3D Modeling and Intelligent Scheduling System for Urea Warehouses

Abstract

This invention discloses a 3D modeling and intelligent scheduling system for urea warehouses, belonging to the field of intelligent modeling and scheduling technology for urea warehouses. Specifically, it includes: real-time acquisition of internal images of the urea warehouse, preprocessing the acquired images, using computer vision technology to generate 3D point cloud data of the urea warehouse, constructing a 3D model of the urea warehouse, building a spatiotemporal heat map, and intelligently generating a scheduling plan based on real-time inventory status, warehouse space usage, task status, and transportation needs; real-time scanning and analysis of the spatial layout, item location, and type of the urea warehouse, enabling dynamic generation of a 3D spatial model of the urea warehouse based on real-time acquired data. The constructed spatiotemporal heat map can detect obstacles in advance and intelligently plan the operating path of the robotic arm, eliminating the need to adjust or replan the robotic arm's path based on obstacles, saving handling time and greatly improving the operating efficiency of the robotic arm.

Description

Technical Field

[0001] This invention belongs to the field of intelligent modeling and scheduling technology for urea warehouses, specifically a 3D modeling and intelligent scheduling system for urea warehouses. Background Technology

[0002] In large chemical warehouses and mineral warehouses, urea, as an important fertilizer raw material, involves complex processes of storage, handling, sorting, and scheduling that require efficient management. Traditional warehouse management typically relies on manual monitoring, paper records, and conventional scheduling methods for urea storage and distribution. This not only leads to low operational efficiency but also results in problems such as outdated inventory information, resource waste, low warehouse space utilization, and inflexible scheduling. Therefore, achieving digitalization, automation, and intelligentization of urea warehouse management has become a crucial research direction.

[0003] AI visual recognition uses cameras and sensors to capture real-time information such as the location, quantity, and type of items in the warehouse, and analyzes this information through image processing and pattern recognition algorithms to provide accurate data support for warehouse management. Meanwhile, 3D modeling technology can construct the 3D spatial layout of the warehouse in real time, helping to accurately plan the storage locations of items and optimize warehouse space utilization. Simultaneously, combined with intelligent scheduling algorithms, it can effectively solve the scheduling and handling path planning of items, improving the flexibility and efficiency of warehouse operations. However, some shortcomings still exist: 1) Traditional 3D modeling technology is usually static, used only in the planning stage, and it is difficult to update the space in a timely manner according to dynamic changes in the warehouse; 2) Existing intelligent scheduling algorithms are often limited to rule-based scheduling methods, lacking flexibility and adaptability, making it difficult to achieve dynamic real-time scheduling in warehouse environments with multiple items, multiple devices, and multiple tasks. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes a 3D modeling and intelligent scheduling system for urea warehouses, which can achieve efficient, accurate, and dynamic warehouse space planning, task scheduling, and resource allocation, thereby significantly improving the management efficiency and operational capabilities of urea warehouses.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] The urea warehouse 3D modeling and intelligent scheduling system includes: a data acquisition and processing module, a 3D modeling module, and an intelligent scheduling module;

[0007] The data acquisition and processing module is used to acquire internal images of the urea warehouse in real time and to preprocess the acquired internal images of the urea warehouse.

[0008] The 3D modeling module is used to generate 3D point cloud data of the urea warehouse based on the preprocessed internal images of the urea warehouse, and to construct a 3D model of the urea warehouse. The 3D model of the urea warehouse derives the space occupancy of the items based on the morphological characteristics and size threshold range of the items.

[0009] The intelligent scheduling module is used to construct a heatmap sequence of spatial occupancy at future time steps, and intelligently generate scheduling schemes based on real-time inventory status, warehouse space usage, task status, and transportation needs.

[0010] Specifically, the 3D modeling module includes: a 3D point cloud generation unit, a spatial reconstruction unit, and a 3D model update unit;

[0011] The three-dimensional point cloud generation unit is used to generate a preliminary three-dimensional point cloud based on the preprocessed internal image of the urea warehouse, refine the preliminary point cloud, optimize the matching accuracy of the point cloud, and remove noise points.

[0012] The spatial reconstruction unit is used to adaptively reconstruct the urea warehouse space based on real-time environmental information and item type, learn the item stacking rules of the urea warehouse, deduce the space occupation of the items, and construct a three-dimensional model of the urea warehouse.

[0013] The three-dimensional model update unit is used to dynamically update the three-dimensional model of the urea warehouse in real time.

[0014] Specifically, the adaptive reconstruction of the urea warehouse space based on real-time environmental information and item type, the learning of item stacking rules in the urea warehouse, the deduction of item space occupancy, and the construction of a 3D model of the urea warehouse include:

[0015] Based on the optimized 3D point cloud, a preliminary 3D model of the urea warehouse is established according to the spatial location and relative relationship of different items in the urea warehouse.

[0016] The morphological features of items in the urea warehouse are automatically extracted and modeled. Item morphological constraints are introduced, taking into account the size, shape and elasticity of the items, to determine whether the items can be stacked together. If so, the gaps and tilt angles of the items during the stacking process are calculated. If not, no constraints are imposed.

[0017] Set a size threshold range for the item, divide the space occupied by the item into multiple closely fitting n-sided units, deduce the space occupied by the item by combining the morphological characteristics and size threshold range, and adaptively adjust the spatial modeling of each type of item.

[0018] By integrating the spatial modeling of each type of item and combining it with the preliminary 3D model of the urea warehouse, a 3D model of the urea warehouse is obtained.

[0019] Specifically, the intelligent scheduling module includes: a task dynamic scheduling unit and a path planning unit;

[0020] The task dynamic scheduling unit is used to analyze historical data of urea warehouse, predict order demand, outbound sequence and transportation route in the future, and dynamically adjust and optimize scheduling tasks based on real-time environmental data and task objectives.

[0021] The path planning unit is used to optimize space utilization and the movement path of the robotic arm.

[0022] Specifically, the analysis of historical data from urea warehouses to predict order demand, outbound sequence, and transportation routes for a future period, and the dynamic adjustment and optimization of scheduling tasks based on real-time environmental data and task objectives, includes:

[0023] A product prediction model is established based on a neural network. The model is trained using historical data. The historical inbound and outbound records, transportation frequency and time intervals of the products are analyzed to obtain the trained product prediction model.

[0024] The real-time inventory and status of the urea warehouse are input into the trained item prediction model to predict the item's outbound priority and inbound / outbound demand patterns.

[0025] Based on real-time order demand, item outbound priority, and time window factors, the priority of scheduling tasks is automatically adjusted, and the execution path and order of tasks are dynamically adjusted.

[0026] Specifically, the optimization of space utilization and the movement path of the robotic arm includes:

[0027] An improved genetic algorithm is used to simulate the stacking layout of goods. A fitness function is designed. Combined with the space occupation of goods in the 3D model of the urea warehouse, the storage requirements of goods, the frequency of outbound delivery, the space utilization of warehouse aisles and the number of stacking layers are considered to generate and select the optimal stacking scheme.

[0028] Set a time step, construct a heatmap sequence of space occupancy for future time steps, calculate the prediction error at the current time step in real time, establish a mapping relationship between prediction error and time step, and based on task scheduling and item outbound priority, trigger a time step shortening mechanism when the error exceeds a set threshold for three consecutive time windows, and extend the time step when the error remains below the set threshold, thus planning the future path for each robotic arm.

[0029] Specifically, the spatiotemporal heat map is a spatiotemporal conflict prediction map, including a sequence of heat map sequences showing the space occupancy of urea warehouses within a preset future time period.

[0030] Specifically, the preprocessing includes: image denoising, illumination compensation and dynamic enhancement, and target recognition and localization;

[0031] Image denoising involves using a trained neural network model to denoise the collected images of the interior of the urea warehouse.

[0032] Illumination compensation and dynamic enhancement: A local contrast enhancement algorithm is introduced to enhance the images inside the urea warehouse.

[0033] Target recognition and localization: Identifying objects and locating item boundaries in the urea warehouse.

[0034] Specifically, the items include: bagged urea, transport trucks, robotic arms, and warehouse tools.

[0035] Specifically, the three-dimensional model update unit adopts an incremental update mechanism.

[0036] Specifically, the system also includes a safety monitoring module that monitors the movement trajectory of robotic arms and transport trucks in the urea warehouse in real time and establishes a dynamic collision warning zone.

[0037] Compared with the prior art, the beneficial effects of the present invention are:

[0038] 1. This invention proposes a 3D modeling and intelligent scheduling system for urea warehouses, which can scan and analyze the spatial layout, item location and type of urea warehouses in real time. By dividing the space occupied by items into multiple closely fitting n-hedral units, and combining the morphological characteristics and size threshold range of the items, the system can deduce the spatial occupancy of the items and dynamically generate a 3D spatial model of the urea warehouse based on the real-time collected data.

[0039] 2. This invention proposes a 3D modeling and intelligent scheduling system for urea warehouses. Based on real-time data, the system calculates task scheduling priorities and automatically adjusts the task execution order. Through the constructed spatiotemporal heat map, obstacles can be detected in advance, and the running path of the robotic arm can be intelligently planned. There is no need to adjust or replan the path of the robotic arm based on obstacles, saving handling time and greatly improving the operating efficiency of the robotic arm. Attached Figure Description

[0040] Figure 1 The diagram architecture of the urea warehouse 3D modeling and intelligent scheduling system provided by this invention;

[0041] Figure 2 A schematic diagram illustrating the spatial occupancy of the item provided by the present invention;

[0042] Figure 3 This is a schematic diagram of path planning provided by the present invention. Detailed Implementation

[0043] The present application will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all fall within the protection scope of the present application.

[0044] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0045] It should be noted that, unless there is a conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. In addition, the terms "first," "second," and "third" used in this application do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0046] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The term "and / or" as used in this specification includes any and all combinations of one or more of the associated listed items.

[0047] Example 1

[0048] Please see Figures 1-3 The present invention provides an embodiment of a urea warehouse 3D modeling and intelligent scheduling system, comprising: a data acquisition and processing module, a 3D modeling module and an intelligent scheduling module;

[0049] The data acquisition and processing module is used to acquire internal images of the urea warehouse in real time and to preprocess the acquired internal images of the urea warehouse.

[0050] High-definition, low-light AI smart cameras with wide dynamic range are evenly installed in a grid pattern on the top of the urea warehouse to ensure coverage of every corner of the warehouse without any blind spots. At the same time, auxiliary cameras are added at key locations such as warehouse entrances and exits and shelf aisles to capture details of goods entering, leaving and being handled.

[0051] In conjunction with the camera, LiDAR sensors are installed at intervals. The laser beams emitted by these sensors can accurately measure the distance to the object's surface and obtain depth information. Through a data fusion algorithm, the camera image data and the LiDAR depth data are synchronized and integrated in real time.

[0052] At the same time, temperature and humidity sensors and gas sensors are introduced to monitor environmental parameters in the warehouse. Urea storage is sensitive to temperature and humidity, and abnormal environment may affect product quality. This data is used to assist inventory management and provide timely warnings of adverse storage conditions.

[0053] The preprocessing includes: image denoising, illumination compensation and dynamic enhancement, and target recognition and localization;

[0054] Image denoising utilizes a trained neural network model to denoise acquired images. The model can automatically adjust the denoising strategy based on the noise characteristics of different regions of the image, and differentiate between periodic noise caused by the flickering of lighting equipment and particulate noise caused by dust during cargo handling. It effectively removes various noise interferences while preserving image details.

[0055] Illumination compensation and dynamic enhancement introduce a local contrast enhancement algorithm to enhance low-contrast areas in the image, improving the recognition accuracy of low-light and shadow areas. Unlike traditional image enhancement methods, it adopts an adaptive illumination compensation strategy: based on real-time changes in ambient light, it automatically adjusts the brightness, contrast and saturation of the image to eliminate inconsistencies caused by changes in lighting conditions, ensuring that targets in the warehouse can be clearly identified regardless of whether it is day or night.

[0056] The specific steps of target identification and localization include:

[0057] 1. Use a convolutional neural network to identify objects within a urea warehouse;

[0058] In this step, before using the convolutional neural network, a custom convolutional neural network (CNN) needs to be trained. By annotating a large amount of data on common objects in the urea warehouse (such as urea bags, pallets, etc.), an efficient neural network model that can recognize various items is trained. A multi-scale detection network is also designed, which can recognize objects of different sizes at different scales, avoiding missed detections in complex scenarios with stacked items.

[0059] 2. Target localization of objects in a urea warehouse based on a region convolutional neural network;

[0060] In this step, the Region Convolutional Neural Network (RCNN) technique is combined. Object candidate boxes are generated through the Region Generation Network (RPN), and then the convolutional neural network is used for classification and precise localization. Compared with the traditional simple classification method, the RCNN method can greatly improve the localization accuracy of the target, especially in the case of stacked or occluded objects, it can better identify the boundaries of the target objects.

[0061] 3. Establish an object relationship model to handle obstructions in the urea warehouse;

[0062] In this step, in a complex warehouse environment, items may have incomplete visual information due to stacking methods or occlusion by other objects. An occlusion separation neural network is used, which can identify and separate occlusion problems caused by overlapping objects. Through multi-angle visual information and training with a deep learning model, the network can accurately identify occluded items and infer their complete spatial location and shape, thereby improving the completeness and accuracy of target recognition.

[0063] This paper employs a Graph Neural Network (GNN) to model the spatial relationships between items in a warehouse. Based on features such as the object's location, size, and shape, this method establishes semantic relationships between objects, such as "stacked on top" or "adjacent," and combines these relationships for more accurate target classification and localization. This approach helps improve recognition accuracy in complex environments with multiple objects, especially in warehouse scenarios where objects are densely packed and irregularly stacked, effectively reducing recognition errors.

[0064] 4. Utilize a self-learning mechanism to update the convolutional neural network and the regional convolutional neural network in real time.

[0065] In this step, self-learning is mainly to cope with the constantly changing types of items in the warehouse or new items that have never been seen before. It can use existing labeled data for rapid fine-tuning and incremental training. Through continuous learning and feedback, it can automatically identify new types of items and update the recognition model in a timely manner, avoiding the tedious process of manual relabeling.

[0066] The items include bagged urea, robotic arms, trucks, and other tools;

[0067] The 3D modeling module is used to generate 3D point cloud data of the urea warehouse based on the preprocessed internal images of the urea warehouse, and to construct a 3D model of the urea warehouse. The 3D model of the urea warehouse derives the space occupancy of the items based on the morphological characteristics and size threshold range of the items.

[0068] The 3D modeling module includes: a 3D point cloud generation unit, a spatial reconstruction unit, and a 3D model update unit;

[0069] The three-dimensional point cloud generation unit is used to generate a preliminary three-dimensional point cloud based on the preprocessed internal image of the urea warehouse, refine the preliminary point cloud, optimize the matching accuracy of the point cloud, and remove noise points.

[0070] The spatial reconstruction unit is used to adaptively reconstruct the urea warehouse space based on real-time environmental information and item type, learn the item stacking rules of the urea warehouse, deduce the space occupation of the items, and construct a three-dimensional model of the urea warehouse.

[0071] The process of adaptively reconstructing the urea warehouse space based on real-time environmental information and item types, learning the item stacking rules of the urea warehouse, deriving the space occupancy of items, and constructing a 3D model of the urea warehouse includes the following steps:

[0072] Step S1: Based on the optimized 3D point cloud, establish a preliminary 3D model of the urea warehouse according to the spatial location and relative relationship of different items in the urea warehouse;

[0073] The initial 3D model of the urea warehouse is a simple outline of the warehouse and its contents;

[0074] Step S2: Automatically extract and model the morphological features of items in the urea warehouse, introduce item morphological constraints, consider factors such as item size, shape, and elasticity, and determine whether items can be stacked together. If so, calculate the item gaps and tilt angles during the stacking process; otherwise, no constraints are imposed.

[0075] For example, when bagged urea is stacked together, it will deform, and there will be gaps between one bag of urea and the stacked urea. Calculate the size of the bagged urea at this time to provide data support for the subsequent automatic inference of the space occupation of the items and modeling.

[0076] Step S3: Set the size threshold range of the item, divide the space occupied by the item into multiple closely fitting n-sided units, deduce the space occupation of the item by combining the shape characteristics of the item and the size threshold range, and adaptively adjust the spatial modeling of each type of item.

[0077] In this embodiment, dividing the space occupied by the item into multiple closely fitting n-hedral units is a two-dimensional surface of the item obtained from different perspectives. The two-dimensional surfaces of the item are combined to obtain n-hedral units. Based on the target recognition results, n-hedral units are constructed for all items.

[0078] By utilizing clues such as the texture direction of objects in the image and the relative position changes of feature points, combined with the tilt angle of the objects, the construction parameters of the three-dimensional n-sided unit are adjusted in real time to ensure the accuracy of space occupation calculation.

[0079] By accurately deriving the space occupancy of each type of item, the actual inventory quantity and space occupied in the warehouse can be monitored in real time, avoiding inventory errors caused by traditional manual estimation or fixed model calculation. Whether it is the number of stacked layers of bagged urea, changes in the placement angle, or mixed placement of different specifications of products, each package of goods can be accurately located, making the inventory data accurate to the single item level.

[0080] Reference Figure 2 By identifying and locating targets, bagged urea was identified. Figure B shows part of the bagged urea identification box, but not all of them are shown. Figure C shows two two-dimensional surfaces. By combining the two two-dimensional surfaces, the space occupied by the bagged urea in Figure D is obtained.

[0081] Step S4: Integrate the spatial modeling of each type of item, and combine it with the preliminary 3D model of the urea warehouse to obtain the 3D model of the urea warehouse.

[0082] The three-dimensional model update unit is used to dynamically update the three-dimensional model of the urea warehouse in real time.

[0083] The 3D model update unit adopts an incremental update mechanism, which focuses on scanning the local areas that have changed within the warehouse to generate incremental local point cloud data; the incremental data is then aligned and fused with the global 3D model through a feature matching algorithm.

[0084] In response to the harsh conditions in urea warehouses, such as frequent changes in goods, complex lighting conditions (with both natural light and artificial lighting switching), and reflective packaging bags, this method derives the spatial occupancy of items. Even if some urea bags are obscured by robotic arms or trucks, the model can still be updated in real time using 3D point clouds, ensuring that the model always reflects the real state of the warehouse. This is different from conventional reconstruction techniques that are suitable for static and ideal environments.

[0085] The intelligent scheduling module is used to construct a heatmap sequence of spatial occupancy at future time steps, and intelligently generate scheduling schemes based on real-time inventory status, warehouse space usage, cargo attributes (such as weight and stacking rules) and transportation requirements.

[0086] The intelligent scheduling module includes: a task dynamic scheduling unit and a path planning unit;

[0087] The task dynamic scheduling unit is used to analyze historical data of urea warehouse, predict order demand, outbound sequence and transportation route in the future, and dynamically adjust and optimize scheduling tasks based on real-time environmental data and task objectives.

[0088] The process of analyzing historical data from urea warehouses to predict order demand, outbound sequence, and transportation routes for a future period, and dynamically adjusting and optimizing scheduling tasks based on real-time environmental data and task objectives, includes the following steps:

[0089] A product prediction model is established based on a neural network. The model is trained using historical data. The historical inbound and outbound records, transportation frequency and time intervals of the products are analyzed to obtain a well-trained product prediction model.

[0090] The real-time inventory and status of the urea warehouse are input into the trained item prediction model to predict the item's outbound priority and inbound / outbound demand patterns.

[0091] Based on real-time order demand, item outbound priority, time window, and other factors, the priority of scheduling tasks is automatically adjusted, and the execution path and order of tasks are dynamically adjusted.

[0092] In this embodiment, when the demand for a certain item increases sharply, the system can prioritize scheduling that item to ensure efficient use of warehouse resources. It not only takes into account the inbound and outbound needs of items, but also generates the optimal scheduling task based on factors such as the status of warehouse equipment and task priority.

[0093] Dynamically adjusting the task execution path and sequence is used to deal with complex situations such as sudden tasks and equipment failures. When equipment in the warehouse fails or cannot work as planned, reinforcement learning is used to automatically adjust the scheduling scheme, replan the task execution path, select backup equipment to execute the task, and reduce task delays. When a delay in task progress or the insertion of a new urgent task is detected, the execution order of the currently scheduled tasks is automatically adjusted to optimize the warehouse resource allocation and avoid long periods of resource idleness or task backlog.

[0094] The path planning unit is used to optimize space utilization and the movement path of the robotic arm.

[0095] The specific steps for optimizing space utilization and the movement path of the robotic arm include:

[0096] Step 1: Use an improved genetic algorithm to simulate the stacking layout of items, design a fitness function, combine the space occupation of items in the 3D model of the urea warehouse, consider the storage requirements of items, outbound frequency, space utilization of warehouse aisles and stacking layers, generate and select the optimal stacking scheme.

[0097] In this embodiment, if only the storage requirements of the items, the frequency of outbound shipments, the space utilization of the warehouse aisles, and the number of stacking layers are considered, it is impossible to deduce the stacking scheme. Different or the same items occupy different spaces. Based on the space occupancy of the items in the 3D model of the urea warehouse, the actual occupancy of the items in the three-dimensional space can be obtained, thereby generating a stacking scheme. Multiple stacking schemes are generated through an improved genetic algorithm, and the fitness is evaluated to select the optimal scheme. Regardless of whether the items are stacked regularly or placed compactly, they can be reasonably arranged based on the accurately calculated space data, effectively avoiding idle waste or overcrowding caused by inaccurate space estimation.

[0098] Step 2: Set the time step, construct a heatmap sequence of future time step space occupancy, calculate the prediction error at the current time step in real time, establish a mapping relationship between prediction error and time step, and based on task scheduling and item outbound priority, trigger the time step shortening mechanism when the error of three consecutive time windows exceeds the set threshold, and extend the time step when the error is consistently below the set threshold, and plan the future path for each robotic arm.

[0099] In this embodiment, the spatiotemporal heatmap is a spatiotemporal conflict prediction map. 3D point cloud data during the historical operation period of the urea warehouse is collected, sliced ​​and labeled according to the time series, and a training dataset is constructed. The input features include the spatial coordinates, point density, and geometric features of the point cloud at the current moment. The output is a space occupancy heatmap every second in the next 5-second time period. Then, an LSTM network is used for multiple rounds of iterative training. The 3D model of the urea warehouse and the real-time 3D point cloud are input into the trained LSTM to output a sequence of space occupancy heatmaps for the next 5 seconds.

[0100] Traditional path planning involves planning the robotic arm's retrieval and unloading of items based on task requirements, while also considering obstacles and other robotic arms along the path to avoid path conflicts. This application analyzes the spatial occupancy heatmap sequence within a future time step. However, the setting of the time step is affected by the robotic arm's speed, the influence of other robotic arms or obstacles, and the task priority and duration. By analyzing the relationship between the time step and errors, including global errors, local area errors, and time series errors, this application addresses complex factors such as multi-task concurrency, dynamic environmental changes, and multi-robotic arm collaboration, such as robotic arm acceleration and cargo collapse. By predicting the spatial and temporal occupancy in advance and avoiding influencing factors, it eliminates the need to adjust or replan the robotic arm's path based on obstacles, saving handling time and improving the robotic arm's operational efficiency.

[0101] The system also includes a safety monitoring module, which detects the movement trajectory of equipment such as robotic arms and transport trucks in the warehouse in real time based on a 3D model and establishes a dynamic collision warning zone; combined with video surveillance and lidar data, it identifies the activity area of ​​personnel in the warehouse, and triggers a safety alarm when the movement trajectory of the equipment overlaps with the activity area of ​​personnel.

[0102] Reference Figure 3 When traditional path planning encounters obstacles, the path needs to be replanned. Using the method in this application, the location of obstacles in the future can be known through a spatiotemporal heat map, allowing for path planning to avoid obstacles in advance.

[0103] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0104] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A three-dimensional modeling and intelligent scheduling system for a urea warehouse, characterized in that, The application relates to a urea warehouse intelligent scheduling system, which comprises a data acquisition and processing module, a three-dimensional modeling module and an intelligent scheduling module. The data acquisition and processing module is used for collecting internal images of a urea warehouse in real time and pre-processing the collected internal images of the urea warehouse. The three-dimensional modeling module is used for generating three-dimensional point cloud data of the urea warehouse based on the pre-processed internal images of the urea warehouse, constructing a three-dimensional model of the urea warehouse, and deriving the space occupation of the articles according to the morphological characteristics and size threshold range of the articles. The intelligent scheduling module is used for constructing a future time step space occupation heat map sequence, intelligently generating a scheduling scheme based on real-time inventory status, warehouse space usage, task status and transportation demand. The three-dimensional modeling module comprises a three-dimensional point cloud generation unit, a space reconstruction unit and a three-dimensional model updating unit. The three-dimensional point cloud generation unit is used for generating a preliminary three-dimensional point cloud according to the pre-processed internal images of the urea warehouse, refining the generated point cloud, optimizing the matching precision of the point cloud and removing noise points. The space reconstruction unit is used for adaptively reconstructing the urea warehouse space according to real-time environment information and article types, learning the article stacking rules of the urea warehouse, deriving the space occupation of the articles and constructing a three-dimensional model of the urea warehouse. The three-dimensional model updating unit is used for dynamically updating the three-dimensional model of the urea warehouse in real time. The adaptive reconstruction of the urea warehouse space according to real-time environment information and article types, the learning of the article stacking rules of the urea warehouse, the derivation of the space occupation of the articles and the construction of the three-dimensional model of the urea warehouse comprise the following steps: Based on the optimized three-dimensional point cloud, a preliminary three-dimensional model of the urea warehouse is established according to the spatial positions and relative relationships of different articles in the urea warehouse. The morphological characteristics of the articles in the urea warehouse are automatically extracted and modeled, the article form constraint is introduced, the size, form and elasticity factors of the articles are considered, it is judged whether the articles can be stacked together, if yes, the article gap and inclination angle in the article stacking process are calculated, and if not, the constraint is not performed. The size threshold range of the articles is set, the space occupied by the articles is divided into multiple closely fitted n-hedron units, the space occupation of the articles is derived according to the morphological characteristics and size threshold range of the articles, and the space modeling of each type of article is adaptively adjusted. The space modeling of each type of article is integrated, the preliminary three-dimensional model of the urea warehouse is combined, and the three-dimensional model of the urea warehouse is obtained. The intelligent scheduling module comprises a task dynamic scheduling unit and a path planning unit. The task dynamic scheduling unit is used for analyzing historical data of the urea warehouse, predicting order demand, warehouse-out sequence and transportation path in a future period of time, dynamically adjusting and optimizing the scheduling task according to real-time environment data and task targets. The path planning unit is used for optimizing space utilization and the movement path of a mechanical arm. The analysis of the historical data of the urea warehouse, the prediction of order demand, warehouse-out sequence and transportation path in a future period of time and the dynamic adjustment and optimization of the scheduling task according to real-time environment data and task targets comprise the following steps: ​ The item prediction model is established based on a neural network, and the item prediction model is trained using historical data to analyze historical warehouse-in and warehouse-out records, transportation frequency and time interval of the item, so as to obtain the trained item prediction model; The real-time inventory and state of the urea warehouse are input into the trained item prediction model to predict the warehouse-out priority of the item and the demand mode of the warehouse-in and warehouse-out; According to the real-time order demand, the warehouse-out priority of the item and the time window factor, the priority of the scheduling task is automatically adjusted, and the path and sequence of the task execution are dynamically adjusted; The optimization of the space utilization and the motion path of the mechanical arm comprises: An improved genetic algorithm is used to simulate the item stacking layout, a fitness function is designed, the space occupation of the item in the three-dimensional model of the urea warehouse is combined, the storage demand, warehouse-out frequency, space utilization of the warehouse channel and stacking layer of the item are considered, and the optimal stacking scheme is generated and selected; A time step is set, a future time step space occupation heat map sequence is constructed, the prediction error at the current time step is calculated in real time, a mapping relationship between the prediction error and the time step is established, based on the task scheduling and the warehouse-out priority of the item, when the error of three consecutive time windows exceeds the set threshold, the time step shortening mechanism is triggered, and when the error continuously falls below the set threshold, the time step is extended, and the future path of each mechanical arm is planned.

2. The urea warehouse three-dimensional modeling and intelligent scheduling system of claim 1, wherein, The future time step space occupation heat map sequence is a space-time conflict prediction situation map, which comprises a space occupation heat map sequence of the urea warehouse in a future preset time.

3. The urea warehouse three-dimensional modeling and intelligent scheduling system of claim 2, wherein, The preprocessing comprises image denoising, illumination compensation and dynamic enhancement, and target recognition and positioning; Image denoising: the collected internal image of the urea warehouse is denoised using a trained neural network model; Illumination compensation and dynamic enhancement: a local contrast enhancement algorithm is introduced to enhance the internal image of the urea warehouse; Target recognition and positioning: the objects in the urea warehouse are recognized and the item boundary is positioned.

4. The urea warehouse three-dimensional modeling and intelligent scheduling system of claim 3, wherein, The item comprises bagged urea, a transport truck, a mechanical arm and a warehouse tool.

5. The urea warehouse three-dimensional modeling and intelligent scheduling system of claim 4, wherein, The three-dimensional model updating unit adopts an incremental updating mechanism.

Images