongs to each material specification. The material specification category with the largest probability value is selected as the final recognition result.

[0066] Step S103 , calculating the number of material boxes on the pallet based on the material box information, preset stacking rules corresponding to the material specifications, and the AI ​​algorithm.

[0067] Specifically, point cloud data of the material palletizing is first acquired. This point cloud data, collected by devices such as LiDAR, records the spatial configuration of the material palletizing in the form of three-dimensional coordinates. This point cloud data is then processed using object detection algorithms, such as the deep learning-based Single Shot MultiBox Detector (SSD) algorithm or the You Only Look Once (YOLO) family of algorithms. By learning from a large number of point cloud samples containing material boxes, the algorithm can identify the location and outline of the material boxes in the material palletizing point cloud data. During the recognition process, the algorithm divides the point cloud data into multiple small regions, extracts and analyzes features in each region, determines whether the region contains a material box, and if so, determines its location and outline information.

[0068] The stacker crane moves along a pre-set path, using its onboard sensors to obtain information about the relative position of the material boxes on the pallet. Combining this information with the box's position and outline, it analyzes the actual arrangement and stacking order of the materials. Once this information is known, the number of boxes is calculated based on the pre-set stacking rules for the material specifications.

[0069] In step S104, the material specifications, the number of material boxes and the pallet code are bound together, and then the pallet code is compared with the upper-level inventory data to obtain the inventory result.

[0070] Specifically, within the local warehouse management system, a data table or data structure is constructed specifically to store pallet-related information. This structure includes fields such as pallet code, material specification, and material case quantity. For example, a table named "Pallet Information Table" is created using a relational database. The "Pallet Code" field serves as the primary key to uniquely identify each pallet; the "Material Specification" field records the specific product category and specification of the goods on the pallet; and the "Material Case Quantity" field stores the calculated actual number of material cases on the pallet. The material specification, material case quantity, and corresponding pallet code information obtained from the previous inventory count process are accurately entered into this data structure, effectively binding the material specification, material case quantity, and pallet code. The acquired higher-level inventory data is typically returned in a specific data format, such as JSON or XML. The local system needs to parse this data to extract the material specification and material case quantity information corresponding to the pallet code. The locally bound material specification and material case quantity are then compared against the data obtained from the higher-level inventory system, one-to-one, based on the pallet code.

[0071] Based on the comparison results, the inventory status is categorized as either "Inventory Consistency" or "Inventory Discrepancy." If the local material specifications match those of the higher-level material, and the number of local material boxes matches the number of boxes of the higher-level material, the situation is considered "Inventory Consistency." Otherwise, it is considered "Inventory Discrepancy."

[0072] The in-stock inventory method provided by this embodiment directly obtains the pallet code and material box information on the target inventory location without the need to continuously remove goods from the target inventory location. This provides comprehensive and fundamental data support for the subsequent inventory process, reduces the time cost of data collection, and ensures the timeliness and integrity of the data. A deep learning recognition algorithm is used to identify material box information and determine material specifications. Compared to traditional manual recognition or simple image recognition methods, deep learning algorithms have stronger learning capabilities and adaptability, can handle complex and changing material box information, improve the accuracy of material specification identification, and reduce inventory errors caused by specification identification errors. The number of material boxes on the pallet is calculated based on material box information, preset stacking rules, and AI algorithms. This method fully utilizes existing rules and intelligent algorithms, not only accurately calculating the number, but also, for some material boxes with irregular stacking or occlusion, intelligent analysis by the AI ​​algorithm can also determine a more accurate number, improving counting efficiency and accuracy. After binding the material specification and material box quantity with the pallet code and comparing them with the upper-level inventory data, the inventory results can be quickly and intuitively obtained. This process realizes automatic matching and difference analysis of data, reduces the workload and error probability of manual data verification, and can promptly discover the differences between inventory data and actual inventory data, providing accurate data basis for inventory management decisions, helping to timely adjust inventory strategies and optimize inventory layout, etc., and solves the problem of reduced inventory efficiency and increased inventory time caused by the need to continuously move goods out of the target inventory location.

[0073] In this embodiment, a method for counting goods in a warehouse is provided, which can be used on the above-mentioned computer device. The computer device is installed with a system for counting goods in a warehouse. The system includes a data acquisition terminal, such as a mobile phone, a tablet computer, etc. Figure 2 FIG. 1 is a flow chart of a method for counting goods in a warehouse according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0074] Step S201: Obtain the pallet code on the target inventory location and the material box information on the corresponding pallet.

[0075] Specifically, the scanning equipment includes long-range and short-range scanners. The short-range scanner uses a highly sensitive optical scanning module, while the long-range scanner utilizes laser scanning technology. When the target inventory location is a single-deep bin, the short-range scanner directly scans the QR code or barcode on the pallet in the bin to obtain the pallet code. A high-definition camera also captures the material box information on the pallet.

[0076] The target inventory locations include single-depth locations and double-depth locations; double-depth locations include deep locations and shallow locations; the above step S201 includes:

[0077] Step S2011: When the target inventory location is a single-deep location, a short-range barcode scanning device is used to scan the pallet code on the single-deep location, and an image of the goods on the pallet is captured to obtain the single-deep location material box information.

[0078] Pallet Code Capture: Workers use specialized close-range scanners, typically equipped with highly sensitive optical scanning modules. When held close to the QR code label on the pallet, the scanner emits light of a specific wavelength, illuminating the QR code pattern. A QR code consists of small black and white squares, with different arrangements representing different information. The scanner identifies the arrangement of the black and white squares based on the intensity of the reflected light, and then interprets the pallet code. The entire process is responsive, typically completing scanning and interpretation within seconds. The data is transmitted in real time to a connected mobile terminal or backend system.

[0079] Material box information collection: As the pallet code is acquired, a high-definition camera, mounted in a fixed position near the single-deep position, begins operation. The camera boasts a wide viewing angle and high resolution, capable of covering the entire pallet area. Following a pre-set program, it captures the cargo on the pallet from multiple angles, ensuring a clear image of each material box (including the top surface and all four sides). The captured images, typically in common JPEG or PNG formats, capture not only the appearance and identification of the material boxes, but also their placement and relationship on the pallet.

[0080] Step S2012: When the target inventory location is a double-deep location and there are goods placed in both the deep and shallow locations, a stacker is used to move the goods in the shallow location to the loading platform, and a long-distance scanning device provided on the stacker is used to scan and obtain the pallet code on the deep location, and an image of the goods on the deep location pallet is captured to obtain the deep location material box information; at the same time, a scanning device of the loading platform detection mechanism is used to scan and obtain the pallet code on the shallow location, and an image of the goods on the shallow location pallet is captured to obtain the shallow location material box information.

[0081] Specifically, when there are goods in both deep and shallow cargo spaces, it includes:

[0082] Shallow-bay cargo handling: As automated warehousing equipment, the stacker crane receives instructions from the warehouse management system, activates its positioning system, and precisely maneuvers to shallow, double-deep bays. Its robotic arm is equipped with specialized cargo-grabbing devices, such as suction cups or grippers, that adjust to the packaging type and size of the goods. To grasp the goods, the robotic arm descends precisely, and the gripper securely grasps the goods on the shallow-bay pallet. The goods are then smoothly lifted and moved to the loading platform along a pre-set path. The entire handling process is precisely controlled by automated programs, ensuring safe and stable transfer of goods.

[0083] Deep-slot information collection: The long-distance scanning equipment installed on the stacker crane begins operation. This type of equipment often uses laser scanning technology. The equipment emits a fan-shaped laser beam to scan the pallet code on the deep-slot cargo area. The pallet code is generally affixed to the side or top of the pallet and has a highly reflective coating. When the laser beam hits the pallet code, some of the light is reflected back to the scanning equipment. Based on the time delay and angle change of the reflected light, the equipment calculates the location of the pallet code and identifies the information contained therein. At the same time, an industrial-grade high-definition camera installed on the stacker crane captures images of the cargo on the deep-slot pallets. The camera has zoom and autofocus functions, and can adjust shooting parameters according to the distance from the cargo to obtain clear images of the cargo. The image data is transmitted to the stacker crane control system via the high-speed data transmission line within the stacker crane, and then forwarded by the control system to the back-end data processing center.

[0084] Shallow-bay Pallet Information Collection: The loading platform inspection unit is equipped with specialized scanning equipment that operates similarly to short-range barcode scanners. When shallow-bay cargo is placed on the loading platform, the scanner rapidly scans the pallet code and obtains the pallet code information. Simultaneously, multiple low-angle cameras positioned on the loading platform capture the cargo on the shallow-bay pallet from different angles. These cameras work together to capture comprehensive image information of the cargo, including the bottom of the material boxes.

[0085] Step S2013: When the target inventory storage location is a double-deep location and there are goods in the deep location but no goods in the shallow location, a long-distance scanning device is directly used to scan the pallet code on the deep location, and an image of the goods on the deep location pallet is captured to obtain the deep location material box information.

[0086] Specifically, when there are goods in the deep cargo space and no goods in the shallow cargo space, there is no need to move the goods, and the long-distance scanning device on the stacker can be used to scan the deep cargo space pallet code directly. The scanning device adjusts the emission angle and scanning range of the laser beam to ensure that it can cover the area where the deep cargo space pallet code is located. Consistent with the scanning principle mentioned above when there are goods in the deep cargo space and there are goods in the shallow cargo space, the pallet code information is identified through laser reflection. Then, the camera on the stacker captures images of the goods on the deep cargo space pallet. Since there is no goods in the shallow cargo space, the camera can focus more on obtaining detailed images of the deep cargo space goods, reduce interference factors, and complete the acquisition of the deep cargo space material box information. The entire material box information includes material LOGO information and / or material text information; the process is simple and efficient, which reduces unnecessary equipment operations and improves information collection efficiency.

[0087] Step S202: Use deep learning recognition algorithm to identify the material box information and determine the material specifications. Specifically, the material box information includes material LOGO information and / or material text information; Figure 4As shown, the data acquisition terminal also includes a high-definition industrial camera 2, a TOF sensor 1, a pan-tilt camera 3 and a code scanning sensor 4.

[0088] The above step S202 includes:

[0089] Step S2021: Use a deep learning recognition algorithm to train the material LOGO information and / or material text information to identify the material's specific pattern features and / or text identification features.

[0090] Specifically, such as Figure 4 As shown in the figure, when the pallet moves to the designated detection starting point, the TOF sensor 1 installed on the edge of the frame will receive the signal, which will be converted into the acquisition start signal of the TOF sensor through the electrical control cabinet. At this time, the TOF sensor starts working and is responsible for collecting the overall data of the box and forming point cloud data. The whole process continues until the entire pallet moves out of the detection area.

[0091] like Figure 4 As shown, a high-definition industrial camera 2 is installed above or on the side of the detection mechanism to capture images of the materials on the pallet (the specific installation method needs to be determined according to the on-site conditions), and a deep learning recognition algorithm is used to identify the characteristic information on the material box. The characteristic information includes but is not limited to the product logo or obvious text identification information. The conveyor belt needs to remain stopped when the material box information is identified. A light source needs to be installed on the data acquisition terminal profile to provide a relatively stable optical environment for the recognition camera. The detection mechanism is also equipped with a pan-tilt camera 3, which is used to view the on-site situation in real time at a fixed point according to the actual needs of the customer.

[0092] A series of preprocessing operations are performed on image data to improve data quality and adapt it to the input requirements of deep learning models. First, image grayscale conversion is performed to reduce the data dimension and reduce the computational complexity of the model. Next, image normalization is performed to adjust the brightness, contrast, and size of the image, bringing all images to a specific size.

[0093] The deep learning recognition algorithm includes a convolutional neural network algorithm. It preprocesses the material box information on the pallet, including images of the top surface and the four side surfaces (front, back, left, and right). This improves data quality and meets the input requirements of the deep learning model. The preprocessed image dataset is divided into a training set and a test set. The constructed convolutional neural network model is trained using the training set. The model parameters are continuously adjusted through the backpropagation algorithm to ensure that the model's predictions on the training set are as close as possible to the annotated material specifications. The trained convolutional neural network model is tested on the test set, and the convolutional neural network model that meets the recognition accuracy threshold is selected as the final deep learning model.

[0094] The preprocessed image dataset is divided into a training set and a test set in a ratio of 70% and 30% respectively. The constructed model is trained using the training set, and the parameters of the model are continuously adjusted through the back-propagation algorithm so that the prediction results of the model on the training set are as close as possible to the annotated material-specific patterns and text identification features and the corresponding material specification information. During the training process, hyperparameters such as the learning rate and the number of iterations are reasonably set. The learning rate determines the step size of the model parameter update and affects the convergence speed of the model; the number of iterations controls the number of learning rounds of the model on the training set. At the same time, the performance of the model is evaluated in real time using the validation set, and the accuracy, loss value and other indicators of the model on the validation set are monitored to prevent the model from overfitting. When the performance indicators of the model on the validation set (such as accuracy) no longer improve or reach the preset performance target, the training is stopped and the trained model weights are saved.

[0095] The deep learning recognition algorithm begins running, performing in-depth analysis and training on the collected material logo and / or text information. The algorithm learns from a large number of samples and identifies specific pattern features and / or textual identification features of the material, such as a brand's unique trademark pattern or the textual features of the material name or model number.

[0096] Step S2022: performing feature comparison based on the material-specific pattern features and / or text identification features with a preset material specification feature library to obtain the material specification.

[0097] Specifically, the identified material's specific pattern features and / or textual identification features are compared one by one with a pre-set material specification feature library. This library stores standard feature information for various known materials. Through algorithmic matching calculations, the material specification that best matches the identified features is found.

[0098] Step S203 , calculating the number of material boxes on the pallet based on the material box information, preset stacking rules corresponding to the material specifications, and the AI ​​algorithm.

[0099] Specifically, the material box information also includes point cloud data of material stacking; the AI ​​algorithm includes a target detection algorithm and an OCR algorithm; the above step S203 includes:

[0100] In step S2031, a target detection algorithm is used to identify the position and outline of the material boxes in the point cloud data of the material palletizing, and based on the position and outline of the material boxes, the actual arrangement and stacking order of the materials are calculated using a stacker full pallet counting method or a loose pallet counting method based on an OCR algorithm.

[0101] Specifically, before inventorying materials, point cloud data of the materials is collected using equipment such as LiDAR. This point cloud data is a collection of a large number of three-dimensional points that reflects the shape and position of the materials.

[0102] The preprocessed point cloud data is input into the object detection algorithm. Common object detection algorithms, such as the deep learning-based PointNet series of algorithms, search for the characteristics of material bins in point cloud data. The algorithm learns from a large number of labeled material bin point cloud data samples to build a model capable of identifying material bins. During operation, the algorithm divides the point cloud data into multiple local regions, extracts features and classifies each region, and determines whether the region belongs to a material bin. By continuously scanning and analyzing the entire point cloud data, the location and outline of the material bin in the point cloud data are ultimately accurately identified. For example, the algorithm can determine the coordinate range of each material bin in three-dimensional space and a rough description of its shape and outline, such as the length, width, and height range of a rectangular parallelepiped.

[0103] The stacker crane's full pallet counting methods include:

[0104] Stacker Crane Operation Data Acquisition: As a stacker crane moves a pallet, its control system records operational data in real time. This data includes the crane's movement path, location information, and fork extension and retraction movements. For example, as the stacker crane moves from the starting position in the aisle to the target location, its displacement in the X, Y, and Z axes is precisely recorded. When the stacker crane picks up a pallet, data such as the fork's extension length and lift height are also collected and transmitted to the system.

[0105] Preset pallet counting rules: Preset pallet counting rules are based on standard palletizing methods for materials. For example, for a specific material, five boxes per layer are stacked on the pallet, for a total of four layers, neatly arranged in rows and columns. Rules may also include spacing requirements between boxes. The system pre-stores these rules in the database, corresponding to different material specifications.

[0106] Calculating the arrangement and stacking order: The location and outline information of the material boxes identified by the object detection algorithm is combined with the stacker crane's operating data and preset pallet counting rules. The position and posture of the stacker crane forks when picking up the pallet, as well as the position of the material boxes in the point cloud data, are used to determine the row and column positions of the material boxes on the pallet. For example, if the stacker crane forks insert horizontally from one side of the pallet, the number of columns of material boxes on the pallet can be determined based on the X-axis position of the material boxes in the point cloud data. The number of stacking layers of material boxes is then calculated based on the height distribution of the material boxes in the Z-axis and the preset stacking height of each layer. In this way, the arrangement and stacking order of the material boxes on the pallet are accurately calculated.

[0107] The bulk pallet counting methods based on OCR algorithms include:

[0108] OCR Algorithm Preparation: Before counting bulk pallets, ensure the OCR algorithm has been trained on the textual markings likely to appear on material boxes. This training data includes common text fonts, sizes, colors, and possible deformations found on various material boxes. Through extensive training, the OCR algorithm can accurately recognize textual information on material boxes, such as part number, batch number, and quantity identification.

[0109] Text Marking Recognition: After capturing an image of a material box, the image data is fed into an OCR algorithm. The algorithm first preprocesses the image, including grayscale conversion, noise reduction, and binarization, to improve text recognition accuracy. Then, a character segmentation algorithm separates the characters from the image and recognizes them individually. For example, if a material box is printed with "Material Model: ABC-123, Quantity: 5," the OCR algorithm can accurately recognize the text and convert it into a computer-readable text format.

[0110] Determine arrangement and stacking based on preset rules: Preset rules specify how to determine the arrangement and stacking of material boxes based on textual identifiers. For example, a rule might dictate that boxes be stacked and categorized by material model, with boxes of the same model placed on the same layer and arranged in chronological order by batch number. Based on information such as the material model and batch number identified by the OCR algorithm and in conjunction with preset rules, the position and stacking order of each box on the pallet are determined. If multiple boxes are identified as having the same material model and consecutive batch numbers, it can be determined that they are stacked and arranged in a specific order on the pallet, thereby determining the arrangement and stacking of the boxes on the entire pallet.

[0111] Step S2032 : Calculate the number of material boxes based on the actual arrangement and stacking order of the materials and the preset stacking rules corresponding to the material specifications.

[0112] Specifically, the system accurately calculates the number of boxes on a pallet based on the material's actual arrangement and stacking sequence, as well as the preset palletizing rules for the material specification. For example, if a certain material specification's palletizing rules require a fixed number of boxes per layer and a fixed number of layers, the total number of boxes can be calculated based on the identified arrangement and stacking.

[0113] Step S204: Bind the material specifications, material box quantity and pallet code, and then compare the pallet code with the upper inventory data to obtain the inventory result. Figure 1 Step S104 of the illustrated embodiment will not be described in detail here.

[0114] The in-warehouse inventory method provided by this embodiment adopts a special information collection method for different types of target inventory locations. For single-depth locations, the short-range code scanning device can accurately and quickly obtain the pallet code, and simultaneously complete the image collection of the goods to obtain the material box information. There is no need for complex equipment scheduling, which greatly shortens the data collection time, adapts to the simple layout characteristics of the single-depth location, and ensures that the data is timely and complete. In the double-depth location scenario, when there are goods in both the deep and shallow locations, the stacker, the long-range scanning device and the loading platform detection mechanism work together. The stacker accurately moves the shallow location goods to create conditions for scanning and image collection at different locations, comprehensively covering the information acquisition of goods at different locations in the double-depth location, and ensuring that no information is missed. When only the deep location in the double-depth location has goods, the long-range scanning device is directly used to flexibly simplify the operating process and improve inventory efficiency. This information acquisition method tailored to the location greatly improves the efficiency and adaptability of the entire inventory process, laying a solid data foundation for subsequent inventory links.

[0115] In this embodiment, a method for counting goods in warehouse is provided, which can be used in the above-mentioned computer device. The computer device is installed with a system for counting goods in warehouse, which includes a data acquisition terminal, Figure 3 FIG. 1 is a flow chart of a method for counting goods in a warehouse according to an embodiment of the present invention. Figure 3 As shown, the process includes the following steps:

[0116] Step S301: Get the pallet code on the target inventory location and the material box information on the corresponding pallet. Figure 2 Step S201 of the illustrated embodiment will not be described in detail here.

[0117] Step S302: Use deep learning recognition algorithm to identify the material box information and determine the material specifications. Figure 2 Step S202 of the illustrated embodiment will not be described in detail here.

[0118] Step S303: Calculate the number of material boxes on the pallet based on the material box information, the preset stacking rules corresponding to the material specifications, and the AI ​​algorithm. Figure 2 Step S203 of the illustrated embodiment will not be described in detail here.

[0119] In step S304, the material specifications, the number of material boxes and the pallet code are bound together, and then the pallet code is compared with the upper-level inventory data to obtain the inventory result.

[0120] Specifically, the above step S304 includes:

[0121] Step S3041: Bind the material specification, material box quantity and pallet code to obtain complete pallet material data.

[0122] Specifically, the determined material specifications and the calculated number of material boxes are associated and bound with the corresponding pallet code to form complete pallet material data, ensuring that each pallet code corresponds to accurate material-related information.

[0123] Step S3042: Obtain the upper-level inventory data corresponding to the pallet code.

[0124] Specifically, the inventory data corresponding to the pallet is accurately obtained from the upper inventory data system based on the pallet code. This data includes the material specifications, quantity and other information of the pallet material recorded in the system.

[0125] In step S3043, the complete data of the pallet material is compared with the upper-level inventory data to obtain the inventory results of the goods that are consistent and the inventory results of the goods that are different.

[0126] Specifically, the complete data of the pallet material is carefully compared with the upper-level inventory data. If all information is completely consistent, the inventory count result is consistent. If there are discrepancies, such as inconsistent material specifications or inconsistent quantities, the inventory count result is discrepant.

[0127] like Figure 4 As shown, in the actual inventory scenario, a barcode scanning sensor 4 is installed at the bottom of the conveyor belt of the detection mechanism. When the signal is received, the pallet code is identified, and the number of material boxes and the material box information are bound to the pallet code. At the data processing terminal, the detected material quantity, material specifications in the material box information and the pallet code are compared with the data sent by the upper level. After the comparison is passed, an OK signal is sent to the electronic control system. If the comparison is abnormal, an NG signal is sent to the electronic control system, and finally it is handed over to manual verification.

[0128] In step S305, when the inventory result of the goods inventory difference is obtained, the current inventory pallet and the goods on the pallet are frozen, and the corresponding handling tasks are retrieved for the goods with inventory difference in the online warehouse management system log. According to the handling tasks, the equipment scheduling operation trajectory and the corresponding timestamp are found in the online equipment scheduling system through the key information matching algorithm. Combined with the images taken by the camera or camera in the current inventory stereo library, the image capture technology is combined with the deep learning algorithm to capture the image features, and the reasons for the goods inventory difference are found based on the image features. The reasons for the goods inventory difference include manual handling, box drop caused by equipment failure in the middle, or other reasons other than manual handling and box drop caused by equipment failure in the middle.

[0129] Specifically, the above step S305 includes:

[0130] Freezing Goods: Once the system detects a discrepancy in the inventory count, it immediately issues a command to freeze the pallet and its contents. This operation is accomplished through communication between the warehouse management system and related equipment control systems. For example, the system controls the loading and unloading equipment in the pallet's shelf area to stop handling the pallet. This prevents the goods from being mishandled or redeployed before the problem is identified, ensuring that the discrepant goods remain in the same state as at the time of the inventory count, allowing for accurate tracing of the issue later.

[0131] Retrieving Handling Tasks: Utilizing the powerful search functionality within the online warehouse management system log, we use material information and pallet codes associated with inventory discrepancies as key search criteria. The log details every handling task from the moment the goods enter the warehouse to the inventory count, including the time, starting and ending locations, and the number of the handling equipment involved. Through precise searching, we can quickly locate all handling task records related to the batch of discrepant goods, providing foundational data for subsequent analysis.

[0132] Matching equipment trajectory: Based on the retrieved handling tasks, the key information matching algorithm comes into play. The algorithm extracts key information from the handling task records, such as the handling equipment number, handling time range, etc. Using this information, the algorithm conducts an in-depth search in the online equipment scheduling system. The equipment scheduling system stores detailed operation trajectory data for each device, including the coordinate position, operating speed, operating actions, etc. of the equipment at different time points. The algorithm accurately matches the key information in the handling task with the data in the equipment scheduling system, and finds the complete equipment scheduling operation trajectory and corresponding timestamp of the corresponding handling equipment during the execution of the handling task. For example, through the equipment number and handling time range, the operation path of a stacker from one cargo location to another within a specific time period and the time point when each key action occurs can be accurately matched.

[0133] Analyze image features: At the same time, the system calls the image data captured by the camera or camera in the current inventory stereo warehouse. These image data cover the storage area and transportation path of the goods in the warehouse. Image capture technology starts to work. It accurately captures key image frames related to cargo handling from a large number of images based on the equipment scheduling operation trajectory and timestamp. The captured image frames are input into the deep learning algorithm, which conducts in-depth analysis and learning through features such as the shape, color, and position changes of the goods in the image. The algorithm has been pre-trained with a large number of image samples containing different situations (such as normal handling, box drop, human intervention, etc.) and has the ability to identify various abnormal image features. For example, by analyzing the integrity of the cargo outline, it can be determined whether a box has been dropped, and by detecting whether there are signs of human activity in the image, it can be determined whether there has been human handling behavior.

[0134] Determining the cause of discrepancies: Based on the results of image feature analysis using deep learning algorithms, combined with information about handling tasks and equipment trajectory, the system can accurately determine the cause of inventory discrepancies. If the image shows equipment failure during handling, resulting in dropped cargo, and the equipment trajectory data also shows abnormal equipment operation at that point in time, the discrepancy is determined to be caused by equipment failure mid-transit, resulting in dropped containers. If the image captures footage of someone manually moving cargo near the cargo area, and the handling task and timestamp match the image time, manual handling can be determined to be the cause of the discrepancy. After eliminating manual handling and mid-transit equipment failure as possible causes of dropped containers, further analysis based on image features and other relevant data can identify other causes beyond the two aforementioned scenarios, such as abnormal damage to the cargo packaging, which may have contributed to the quantity discrepancy.

[0135] In step S306, a global matching retrieval algorithm is used in the online warehouse management system to perform a global search on the virtual storage location. If the virtual storage location at the time when the goods first enter the system does not change from the actual physical storage location at the current inventory time, and the storage location does not participate in order allocation, it is considered that the goods in the virtual storage location in the warehouse management system are consistent with the goods in the actual physical storage location, and the inventory result is directly based on the stored data of the warehouse management system.

[0136] Specifically, this step belongs to the case where the virtual cargo location has not changed and has not participated in order allocation, and includes the following steps:

[0137] Information Comparison: The algorithm performs a detailed comparison of the extracted virtual location information with the corresponding physical location information. During this comparison, the algorithm first checks whether the virtual location number matches the physical location number. If so, the algorithm then verifies whether the virtual location at the time of the initial inventory entry is the same as the physical location at the time of the current inventory count, and verifies whether the location has not been used for any order allocation. For example, the comparison reveals that the virtual location C-003 corresponds to the actual physical location 3, and that this location has not changed since the initial entry and has not been used for any order allocation.

[0138] Directly use system data: When the above conditions are met, the system directly determines that the goods in the virtual location in the warehouse management system are consistent with the goods in the actual physical location. In this case, the inventory result does not require additional physical inventory operations, but directly uses the storage data corresponding to the virtual location in the warehouse management system. This storage data contains detailed information such as material specifications and quantity. The system directly outputs this data as the inventory result, greatly improving inventory efficiency and reducing unnecessary labor and material resources.

[0139] Step S307: When the virtual cargo location at the time of the first online warehousing entry changes from the actual physical cargo location at the current inventory time and / or the cargo location participates in order allocation, the goods at the actual physical cargo location are counted using a stacker crane or a loose pallet counting method based on an OCR algorithm. The inventory result is subject to the actual inventory result.

[0140] Specifically, if the information comparison process reveals a change between the virtual location at the time of the first online entry and the actual physical location at the time of the current inventory count—for example, virtual location D-004 originally corresponded to physical location 4, but now corresponds to physical location 8—or if the location is involved in order allocation, such as when some of the goods in the location are partially withdrawn for a specific order shipment—then an inventory process is immediately triggered for the goods in the actual physical location, including either a full pallet count using a stacker crane or a loose pallet count based on an OCR algorithm.

[0141] Execute the stacker crane full pallet counting method: If the stacker crane full pallet counting method is selected, the stacker crane moves along a pre-set path to the target physical location. The stacker crane's scanner reads the pallet code and uses its onboard image acquisition device to capture images of the goods on the pallet to obtain the material box information. As previously mentioned, the target detection algorithm analyzes the point cloud data of the material stack (if available). Combined with the stacker crane's operating data and the pre-set full pallet counting rules, the arrangement and stacking order of the material boxes on the pallet are calculated. The number of material boxes is then calculated based on the pre-set stacking rules corresponding to the material specifications.

[0142] Executing an OCR-based bulk pallet counting method: If using an OCR-based bulk pallet counting method, first capture images of the goods on the actual physical storage location. This captured image is fed into the OCR algorithm, which recognizes textual information on the material boxes, such as the material model, batch number, and quantity. Combined with pre-set rules (e.g., stacking by material model, or sorting by batch number for the same model), the arrangement and stacking of the material boxes are determined. Finally, the number of material boxes is calculated based on the pre-set stacking rules corresponding to the material specifications.

[0143] Actual inventory count results prevail: Regardless of the actual inventory count method used, the final inventory count results are based on the information such as material specifications and number of boxes and cases obtained from the actual inventory count. The system updates the actual inventory count results to the online warehouse management system, ensuring consistency between system data and actual inventory status, providing an accurate basis for subsequent warehouse management decisions.

[0144] like Figure 5As shown, in an actual inventory count scenario, after the stacker is in place, the high-definition camera installed on the stacker takes a photo of the goods in the shallow storage area. Once the photo is taken, a photo completion command is sent to the upper control system, which then controls the stacker's shifting. Simultaneously, the control system sends a scan code command (operating in inventory mode) to the stacker. The short-range scanning device installed on the stacker scans the pallet code of the goods in the shallow storage area, obtaining a pallet code. After decoding, the pallet code is compared with the inventory information code. If the comparison results are consistent, the inventory count is complete. If the comparison results are inconsistent, an error signal is sent.

[0145] The comparison results generate an inventory effect diagram as follows Figure 6 As shown, inconsistent anomalies are also visually rechecked. If the inventory scan results are consistent, the inventory location is displayed in different colors, indicating a consistent inventory result (TRUE) and an abnormal result (FLASE). After the inventory is completed, click the abnormal cell to recheck the inventory information based on the inventory result image information.

[0146] The inventory review interface generated by the inventory system is shown in the following diagram: Figure 7 As shown, the current location pallet picture can be viewed in the right cell. The current location pallet picture contains pallet decoding information, that is, the pallet code picture on the left corresponds to the pallet one-dimensional code 3 on the right, the execution result (TRUE or FLASE), barcode angle, symbol level, barcode center address, coordinate system value, and pallet code string number and other information.

[0147] It should be noted that, depending on the inventory business type, inventory tasks include annual inventory (credit location designation or goods designation), periodic inventory (goods designation), random inventory (credit location designation or goods designation), shelf inventory (location designation), low inventory inventory (location designation), zero inventory inventory (location designation) and dynamic inventory (location designation).

[0148] Among them, annual inventory count (credit position or goods specified) and cycle inventory count (goods specified) are regular inventory counts. Regular inventory counts can be used for existing commodity inventory. Regular inventory counts include:

[0149] Inventory method: Inventory by area or by product category.

[0150] Inventory execution: supports written inventory lists and scanning inventory using RF (Radio Frequency) handheld devices.

[0151] Cycle Counting: To reduce the workload of regular inventory counting and promptly identify inventory discrepancies, Westernacher recommends categorizing different products into ABC categories and setting up periodic inventory counting for selected products based on these categories.

[0152] When inventory operations conflict with other warehouse operations, in order to reduce the conflict with other warehouse operations, EWM (Extended Warehouse Management) provides two different methods to resolve the conflict.

[0153] When creating a stocktaking voucher, the system automatically locks the target location, preventing any operations from involving that location. If a location is unlocked and a corresponding picking task is generated for that location after the stocktaking voucher is created, and the system posts the stocktaking results, if the posted quantity is less than the picked quantity, the system will prompt that the posting cannot proceed, requiring the operator to consider whether to cancel the stocktaking voucher or the picking task for further analysis.

[0154] After the inventory count is complete, the warehouse manager can review the results. If any discrepancies occur, they can enter an exception code to trigger a workflow, which sends a discrepancy report to the inventory controller via email. After obtaining the inventory controller's approval, the inventory count is posted and the inventory in EWM is updated. The workflow triggered by the exception code only provides a simple information sending function; if an approval process is required, enhanced development is required.

[0155] Integrated inventory with automated equipment: The three-dimensional warehouse is integrated with the camera of the hardware equipment. After the EWM generates a warehouse task, it is distributed to the WCS (Warehouse Control System) for scanning and data collection. The data is automatically fed back to the EWM for inventory result entry, generating the corresponding inventory actual number record, performing difference comparison and final inventory posting.

[0156] The AGV (Automated Guided Vehicle) warehouse area can use PDA (Personal Digital Assistant) inventory or goods-to-person inventory. If it is goods-to-person inventory, the inventory task can be triggered in the form of a workbench to allow the AGV to go straight.

[0157] The in-stock inventory method provided by this embodiment binds the material specifications, the number of material boxes and the pallet code to form complete data for the pallet material, and establishes a closely related and comprehensive information set. This allows the material information of each pallet to have a unique identifier and detailed attribute record, which is convenient for subsequent query, traceability and management. By obtaining the upper-level inventory data corresponding to the pallet code and comparing it with the complete data, inventory differences can be quickly and intuitively discovered. Clearly distinguishing between the inventory results of goods that are consistent and the inventory results of goods that are different provides clear inventory status information for the decision-making level of the enterprise. Once a difference in the inventory of goods is found, the current inventory pallet and goods are immediately frozen, which can effectively prevent the problem goods from continuing to participate in the warehouse circulation and prevent the difference problem from further expanding. The handling tasks corresponding to the inventory differences of goods are retrieved through the online warehouse management system log, and the equipment scheduling operation track and timestamp are obtained from the online equipment scheduling system with the help of the key information matching algorithm. This traceability method can accurately restore the entire transportation process of goods from the time they enter the warehouse to the time before inventory. The time and path of each operation link are clearly traceable. By using images taken by cameras or cameras in the stereoscopic warehouse, combined with image capture technology and deep learning algorithms to extract image features, the reasons for differences in goods inventory can be accurately located.

[0158] In this embodiment, a system for counting goods in warehouse is also provided, which includes:

[0159] The data collection terminal is used to obtain the pallet code on the target inventory location and the material box information on the corresponding pallet. Figure 4 As shown, the data acquisition terminal further includes a high-definition industrial camera 2, a TOF sensor 1, a pan-tilt camera 3, and a barcode scanning sensor 4. The further functional description of each of the above components is the same as that of the corresponding embodiment above, and will not be repeated here.

[0160] The data processing industrial computer is connected to the data acquisition terminal. The industrial computer is used to use a deep learning recognition algorithm to identify the material box information and determine the material specifications; calculate the number of material boxes on the pallet based on the preset stacking rules corresponding to the material box information and material specifications and the AI ​​algorithm; bind the material specifications and the number of material boxes with the pallet code, and then compare the pallet code with the upper inventory data to obtain the goods inventory result.

[0161] The data processing industrial computer needs to be placed near the data acquisition terminal. The data acquisition terminal equipment does not contain a data processing center. The collected data needs to be converted from digital to analog (data acquisition module) and then transmitted to the software algorithm in the industrial computer for processing and output of the final result. The result is compared with the information in the upper database, and the abnormal situation is sent to the electronic control system to facilitate the subsequent operation of the electronic control system. At the same time, when the box information is recognized, the assembly line needs to stop and can operate normally after taking pictures.

[0162] This embodiment also provides a device for inventorying goods in a warehouse. This device is used to implement the above-mentioned embodiments and preferred embodiments. Details already described will not be repeated here. As used below, the term "module" may refer to a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.

[0163] This embodiment provides a device for counting goods in a warehouse, such as Figure 8 Shown, including:

[0164] The data collection module 801 is used to obtain the pallet code on the target inventory location and the material box information on the corresponding pallet.

[0165] The material box information recognition module 802 is used to use a deep learning recognition algorithm to recognize the material box information and determine the material specifications.

[0166] The material box quantity calculation module 803 is used to calculate the number of material boxes on the pallet based on the material box information, the preset stacking rules corresponding to the material specifications and the AI ​​algorithm.

[0167] The goods comparison and inventory module 804 is used to bind the material specifications, material box quantity and pallet code, and then compare the pallet code with the upper inventory data to obtain the goods inventory result.

[0168] In some optional embodiments, the target inventory storage locations include single-depth locations and double-depth locations; the double-depth locations include deep locations and shallow locations; and the data collection module 801 includes:

[0169] The first data acquisition unit is used to use a short-range scanning device to scan the pallet code on the single-depth position when the target inventory location is a single-depth position, and to capture images of the goods on the pallet to obtain the single-depth material box information.

[0170] The second data acquisition unit is used to use a stacker to move the goods on the shallow cargo space to the loading platform when the target inventory location is a double-deep location and there are goods placed in both the deep cargo space and the shallow cargo space, and use a long-distance scanning device installed on the stacker to scan and obtain the pallet code on the deep cargo space, and perform image acquisition on the goods on the deep cargo space pallet to obtain the deep cargo space material box information; at the same time, use the scanning device of the loading platform detection mechanism to scan and obtain the pallet code on the shallow cargo space, and perform image acquisition on the goods on the shallow cargo space pallet to obtain the shallow cargo space material box information.

[0171] The third data acquisition unit is used to directly use a long-distance scanning device to scan the pallet code on the deep cargo position when the target inventory location is a double-deep position and there are goods in the deep cargo position but no goods in the shallow cargo position, and to capture images of the goods on the deep cargo position pallet to obtain the deep cargo position material box information.

[0172] The material box information identification module 802 includes:

[0173] The pattern feature and text identification feature recognition unit is used to train the material LOGO information and / or material text information using a deep learning recognition algorithm to identify the material's specific pattern features and / or text identification features.

[0174] The material specification determination unit is used to compare the material's specific pattern features and / or text identification features with a preset material specification feature library to obtain the material specification.

[0175] In an optional embodiment, the material box information further includes point cloud data of material stacking; the AI ​​algorithm includes a target detection algorithm and an OCR algorithm; and the material box quantity calculation module 803 includes:

[0176] The AI ​​algorithm recognition unit is used to use the target detection algorithm to identify the position and outline of the material boxes in the point cloud data of material palletizing, and based on the position and outline of the material boxes, use the stacker crane whole pallet counting method or the OCR algorithm-based loose pallet counting method to calculate the actual arrangement and stacking order of the materials.

[0177] The material box quantity calculation unit is used to calculate the number of material boxes based on the actual material arrangement, stacking order and preset stacking rules corresponding to the material specifications.

[0178] In an optional embodiment, the goods comparison and inventory module 804 includes:

[0179] The binding unit is used to bind the material specifications, material box quantity and pallet code to obtain complete pallet material data.

[0180] The upper-level inventory data acquisition unit is used to acquire the upper-level inventory data corresponding to the pallet code.

[0181] The inventory comparison unit is used to compare the complete data of the pallet material with the upper inventory data to obtain the inventory results of the goods that are consistent and the inventory results of the goods that are different.

[0182] In an optional embodiment, the goods inventory counting device further includes:

[0183] The inventory module for the unchanged storage location and the storage location that does not participate in order allocation is used to perform a global search on the virtual storage location using a global matching retrieval algorithm in the online warehouse management system. When the virtual storage location at the time when the goods in the system are first put into storage does not change from the actual physical storage location at the current inventory time, and the storage location does not participate in order allocation, it is considered that the goods in the virtual storage location in the warehouse management system are consistent with the goods in the actual physical storage location, and the inventory result is directly based on the storage data of the warehouse management system.

[0184] The inventory counting module for changes in storage locations and / or storage locations participating in order allocation is used to execute the stacker crane full pallet counting method or the OCR algorithm-based loose pallet counting method on the actual physical storage location's goods when the virtual storage location at the time of the first online entry changes and / or the storage location participates in order allocation. The inventory counting result is based on the actual inventory counting result.

[0185] The inventory reason tracking code module is used to freeze the current inventory pallet and the goods on the pallet when the inventory result of the goods inventory difference is obtained, and at the same time retrieve the corresponding handling tasks for the goods with inventory difference in the online warehouse management system log, and find the equipment scheduling operation trajectory and the corresponding timestamp in the online equipment scheduling system through the key information matching algorithm according to the handling task. Combined with the images taken by the camera or camera in the current inventory stereo library, the image capture technology is combined with the deep learning algorithm to capture the image features, and the reasons for the inventory differences are found based on the image features. The reasons for the inventory differences include manual handling, box drop caused by equipment failure in the middle, or other reasons other than manual handling and box drop caused by equipment failure in the middle.

[0186] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0187] The inventory counting device for goods in the warehouse in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0188] The embodiment of the present invention also provides a computer device having the above Figure 8 The device for counting goods in the warehouse is shown.

[0189] See also Figure 9 , Figure 9 is a structural diagram of a computer device provided by an optional embodiment of the present invention, such as Figure 9As shown, the computer device includes: one or more processors 10, memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components utilize different buses to communicate with each other and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in the memory or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Equally, multiple computer devices can be connected, and each device provides part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 9 A processor 10 is taken as an example.

[0190] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0191] The memory 20 stores instructions that can be executed by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0192] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created based on the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0193] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0194] The computer device further includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 9 The bus connection is taken as an example.

[0195] The input device 30 can receive input digital or character information and generate key signal input related to user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a trackpad, a touch pad, an indicator stick, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 can include a display device, an auxiliary lighting device (e.g., an LED), and a tactile feedback device (e.g., a vibration motor). The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display, and a plasma display. In some optional embodiments, the display device can be a touch screen.

[0196] The embodiment of the present invention also provides a computer-readable storage medium. The above-mentioned method according to the embodiment of the present invention can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0197] A portion of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the form in which the computer program instruction exists in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc. Accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium that can be accessed by the computer.

[0198] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for inventorying goods in warehouse, characterized in that: The method comprises: Obtain the pallet code on the target inventory location and the material box information on the corresponding pallet; Using a deep learning recognition algorithm to identify the material box information and determine the material specifications; Calculate the number of material boxes on the pallet based on the material box information, preset stacking rules corresponding to the material specifications, and AI algorithms; After binding the material specifications and the number of material boxes with the pallet code, the pallet code is compared with the upper inventory data to obtain the inventory result.

2. The method according to claim 1, characterized in that The target inventory storage locations include single-depth locations and double-depth locations; the double-depth locations include deep locations and shallow locations; The step of obtaining the pallet code on the target inventory location and the material box information on the corresponding pallet includes: When the target inventory location is a single-deep location, a close-range barcode scanner is used to scan the pallet code on the single-deep location, and an image of the goods on the pallet is captured to obtain the single-deep location material box information; When the target inventory storage location is a double-deep location and there are goods placed in both the deep and shallow locations, a stacker is used to move the goods in the shallow location to the loading platform, and a long-distance scanning device provided on the stacker is used to scan and obtain the pallet code on the deep location, and an image of the goods on the deep location pallet is captured to obtain the deep location material box information; at the same time, a scanning device of the loading platform detection mechanism is used to scan and obtain the pallet code on the shallow location, and an image of the goods on the shallow location pallet is captured to obtain the shallow location material box information; When the target inventory location is double-deep and there are goods in the deep location but no goods in the shallow location, a long-distance scanning device is directly used to scan the pallet code on the deep location, and the image of the goods on the deep location pallet is captured to obtain the deep location material box information.

3. The method according to claim 1 or 2, characterized in that The material box information includes material LOGO information and / or material text information; The deep learning recognition algorithm is used to identify the material box information and determine the material specifications, including: Use deep learning recognition algorithms to train material logo information and / or material text information to identify material-specific pattern features and / or text identification features; The material specifications are obtained by performing feature comparison based on the material-specific pattern features and / or text identification features and a preset material specification feature library.

4. The method according to claim 1, wherein The material box information also includes point cloud data of material stacking; the AI ​​algorithm includes a target detection algorithm and an OCR algorithm; The calculation of the number of material boxes on the pallet based on the material box information, the preset stacking rules corresponding to the material specifications and the AI ​​algorithm includes: Use target detection algorithms to identify the location and outline of material boxes in the point cloud data of material palletizing, and calculate the actual arrangement and stacking order of materials based on the location and outline of the material boxes using the stacker crane full pallet counting method or the OCR algorithm-based loose pallet counting method; The number of material boxes is calculated based on the actual material arrangement, stacking order and preset stacking rules corresponding to the material specifications.

5. The method according to claim 1, wherein The method further comprises: In the online warehouse management system, a global matching retrieval algorithm is used to perform a global search of virtual storage locations. If the virtual storage location at the time the goods first enter the system does not change from the actual physical storage location at the time of the current inventory count, and the storage location has not been used for order allocation, the goods in the virtual storage location in the warehouse management system are considered to be consistent with the goods in the actual physical storage location. The inventory result is directly based on the data stored in the warehouse management system. When the virtual storage location at the time of first online warehousing changes from the actual physical storage location at the current inventory time and / or the storage location is involved in order allocation, the goods at the actual physical storage location will be counted using the stacker crane full pallet method or the OCR algorithm-based loose pallet method. The inventory result will be based on the actual inventory result.

6. The method according to claim 1, characterized in that After binding the material specifications and material box quantity with the pallet code, the pallet code is compared with the upper inventory data to obtain the inventory results, including: Bind the material specification, material box quantity and pallet code to obtain complete pallet material data; Obtaining upper-level inventory data corresponding to the pallet code; The complete data of the pallet material is compared with the upper inventory data to obtain the inventory results of the goods that are consistent and the inventory results of the goods that are different.

7. The method according to claim 6, characterized in that The method further comprises: When the inventory results of the goods inventory differences are obtained, the current inventory pallet and the goods on the pallet are frozen, and the corresponding handling tasks are retrieved for the goods with inventory differences in the online warehouse management system log. According to the handling tasks, the equipment scheduling operation trajectory and the corresponding timestamp are found in the online equipment scheduling system through the key information matching algorithm. Combined with the images taken by the camera or camera in the current inventory stereo library, the image capture technology is combined with the deep learning algorithm to capture the image features, and the reasons for the differences in the goods inventory are found based on the image features. The reasons for the differences in the goods inventory include manual handling, box drop caused by equipment failure in the middle, or other reasons other than manual handling and box drop caused by equipment failure in the middle.

8. A system for counting goods in warehouse, characterized in that: The system comprises: Data collection terminal, used to obtain the pallet code on the target inventory location and the material box information on the corresponding pallet; A data processing industrial computer is connected to the data acquisition terminal, and is used to use a deep learning recognition algorithm to identify the material box information and determine the material specifications; calculate the number of material boxes on the pallet based on the preset stacking rules corresponding to the material box information and material specifications and the AI ​​algorithm; and after binding the material specifications and the number of material boxes with the pallet code, compare the pallet code with the upper inventory data to obtain the goods inventory result.

9. A device for counting goods in warehouse, characterized in that: The device comprises: Data collection module, used to obtain the pallet code on the target inventory location and the material box information on the corresponding pallet; A material box information recognition module is used to identify the material box information using a deep learning recognition algorithm to determine the material specifications; The material box quantity calculation module is used to calculate the number of material boxes on the pallet based on the material box information, preset stacking rules corresponding to the material specifications, and AI algorithms; The goods comparison and inventory module is used to bind the material specifications, the number of material boxes and the pallet code, and then compare the pallet code with the upper inventory data to obtain the goods inventory result.

10. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the inventory counting method for goods in warehouse according to any one of claims 1 to 7 by executing the computer instructions.

Images