A method for measuring fertilizer material volume change based on multi-sensor information fusion

Abstract

The present invention is applicable to the field of agricultural information technology and provides a method for measuring fertilizer material volume changes based on multi-sensor information fusion. The method comprises the following steps: physical modeling of the detection area; selection of a depth camera and determination of its angle; point cloud data acquisition and processing; and splicing and quantification of target areas of previous and subsequent point clouds. This method maps the storage space through physical modeling, utilizes multiple sensors for data acquisition and reconstruction, and accurately extracts point cloud data of the fertilizer area using a point cloud segmentation algorithm. The system calculates the change in volume through periodic data acquisition and accurately quantifies the volume change using voxel technology. This method not only provides real-time volume change measurement but also allows for adjustment of measurement accuracy as needed, making it suitable for storage spaces containing a variety of materials, including fertilizers and grains.

Description

Technical Field

[0001] The invention belongs to the field of agricultural information technology, and in particular relates to a fertilizer material volume change measurement method based on multi-sensor information fusion. Background Art

[0002] Currently, a common method for fertilizer storage is to estimate the amount of fertilizer discharged based on the speed and area of ​​the turntable. However, this method has significant limitations. Because the fertilizer accumulation in each discharge hopper does not fully fill the entire storage space, there is a large error in calculating the volume change. This traditional calculation method cannot accurately reflect the actual remaining volume of fertilizer, especially when the material is unevenly accumulated, where the error is particularly significant, leading to uncertainty in warehouse management and logistics scheduling.

[0003] To address this shortcoming, many studies in recent years have attempted to use alternative algorithms to measure fertilizer volume. Existing technologies include LiDAR systems (also known as laser radar systems), which use visual algorithms to estimate object volume. This approach provides relatively accurate results, especially in relatively regular storage areas. However, these devices still have several limitations:

[0004] (1) Inconvenient installation and use of the equipment: The LiDAR system requires the equipment to be wrapped in the unloading area, and the equipment needs to be disassembled and reinstalled each time the material is unloaded, which greatly reduces the convenience of operation.

[0005] (2) Only calculating the existing storage volume, unable to measure volume changes: Existing LiDAR systems and related technologies are generally only used to model and measure the existing storage volume, and cannot effectively track the changes in the amount of materials at different time points. When calculating the remaining volume of fertilizers or other materials, these systems mainly focus on the space occupancy at a fixed moment, and lack the dynamic monitoring and calculation of volume changes in time series. Therefore, it is impossible to provide accurate remaining volume measurement or prediction, resulting in an inability to effectively support real-time inventory management and replenishment decisions.

[0006] (3) Insufficient coverage: This device focuses on the volume changes of fertilizers in the center of the storage area and cannot fully cover the entire storage area. Since the measurement is limited to a certain area, this method cannot accurately reflect the actual situation of fertilizer changes in the entire storage space.

[0007] Therefore, existing technologies cannot meet the needs of accurate, real-time and comprehensive volume measurement in environments such as fertilizer silos and granaries. Summary of the Invention

[0008] The purpose of the embodiments of the present invention is to provide a method for measuring the volume change of fertilizer materials based on multi-sensor information fusion, aiming to solve the problems raised in the above background technology.

[0009] The embodiment of the present invention is implemented as follows: a method for measuring the volume change of fertilizer materials based on multi-sensor information fusion, comprising the following steps:

[0010] S1: Physical modeling of the detection area (fertilizer silo or granary);

[0011] In fertilizer or grain silos, first determine the boundaries and structure of the measurement area, taking into account the warehouse's shape, size, and material stacking characteristics. Based on the warehouse's structure, create a standardized physical model that adapts to different storage environments, providing a foundation for subsequent measurement, camera configuration, and data acquisition.

[0012] S2: Select the depth camera and determine the angle (based on the simulation of the depth camera FOV);

[0013] Select the appropriate depth camera type based on the size of the measurement area and the required accuracy. Choose the appropriate camera configuration based on the depth camera's FOV (field of view) and the height and coverage of the measurement area. Simulate the angle and placement of the depth camera through simulation, adjust the angle to avoid blind spots, and ensure that the depth camera can fully cover the measurement area.

[0014] S3: point cloud data acquisition and processing;

[0015] During the point cloud data acquisition and processing process, multiple depth cameras are first controlled synchronously to collect data to ensure the temporal consistency of the collected data. Then, the point cloud data collected by multiple depth cameras are registered and reconstructed in the same coordinate system to generate a complete 3D model. The point cloud data is then optimized, including denoising and removing blank areas.

[0016] S4: stitching and quantification of target areas of previous and subsequent point clouds;

[0017] The point cloud data collected at different time points are compared and processed. By comparing the data before and after, the changes in the material can be understood in real time. The volume change is calculated based on the difference, and the change in the storage medium volume is quantified using voxel technology.

[0018] A further technical solution is that in said S1, the physical model includes the basic structure of the warehouse, which includes the height, width and actual distribution of stored materials in the warehouse, as well as the stacking method of the stored materials in the warehouse, the looseness of the stacking of the materials, and the shape and size of the materials.

[0019] A further technical solution is that in S2, the optimal placement position of the depth camera is simulated according to the physical model created in step S1 to ensure that the field of view of the depth camera covers the entire measurement area; the effects under different depth camera configurations are simulated by simulation software, and the placement angle and height of the depth camera are adjusted to avoid measurement blind spots.

[0020] A further technical solution is that in said S3, after the first scan is classified, the target point cloud is , where each point is a point in three-dimensional space; after the second scan and classification, the target point cloud is , where each point is a point in three-dimensional space.

[0021] Through the above steps, the point cloud data obtained and , used for subsequent mesh reconstruction and volume calculation.

[0022] A further technical solution is that in said S4, a local incremental triangular mesh construction method is used to reconstruct the mesh; firstly, point cloud data is selected Three points in 、 and As the initial triangle , and expand the grid by gradually selecting points; each time a candidate point is selected , you need to check its projection relationship with the current triangle to ensure that it meets the following conditions:

[0023] Projection distance condition: Calculate candidate points The projection distance to the plane where the current triangle is located ,If the projection distance is less than the set threshold, the point is considered close to the current plane and can be added to the grid;

[0024] Edge length condition: The difference in the length of the three sides of the generated new triangle should be less than the set maximum difference ;

[0025] Internal angle condition: The internal angle of the newly generated triangle should be within the preset range;

[0026] Repeat the above process until all point cloud data and are connected into a complete 3D mesh model; through this process, an accurate triangular mesh model can be obtained from the point cloud data M 1 and M 2;

[0027] Next, the generated triangular mesh model is voxelized; the side length of the voxel unit is set to , divide the three-dimensional space into uniform cubic voxel units , determine whether each voxel is located inside the grid model by using the ray method or breadth-first search method. If the center of the voxel is inside the grid model, it is marked as an occupied voxel;

[0028] Count the number of voxels inside the model N 1 and N 2, corresponding to the mesh models obtained by the first scan and the second scan respectively; according to the volume of each voxel , calculate the volume of the model after two scans:

[0029] ;

[0030] By comparing the volumes of the two scans, the volume difference is obtained , that is, the volume change of the model twice:

[0031] ;

[0032] Through the above process, the volume change of the storage medium or material between two scans is estimated, that is, the volume monitoring result.

[0033] The embodiment of the present invention provides a method for measuring fertilizer material volume change based on multi-sensor information fusion, which has the following beneficial effects:

[0034] (1) No need for complex equipment installation and disassembly: By installing multiple depth cameras in fixed positions, the tedious operation of frequently disassembling and installing equipment in traditional methods is avoided, which greatly simplifies the operation process of the equipment and improves the degree of automation of the system, allowing users to continuously perform efficient volume measurement without interfering with daily operations. In addition, the system's automation function reduces interference from human factors and ensures the efficiency and consistency of data collection.

[0035] (2) Comprehensive coverage of the storage area: Multiple depth cameras collect data simultaneously from different angles, ensuring comprehensive coverage of the storage area. Regardless of the size or shape of the storage space, multi-camera fusion technology integrates the data collected by different depth cameras to form a complete three-dimensional model. This comprehensive coverage capability ensures the accuracy of volume calculation results, especially in cases where materials are unevenly stacked or the storage area is complex. It can effectively avoid omissions or errors in traditional methods and significantly improve the reliability and accuracy of measurements.

[0036] (3) Adjustable measurement accuracy: Through depth cameras and point cloud data processing technology, the measurement accuracy can be adjusted according to actual needs, ensuring efficient and accurate volume measurement in storage spaces of various sizes. Whether it is a small fertilizer silo or a large grain silo, it can be optimized according to different size and accuracy requirements, making this method more adaptable in practical applications and able to meet the diverse needs of agriculture, warehousing, logistics and other fields.

[0037] Through the above three improvements, this method not only improves the efficiency and accuracy of measurement, but also enhances its applicability in practical applications, especially in large-scale warehouse management and logistics scheduling, and has significant promotion prospects and market value. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A flow chart of a method for measuring fertilizer material volume change based on multi-sensor information fusion provided by an embodiment of the present invention;

[0039] Figure 2 Schematic diagram of the depth camera projection (a is vertical installation, b is installation at a specific angle);

[0040] Figure 3 For visualization of the reconstructed image (where a is the overall side view, b is the top view, c is the main view, and d is the left view). DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0042] The specific implementation of the present invention is described in detail below with reference to specific embodiments.

[0043] like Figure 1 As shown in FIG, a method for measuring the volume change of fertilizer materials based on multi-sensor information fusion is provided in one embodiment of the present invention, comprising the following steps:

[0044] S1: Physical modeling of the detection area (fertilizer silo or granary);

[0045] In a fertilizer or grain silo, the measurement area's boundaries and structure are first determined, taking into account the warehouse's shape, size, and material storage characteristics. Based on the warehouse's structure, a standardized physical model is created to adapt to different storage environments. This provides a foundation for subsequent measurement, camera configuration, and data collection, ensuring high-precision measurement in warehouses of varying sizes. Through detailed modeling of the storage environment, the system determines camera placement based on the physical model and optimizes camera configuration based on the storage environment's characteristics, ensuring there are no blind spots within the measurement area. This provides an accurate reference framework for subsequent point cloud data collection and processing.

[0046] S2: Select the depth camera and determine the angle (based on the simulation of the depth camera FOV);

[0047] Based on the size of the measurement area and the required accuracy, select the appropriate depth camera type (such as a structured light camera, LiDAR, or a binocular camera). Choose the appropriate camera configuration based on the depth camera's FOV (field of view) and the height and coverage of the measurement area. By default, the cameras are placed at two of the four corners of the top of the fertilizer or grain silo to ensure full coverage of the measurement area. In real-world applications, simulations can be used to simulate the angle and placement of the depth camera, allowing adjustments to avoid blind spots and ensure optimal measurement results.

[0048] S3: point cloud data acquisition and processing;

[0049] During the point cloud data acquisition and processing process, data is first collected by synchronously controlling multiple depth cameras. The synchronous operation of multiple depth cameras ensures the temporal consistency of the collected data. Then, the point cloud data collected by multiple depth cameras will be aligned and reconstructed in the same coordinate system to generate a complete three-dimensional model, and the point cloud data will be optimized, including denoising and removing blank areas. Parallel computing and streaming processing technologies are used to ensure that point cloud data can enter the processing pipeline in real time, avoiding delays in large-scale data processing. Specifically, point cloud acquisition, alignment, and reconstruction are executed in parallel on multiple processing units, and the current model is updated in real time. By optimizing the algorithm, the processing time can be effectively reduced, ensuring that data can be processed synchronously and output in real time in a multi-sensor environment.

[0050] S4: stitching and quantification of target areas of previous and subsequent point clouds;

[0051] Point cloud data collected at different time points is compared and processed to accurately estimate changes in material volume. By comparing data from previous and subsequent times, material changes can be understood in real time. Volume changes can be calculated based on the differences, and voxel technology is used to quantify the volume change of the storage medium. During the volume change calculation process, voxelization technology dynamically adjusts based on the set voxel size. In critical areas, the system achieves higher calculation accuracy using smaller voxel sizes; in less important areas or areas with less change, larger voxel sizes are used to increase calculation speed. The error control module corrects measurement errors in the point cloud data in real time, ensuring high accuracy and stability of the volume calculation results. Through this adaptive precision control strategy, the system effectively reduces the computational burden and improves efficiency while ensuring accuracy.

[0052] As a preferred embodiment of the present invention, in S1, the required physical model includes not only the basic structure of the warehouse but also considers the stacking pattern and characteristics of the stored materials within the warehouse, such as the shape, size, and looseness of the materials. By analyzing the material stacking characteristics, the physical model can be better adjusted to suit different storage environments. During the physical modeling process, model parameters such as the warehouse height and width and the actual distribution of stored materials need to be adjusted according to the specific environment.

[0053] As a preferred embodiment of the present invention, in step S2, the type and configuration of the depth camera should take into account the storage environment defined in the physical model in step S1. Specifically, the optimal placement of the depth camera is simulated based on the physical model created in step S1 to ensure that the depth camera's field of view covers the entire measurement area. By using simulation software, the effects of different depth camera configurations can be simulated, the placement angle and height of the depth camera can be adjusted to avoid any potential measurement blind spots, and the installation and configuration of the depth camera can be further optimized to ensure that the collection of point cloud data covers the entire storage area. Figure 2 (a) Two depth cameras are mounted vertically, but the FOV cannot fully cover the bottom surface. (See Figure 2b) The two top depth cameras are mounted at a special angle, and the FOV fully covers the entire bottom surface.

[0054] As a preferred embodiment of the present invention, in said S3, during the point cloud data acquisition and processing process, the synchronous operation of multiple depth cameras is based on the depth camera configuration and simulation results in step S2. The synchronous control of the depth camera can ensure that the data collected by each depth camera is completely consistent in time, which is crucial for the subsequent point cloud reconstruction. In order to improve the quality of point cloud data, the collected data needs to undergo a series of preprocessing and optimization operations, including noise removal, blank area filling, etc. After data preprocessing, point cloud segmentation is performed to separate point clouds in different areas, such as walls, storage media, etc., to ensure the accuracy of subsequent measurements. In this way, step S3 is closely related to the previous steps S1 and S2, ensuring that the collected data can be efficiently and accurately processed with the support of the physical model and camera configuration.

[0055] Assume that after the first scan is classified, the target point cloud is , where each point is a point in three-dimensional space. After the second scan and classification, the target point cloud is , where each point is a point in three-dimensional space.

[0056] Through the above steps, the point cloud data obtained and Used for subsequent mesh reconstruction and volume calculation.

[0057] As a preferred embodiment of the present invention, in said S4, the previous and next point cloud data used are derived from the point cloud data collected synchronously in S3. By comparing and splicing the previous and next point cloud data, the changes in the storage medium area can be observed intuitively. Voxel calculation technology can quantify the amount of change and calculate the precise volume change. In this process, the optimization of the voxel size depends on the physical model and camera configuration defined in S1 and S2. The size of the voxel needs to be dynamically adjusted according to the characteristics of the storage environment and the required measurement accuracy, so as to avoid excessive computational burden while ensuring measurement accuracy. First, the point cloud data obtained by the two scans are combined. and Preprocessing is performed separately, including denoising, filling blank areas, etc. After these steps, the target point cloud set is obtained.

[0058] On this basis, a local incremental triangular mesh construction method is used to reconstruct the mesh. Specifically, first select the point cloud data Three points in 、 and As the initial triangle , and expand the grid by gradually selecting points. Each time a candidate point is selected , you need to check its projection relationship with the current triangle to ensure that it meets the following conditions:

[0059] Projection distance condition: Calculate candidate points The projection distance to the plane where the current triangle is located ,If the projection distance is less than the set threshold, the point is considered close to the current plane and can be added to the grid;

[0060] Edge length condition: The difference in the length of the three sides of the generated new triangle should be less than the set maximum difference ;

[0061] Inner angle condition: The inner angle of the newly generated triangle should be within the preset range to avoid generating unreasonable triangles.

[0062] Repeat the above process until all point cloud data and All of them are connected into a complete 3D mesh model. Through this process, an accurate triangular mesh model can be obtained from point cloud data. M 1 and M 2. See Figure 3 As shown, it is a visual reconstructed image. Figure 3 a is the overall side view, Figure 3 b is the top view among the three views, Figure 3 c is the main view among the three views, Figure 3 d is the left view among the three views.

[0063] Next, the generated triangular mesh model is voxelized. Set the side length of the voxel unit to , divide the three-dimensional space into uniform cubic voxel units Each voxel is determined to be inside the mesh by using ray tracing or breadth-first search (BFS). If the center of the voxel is inside the mesh, it is marked as occupied.

[0064] Count the number of voxels inside the model N 1 and N 2, corresponding to the mesh models obtained by the first and second scans. According to the volume of each voxel , the volume of the twice scanned model can be calculated:

[0065] ;

[0066] By comparing the volumes of the two scans, the volume difference can be obtained , that is, the volume change of the model twice:

[0067] ;

[0068] Through the above process, the volume change of the storage medium or material between two scans can be effectively estimated, providing accurate volume monitoring results.

[0069] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for measuring fertilizer material volume change based on multi-sensor information fusion, characterized in that: The following steps are involved: S1: Physical modeling of the detection area; In fertilizer or grain silos, first determine the boundaries and structure of the measurement area, taking into account the warehouse's shape, size, and material stacking characteristics. Based on the warehouse's structure, create a standardized physical model that adapts to different storage environments. S2: Select the depth camera and determine the angle; Select the depth camera type based on the size of the measurement area and the required accuracy. Choose the camera configuration based on the depth camera's FOV and the height and coverage of the measurement area. Use simulation to simulate the depth camera's angle and placement, adjusting the angle to ensure the depth camera fully covers the measurement area. S3: point cloud data acquisition and processing; During the point cloud data acquisition and processing process, multiple depth cameras are first controlled synchronously to collect data to ensure the temporal consistency of the collected data. Then, the point cloud data collected by multiple depth cameras are registered and reconstructed in the same coordinate system to generate a complete 3D model. The point cloud data is then optimized, including denoising and removing blank areas. S4: stitching and quantification of target areas of previous and subsequent point clouds; Compare and process point cloud data collected at different time points. By comparing the data before and after, understand the changes in materials in real time, calculate the volume change based on the difference, and use voxel technology to quantify the change in volume of the storage medium. In S3, it is assumed that after the first scan is classified, the target point cloud is , where each point is a point in three-dimensional space; after the second scan and classification, the target point cloud is , where each point is a point in three-dimensional space; Through the above steps, the point cloud data obtained and , used for subsequent mesh reconstruction and volume calculation; In the S4, a local incremental triangular mesh construction method is used to reconstruct the mesh; first, point cloud data is selected Three points in and As the initial triangle , and expand the grid by gradually selecting points; each time a candidate point is selected , you need to check its projection relationship with the current triangle to ensure that it meets the following conditions: Projection distance condition: Calculate candidate points The projection distance to the plane where the current triangle is located ,If the projection distance is less than the set threshold, the point is considered close to the current plane and can be added to the grid; Edge length condition: The difference in the length of the three sides of the generated new triangle should be less than the set maximum difference ; Internal angle condition: The internal angle of the newly generated triangle should be within the preset range; Repeat the above process until all point cloud data and are connected into a complete 3D mesh model; through this process, an accurate triangular mesh model can be obtained from the point cloud data M 1 and M 2； Next, the generated triangular mesh model is voxelized; the side length of the voxel unit is set to , divide the three-dimensional space into uniform cubic voxel units , determine whether each voxel is located inside the grid model by using the ray method or breadth-first search method. If the center of the voxel is inside the grid model, it is marked as an occupied voxel; Count the number of voxels inside the model N 1 and N 2, corresponding to the mesh models obtained by the first scan and the second scan respectively; according to the volume of each voxel , calculate the volume of the model after two scans: ； By comparing the volumes of the two scans, the volume difference is obtained , that is, the volume change of the model twice: ； Through the above process, the volume change of the storage medium or material between two scans is estimated, that is, the volume monitoring result.

2. The method for measuring fertilizer volume change based on multi-sensor information fusion according to claim 1, characterized in that: In S1, the physical model includes the basic structure of the warehouse, which includes the height, width and actual distribution of stored materials, as well as the stacking method of stored materials in the warehouse, the looseness of the stacking of materials, and the shape and size of the materials.

3. The method for measuring fertilizer volume change based on multi-sensor information fusion according to claim 1, characterized in that: In step S2, the optimal placement of the depth camera is simulated according to the physical model created in step S1 to ensure that the field of view of the depth camera covers the entire measurement area; the effects under different depth camera configurations are simulated by simulation software, and the placement angle and height of the depth camera are adjusted to avoid measurement blind spots.

Images