Inventory method for three-dimensional storage of electric power materials
Through the integrated image acquisition and weight acquisition device, combined with the convolutional neural network model and correction algorithm, the intelligent automatic inventory of the three-dimensional warehouse of power materials is realized, solving the problems of low efficiency and safety risks of traditional inventory methods, and improving inventory efficiency and accuracy optimized inventory management.
Patent Information
- Application Number
- CN202411085453.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-08
- Publication Date
- 2024-11-12
AI Technical Summary
The traditional three-dimensional warehouse inventory method of power supplies relies on manual operations, is inefficient, error-prone and has security risks, and the existing intelligent methods have not been fully automated.
The integrated image acquisition device, weight acquisition device and data processing device are adopted, combined with the pre-trained convolutional neural network model and correction algorithm to realize intelligent and automated inventory of materials in the warehouse, automatically identify the types of materials and accurately weigh them.
It significantly improves inventory efficiency and accuracy, reduces labor costs and security risks, optimizes inventory management and operation decisions, and provides strong technical support.
Smart Images

Figure CN118929039A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of electric power materials, and in particular to an inventory method for three-dimensional storage of electric power materials. Background Art
[0002] The inventory of power material warehouses is a key link in the operation and management of power companies. Through inventory, companies can ensure the accuracy and real-time nature of material information in the warehouse, thereby optimizing inventory management and avoiding excess or shortage of materials. This can not only reduce inventory costs and improve capital turnover, but also protect corporate property safety by promptly discovering and handling material losses and thefts. In addition, inventory data can also provide strong support for corporate procurement plans, budget formulation, and management decisions, helping companies better predict material needs, optimize resource allocation, and improve operational efficiency, thereby promoting the stable operation and sustainable development of power companies.
[0003] The traditional inventory method mainly relies on manual work, and the staff needs to check the materials in each warehouse one by one. This method has obvious defects and shortcomings: first, manual inventory is time-consuming and labor-intensive, and the efficiency is low; second, manual operation is prone to errors, resulting in inaccurate data; finally, there are certain safety risks in the high-altitude operations and other links during the inventory process. With the advancement of science and technology, intelligent material inventory methods have emerged and become an inevitable trend in the three-dimensional storage management of power materials. The intelligent inventory method does not require direct human participation, which not only ensures the safety of personnel, but also realizes the automation and precision of the inventory of materials in the warehouse, greatly improving the efficiency and accuracy of the inventory.
[0004] In the prior art, CN109902999A proposed an intelligent inventory method for power materials based on AR technology, which uses AR glasses to identify QR code labels and real-life images, guide inventory personnel to conduct a streamlined inventory, and archive inventory information. Although this method has improved management efficiency to a certain extent, it still relies on manual operation and has not been fully automated. In addition, the use of AR glasses has high requirements on ambient light, image clarity and other conditions, and is more restricted. Summary of the invention
[0005] The problem to be solved by the present invention is an inventory method for three-dimensional storage of power materials. By integrating an image acquisition device, a weight acquisition device and a data processing device, an intelligent and automated inventory of materials in the warehouse is realized. Through the pre-trained convolutional neural network model and correction algorithm, the types of materials can be automatically identified and accurately weighed, which greatly improves the efficiency and accuracy of inventory, reduces labor costs and safety risks, and provides strong technical support for the inventory management and operational decisions of power companies.
[0006] In order to solve the above problems, the present invention provides an inventory method for three-dimensional storage of electric power materials, comprising the following steps:
[0007] Image acquisition and transmission: Use the image acquisition device installed on the stacker loading platform to collect the material box image, and transmit the image data to the data processing device through the wireless communication module;
[0008] Image processing and recognition: The data processing device processes the received images, uses the pre-trained convolutional neural network model to automatically identify the type of material box, and obtains material information through the QR code parsing algorithm;
[0009] Weighing and data transmission: Use the weight acquisition device installed at the end of the stacker robot arm to weigh the material box and transmit the weighing data to the data processing device. The weight acquisition device performs error calibration before inventory counting. The error coefficient is determined by weighing a variety of standard weight objects, so that the stacker can accurately obtain the weight of the object;
[0010] Weight calculation and comparison: The data processing device calculates the actual weight of the material using the weight calculation and correction algorithm based on the image analysis and weighing data; the data processing device compares the calculated material weight with the material quantity in the order to realize the material inventory;
[0011] Data analysis: The data analysis of the data processing device includes three parts: image data analysis, weight data analysis and inventory analysis. These analyses ensure the accuracy of the inventory results. Specifically, the following steps are included:
[0012] 1. Image data analysis: using deep learning models to analyze image data;
[0013] 2.Weight data analysis: analyzing weighing data through weight calculation and correction algorithms;
[0014] 3. Inventory analysis: compare and confirm the inventory results by integrating image data and weight data.
[0015] Preferably, when taking inventory of materials, the stacker lifts up the material box, measures the overall weight of the material box when it is in a stable state, then puts the material box back to measure the reset weight, and ensures the accuracy of the weighing module by comparing the two weighing results.
[0016] Preferably, the data processing device uses a convolutional neural network model to automatically identify the type of the material box on the returned material box image, specifically comprising the following steps:
[0017] a. Image preprocessing, including image denoising, enhancement and edge detection;
[0018] b. Feature extraction, extracting high-level features of the image through multiple convolutional layers;
[0019] c. Image matching: matching the extracted image features with known features in the database to identify the type of bin;
[0020] d. Use the trained CNN model to classify the images collected by the image acquisition device and output the corresponding bin type label.
[0021] Preferably, the two-dimensional code parsing algorithm is used to quickly and accurately locate and decode the two-dimensional code from the high-definition image to obtain material information and order information, which specifically includes the following steps:
[0022] a. QR code positioning: quickly determine the QR code position through image processing algorithms;
[0023] b. QR code decoding: use the QR code parsing function to parse the data in the QR code.
[0024] The two-dimensional code parsing algorithm specifically includes the following steps:
[0025] a. Use OpenCV library to perform image preprocessing, including grayscale, binarization, morphological transformation and other steps to improve the accuracy of QR code recognition;
[0026] b. Use ZBar 2D code recognition library to decode the preprocessed image and extract the information in the 2D code;
[0027] c. Parse the data in the QR code and convert it into usable material and order information.
[0028] Preferably, the weight calculation and correction algorithm comprises the following steps:
[0029] a. Data calibration: calibrate the weighing module before weighing, and determine the error coefficient by weighing a variety of standard weight objects;
[0030] b. Weight calculation, use the following formula to calculate the actual weight of the material:
[0031] Wactual=WVmea.sured-Wtare
[0032] Among them, Wactual is the actual weight of the material, Wmeasured is the weight measured by the weighing module, and Wtare is the tare weight of the material box;
[0033] c. Error correction, correct the calculation results by the following correction formula:
[0034] Wcorrecteedd=Wactual×(1+∈)
[0035] Among them, ∈ is the error coefficient.
[0036] Preferably, the convolutional neural network model is a target detection model YOLO or SSD, which is used to locate and identify multiple bins in a high-definition image, and outputs the location information and type labels of the bins by inputting images collected by an image acquisition device, specifically including the following steps:
[0037] a. Use the YOLO model to detect objects in the image and locate multiple bins;
[0038] b. Use the SSD model to classify the located bins and output category labels.
[0039] Preferably, the data processing device compares the inventory result with the quantity of materials in the order to determine the accuracy of the inventory and generates an inventory report, which specifically includes the following steps:
[0040] a. Compare the inventory results and compare the calculated material weight with the order quantity;
[0041] b. Generate an inventory report, including inventory results, comparison results, error analysis and other information.
[0042] Preferably, the CNN model for material type identification specifically includes the following steps:
[0043] a. Use a large-scale bin image dataset for pre-training and use data augmentation techniques to expand the dataset, including operations such as rotation, scaling, and translation;
[0044] b. Use ResNet and Inception advanced deep convolutional neural network architectures for model training and optimize loss functions to improve recognition accuracy;
[0045] c. Through transfer learning methods, the pre-trained model is applied to the specific bin image classification task and fine-tuned to adapt to the actual application scenario.
[0046] Preferably, the correction algorithm for weighing data analysis specifically includes the following steps:
[0047] a. Establish the error model of the weighing module and calculate the error coefficient by weighing standard weight objects multiple times;
[0048] b. Use polynomial fitting method to model and correct weighing errors and optimize weighing results;
[0049] c. Compare the corrected weighing results with the actual material weight to evaluate the accuracy and robustness of the correction algorithm.
[0050] Preferably, the data processing device receives and processes data via the wireless communication module, specifically comprising the following steps:
[0051] a. Use the MQTT protocol to achieve real-time data transmission between the data processing device and the stacker to ensure the efficiency and stability of data transmission;
[0052] b. Use data encryption technology to ensure the security of data transmission and prevent data from being tampered with or leaked during transmission;
[0053] c. The data processing device adopts a distributed architecture and uses load balancing and data redundancy technology to improve the system's reliability and processing capabilities.
[0054] Compared with the prior art, the present invention has the following beneficial technical effects:
[0055] 1. Improve inventory efficiency: The present invention adopts an intelligent and automated material inventory method, and realizes rapid and accurate inventory of materials in the warehouse by integrating an image acquisition device and a weight acquisition device, thereby significantly improving inventory efficiency and reducing manual intervention.
[0056] 2. Improve inventory accuracy: The present invention uses a pre-trained convolutional neural network model and a two-dimensional code parsing algorithm to automatically identify the type of material box and obtain material information, combined with data from a high-precision weight sensor to ensure the high accuracy of the material inventory results.
[0057] 3. Reduce labor costs and safety risks: The present invention reduces manual participation through automated inventory counting equipment, which not only reduces labor costs but also improves personnel safety, solving the problem that traditional inventory counting methods require a large amount of manpower input and have certain safety risks.
[0058] 4. Optimize inventory management and support management decisions: The automated inventory method of the present invention can provide real-time and accurate information on materials in the warehouse, which helps to optimize inventory management, avoid excess or shortage of materials, improve capital turnover, and reduce inventory costs. Through real-time analysis and archiving of inventory data, managers can promptly discover and deal with material losses and thefts to protect corporate property safety. Inventory data also provides strong support for corporate procurement plans, budget formulation, and management decisions, helping companies better predict material needs and optimize resource allocation.
[0059] 5. System stability and data security: The present invention uses wireless communication modules and data encryption technology to ensure the efficiency and security of data transmission and prevent data from being tampered with or leaked during transmission. The data processing device adopts a distributed architecture and load balancing technology to improve the reliability and processing capacity of the system.
[0060] 6. Combined application of multiple algorithms to reduce weighing errors: This invention combines the YOLO or SSD target detection model to realize the positioning and identification of multiple bins, and uses the advanced deep convolutional neural network architecture of ResNet and Inception for image classification, which improves the accuracy of material type identification. The weight acquisition device is used to calibrate errors before inventory to ensure the accuracy of weighing data. The comparison of multiple weighing and reset weights of bins further ensures the stability and accuracy of the weighing module. BRIEF DESCRIPTION OF THE DRAWINGS
[0061] Figure 1 This is a module diagram of the inventory system for three-dimensional storage of power materials.
[0062] Figure 2 It is a schematic diagram of an image acquisition device and a weight acquisition device installed on a stacker.
[0063] Figure 3 How to use the inventory system for three-dimensional storage of power materials
[0064] Figure 4 The figure is a flowchart of the inventory method for three-dimensional storage of electric power materials. DETAILED DESCRIPTION
[0065] The present invention will be further explained below in conjunction with the accompanying drawings and embodiments.
[0066] Reference Figure 1 , Figure 2 As shown, an inventory system for three-dimensional storage of electric power materials includes:
[0067] The image acquisition device 1 is used to photograph the cargo boxes stored in the stereoscopic warehouse and transmit the images of the cargo boxes acquired by the photography to the data processing device; further, the image acquisition device 1 is preferably a high-definition camera on the top of an automatic stacker;
[0068] The weight acquisition device 2 is a weight sensor arranged on the mechanical arm of the automatic stacker. When the automatic stacker runs to the position to be counted in the stereoscopic warehouse, the goods at the position to be counted are forked and the weight of the goods is transmitted to the data processing device. Furthermore, the weight acquisition device 2 is preferably a weight sensor at the front end of the mechanical arm.
[0069] The data processing device 3 is used to receive the image transmitted by the cargo image acquisition device, receive the weight transmitted by the weight acquisition device, extract the image and weight characteristics of the received cargo as the external characteristics of the material to be counted, and match the external characteristics of the material to be counted with the original image and weight characteristics stored in the system, and obtain the name and quantity of the cargo at the location to be counted based on the matching results. Furthermore, the data processing device 3 is preferably a server.
[0070] Reference Figure 2 As shown, the installation position of the image acquisition device 1 is: the stacker loading platform is usually composed of a cubic structural frame, including vertical and horizontal frame components. In order to meet the comprehensiveness of monitoring and the convenience of operation, a special mounting box or bracket is set in the middle of the top of the stacker loading platform to place the camera. An industrial camera with a 120° wide-angle viewing angle is used to ensure sufficient field of view. Through the above structural measures, the camera can adjust the angle within a certain range to obtain the best image of the object. Considering the external protection and structural beauty, the data transmission line of the camera is routed through a specific structure (such as the cavity of the sphere and the column base) and connected to the operating table to transmit the captured video recording to the server. The video recognition algorithm is trained so that the algorithm can recognize the QR code from the video, and then identify different goods. The identified goods must match the goods in the record, so as to facilitate the subsequent algorithm to compare directly with the inventory records.
[0071] The camera position design on the stacker is aimed at achieving real-time monitoring and data collection of goods through top monitoring, and ensuring the clarity of the camera image and the stability of transmission through a special installation structure and data transmission design.
[0072] Installation location of weight acquisition device 2: According to the structure and size of the stacker, a high-precision weight sensor is installed at the end of the robotic arm. The support surface of each sensor module is in good contact with the support surface of the stacker and is connected through a wireless transmission module. Use bolts or other fixings to fix the weighing module to the stacker to ensure the stability and reliability of the module. Before the material inventory, the weighing module of the stacker needs to be calibrated for error. The error coefficient is determined by weighing a variety of standard weight objects so that the stacker can accurately measure the weight of the material box that needs to be counted. When taking inventory of power materials, the stacker is used to measure the overall weight of the material box, and the measured value includes the weight of the material box and the weight of the material. When the stacker puts the material box back to its original position, the reset weight is measured again. Then the overall weight and reset weight of the material box are transmitted to the server through the wireless module. The background software calculates the number of materials entering the warehouse based on the weight and item category through an algorithm to form an inventory list.
[0073] Reference Figure 1 As shown, the data processing device includes: an image receiving unit, an image feature extraction unit, a weight receiving unit, a weight feature extraction unit, a matching processing unit and a calculation processing unit;
[0074] An image receiving unit, configured to receive an image transmitted by the first image acquisition device;
[0075] An image feature extraction unit, configured to extract image features from the image received by the first receiving unit as material box image features;
[0076] a weight receiving unit, configured to receive the image transmitted by the second weight obtaining device;
[0077] A weight feature extraction unit, used to extract features from the weight of the material box received by the second weight receiving unit as weight features of the material to be counted;
[0078] A matching processing unit, used for matching the image and weight characteristics of the material to be counted with the original data already existing / planned in the system;
[0079] The calculation processing unit calculates the name and quantity of the goods at the location to be counted according to the matching result of the matching processing unit.
[0080] Reference Figure 3 As shown, this embodiment provides a method for using an inventory system for three-dimensional storage of electric power materials, which is used to inventory the three-dimensional storage goods, including the following steps:
[0081] Step 1: The automatic stacker receives the cargo inventory instruction and moves to the designated location according to the optimal planned route;
[0082] Step 2, the high-definition camera of the automatic stacker is installed on the stacker loading platform and connected to the wireless transmission module. The high-definition camera records the cargo boxes that need to be counted, extracts image features from the video images obtained by the camera as the feature map of the goods to be counted and sends it to the data processing device. When the stacker moves to the vicinity of the designated warehouse according to the instructions, the high-definition camera automatically starts and adjusts to the best shooting angle and focal length. The camera collects images of the material boxes in real time and quickly transmits the high-definition image data to the server through the wireless communication module. After receiving the image, the server will identify the type of material box based on the pictures of the material box at various angles, so as to determine the weight of the empty material box. At the same time, the server will select the image with the best quality from the video images sent back for QR code recognition, and obtain the category, order and other information of the materials in the material box through QR code analysis.
[0083] Step 3, control the automatic stacker to move to the appropriate position of the material box to be counted, extend the mechanical arm to fork the required inventory material box and deliver it to the designated position, the mechanical arm sensor receives the weight of the material box and sends it to the data processing device, and the high-precision weight sensor is installed at the end of the mechanical arm and connected to the wireless transmission module. Before the material inventory is taken, the stacker's weighing module needs to be calibrated for error. The error coefficient is determined by weighing a variety of standard weight objects so that the stacker can accurately obtain the weight of the inserted object. When taking inventory of power materials, the stacker picks up the material box on the inventory position, measures the overall weight of the material box (material box + material) when it is in a stable state, and measures the reset weight after the stacker puts the material box back. Then the overall weight of the material box and the reset weight are transmitted back to the server through the wireless module. The purpose of testing the reset weight is to ensure the stability of the stacker weighing module. In theory, the reset weight after each weighing should always be 0. If the measured reset weight deviation is large, it means that the weighing error of the stacker weighing module is large, and there may even be a fault, and the staff needs to be notified for maintenance.
[0084] Step 4: Perform data analysis on the video images captured by the camera. The server inputs the returned container images into the trained container classification deep model to obtain the container type, and further obtain the weight of the empty container based on the container type. The weight of the material in the container is calculated based on the overall weight of the container and the weight of the empty container obtained through image data analysis.
[0085] Step 5: Use QR code recognition technology to quickly obtain cargo information from a clear image containing the QR code. Through the material information obtained from the QR code (the weight of a single piece of material is queried) and the material weight obtained by weighing, the quantity of materials in the material box can be calculated and compared with the quantity of materials in the order to achieve the effect of material inventory.
[0086] Specifically, the data analysis in step 4 includes image data analysis, weight data analysis and inventory analysis. Image data analysis is to analyze the video images collected by the camera. The server inputs the returned material box image into the trained material box classification deep model to obtain the type of material box, and further obtains the weight of the empty material box according to the type of material box. In addition, the outside of the material box is affixed with a QR code containing the material information and order information in the material box. Through the QR code recognition technology, this information can be quickly obtained from the clear image containing the QR code. Weight data analysis is to analyze the data related to the weight of the material. The weight of the material in the material box is calculated according to the overall weight of the material box returned and the weight of the empty material box obtained by image data analysis. According to the results obtained by image data analysis and weight data analysis, the inventory analysis of materials is carried out. Due to the high standard requirements of power materials, the weight of each type of material is basically fixed with a small error. Through the material information obtained from the QR code (the weight of a single piece of material is queried) and the material weight obtained by weighing, the quantity of materials in the material box can be calculated and compared with the quantity of materials in the order to achieve the effect of material inventory. In addition, the detected material box type is compared with the material box to which the material belongs to check whether the wrong material box type is placed.
[0087] Reference Figure 4 As shown, a method for inventorying electric power materials in three-dimensional warehousing includes the following steps:
[0088] S1. Use an image acquisition device mounted on the stacker loading platform to collect box images, and transmit the image data to a data processing device via a wireless communication module;
[0089] S2. The data processing device processes the received image, uses a pre-trained convolutional neural network model to automatically identify the type of material box, and obtains material information through a QR code parsing algorithm;
[0090] S3. Use a weight acquisition device installed at the end of the stacker robot arm to weigh the material box and transmit the weighing data to the data processing device, wherein the weight acquisition device performs error calibration before inventory counting, and determines the error coefficient by weighing a variety of standard weight objects, so that the stacker can accurately obtain the weight of the object;
[0091] S4. The data processing device calculates the actual weight of the material using a weight calculation and correction algorithm based on the image analysis and weighing data; the data processing device compares the calculated material weight with the material quantity in the order to implement material inventory;
[0092] S5. The data analysis of the data processing device includes three parts: image data analysis, weight data analysis and inventory analysis. These analyses ensure the accuracy of the inventory results, which specifically include the following steps:
[0093] S5-1. Image data analysis, using deep learning models to analyze image data;
[0094] S5-2. Weight data analysis, analyzing weighing data through weight calculation and correction algorithm;
[0095] S5-3. Inventory analysis: compare and confirm the inventory results by integrating image data and weight data.
[0096] When taking inventory of materials, the stacker lifts up the material box, measures the overall weight of the material box when it is in a stable state, then puts the material box back and measures the reset weight. The accuracy of the weighing module is ensured by comparing the two weighing results.
[0097] The data processing device uses a convolutional neural network model to automatically identify the type of the box on the returned box image, specifically including the following steps:
[0098] a. Image preprocessing, including image denoising, enhancement and edge detection;
[0099] b. Feature extraction, extracting high-level features of the image through multiple convolutional layers;
[0100] c. Image matching: matching the extracted image features with known features in the database to identify the type of bin;
[0101] d. Use the trained CNN model to classify the images collected by the image acquisition device and output the corresponding bin type label.
[0102] The QR code parsing algorithm is used to quickly and accurately locate and decode the QR code from the high-definition image to obtain material information and order information. The specific steps include:
[0103] a. QR code positioning: quickly determine the QR code position through image processing algorithms;
[0104] b. QR code decoding: use the QR code parsing function to parse the data in the QR code.
[0105] The two-dimensional code parsing algorithm specifically includes the following steps:
[0106] a. Use OpenCV library to perform image preprocessing, including grayscale, binarization, morphological transformation and other steps to improve the accuracy of QR code recognition;
[0107] b. Use ZBar 2D code recognition library to decode the preprocessed image and extract the information in the 2D code;
[0108] c. Parse the data in the QR code and convert it into usable material and order information.
[0109] Preferably, the weight calculation and correction algorithm comprises the following steps:
[0110] a. Data calibration: calibrate the weighing module before weighing, and determine the error coefficient by weighing a variety of standard weight objects;
[0111] b. Weight calculation, use the following formula to calculate the actual weight of the material:
[0112] Wactual=WVmeasured-Wtare
[0113] Among them, Wactual is the actual weight of the material, Wmeasured is the weight measured by the weighing module, and Wtare is the tare weight of the material box;
[0114] c. Error correction, correct the calculation results by the following correction formula:
[0115] Wcorrected=Wa.actual×(1+e)
[0116] Among them, ∈ is the error coefficient.
[0117] The convolutional neural network model is a target detection model YOLO or SSD, which is used to locate and identify multiple bins in a high-definition image. The image captured by the image acquisition device is input and the location information and type label of the bin are output. Specifically, the following steps are included:
[0118] a. Use the YOLO model to detect objects in the image and locate multiple bins;
[0119] b. Use the SSD model to classify the located bins and output category labels.
[0120] The data processing device compares the inventory result with the quantity of materials in the order to determine the accuracy of the inventory and generates an inventory report, which specifically includes the following steps:
[0121] a. Compare the inventory results and compare the calculated material weight with the order quantity;
[0122] b. Generate an inventory report, including inventory results, comparison results, error analysis and other information.
[0123] The CNN model used for material type identification specifically includes the following steps:
[0124] a. Use a large-scale bin image dataset for pre-training and use data augmentation techniques to expand the dataset, including operations such as rotation, scaling, and translation;
[0125] b. Use ResNet and Inception advanced deep convolutional neural network architectures for model training and optimize loss functions to improve recognition accuracy;
[0126] c. Through transfer learning methods, the pre-trained model is applied to the specific bin image classification task and fine-tuned to adapt to the actual application scenario.
[0127] The correction algorithm used for weighing data analysis specifically includes the following steps:
[0128] a. Establish the error model of the weighing module and calculate the error coefficient by weighing standard weight objects multiple times;
[0129] b. Use polynomial fitting method to model and correct weighing errors and optimize weighing results;
[0130] c. Compare the corrected weighing results with the actual material weight to evaluate the accuracy and robustness of the correction algorithm.
[0131] The data processing device receives and processes data through the wireless communication module, specifically including the following steps:
[0132] a. Use the MQTT protocol to achieve real-time data transmission between the data processing device and the stacker to ensure the efficiency and stability of data transmission;
[0133] b. Use data encryption technology to ensure the security of data transmission and prevent data from being tampered with or leaked during transmission;
[0134] c. The data processing device adopts a distributed architecture and uses load balancing and data redundancy technology to improve the system's reliability and processing capabilities.
[0135] In summary, the present invention realizes efficient, accurate and safe inventory of three-dimensional storage of electric power materials through the combination of intelligent and automated technologies, which not only greatly improves the inventory efficiency and accuracy, but also reduces labor costs and safety risks.
[0136] In the description of the present invention, it should be noted that the terms "upper", "lower", etc. indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific position, be constructed and operated in a specific position, and therefore cannot be understood as a limitation on the present invention. Unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be a connection between the two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances.
[0137] It should be noted that, in the present invention, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.
[0138] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features of the invention described herein.
Claims
1. An inventory method for three-dimensional storage of power materials, characterized in that: The following steps are involved: S1. Use an image acquisition device mounted on the stacker loading platform to collect box images, and transmit the image data to a data processing device via a wireless communication module; S2. The data processing device processes the received image, uses a pre-trained convolutional neural network model to automatically identify the type of material box, and obtains material information through a QR code parsing algorithm; S3. Use a weight acquisition device installed at the end of the stacker robot arm to weigh the material box and transmit the weighing data to the data processing device, wherein the weight acquisition device performs error calibration before inventory counting, and determines the error coefficient by weighing a variety of standard weight objects, so that the stacker can accurately obtain the weight of the object; S4. The data processing device calculates the actual weight of the material using a weight calculation and correction algorithm based on the image analysis and weighing data; the data processing device compares the calculated material weight with the material quantity in the order to implement material inventory; S5. The data analysis of the data processing device includes three parts: image data analysis, weight data analysis and inventory analysis. These analyses ensure the accuracy of the inventory results, which specifically include the following steps: S5-1. Image data analysis, using deep learning models to analyze image data; S5-2. Weight data analysis, analyzing weighing data through weight calculation and correction algorithm; S5-3. Inventory analysis: compare and confirm the inventory results by integrating image data and weight data.
2. The inventory method for three-dimensional storage of electric power materials according to claim 1 is characterized in that: When taking inventory of materials, the stacker lifts up the material box, measures the overall weight of the material box when it is in a stable state, then puts the material box back and measures the reset weight. The accuracy of the weighing module is ensured by comparing the two weighing results.
3. The inventory method for three-dimensional storage of electric power materials according to claim 1 is characterized in that: The data processing device uses a convolutional neural network model to automatically identify the type of the box on the returned box image, specifically including the following steps: a. Image preprocessing, including image denoising, enhancement and edge detection; b. Feature extraction, extracting high-level features of the image through multiple convolutional layers; c. Image matching: matching the extracted image features with known features in the database to identify the type of bin; d. Use the trained CNN model to classify the images collected by the image acquisition device and output the corresponding bin type label.
4. The inventory method for three-dimensional storage of electric power materials according to claim 1 is characterized in that: The QR code parsing algorithm is used to quickly and accurately locate and decode the QR code from the high-definition image to obtain material information and order information. The specific steps include: QR code positioning: quickly determine the QR code position through image processing algorithms; QR code decoding, using the QR code parsing function to parse the data in the QR code; The two-dimensional code parsing algorithm specifically includes the following steps: a. Use OpenCV library to perform image preprocessing, including grayscale, binarization, morphological transformation and other steps to improve the accuracy of QR code recognition; b. Use ZBar 2D code recognition library to decode the preprocessed image and extract the information in the 2D code; c. Parse the data in the QR code and convert it into usable material and order information.
5. The inventory method for three-dimensional storage of electric power materials according to claim 1 is characterized in that: The weight calculation and correction algorithm includes the following steps: a. Data calibration: calibrate the weighing module before weighing, and determine the error coefficient by weighing a variety of standard weight objects; b. Weight calculation, use the following formula to calculate the actual weight of the material: IN actual =In measured -IN tare Among them, Wactual is the actual weight of the material, Wmeasured is the weight measured by the weighing module, and Wtare is the tare weight of the material box; c. Error correction, correct the calculation results by the following correction formula: IN corrected =In actual ×(1+∈) Among them, ∈ is the error coefficient.
6. The inventory method for three-dimensional storage of electric power materials according to claim 1 is characterized in that: The convolutional neural network model is a target detection model YOLO or SSD, which is used to locate and identify multiple bins in a high-definition image. The image captured by the image acquisition device is input and the location information and type label of the bin are output. Specifically, the following steps are included: a. Use the YOLO model to detect objects in the image and locate multiple bins; b. Use the SSD model to classify the located bins and output category labels.
7. The inventory method for three-dimensional storage of electric power materials according to claim 1 is characterized in that: The data processing device compares the inventory result with the quantity of materials in the order to determine the accuracy of the inventory and generates an inventory report, which specifically includes the following steps: a. Compare the inventory results and compare the calculated material weight with the order quantity; b. Generate an inventory report, including inventory results, comparison results, error analysis and other information.
8. The inventory method for three-dimensional storage of electric power materials according to claim 3 is characterized in that: The CNN model used for material type identification specifically includes the following steps: a. Use a large-scale bin image dataset for pre-training and use data augmentation techniques to expand the dataset, including operations such as rotation, scaling, and translation; b. Use ResNet and Inception advanced deep convolutional neural network architectures for model training and optimize loss functions to improve recognition accuracy; c. Through transfer learning methods, the pre-trained model is applied to the specific bin image classification task and fine-tuned to adapt to the actual application scenario.
9. The inventory method for three-dimensional storage of electric power materials according to claim 5 is characterized in that: The correction algorithm used for weighing data analysis specifically includes the following steps: a. Establish the error model of the weighing module and calculate the error coefficient by weighing standard weight objects multiple times; b. Use polynomial fitting method to model and correct weighing errors and optimize weighing results; c. Compare the corrected weighing results with the actual material weight to evaluate the accuracy and robustness of the correction algorithm.
10. The inventory method for three-dimensional storage of electric power materials according to claim 1, characterized in that: The data processing device receives and processes data through the wireless communication module, specifically including the following steps: a. Use the MQTT protocol to achieve real-time data transmission between the data processing device and the stacker to ensure the efficiency and stability of data transmission; b. Use data encryption technology to ensure the security of data transmission and prevent data from being tampered with or leaked during transmission; c. The data processing device adopts a distributed architecture and uses load balancing and data redundancy technology to improve the system's reliability and processing capabilities.
Citation Information
Patent Citations
Warehouse self-service material taking method, warehouse self-service material taking device and system
CN111476530A
Electric power material storage system analysis system
CN112465429A
Electric power material intelligent acceptance system based on image recognition
CN115689441A
Unmanned intelligent checking system and method for electric power material storage
CN115759892A
Inventory system and method, stereoscopic warehouse, electronic equipment and computer medium
CN116468366A
