A method and system for detecting overloading and unbalanced loading of a vehicle compartment based on a drone

Abstract

The present application is suitable for the technical field of overload and eccentric load detection, and particularly relates to a carriage overload and eccentric load detection method and system based on a UAV. The present application obtains three-dimensional point cloud data, visible light images and spectral scattering images of a target carriage, constructs a three-dimensional model, divides the three-dimensional model into multiple carriage slices, determines the actual accumulation volume, porosity and gravity center coordinates of the goods of each carriage slice, determines the actual accumulation density of the goods of each carriage slice in combination with the standard density and porosity of the goods, determines the total mass of the goods of the target carriage and the actual gravity center coordinates of the goods according to the actual accumulation volume, actual accumulation density and gravity center coordinates of the goods, and compares the total mass of the goods and the actual gravity center coordinates of the goods with the rated load and the preset ideal gravity center range respectively to determine the loading state and the transfer balancing state of the target carriage, thereby significantly improving the accuracy, detection precision and efficiency of carriage overload and eccentric load detection and being suitable for dynamic scene detection.

Description

Technical Field

[0001] This invention relates to the field of overload and off-center load detection technology, and in particular to a method and system for detecting overload and off-center load in a vehicle compartment based on an unmanned aerial vehicle (UAV). Background Technology

[0002] In existing technologies, a gantry crane equipped with lidar and image acquisition equipment is typically installed on the ground to detect overload and off-center loading in railway freight cars. The cars are scanned as the train passes over the gantry. However, because the train is in motion when passing the gantry, vibrations and tilting interference are transmitted to the gantry and its mounted detection equipment through multiple paths, including the track, air, and foundation. This increases the measurement error of the lidar and image acquisition equipment under dynamic detection conditions, resulting in low accuracy and poor reliability in overload and off-center loading detection. Furthermore, existing technologies use multi-rotor UAV systems carrying laser payloads to scan and reconstruct the materials in the cars, calculating the volume of the cargo. The total mass of the cargo is then determined using the standard density of the cargo and the calculated volume, thus assessing the loading status of the cars. However, this approach suffers from problems such as relying on a single source of data for calculating the cargo volume and significant calculation errors. In addition, existing technologies collect the weight data of the car body by installing a weighing and weighing module in a specific off-center load testing site, and use the torque balance principle to calculate the center of gravity coordinates of the cargo loaded in the car body to determine the load balance status of the car body. However, this method can only achieve static weighing and is easily limited by the testing site. Summary of the Invention

[0003] This invention provides a method and system for detecting overloaded carriages based on unmanned aerial vehicles (UAVs), to solve the problems of low accuracy and poor reliability in the existing technology for detecting overloaded carriages; this invention also provides a method and system for detecting off-center loads on carriages based on unmanned aerial vehicles (UAVs), to solve the problems of large errors and poor reliability in the existing technology for detecting off-center loads on carriages.

[0004] In a first aspect, the present invention provides a method for detecting overloaded carriages based on unmanned aerial vehicles (UAVs), the method comprising: Step 100: Obtain the 3D point cloud data, visible light image, spectral scattering image, rated load, and standard density of the cargo in the target carriage loaded with cargo at the current moment; Step 200: Construct a three-dimensional model of the target carriage based on the three-dimensional point cloud data, the visible light image, and the spectral scattering image; Step 300: Along the length of the target carriage, the three-dimensional model is divided into N carriage slices according to the preset carriage slice width. The actual stacking volume of the cargo in the N carriage slices and the porosity of the cargo in the N carriage slices are determined, where N is a positive integer greater than 1. Step 400: Determine the actual bulk density of the N car body slices based on the standard density of the cargo and the porosity of the cargo in the N car body slices. Step 500: Determine the total weight of the cargo in the target car based on the actual stacking volume of the cargo in the N car body slices and the actual stacking density of the cargo in the N car body slices. Step 600: When the total weight of the cargo is less than or equal to the rated load, the loading status of the target carriage is determined to be rated load status; when the total weight of the cargo is greater than the rated load, the loading status of the target carriage is determined to be overload status.

[0005] Secondly, the present invention provides a method for detecting off-center load on a train carriage based on unmanned aerial vehicles (UAVs), the method comprising: Step S100: Obtain the 3D point cloud data, visible light image, spectral scattering image, standard density of the cargo, X-axis tilt angle, and Y-axis tilt angle of the target carriage loaded with cargo at the current moment; Step S200: Construct a three-dimensional model of the target carriage based on the three-dimensional point cloud data, the visible light image, and the spectral scattering image; Step S300: Along the length of the target carriage, the three-dimensional model is divided into N carriage slices according to the preset carriage slice width. The actual stacking volume of the goods in the N carriage slices, the porosity of the goods in the N carriage slices, and the center of gravity coordinates of the goods in the N carriage slices are determined, where N is a positive integer greater than 1. Step S400: Determine the actual bulk density of the N car body slices based on the standard density of the cargo and the porosity of the cargo in the N car body slices. Step S500: Determine the actual center-of-gravity coordinates of the cargo in the target car based on the actual stacking volume of the cargo in the N car body slices, the actual stacking density of the cargo in the N car body slices, the center-of-gravity coordinates of the cargo in the N car body slices, the X-axis tilt angle, and the Y-axis tilt angle. Step S600: When the actual center of gravity coordinates of the cargo are within the preset ideal center of gravity range, the loading balance state of the target car is determined to be a balanced loading state; when the actual center of gravity coordinates of the cargo exceed the ideal center of gravity range, the loading balance state of the target car is determined to be an off-center loading state.

[0006] Thirdly, the present invention provides a vehicle overload detection system based on unmanned aerial vehicles (UAVs). The UAV-based vehicle overload detection system includes a ground processing terminal and a UAV. The ground processing terminal and the UAV communicate with each other, and the ground processing terminal and the UAV cooperate to implement the UAV-based vehicle overload detection method as described in the first aspect.

[0007] Fourthly, the present invention provides a vehicle carriage overload detection system based on unmanned aerial vehicles (UAVs). The UAV-based vehicle carriage overload detection system includes a ground processing terminal and a UAV. The ground processing terminal and the UAV communicate with each other, and the ground processing terminal and the UAV cooperate to implement the UAV-based vehicle carriage overload detection method as described in the second aspect.

[0008] The aforementioned UAV-based method and system for detecting overloaded train carriages acquires 3D point cloud data, visible light images, spectral scattering images, rated load, and standard density of the cargo in the target carriage currently loaded with goods. Then, based on the 3D point cloud data and visible light images, a 3D model is constructed and divided into N carriage slices. The actual bulk volume and porosity of the cargo in each carriage slice are determined. Next, based on the standard density and porosity of the cargo, the actual bulk density of the cargo in each carriage slice is determined. Finally, based on the actual bulk volume and density of each cargo, the total weight of the cargo in the target carriage is determined. When the total weight is less than or equal to the rated load, the target carriage is considered to be under rated load; when the total weight exceeds the rated load, the target carriage is considered to be overloaded. Compared to existing technologies, this invention utilizes the standard density and porosity of the cargo to determine the actual bulk density of the cargo in the target carriage, thereby determining the total weight of the cargo. This achieves high-precision detection under normal train operation, significantly improving the accuracy and precision of overload detection.

[0009] The aforementioned UAV-based method and system for detecting off-center loading of cargo compartments first acquires the 3D point cloud data, visible light image, standard density of the cargo, X-axis tilt angle, and Y-axis tilt angle of the target cargo compartment currently loaded with cargo. Based on the 3D point cloud data and visible light image, a 3D model is constructed and divided into N compartment slices. The actual bulk volume, porosity, and initial center of gravity coordinates of the cargo in each compartment slice are determined. Then, based on the standard density and porosity of the cargo, the actual bulk density of the cargo in each compartment slice is determined. Finally, based on the actual bulk volume, porosity, initial center of gravity coordinates, X-axis tilt angle, and Y-axis tilt angle, the actual center of gravity coordinates of the cargo in the target cargo compartment are determined. When the actual center of gravity coordinates are within a preset ideal center of gravity range, the loading state of the target cargo compartment is determined to be balanced loading; when the actual center of gravity coordinates exceed the preset ideal center of gravity range, the loading state of the target cargo compartment is determined to be off-center loading. Compared with existing technologies, this invention uses the standard density, porosity, and initial center of gravity coordinates of the cargo to determine the actual center of gravity coordinates of the cargo in the target carriage, achieving accurate calculation of the actual three-dimensional center of gravity of the cargo in the carriage. This breaks through the limitation of traditional weighing methods that cannot sense the height of the center of gravity, significantly improving the accuracy and anti-interference ability of off-center load detection, and ensuring the efficiency and non-contact nature of the detection process. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of an application environment for a method for detecting overload and off-center load in a carriage based on an unmanned aerial vehicle (UAV) according to an embodiment of the present invention. Figure 2 This is a flowchart of a method for detecting overloaded carriages based on unmanned aerial vehicles (UAVs) according to an embodiment of the present invention; Figure 3 This is a flowchart of a method for detecting off-center load in a vehicle compartment based on an unmanned aerial vehicle (UAV) according to an embodiment of the present invention. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] The UAV-based method for detecting overloaded carriages provided in this invention can be applied to, for example... Figure 1 The application environment is shown. Specifically, this UAV-based method for detecting overloaded carriages is applied in a UAV-based carriage overload detection system, which includes, as shown in the example... Figure 1 The diagram shows a ground processing terminal and a drone. The ground processing terminal communicates with the drone via a network to achieve real-time, drone-based detection of overloaded cargo compartments. The ground processing terminal, also known as the user terminal, refers to the program that provides local services to customers, corresponding to the drone. The ground processing terminal can be installed on, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The drone can use an integrated 3D LiDAR and a high-resolution SLR camera to simultaneously acquire 3D point cloud and visible light images of the cargo compartment.

[0014] In one embodiment, such as Figure 2 As shown, this embodiment provides a method for detecting overloaded carriages based on unmanned aerial vehicles (UAVs), which is applied to... Figure 1 Taking the ground processing terminal as an example, the UAV-based method for detecting overloaded carriages includes: Step 100: Obtain the 3D point cloud data, visible light image, spectral scattering image, rated load, and standard density of the cargo in the target carriage loaded with cargo at the current moment; The target car refers to a railway freight car, road freight car, or mining transport car loaded with goods and subject to overload detection. The target car has known fixed dimensional parameters, including but not limited to its internal length, width, and height. The 3D point cloud data refers to a set of 3D coordinate points collected by a lidar scanning system mounted on a drone, reflecting the spatial position information of the target car and its cargo surface. Each point contains X, Y, and Z axis coordinate values ​​in a Cartesian coordinate system, used to characterize the geometric contour and volume distribution of the car and its cargo. The visible light image refers to a 2D digital image collected by a high-definition visible light camera mounted on a drone, reflecting the surface texture, color, and edge information of the target car and its cargo surface, used to assist in the segmentation, recognition, and modeling of the 3D point cloud data, distinguishing the inner wall of the car from the cargo area. The term "domain" refers to the image data acquired by a hyperspectral imager within a specific wavelength range (typically visible to near-infrared, such as 450–2500 nm), containing information on the distribution of scattered light intensity at different spatial locations and wavelengths on the surface of cargo. This image records the two-dimensional or three-dimensional distribution of scattered light intensity generated by the interaction of light with particles on the cargo surface, varying with spatial distance and wavelength. It is used to characterize the physical structural properties of the cargo (such as particle packing state, surface roughness, density distribution, etc.). "Rated load capacity" refers to the maximum cargo mass that the target wagon is designed and manufactured to carry under safe operating conditions, serving as the benchmark threshold for overload judgment. "Standard density of cargo" refers to the inherent physical density value of the cargo loaded in the target wagon under natural packing conditions, expressed in kg / m³. 3 Standard density can be derived from cargo attribute databases, material manuals, or pre-calibrated.

[0015] For example, before conducting overweight detection of the carriages, the equipment selection process begins with choosing a multi-rotor drone (DJI M300 RTK) equipped with a LiDAR scanning system (model: Velodyne VLP-16) and a high-definition visible light camera (resolution: 4K; frame rate: 30f / s). The drone's flight stability is then tested, the LiDAR scanning accuracy is calibrated (error less than or equal to ±2cm), and the synchronization accuracy of the high-definition visible light camera and LiDAR scanning system is adjusted (time synchronization error less than or equal to 50ms) to ensure proper equipment operation. Subsequently, manual inspection is used to determine the carriage model and cargo type indicated on the carriage nameplate. The rated load capacity and internal dimensions (length, width, and height) of the carriage are further determined by consulting railway freight car technical manuals. The standard density value of the loaded cargo is obtained by consulting relevant national standards, and the carriage's rated load capacity, internal dimensions, cargo type, and standard density are then input into the ground processing terminal.

[0016] Based on the received information about the carriages and cargo, the ground processing terminal plans the UAV's flight path and data acquisition parameters, and sends the operation instructions to the UAV. For example, the UAV flies above the center line of the target carriage, with the flight direction consistent with the train's direction of travel; the flight altitude is 3m above the top of the carriage, and the horizontal offset is 0 (i.e., flying directly above), ensuring that the scanning range of the lidar scanning system completely covers the interior of the carriage and the surface of the cargo, with no blind spots; the preset flight speed, taking into account the UAV's flight stability, the lidar scanning system's scanning frequency (20 Hz), the image acquisition frame rate (30 f / s), and the ambient vibration (wind less than or equal to 5 m / s), determines the flight speed to be 5 m / s, ensuring that the acquired 3D point cloud data and image data are complete, clear, and without distortion.

[0017] The UAV receives operational commands from the ground processing terminal and flies at a constant speed above the target carriage according to the preset flight path and data acquisition parameters. During flight, the UAV maintains a stable relative position to the carriage. The lidar scanning system continuously acquires 3D point cloud data of the carriage's interior and the loaded cargo. Simultaneously, a high-definition visible light camera acquires visible light images of the target carriage and its cargo, with each frame covering a 1-meter area along the carriage's length, ensuring complete image coverage and clear, discernible images. During the acquisition process, the ground processing terminal monitors the data transmission status in real time, ensuring that the 3D point cloud data and visible light images acquired by the UAV are transmitted synchronously to the ground processing terminal without data loss or transmission delay. The ground processing terminal then performs preliminary verification of the acquired raw data, discarding blurry, distorted images and invalid point cloud data.

[0018] Step 200: Construct a three-dimensional model of the target carriage based on the three-dimensional point cloud data, the visible light image, and the spectral scattering image; Among them, the three-dimensional model refers to a three-dimensional digital representation that can realistically reflect the spatial state of the target carriage and its loaded cargo from both geometric shape and texture features. It is constructed by fusing three-dimensional point cloud data collected by a lidar scanning system, visible light images collected by a high-definition visible light camera, and spectral scattering images collected by a hyperspectral imager.

[0019] For example, the ground processing terminal uses a Gaussian filtering algorithm to denoise the original 3D point cloud data, removing noise points caused by environmental interference (such as dust and light), and uses a pass-through filtering algorithm to remove irrelevant 3D point clouds outside the carriage (such as the surrounding environment and other train components), retaining the 3D point cloud data of the carriage interior and loaded cargo. Subsequently, coordinate transformation is performed to convert the relative coordinates collected by the lidar scanning system into absolute coordinates in the carriage coordinate system with a specified corner of the carriage as the origin, to ensure the spatial accuracy of the 3D point cloud data. The carriage coordinate system is constructed as follows: the corner where the carriage floor intersects with the left side wall and the rear end wall is the origin; the width direction of the carriage (from the left side wall to the right side wall) is the positive X-axis; the length direction of the carriage (from the rear end wall to the front end wall) is the positive Y-axis; and the height direction of the carriage (vertically upward from the floor) is the positive Z-axis.

[0020] The ground processing terminal preprocesses the visible light images, including grayscale conversion and edge enhancement, and extracts the texture features and edge contours of the loaded cargo surface. Then, it registers the preprocessed image data with the preprocessed 3D point cloud data, and uses the image texture information to assist in filling the holes in the 3D point cloud data, filling the point cloud holes caused by occlusion and scanning angle, and improving the integrity of the point cloud data.

[0021] For the acquired spectral scattering images, spectral data of the loaded cargo surface is obtained, and the digital signal of the original image is converted into physically meaningful reflectance. Then, after radiometric calibration and reflectance conversion, the reflectance spectral curve of the cargo surface is obtained. The acquired original spectral signal is denoised and enhanced; common preprocessing methods include multivariate scattering correction and standard normal variable transformation. To eliminate spectral redundancy and improve model efficiency, a feature wavelength screening algorithm is used. Common algorithms in hyperspectral inversion of loaded cargo include continuous projection algorithms and competitive adaptive reweighting algorithms. Using pre-calibrated spatial transformation parameters and each pixel in the hyperspectral image, the corresponding 3D point cloud points are found through the calibration parameters. The point cloud is divided into a regular voxel grid, and the storage location, point cloud density, and average reflectance spectral curve of each voxel are determined. The average reflectance spectral curve of the hyperspectral pixels is directly assigned to the voxels using direct assignment or weighted averaging methods.

[0022] Based on the 3D point cloud data (containing voxel information) after void filling, a 3D geometric model of the target carriage and loaded cargo is constructed using a triangular meshing algorithm. Using image texture mapping technology, texture information, spectral information, and granularity features of high-definition visible light images and spectral scattering images are mapped to the 3D geometric model to obtain a 3D model that can clearly present the stacking shape of the loaded cargo and the loading situation inside the carriage.

[0023] Step 300: Along the length of the target carriage, the three-dimensional model is divided into N carriage slices according to the preset carriage slice width. The actual stacking volume of the cargo in the N carriage slices and the porosity of the cargo in the N carriage slices are determined, where N is a positive integer greater than 1. The width of the wagon slice refers to the thickness of each wagon slice along the length (Y-axis) of the 3D model. A wagon slice is an independent sub-3D model block obtained by equidistantly cutting the 3D model of the target wagon along its length according to a preset wagon slice width. The actual stacking volume of the cargo refers to the actual 3D spatial volume occupied by the cargo within the space corresponding to a single wagon slice, measured in meters (m). 3 Cargo porosity refers to the ratio of the void volume between cargo particles to the actual accumulated volume of cargo within the space corresponding to a single car body slice. It is dimensionless and usually expressed as a percentage or decimal (ranging from 0 to 1). The preset car body slice width is determined based on a combination of detection accuracy requirements and computational efficiency. The smaller the car body slice width, the more slices (N) are divided, resulting in higher volume calculation accuracy, but the computational workload increases accordingly. Conversely, the larger the car body slice width, the higher the computational efficiency, but the lower the volume calculation accuracy.

[0024] For example, along the length of the target carriage, the 3D model is divided into N carriage slices according to a preset carriage slice width. For any nth carriage slice (n=1, 2, ..., N), the median cross-section of this slice along the length of the carriage is used as the cross-section. Image processing algorithms or point cloud segmentation algorithms are used to extract the contour boundary of the cargo area on this cross-section, and the area enclosed by the contour is calculated as the cross-sectional area of ​​the slice. A n The cross-sectional area of ​​each carriage slice is calculated using the in-slice cross-sectional area integration method. The cross-sectional area of ​​the nth carriage slice is... A n This can be expressed by the formula: , in, n This represents a slice of the nth carriage; A n This represents the cross-sectional area of ​​the nth carriage slice; h cargo,n ( x ) represents the height of the upper surface of the cargo at position x in the cross section of the nth carriage slice; h floor,n This represents the height of the car floor of the nth car slice, i.e., the coordinate of the floor on the Z-axis. According to the preset car coordinate system, the height of the car floor of the nth car slice is usually set to 0. x max,nThis represents the coordinate boundary of the right side wall of the nth carriage slice in the width direction; x min,n This represents the coordinate boundary of the left side wall of the nth car body slice in the width direction, usually set to 0. Based on the cross-sectional area of ​​each car body slice and the preset width of the car body slice, the volume of each car body slice, i.e., the actual stacking volume of the goods, is calculated.

[0025] Step 400: Determine the actual bulk density of the N car body slices based on the standard density of the cargo and the porosity of the cargo in the N car body slices. The actual bulk density of the cargo refers to the mass per unit volume of the cargo (including its internal pores) in a natural stacking state within the spatial range corresponding to a single wagon section, and is measured in kg / m³. 3 .

[0026] For example, the actual bulk density of the cargo in the nth carriage slice. r cal,n It can be expressed by the formula as follows r cal,n = r ×(1- f cal,n ),in n This represents a slice of the nth carriage; r cal,n This represents the actual bulk density of the cargo in the nth carriage slice; r Indicates the standard density of goods; f cal,n This represents the porosity of the cargo in the nth carriage slice.

[0027] Step 500: Determine the total weight of the cargo in the target car based on the actual stacking volume of the cargo in the N car body slices and the actual stacking density of the cargo in the N car body slices. The total weight of the cargo refers to the total mass of the cargo loaded in the target carriage, expressed in tons (t) or kilograms (kg).

[0028] For example, the total weight of the cargo in the target carriage. M This can be expressed by the formula: , in, n This represents a slice of the nth carriage; M Indicates the total weight of the cargo in the target carriage; V cal,n This represents the actual stacked volume of the goods in the nth carriage slice; r cal,n This represents the actual bulk density of the cargo in the nth carriage slice.

[0029] Step 600: When the total weight of the cargo is less than or equal to the rated load, the loading status of the target carriage is determined to be rated load status; when the total weight of the cargo is greater than the rated load, the loading status of the target carriage is determined to be overload status.

[0030] Among them, loading status refers to the classification and judgment result of whether the target car is overloaded after comparing the total weight of the goods with the rated load capacity of the car; rated load status refers to the loading status when the total weight of the goods in the target car does not exceed the rated load capacity of the car, indicating that the loading meets safety and regulatory requirements; overload status refers to the loading status when the total weight of the goods in the target car exceeds the rated load capacity of the car, indicating that there is a safety hazard in the loading, and measures such as unloading or refusal to deliver should be taken.

[0031] For example, when the total weight of the cargo is less than or equal to the rated load capacity of the wagon, the loading status of the target wagon is determined to be the rated load status; when the total weight of the cargo is greater than the rated load capacity of the wagon, but less than or equal to the rated load capacity... a When the load capacity of the target car is multiples of the car's rated load (where 'a' is a real number greater than 1), the loading status of the target car is determined to be slightly overloaded, and a slight overload warning is output, suggesting adjustments to be made during subsequent car loading; when the total weight of the goods exceeds... a When the load capacity of a car is twice the rated load of the car body, the loading status of the target car body is determined to be severely overloaded, immediately triggering a high-level warning. Simultaneously, the overload weight and overload rate are calculated to provide dispatchers with accurate data. At the same time, an overload tracing function is added to the ground processing terminal, using a 3D model to locate the overloaded area and clearly mark the location and volume of the overloaded goods, facilitating rapid unloading and adjustment.

[0032] This embodiment of the UAV-based vehicle overload detection method acquires 3D point cloud data, visible light images, spectral scattering images, rated load, and standard density of the cargo in the target vehicle currently loaded with goods. Then, based on the 3D point cloud data and visible light images, a 3D model is constructed and divided into N vehicle slices. The actual bulk volume and porosity of the cargo in each slice are determined. Next, based on the standard density and porosity of the cargo, the actual bulk density of the cargo in each slice is determined. Finally, based on the actual bulk volume and density of each cargo, the total weight of the cargo in the target vehicle is determined. When the total weight is less than or equal to the rated load, the target vehicle is determined to be in a rated load state; when the total weight is greater than the rated load, the target vehicle is determined to be in an overload state. Compared to existing technologies, this invention utilizes the standard density and porosity of cargo to determine the actual bulk density of cargo in the target carriage, thereby determining the total weight of cargo. It achieves high-precision detection under normal train operation conditions. By flexibly adjusting the flight path and detection angle, it is applicable to overload detection of freight carriages in various types and scenarios, significantly improving the accuracy and precision of carriage overload detection.

[0033] In one embodiment, step 300 includes: Step 301: Determine the initial stacking volume of the goods in the N car body slices based on the width of the car body slice and the N car body slices; The initial stacked volume of the cargo refers to the apparent volume of the cargo directly calculated from the 3D model without considering the influence of porosity between cargo particles, and the unit is m. 3 .

[0034] For example, the initial stacking volume of the cargo in the nth carriage slice. V original,n It can be expressed by the formula as follows V original,n = A n ×Δ y ,in n This represents a slice of the nth carriage; V original,n This represents the initial stacking volume of the goods in the nth carriage slice; A n Δ represents the cross-sectional area of ​​the nth carriage slice; y This indicates the preset width of the carriage slice.

[0035] Step 302: According to the preset voxel side length, divide the nth carriage slice into M voxels, and determine the number of 3D point clouds in the mth voxel within the nth carriage slice, where n = 1, 2, ..., N, M is a positive integer greater than 1, and m = 1, 2, ..., M; Here, a voxel refers to the smallest volume unit obtained by regularly discretizing three-dimensional space, used to discretize continuous space; the voxel side length refers to the side length of a single voxel in each direction of three-dimensional space, in meters. Since voxels usually adopt a regular hexahedral structure, the voxel side length is the uniform side length of a cube in the X, Y, and Z directions; the number of three-dimensional point clouds refers to the total number of three-dimensional point cloud data points falling within the space of the m-th voxel in a single car slice.

[0036] Step 303: For N carriage slices, determine the number of entity voxels and the number of boundary voxels in N carriage slices based on the number of 3D point clouds in the m-th voxel within the n-th carriage slice. Among them, the number of entity voxels refers to the total number of voxels in the nth car slice that have sufficient internal point cloud to be identified as cargo entity regions; the number of boundary voxels refers to the total number of voxels in the nth car slice that are located on the cargo surface or at the edge of internal pores and have sparse but not empty point clouds.

[0037] For example, within the nth carriage slice, if the number of 3D point clouds in the mth voxel is greater than or equal to a preset threshold for the number of 3D point clouds, the voxel is determined to be a solid voxel; if the number of 3D point clouds in the mth voxel is less than the preset threshold for the number of 3D point clouds but greater than 0, the voxel is determined to be a boundary voxel; if the number of 3D point clouds in the mth voxel is equal to 0, the voxel is determined to be an empty voxel.

[0038] Step 304: Determine the actual stacking volume of the goods in the N carriage slices based on the number of solid voxels in the N carriage slices, the number of boundary voxels in the N carriage slices, the side length of the voxels, and the preset boundary voxel weights. For example, the actual stacking volume of the cargo in the nth carriage slice. V cal,n It can be expressed by the formula as follows V cal,n =( N 实体,n + c × N 边界,n )× L 3 体素 ,in n This represents a slice of the nth carriage; V cal,nThis represents the actual stacked volume of the goods in the nth carriage slice; N 实体,n This represents the number of entity voxels in the nth carriage slice; N 边界,n This represents the boundary voxel number of the nth carriage slice; c This represents the preset boundary voxel weights, with a value range of 0 to 1; L 体素 This indicates the preset voxel side length.

[0039] Step 305: Divide the difference between the initial stacked volume of the goods in the N carriage slices and the actual stacked volume of the goods in the N carriage slices by the initial stacked volume of the goods in the N carriage slices to determine the porosity of the goods in the N carriage slices. For example, the cargo porosity of the nth carriage slice. f cal,n It can be expressed by the formula as follows f cal,n =( V original,n - V cal,n )÷ V original,n ,in n This represents a slice of the nth carriage; f cal,n This represents the porosity of the cargo in the nth carriage slice; V original,n This represents the initial stacking volume of the goods in the nth carriage slice; V cal,n This represents the actual stacked volume of the cargo in the nth carriage slice.

[0040] Step 400 includes: The actual bulk density of the N car body slices is determined by subtracting the product of the cargo porosity and the standard density of the cargo from the standard density of the cargo.

[0041] For example, the actual bulk density of the cargo in the nth carriage slice. r cal,n It can be expressed by the formula as follows r cal,n = r - f cal,n × r ,in n This represents a slice of the nth carriage; r cal,n This represents the actual bulk density of the cargo in the nth carriage slice; r Indicates the standard density of goods; fcal,n This represents the porosity of the cargo in the nth carriage slice.

[0042] The UAV-based vehicle overload detection method in this embodiment determines the initial stacking volume of goods in each vehicle slice using the width of the vehicle slice; divides each vehicle slice into multiple voxels according to a preset voxel side length, and determines the number of 3D point clouds in each voxel; then, based on the number of 3D point clouds in each voxel, determines the number of solid voxels and edge voxels in each vehicle slice; using the number of solid voxels, edge voxels, voxel side lengths, and preset boundary voxel weights in each vehicle slice, determines the actual stacking volume of goods in each vehicle slice; based on the actual stacking volume and the initial stacking volume of goods, determines the porosity of goods in each vehicle slice; finally, using the standard density of goods and the porosity of goods in each vehicle slice, determines the actual stacking density of goods in each vehicle slice. Compared with existing technologies, this invention divides the carriage slice into voxels, distinguishes between solid voxels and boundary voxels based on the number of point clouds, and corrects the actual stacking volume of the cargo by combining the boundary voxel weights, thereby calculating the porosity and correcting the actual stacking density of the cargo. This achieves refined and dynamic calculation of cargo volume and porosity, improving the accuracy of detection and environmental adaptability.

[0043] In one embodiment, step 300 includes: Step 301: Determine the initial stacking volume of the goods in the N car body slices based on the width of the car body slice and the N car body slices; Step 302: According to the preset voxel side length, divide the nth carriage slice into M voxels, and determine the number of 3D point clouds in the mth voxel within the nth carriage slice, where n = 1, 2, ..., N, M is a positive integer greater than 1, and m = 1, 2, ..., M; Step 303: For N carriage slices, determine the number of entity voxels and the number of boundary voxels in N carriage slices based on the number of 3D point clouds in the m-th voxel within the n-th carriage slice. Step 304: Determine the actual stacking volume of the goods in the N carriage slices based on the number of solid voxels in the N carriage slices, the number of boundary voxels in the N carriage slices, the side length of the voxels, and the preset boundary voxel weights. Step 305: Divide the difference between the initial stacked volume of the goods in the N carriage slices and the actual stacked volume of the goods in the N carriage slices by the initial stacked volume of the goods in the N carriage slices to determine the porosity of the goods in the N carriage slices. Step 400 includes: Step 401: Obtain the pitch angle, roll angle, and speed fluctuation value of the UAV at the current moment; Among them, pitch angle refers to the angular deviation of the UAV's body coordinate system relative to the ground coordinate system along the vertical axis (X-axis), used to describe the attitude angle of the UAV rotating around the horizontal axis (Y-axis), and the unit is ° or rad; roll angle refers to the angular deviation of the UAV's body coordinate system relative to the horizontal axis (Y-axis) of the ground coordinate system, used to describe the attitude angle of the UAV rotating around the vertical axis (X-axis), and the unit is ° or rad; speed fluctuation value refers to the instantaneous deviation of the UAV's flight speed from the preset target speed during data acquisition, used to describe the stability of the UAV's speed, and the unit is m / s.

[0044] Step 402: Subtract the product of the porosity of the N car body slices and the standard density of the cargo from the standard density of the cargo to determine the initial bulk density of the N car body slices. The initial bulk density of the cargo refers to the apparent density of the cargo calculated based on standard density and porosity, taking into account the influence of inter-particle porosity, and is expressed in kg / m³. 3 .

[0045] For example, the initial bulk density of the cargo in the nth carriage slice. r original,n It can be expressed by the formula as follows r original,n = r - f cal,n × r ,in n This represents a slice of the nth carriage; r original,n This represents the initial bulk density of the cargo in the nth carriage slice; r Indicates the standard density of goods; f cal,n This represents the porosity of the cargo in the nth carriage slice.

[0046] Step 403: Determine the cargo porosity correction coefficient of the target carriage based on the pitch angle, the roll angle, the speed fluctuation value, and preset pitch angle weight, roll angle weight, and speed fluctuation value weight. Among them, pitch angle weight refers to the contribution coefficient of the UAV pitch angle parameter when calculating the cargo porosity correction coefficient, which is used to quantify the influence of pitch angle deviation on the porosity estimation result; roll angle weight refers to the contribution coefficient of the UAV roll angle parameter when calculating the cargo porosity correction coefficient, which is used to quantify the influence of roll angle deviation on the porosity estimation result; velocity fluctuation value weight refers to the contribution coefficient of the UAV velocity fluctuation value parameter when calculating the cargo porosity correction coefficient, which is used to quantify the influence of velocity fluctuation on the porosity estimation result; the cargo porosity correction coefficient is a correction factor used to dynamically correct the cargo porosity to eliminate the porosity estimation deviation caused by changes in UAV attitude and uneven flight speed.

[0047] For example, the cargo porosity correction factor of the target carriage. m It can be expressed by the formula as follows m =1- k 1×| α |- k 2×| β |- k 3×|Δ v |, among which m This represents the cargo porosity correction factor for the target carriage; α Indicates the pitch angle of the drone; β Indicates the roll angle of the drone; Δ v This indicates the speed fluctuation value of the drone; k 1 indicates the preset pitch angle weight; k 2 indicates the preset roll angle weight; k 3 indicates the preset speed fluctuation value weight.

[0048] Among them, the pitch angle of the drone α The roll angle of the drone β and the speed fluctuation value Δ of the drone v All parameters are normalized dimensionless parameters to eliminate the influence of dimensional differences on the contribution weight coefficients. The normalization formula is as follows: x’ =2×arctg( x )÷π, where x This represents the raw data, such as the actual pitch angle of the drone (unit: ° or rad), the roll angle of the drone (unit: ° or rad), and the speed fluctuation value of the drone (unit: m / s). x’ This represents dimensionless parameters obtained after normalizing the original data, such as the pitch angle, roll angle, and speed fluctuation values ​​of a UAV, all of which are unitless.

[0049] Step 404: Determine the actual bulk density of the N pieces of cargo in the N car body slices based on the product of the initial bulk density of the cargo and the cargo porosity correction coefficient.

[0050] For example, the actual bulk density of the cargo in the nth carriage slice. r cal,n It can be expressed by the formula as follows r cal,n = r original,n × m ,in n This represents a slice of the nth carriage; r cal,n This represents the actual bulk density of the cargo in the nth carriage slice; r original,n This represents the initial bulk density of the cargo in the nth carriage slice; m This represents the cargo porosity correction factor for the target carriage.

[0051] The UAV-based overload detection method for cargo compartments in this embodiment determines the actual stacking volume of goods in each cargo compartment slice using the width of the cargo compartment slice, preset voxel side length, 3D point cloud, number of entity voxels, number of edge voxels, voxel side length, and preset boundary voxel weights. Based on the actual stacking volume and the initial stacking volume of the goods, the porosity of the goods in each cargo compartment slice is determined. Then, using the standard density of the goods and the porosity of the goods in each cargo compartment slice, the initial stacking density of the goods in each cargo compartment slice is determined. Using the obtained pitch angle, roll angle, and speed fluctuation value of the UAV at the current moment, along with preset pitch angle weights, roll angle weights, and speed fluctuation value weights, the porosity correction coefficient for the target cargo compartment is determined. Finally, based on the initial stacking density of the goods in each cargo compartment slice and the porosity correction coefficient, the actual stacking density of the goods in each cargo compartment slice is determined. Compared with existing technologies, this invention dynamically determines the cargo porosity correction coefficient of the target carriage by utilizing the pitch angle, roll angle and speed fluctuation value of the UAV at the current moment, thereby calculating the actual bulk density of the cargo in each carriage slice. By adopting a weighted fusion strategy, the actual bulk density of the cargo is dynamically corrected, and comprehensive compensation for multi-source interference factors is achieved, thereby improving the accuracy and precision of the calculation of the actual bulk density of the cargo.

[0052] In one embodiment, step 300 includes: Step 301: Determine the initial stacking volume of the goods in the N car body slices based on the width of the car body slice and the N car body slices; Step 302: According to the preset voxel side length, divide the nth carriage slice into M voxels, and determine the number of 3D point clouds in the mth voxel within the nth carriage slice, where n = 1, 2, ..., N, M is a positive integer greater than 1, and m = 1, 2, ..., M; Step 303: For N carriage slices, determine the number of entity voxels and the number of boundary voxels in N carriage slices based on the number of 3D point clouds in the m-th voxel within the n-th carriage slice. Step 304: Determine the actual stacking volume of the goods in the N carriage slices based on the number of solid voxels in the N carriage slices, the number of boundary voxels in the N carriage slices, the side length of the voxels, and the preset boundary voxel weights. Step 305: Divide the difference between the initial stacked volume of the goods in the N carriage slices and the actual stacked volume of the goods in the N carriage slices by the initial stacked volume of the goods in the N carriage slices to determine the porosity of the goods in the N carriage slices. Step 400 includes: Step 401: Obtain the pitch angle, roll angle, and speed fluctuation value of the UAV at the current moment; Step 402: Subtract the product of the porosity of the N car body slices and the standard density of the cargo from the standard density of the cargo to determine the initial bulk density of the N car body slices. Step 403: Determine the cargo porosity correction coefficient of the target carriage based on the pitch angle, the roll angle, the speed fluctuation value, and preset pitch angle weight, roll angle weight, and speed fluctuation value weight. Step 404: Determine the corrected bulk density of the N car body slices based on the product of the initial bulk density of the cargo in the N car body slices and the cargo porosity correction coefficient. Among them, the corrected bulk density of goods refers to the bulk density value of goods obtained after correction by the porosity correction factor based on the initial bulk density of goods.

[0053] For example, the corrected bulk density of the cargo in the nth carriage slice. r real,n It can be expressed by the formula as follows r real,n = r original,n × m ,in n This represents a slice of the nth carriage; r real,n This represents the corrected bulk density of the cargo in the nth carriage slice; r original,n This represents the initial bulk density of the cargo in the nth carriage slice; m This represents the cargo porosity correction factor for the target carriage.

[0054] Step 405: Determine the actual bulk density of the goods in the N car body slices based on the corrected bulk density of the goods, the standard density of the goods, and the preset cargo density weight.

[0055] Among them, cargo density weight refers to the contribution coefficient of the corrected bulk density of cargo when calculating the actual bulk density of cargo, and is used to balance the confidence level distribution between the corrected density and the standard density.

[0056] For example, the actual bulk density of the cargo in the nth carriage slice. r cal,n It can be expressed by the formula as follows r cal,n = l × r real,n +(1- l )× r ,in n This represents a slice of the nth carriage; r cal,n This represents the actual bulk density of the cargo in the nth carriage slice; l This indicates the preset cargo density weight; r real,n This represents the corrected bulk density of the cargo in the nth carriage slice; r Indicates the standard density of goods.

[0057] This embodiment of the UAV-based overload detection method for cargo compartments determines the actual stacking volume of goods in each compartment slice using the width of the compartment slice, a preset voxel side length, 3D point cloud, number of entity voxels, number of edge voxels, voxel side length, and preset boundary voxel weights. Based on the actual stacking volume and the initial stacking volume of the goods, the porosity of the goods in each compartment slice is determined. Then, using the standard density of the goods and the porosity of the goods in each compartment slice, the initial stacking density of the goods in each compartment slice is determined. Using the acquired pitch angle, roll angle, and speed fluctuation value of the UAV at the current moment, along with preset pitch angle weights, roll angle weights, and speed fluctuation value weights, a porosity correction coefficient for the target compartment is determined. Next, based on the initial stacking density of the goods in each compartment slice and the porosity correction coefficient, the corrected stacking density of the goods in each compartment slice is determined. Finally, using the corrected stacking density of the goods, the standard density of the goods, and preset cargo density weights, the actual stacking density of the goods in each compartment slice is determined. Compared with existing technologies, this invention significantly improves the accuracy and precision of cargo bulk density calculation by introducing the corrected bulk density of the cargo and dynamically correcting the actual bulk density of the cargo. It achieves a comprehensive balance of multi-source density information and provides an accurate data foundation for detecting overloaded cargo compartments.

[0058] The UAV-based method for detecting off-center load in train carriages provided in this invention can be applied to, for example... Figure 1 The application environment is shown. Specifically, this UAV-based vehicle off-center load detection method is applied in a UAV-based vehicle off-center load detection system, which includes, as shown in the example... Figure 1 The diagram shows a ground processing terminal and a drone. The ground processing terminal communicates with the drone via a network to achieve real-time, drone-based detection of vehicle eccentricity. The ground processing terminal, also known as the user terminal, refers to the program that provides local services to customers, corresponding to the drone. The ground processing terminal can be installed on, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The drone can use an integrated 3D LiDAR and a high-resolution SLR camera to simultaneously acquire 3D point cloud and visible light images of the vehicle compartment.

[0059] In one embodiment, such as Figure 3 As shown, this embodiment provides a method for detecting off-center loading in a vehicle compartment based on an unmanned aerial vehicle (UAV), which is then applied to... Figure 1 Taking the ground processing terminal as an example, the UAV-based method for detecting off-center loading in a carriage includes: Step S100: Obtain the 3D point cloud data, visible light image, spectral scattering image, standard density of the cargo, X-axis tilt angle, and Y-axis tilt angle of the target carriage loaded with cargo at the current moment; Among them, the X-axis tilt angle refers to the lateral tilt angle, that is, the tilt angle of the train perpendicular to the direction of travel (left and right direction), with left tilt being positive and right tilt being negative; the Y-axis tilt angle refers to the longitudinal tilt angle, that is, the tilt angle of the train along the direction of travel (from the front to the rear of the train), with uphill being positive and downhill being negative.

[0060] Step S200: Construct a three-dimensional model of the target carriage based on the three-dimensional point cloud data, the visible light image, and the spectral scattering image; Step S300: Along the length of the target carriage, the three-dimensional model is divided into N carriage slices according to the preset carriage slice width. The actual stacking volume of the goods in the N carriage slices, the porosity of the goods in the N carriage slices, and the center of gravity coordinates of the goods in the N carriage slices are determined, where N is a positive integer greater than 1. The initial center of gravity coordinates of the cargo refer to the coordinates of the center of gravity of each car body slice directly calculated based on the origin of the 3D model coordinates without considering the tilt of the car body (X-axis tilt angle and Y-axis tilt angle).

[0061] For example, the centroid coordinates of the cargo in the nth carriage slice ( x (0)c,n , y (0) c,n , z (0) c,n This can be expressed by the formula: , in, n This represents a slice of the nth carriage; M This represents the number of 3D point clouds in the nth carriage slice; m This represents the m-th 3D point cloud in the n-th carriage slice; x (0) c,n , y (0) c,n , z (0) c,n () represents the centroid coordinates of the cargo in the nth carriage slice; x m,n This represents the X-axis coordinate of the m-th 3D point cloud in the n-th carriage slice; y m,n This represents the Y-axis coordinate of the m-th 3D point cloud in the n-th carriage slice; z m,n This represents the Z-axis coordinate of the m-th 3D point cloud in the n-th carriage slice. 。

[0062] Step S400: Determine the actual bulk density of the N car body slices based on the standard density of the cargo and the porosity of the cargo in the N car body slices. Step S500: Determine the actual center-of-gravity coordinates of the cargo in the target car based on the actual stacking volume of the cargo in the N car body slices, the actual stacking density of the cargo in the N car body slices, the center-of-gravity coordinates of the cargo in the N car body slices, the X-axis tilt angle, and the Y-axis tilt angle. The actual center of gravity coordinates of the cargo refer to the true center of gravity coordinates of the entire cargo within the target carriage, obtained after coordinate transformation or compensation calculation, taking into account the effects of carriage tilt (X-axis tilt angle and Y-axis tilt angle).

[0063] Step S600: When the actual center of gravity coordinates of the cargo are within the preset ideal center of gravity range, the loading balance state of the target car is determined to be a balanced loading state; when the actual center of gravity coordinates of the cargo exceed the ideal center of gravity range, the loading balance state of the target car is determined to be an off-center loading state.

[0064] The ideal center of gravity range refers to the maximum allowable range of variation in the actual center of gravity coordinates of the cargo along the length (X-axis) and width (Y-axis) of the carriage to ensure safe and stable vehicle operation. The balanced loading state refers to the conclusion drawn from the comparison between the actual center of gravity coordinates of the cargo and the ideal center of gravity range, assessing the load distribution balance of the target carriage. A balanced loading state refers to the loading distribution when the actual center of gravity coordinates of the cargo in the target carriage are within the ideal center of gravity range, indicating uniform load distribution, center of gravity offset within safe limits, no risk of off-center loading, and compliance with safe operation requirements. An off-center loading state refers to the loading distribution when the actual center of gravity coordinates of the cargo in the target carriage exceed the ideal center of gravity range, indicating uneven load distribution, center of gravity offset exceeding safe limits, and potential safety hazards such as tilting, derailment, or wheel and axle overload, requiring readjustment of the cargo layout or corrective measures. The ideal center of gravity range is determined based on the carriage's rated load, vehicle dynamics parameters, and railway transportation safety regulations, typically using the carriage's geometric center or design center of gravity as a benchmark, and setting allowable longitudinal and lateral offset thresholds.

[0065] For example, the actual center of gravity coordinates of the cargo in the target carriage ( x g , y g , z g ) and the range of the center of gravity of the carriage during operation [ x min , x max ]and[ y min , y max If the load is compared and exceeds any range, the target car's loading balance is determined to be unbalanced.

[0066] The actual center of gravity coordinates of the cargo in the target carriage ( x g , y g , z g Offset Δ on the X-axis x This can be expressed by the formula Δ x =| x g -( x min + x max )÷2|, where Δ x Indicates the actual center of gravity coordinates of the cargo in the target carriage. x g , y g , z gThe offset on the X-axis; x g The X-axis coordinate represents the actual center of gravity of the cargo in the target carriage. x min This represents the minimum X-axis coordinate of the ideal center of gravity of the target carriage. x max This represents the maximum X-axis coordinate of the ideal center of gravity of the target carriage.

[0067] The actual center of gravity coordinates of the cargo in the target carriage ( x g , y g , z g The offset Δ on the Y-axis y This can be expressed by the formula Δ y =| y g -( y min + y max )÷2|, where Δ y Indicates the actual center of gravity coordinates of the cargo in the target carriage. x g , y g , z g The offset on the Y-axis; y g The Y-axis coordinate represents the actual center of gravity of the cargo in the target carriage. y min This represents the minimum Y-axis coordinate of the ideal center of gravity of the target carriage. y max This represents the maximum value of the ideal center of gravity Y-axis coordinate of the target carriage.

[0068] Therefore, the actual center of gravity coordinates of the cargo in the target carriage ( x g , y g , z g Offset rate on the X-axis or x It can be expressed by the formula as follows or x =Δ x ÷[( x min + x max )÷2], where or x Indicates the actual center of gravity coordinates of the cargo in the target carriage. xg , y g , z g The offset rate on the X-axis; Δ x Indicates the actual center of gravity coordinates of the cargo in the target carriage. x g , y g , z g The offset on the X-axis; x min This represents the minimum X-axis coordinate of the ideal center of gravity of the target carriage. x max This represents the maximum X-axis coordinate of the ideal center of gravity of the target carriage.

[0069] The actual center of gravity coordinates of the cargo in the target carriage ( x g , y g , z g Offset rate on the Y-axis or y It can be expressed by the formula as follows or y =Δ y ÷[( y min + y max )÷2], where or y Indicates the actual center of gravity coordinates of the cargo in the target carriage. x g , y g , z g The offset rate on the Y-axis; Δ y Indicates the actual center of gravity coordinates of the cargo in the target carriage. x g , y g , z g The offset on the Y-axis; y min This represents the minimum Y-axis coordinate of the ideal center of gravity of the target carriage. y max This represents the maximum value of the ideal center of gravity Y-axis coordinate of the target carriage.

[0070] When the actual center of gravity coordinates of the cargo in the target carriage ( x g , y g ,z g Offset rate on the X-axis or x Or the actual center of gravity coordinates of the cargo in the target carriage ( x g , y g , z g Offset rate on the Y-axis or y When the load is less than the preset first offset rate threshold, the loading balance of the target car is determined to be balanced; when the actual center of gravity coordinates of the cargo in the target car are ( x g , y g , z g Offset rate on the X-axis or x Or the actual center of gravity coordinates of the cargo in the target carriage ( x g , y g , z g Offset rate on the Y-axis or y When the load balance of the target car is greater than or equal to the preset first offset rate threshold, the loading balance of the target car is determined to be an off-center load.

[0071] This embodiment of the UAV-based method for detecting cargo eccentricity in a target cargo compartment acquires 3D point cloud data, visible light images, spectral scattering images, standard density of the cargo, X-axis tilt angle, and Y-axis tilt angle of the target cargo compartment at the current moment. Using the acquired 3D point cloud data and visible light images of the target cargo compartment, a 3D model is constructed. Then, according to a preset compartment slice width, the 3D model is divided into multiple compartment slices, and the actual bulk volume of the cargo in each compartment slice, the porosity of the cargo in each compartment slice, and the centroid coordinates of the cargo in each compartment slice are determined. Based on the standard density of the cargo and the... The porosity of the cargo in each car section is used to determine the actual bulk density of the cargo. Finally, based on the actual bulk volume, actual bulk density, center-of-gravity coordinates, X-axis tilt angle, and Y-axis tilt angle of the cargo in each car section, the actual center-of-gravity coordinates of the cargo in the target car section are determined. When the actual center-of-gravity coordinates are within a preset ideal center-of-gravity range, the target car section is considered to be in a balanced loading state; when the actual center-of-gravity coordinates exceed the preset ideal center-of-gravity range, the target car section is considered to be in an off-center loading state. Compared to existing technologies, this invention dynamically determines the actual bulk density of the cargo using the standard density and porosity of the cargo, and uses the tilt angles of the train along the X and Y axes during travel to determine the actual center-of-gravity coordinates of the cargo in the target car section. By flexibly adjusting the flight path and detection angle, it is applicable to off-center loading detection of freight cars in various types and scenarios, achieving automated and high-precision detection of off-center loading in railway freight cars, with higher environmental adaptability and detection reliability.

[0072] In one embodiment, step S300 includes: Step S301: Determine the initial stacking volume of the cargo in the N car body slices based on the width of the car body slice and the N car body slices. Step S302: According to the preset voxel side length, the nth carriage slice is divided into M voxels, and the number of three-dimensional point clouds in the mth voxel within the nth carriage slice is determined, where n = 1, 2, ..., N, M is a positive integer greater than 1, and m = 1, 2, ..., M. Step S303: For N carriage slices, determine the number of entity voxels and the number of boundary voxels in N carriage slices based on the number of 3D point clouds in the m-th voxel within the n-th carriage slice. Step S304: Determine the actual stacking volume of goods in the N car body slices based on the number of solid voxels in the N car body slices, the number of boundary voxels in the N car body slices, the side length of the voxels, and the preset boundary voxel weights. Step S305: Divide the difference between the initial stacked volume of the goods in the N carriage slices and the actual stacked volume of the goods in the N carriage slices by the initial stacked volume of the goods in the N carriage slices to determine the porosity of the goods in the N carriage slices. Step S400 includes: S401, obtain the current pitch angle, roll angle and speed fluctuation value of the UAV; S402, subtract the product of the porosity of the N car body slices and the standard density of the cargo from the standard density of the cargo to determine the initial bulk density of the N car body slices. S403, determine the cargo porosity correction coefficient of the target carriage based on the pitch angle, the roll angle, the speed fluctuation value, and preset pitch angle weight, roll angle weight, and speed fluctuation value weight; S404, determine the corrected bulk density of the N car body slices based on the product of the initial bulk density of the cargo in the N car body slices and the cargo porosity correction coefficient; S405, determine the actual bulk density of the goods in the N car body slices based on the corrected bulk density of the goods in the N car body slices, the standard density of the goods, and the preset cargo density weight.

[0073] The UAV-based unmanned aerial vehicle (UAV)-based method for detecting cargo eccentricity in a wagon compartment determines the actual stacking volume of cargo in each wagon compartment slice using the wagon compartment slice width, preset voxel side length, 3D point cloud, number of entity voxels, number of edge voxels, voxel side length, and preset boundary voxel weights. Based on the actual stacking volume and the initial stacking volume of cargo, the porosity of cargo in each wagon compartment slice is determined. Then, using the standard density of cargo and the porosity of cargo in each wagon compartment slice, the initial stacking density of cargo in each wagon compartment slice is determined. Using the acquired pitch angle, roll angle, and velocity fluctuation value of the UAV at the current moment, along with preset pitch angle weights, roll angle weights, and velocity fluctuation value weights, a cargo porosity correction coefficient for the target wagon compartment is determined. Next, based on the initial stacking density of cargo in each wagon compartment slice and the cargo porosity correction coefficient, the corrected stacking density of cargo in each wagon compartment slice is determined. Finally, using the corrected stacking density of cargo, the standard density of cargo, and preset cargo density weights, the actual stacking density of cargo in each wagon compartment slice is determined. Compared with existing technologies, this invention significantly improves the accuracy and precision of cargo bulk density calculation by introducing the corrected bulk density of the cargo and dynamically correcting the actual bulk density of the cargo. It achieves a comprehensive balance of multi-source density information and provides an accurate data foundation for the detection of off-center loading of the carriage.

[0074] In one embodiment, step S500 includes: S501, for N car body slices, sum the product of the actual bulk density of the goods in the nth car body slice, the actual bulk volume of the goods in the nth car body slice, and the center of gravity coordinates of the goods in the nth car body slice, and divide by the sum of the products of the actual bulk density of the goods in the nth car body slice and the actual bulk volume of the goods in the nth car body slice to determine the initial center of gravity coordinates of the goods in the target car body; The initial center of gravity coordinates of the cargo refer to the overall center of gravity coordinates of the cargo in the entire target car body, calculated based on the actual bulk density, actual bulk volume, and center of gravity coordinates of each car body slice, without considering the car body tilt (X-axis tilt angle and Y-axis tilt angle).

[0075] For example, the initial center of gravity coordinates of the cargo in the target carriage ( x (0) g , y (0) g , z (0) g This can be expressed by the formula: , in, n This represents the slice of the nth carriage; x (0) g , y (0) g , z (0) g () represents the initial center of gravity coordinates of the cargo in the target carriage; r cal,n This represents the actual bulk density of the cargo in the nth carriage slice; V cal,n This represents the actual stacked volume of the goods in the nth carriage slice; x (0) c,n This represents the initial centroid X-axis coordinate of the nth carriage slice; y (0) c,n This represents the initial Y-axis coordinate of the centroid of the nth carriage slice; z (0) c,n This represents the initial Z-axis coordinate of the centroid of the nth carriage slice.

[0076] S502, Based on the initial center of gravity coordinates of the cargo, the X-axis tilt angle, and the Y-axis tilt angle, determine the X-axis tilt amount and the Y-axis tilt amount of the target carriage; Among them, the X-axis tilt refers to the offset distance of the actual center of gravity of the cargo relative to the initial center of gravity coordinate on the horizontal plane along the X-axis direction due to the tilt of the carriage along the X-axis direction (i.e., the left-right direction of the train); the Y-axis tilt refers to the offset distance of the actual center of gravity of the cargo relative to the initial center of gravity coordinate on the horizontal plane along the Y-axis direction due to the tilt of the carriage along the Y-axis direction (i.e., the direction of train travel) (uphill or downhill).

[0077] For example, the X-axis tilt Δ of the target carriage x This can be expressed by the formula Δ x '= z (0) g ×tan i , where Δ x 'Indicates the X-axis tilt of the target carriage; z (0) g This represents the initial Z-axis coordinate of the center of gravity of the cargo in the target carriage; i This indicates the X-axis tilt angle of the target carriage.

[0078] Y-axis tilt Δ of the target carriage y This can be expressed by the formula Δ y '= z (0) g ×tan ψ , where Δ y 'Indicates the Y-axis tilt of the target carriage; z (0) g This represents the initial Z-axis coordinate of the center of gravity of the cargo in the target carriage; ψ This indicates the Y-axis tilt angle of the target carriage.

[0079] S503, the initial center of gravity coordinates of the cargo are subtracted from the X-axis tilt and the Y-axis tilt to determine the actual center of gravity coordinates of the cargo in the target carriage.

[0080] For example, the actual center of gravity coordinates of the cargo in the target carriage ( x g , y g , z g ) can be expressed by the formula ( x g , y g , z g )=( x (0) g -Δx ', y (0) g -Δ y ', z (0) g ),in( x g , y g , z g () indicates the actual center of gravity coordinates of the cargo in the target carriage; x (0) g The X-axis coordinate represents the initial center of gravity of the cargo in the target carriage. y (0) g The Y-axis coordinate represents the initial center of gravity of the cargo in the target carriage. z (0) g This represents the initial Z-axis coordinate of the center of gravity of the cargo in the target carriage; Δ x 'Indicates the X-axis tilt of the target carriage; Δ y 'Indicates the Y-axis tilt of the target carriage.

[0081] This embodiment of the UAV-based method for detecting off-center loading in train carriages determines the initial center-of-gravity coordinates of the cargo in the target carriage by utilizing the actual bulk density, actual bulk volume, and center-of-gravity coordinates of the cargo in each carriage slice. Based on the initial center-of-gravity coordinates, X-axis tilt angle, and Y-axis tilt angle, the X-axis tilt and Y-axis tilt of the target carriage are determined. Finally, the difference between the initial center-of-gravity coordinates and the X-axis and Y-axis tilts is used to determine the actual center-of-gravity coordinates of the cargo in the target carriage. Compared to existing technologies, this invention introduces X-axis and Y-axis tilts to dynamically compensate and correct the initial center-of-gravity coordinates of the cargo, eliminating the influence of carriage tilt on the center-of-gravity calculation when the train is traveling on slopes or curves. This achieves accurate calculation of the true center of gravity of the cargo under non-horizontal conditions, significantly improving the environmental adaptability and accuracy of off-center loading detection.

[0082] In one embodiment, after step S600, the method further includes: S701, Obtain the current speed of the target carriage; Among them, the operating speed refers to the instantaneous speed at which the target carriage travels on the track at the current moment, and is used to characterize the motion state of the train.

[0083] S702, based on the ideal center of gravity range, determine the maximum permissible lateral offset, maximum permissible longitudinal offset, maximum permissible vertical offset, and ideal cargo center of gravity coordinates of the target car body; Among them, the maximum permissible lateral offset of the center of gravity refers to the maximum distance that the actual center of gravity of the cargo deviates from the ideal center of gravity position in the width direction (X-axis) of the car body to ensure driving safety; the maximum permissible longitudinal offset of the center of gravity refers to the maximum distance that the actual center of gravity of the cargo deviates from the ideal center of gravity position in the length direction (Y-axis) of the car body to ensure uniform load on the vehicle bogie and smooth operation; the maximum permissible vertical offset of the center of gravity refers to the maximum distance that the actual center of gravity of the cargo deviates from the ideal center of gravity height in the height direction (Z-axis) of the car body to ensure the vehicle's anti-overturning stability and operational safety; the ideal cargo center of gravity coordinates refer to the theoretically optimal center of gravity position coordinates that the target car body should have under ideal conditions where the car body is not tilted and the load distribution is completely balanced.

[0084] S703, based on the actual center of gravity coordinates of the cargo and the ideal center of gravity coordinates of the cargo, determine the cargo center of gravity X-axis offset, cargo center of gravity Y-axis offset, and cargo center of gravity Z-axis offset of the target carriage. Among them, the X-axis offset of the cargo center of gravity refers to the difference between the actual center of gravity and the ideal center of gravity of the cargo on the X-axis, reflecting the degree of left-right load imbalance; the Y-axis offset of the cargo center of gravity refers to the difference between the actual center of gravity and the ideal center of gravity of the cargo on the Y-axis, reflecting the degree of front-rear load imbalance; and the Z-axis offset of the cargo center of gravity refers to the difference between the actual center of gravity and the ideal center of gravity of the cargo on the Z-axis, reflecting the change in center of gravity height, which affects vehicle stability.

[0085] S704, based on the cargo center of gravity X-axis offset, cargo center of gravity Y-axis offset, cargo center of gravity Z-axis offset, running speed, maximum permissible lateral offset of the center of gravity, maximum permissible longitudinal offset of the center of gravity, maximum permissible vertical offset of the center of gravity, and preset running speed threshold, Y-axis tilt angle threshold, offset weight, running speed weight, and Y-axis tilt angle weight, determine the cargo off-center loading risk coefficient of the target car; Among them, the operating speed threshold refers to the preset critical value of train operating speed, used to determine whether the current speed has a significant impact on the risk of off-center loading; the Y-axis tilt angle threshold refers to the preset critical value of the longitudinal tilt angle of the carriage (i.e., the Y-axis tilt angle), used to determine whether the current tilt degree has a significant impact on the risk of off-center loading; the offset weight refers to the proportional coefficient allocated to the comprehensive evaluation item of the three-dimensional offset of the center of gravity (X-axis offset, Y-axis offset, Z-axis offset) when calculating the cargo off-center loading risk coefficient, reflecting the degree of contribution of the offset degree to the overall off-center loading risk; the operating speed weight refers to the proportional coefficient allocated to the current operating speed item when calculating the cargo off-center loading risk coefficient, reflecting the degree of contribution of train speed to the overall off-center loading risk; the Y-axis tilt angle weight refers to the proportional coefficient allocated to the longitudinal tilt angle (Y-axis tilt angle) item when calculating the cargo off-center loading risk coefficient, reflecting the degree of contribution of the car body tilt to the overall off-center loading risk; the cargo off-center loading risk coefficient is a comprehensive evaluation index used to quantify the risk level of an off-center loading accident in the current carriage.

[0086] For example, the risk coefficient of cargo off-center loading in the target carriage at time t. K ( t This can be expressed by the formula: , in, t Indicates the travel time of the target carriage; K ( t ) represents the risk coefficient of cargo off-center loading in the target carriage at time t; Δ d x ( t ) represents the X-axis offset of the cargo center of gravity of the target carriage at time t; Δ d y ( t ) represents the Y-axis offset of the cargo center of gravity of the target carriage at time t; Δ d z ( t () represents the Z-axis offset of the cargo center of gravity of the target carriage at time t; D x This indicates the maximum permissible lateral offset of the target carriage's center of gravity; D y This indicates the maximum permissible longitudinal offset of the target carriage's center of gravity; D x This indicates the maximum permissible vertical offset of the target carriage's center of gravity; v ( t () represents the speed of the target carriage at time t; v max This indicates the preset target vehicle speed threshold. ψ ( t () represents the Y-axis tilt angle of the target carriage at time t; ψmax This represents the preset threshold for the Y-axis tilt angle of the target carriage; oh 1 indicates the preset offset weight of the target carriage; oh 2 indicates the preset speed weight of the target carriage; oh 3 indicates the preset Y-axis tilt angle weight of the target carriage.

[0087] S705, when the cargo off-center loading risk coefficient is greater than or equal to the preset off-center loading risk coefficient threshold, a risk warning is immediately triggered, and a status adjustment suggestion is output simultaneously.

[0088] The off-center loading risk coefficient threshold refers to a preset critical value for the cargo off-center loading risk coefficient, used to determine whether the current off-center loading risk level of the target car has reached the threshold requiring an early warning. The status adjustment suggestion refers to specific handling measures automatically generated by the system when the cargo off-center loading risk coefficient reaches or exceeds the preset threshold, guiding on-site operators or the dispatch center to eliminate or reduce the off-center loading risk. For example, when a lateral off-center loading occurs in the target car, the cargo on the offset side can be appropriately transferred to the other side of the target car, or some cargo can be unloaded.

[0089] This embodiment of the UAV-based method for detecting off-center loading in train carriages first acquires the target carriage's operating speed. Based on a preset ideal center of gravity range, it determines the maximum permissible lateral, longitudinal, and vertical offsets of the target carriage's center of gravity, as well as the ideal cargo center of gravity coordinates. Then, based on the actual and ideal cargo center of gravity coordinates, it determines the cargo center of gravity offsets along the X, Y, and Z axes. Finally, based on the cargo center of gravity offsets in each direction, the operating speed, and the maximum permissible offset thresholds in each direction, it determines the cargo off-center loading risk coefficient of the target carriage. When the cargo off-center loading risk coefficient is greater than or equal to a preset threshold, a risk warning is immediately triggered, and a status adjustment suggestion is simultaneously output. Compared to existing technologies, this invention integrates the cargo center of gravity offsets in the lateral, longitudinal, and vertical dimensions with train operating speed and Y-axis tilt angle to construct a quantitative assessment model for off-center loading risk coefficients. Based on preset thresholds, it automatically outputs graded warnings and status adjustment suggestions, significantly improving the comprehensive assessment capability, accuracy, and intelligence level of off-center loading detection.

[0090] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0091] In one embodiment, a vehicle overload detection system based on unmanned aerial vehicles is provided; please refer to [reference needed]. Figure 1This includes a ground processing terminal and a drone. The ground processing terminal and the drone communicate with each other, and the ground processing terminal and the drone work together to implement the drone-based vehicle overload detection method in the above embodiments. For example... Figure 2 The steps 100 to 600 shown are as follows: the UAV is used to implement step 100 and send the collected data to the ground processing terminal; the ground processing terminal receives the data sent by the UAV and implements steps 200 to 600. To avoid repetition, these steps will not be described again here.

[0092] In one embodiment, the ground processing terminal in the UAV-based carriage overload detection system is also used to implement steps 301 to 305 and step 400. To avoid repetition, these steps will not be described again here.

[0093] In one embodiment, the drone in the unmanned aerial vehicle (UAV)-based overload detection system is also used to implement step S401 and send the collected data to the ground processing terminal; the ground processing terminal receives the data sent by the UAV and implements steps 301 to 305 and steps 402 to 404. To avoid repetition, these steps will not be described again here.

[0094] In one embodiment, the drone in the unmanned aerial vehicle (UAV)-based overload detection system is also used to implement step S401 and send the collected data to the ground processing terminal; the ground processing terminal receives the data sent by the UAV and implements steps 301 to 305 and steps 402 to 405. To avoid repetition, these steps will not be described again here.

[0095] In one embodiment, a vehicle compartment off-center load detection system based on unmanned aerial vehicles is provided. Please refer to [reference needed]. Figure 1 This includes a ground processing terminal and a drone. The ground processing terminal and the drone communicate with each other, and they work together to implement the drone-based carriage off-center load detection method described in the above embodiments. Figure 3 The steps S100 to S600 shown are as follows: the UAV is used to implement step S100 and send the collected data to the ground processing terminal; the ground processing terminal receives the data sent by the UAV and implements steps S200 to S600. To avoid repetition, these steps will not be described again here.

[0096] In one embodiment, the drone in the unmanned vehicle off-center load detection system is also used to implement step S401 and send the collected data to the ground processing terminal; the ground processing terminal receives the data sent by the drone and implements steps S301 to S305 and steps S402 to S405. To avoid repetition, these steps will not be described again here.

[0097] In one embodiment, the ground processing terminal in the UAV-based carriage off-center load detection system is also used to implement steps S501 to S503. To avoid repetition, these steps will not be described again here.

[0098] In one embodiment, the ground processing terminal in the UAV-based carriage off-center load detection system is also used to implement steps S701 to S705. To avoid repetition, these steps will not be described again here.

[0099] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0100] While specific embodiments of the invention have been described in detail by way of example, those skilled in the art should understand that the examples are for illustrative purposes only and not intended to limit the scope of the invention. Those skilled in the art should also understand that various modifications can be made to the embodiments without departing from the scope and spirit of the invention. The scope of this invention is defined by the appended claims.

Claims

1. A method for detecting overloaded carriages based on unmanned aerial vehicles (UAVs), characterized in that, The drone-based method for detecting overloaded carriages includes: Step 100: Obtain the 3D point cloud data, visible light image, spectral scattering image, rated load, and standard density of the cargo in the target carriage loaded with cargo at the current moment. Step 200: Construct a three-dimensional model of the target carriage based on the three-dimensional point cloud data, the visible light image, and the spectral scattering image; Step 300: Along the length of the target carriage, the three-dimensional model is divided into N carriage slices according to the preset carriage slice width. The actual stacking volume of the cargo in the N carriage slices and the porosity of the cargo in the N carriage slices are determined, where N is a positive integer greater than 1. Step 400: Determine the actual bulk density of the N car body slices based on the standard density of the cargo and the porosity of the cargo in the N car body slices. Step 500: Determine the total weight of the cargo in the target car based on the actual stacking volume of the cargo in the N car body slices and the actual stacking density of the cargo in the N car body slices. Step 600: When the total weight of the cargo is less than or equal to the rated load, the loading status of the target carriage is determined to be rated load status; when the total weight of the cargo is greater than the rated load, the loading status of the target carriage is determined to be overload status.

2. The method for detecting overloaded carriages based on unmanned aerial vehicles according to claim 1, characterized in that, Step 300 includes: Step 301: Determine the initial stacking volume of the cargo in the N car body slices based on the width of the car body slice and the N car body slices; Step 302: According to the preset voxel side length, divide the nth carriage slice into M voxels, and determine the number of 3D point clouds in the mth voxel within the nth carriage slice, where n = 1, 2, ..., N, M is a positive integer greater than 1, and m = 1, 2, ..., M; Step 303: For N carriage slices, determine the number of entity voxels and the number of boundary voxels in N carriage slices based on the number of 3D point clouds in the m-th voxel within the n-th carriage slice. Step 304: Determine the actual stacking volume of the goods in the N carriage slices based on the number of solid voxels in the N carriage slices, the number of boundary voxels in the N carriage slices, the side length of the voxels, and the preset boundary voxel weights. Step 305: Divide the difference between the initial stacked volume of the goods in the N carriage slices and the actual stacked volume of the goods in the N carriage slices by the initial stacked volume of the goods in the N carriage slices to determine the porosity of the goods in the N carriage slices. Step 400 includes: The actual bulk density of the N car body slices is determined by subtracting the product of the cargo porosity and the standard density of the cargo from the standard density of the cargo.

3. The method for detecting overloaded carriages based on unmanned aerial vehicles according to claim 1, characterized in that, Step 300 includes: Step 301: Determine the initial stacking volume of the goods in the N car body slices based on the width of the car body slice and the N car body slices; Step 302: According to the preset voxel side length, divide the nth carriage slice into M voxels, and determine the number of 3D point clouds in the mth voxel within the nth carriage slice, where n = 1, 2, ..., N, M is a positive integer greater than 1, and m = 1, 2, ..., M; Step 303: For N carriage slices, determine the number of entity voxels and the number of boundary voxels in N carriage slices based on the number of 3D point clouds in the m-th voxel within the n-th carriage slice. Step 304: Determine the actual stacking volume of the goods in the N carriage slices based on the number of solid voxels in the N carriage slices, the number of boundary voxels in the N carriage slices, the side length of the voxels, and the preset boundary voxel weights. Step 305: Divide the difference between the initial stacked volume of the goods in the N carriage slices and the actual stacked volume of the goods in the N carriage slices by the initial stacked volume of the goods in the N carriage slices to determine the porosity of the goods in the N carriage slices. Step 400 includes: Step 401: Obtain the pitch angle, roll angle, and speed fluctuation value of the UAV at the current moment; Step 402: Subtract the product of the porosity of the N car body slices and the standard density of the cargo from the standard density of the cargo to determine the initial bulk density of the N car body slices. Step 403: Determine the cargo porosity correction coefficient of the target carriage based on the pitch angle, the roll angle, the speed fluctuation value, and preset pitch angle weight, roll angle weight, and speed fluctuation value weight. Step 404: Determine the actual bulk density of the N pieces of cargo in the N car body slices based on the product of the initial bulk density of the cargo and the cargo porosity correction coefficient.

4. The method for detecting overloaded carriages based on unmanned aerial vehicles according to claim 1, characterized in that, Step 300 includes: Step 301: Determine the initial stacking volume of the goods in the N car body slices based on the width of the car body slice and the N car body slices; Step 302: According to the preset voxel side length, divide the nth carriage slice into M voxels, and determine the number of 3D point clouds in the mth voxel within the nth carriage slice, where n = 1, 2, ..., N, M is a positive integer greater than 1, and m = 1, 2, ..., M; Step 303: For N carriage slices, determine the number of entity voxels and the number of boundary voxels in N carriage slices based on the number of 3D point clouds in the m-th voxel within the n-th carriage slice. Step 304: Determine the actual stacking volume of the goods in the N carriage slices based on the number of solid voxels in the N carriage slices, the number of boundary voxels in the N carriage slices, the side length of the voxels, and the preset boundary voxel weights. Step 305: Divide the difference between the initial stacked volume of the goods in the N carriage slices and the actual stacked volume of the goods in the N carriage slices by the initial stacked volume of the goods in the N carriage slices to determine the porosity of the goods in the N carriage slices. Step 400 includes: Step 401: Obtain the pitch angle, roll angle, and speed fluctuation value of the UAV at the current moment; Step 402: Subtract the product of the porosity of the N car body slices and the standard density of the cargo from the standard density of the cargo to determine the initial bulk density of the N car body slices. Step 403: Determine the cargo porosity correction coefficient of the target carriage based on the pitch angle, the roll angle, the speed fluctuation value, and preset pitch angle weight, roll angle weight, and speed fluctuation value weight. Step 404: Determine the corrected bulk density of the N car body slices based on the product of the initial bulk density of the cargo in the N car body slices and the cargo porosity correction coefficient. Step 405: Determine the actual bulk density of the goods in the N car body slices based on the corrected bulk density of the goods, the standard density of the goods, and the preset cargo density weight.

5. A method for detecting off-center load in a train carriage based on unmanned aerial vehicles (UAVs), characterized in that, The unmanned aerial vehicle (UAV)-based method for detecting off-center loading in a vehicle includes: Step S100: Obtain the 3D point cloud data, visible light image, spectral scattering image, standard density of the cargo, X-axis tilt angle, and Y-axis tilt angle of the target carriage loaded with cargo at the current moment; Step S200: Construct a three-dimensional model of the target carriage based on the three-dimensional point cloud data, the visible light image, and the spectral scattering image; Step S300: Along the length of the target carriage, the three-dimensional model is divided into N carriage slices according to the preset carriage slice width. The actual stacking volume of the goods in the N carriage slices, the porosity of the goods in the N carriage slices, and the center of gravity coordinates of the goods in the N carriage slices are determined, where N is a positive integer greater than 1. Step S400: Determine the actual bulk density of the N car body slices based on the standard density of the cargo and the porosity of the cargo in the N car body slices. Step S500: Determine the actual center-of-gravity coordinates of the cargo in the target car based on the actual stacking volume of the cargo in the N car body slices, the actual stacking density of the cargo in the N car body slices, the center-of-gravity coordinates of the cargo in the N car body slices, the X-axis tilt angle, and the Y-axis tilt angle. Step S600: When the actual center of gravity coordinates of the cargo are within the preset ideal center of gravity range, the loading balance state of the target car is determined to be a balanced loading state; when the actual center of gravity coordinates of the cargo exceed the ideal center of gravity range, the loading balance state of the target car is determined to be an off-center loading state.

6. The method for detecting off-center loading of a carriage based on an unmanned aerial vehicle (UAV) according to claim 5, characterized in that, Step S300 includes: Step S301: Determine the initial stacking volume of the cargo in the N car body slices based on the width of the car body slice and the N car body slices. Step S302: According to the preset voxel side length, the nth carriage slice is divided into M voxels, and the number of three-dimensional point clouds in the mth voxel within the nth carriage slice is determined, where n = 1, 2, ..., N, M is a positive integer greater than 1, and m = 1, 2, ..., M. Step S303: For N carriage slices, determine the number of entity voxels and the number of boundary voxels in N carriage slices based on the number of 3D point clouds in the m-th voxel within the n-th carriage slice. Step S304: Determine the actual stacking volume of goods in the N car body slices based on the number of solid voxels in the N car body slices, the number of boundary voxels in the N car body slices, the side length of the voxels, and the preset boundary voxel weights. Step S305: Divide the difference between the initial stacked volume of the goods in the N carriage slices and the actual stacked volume of the goods in the N carriage slices by the initial stacked volume of the goods in the N carriage slices to determine the porosity of the goods in the N carriage slices. Step S400 includes: S401, obtain the current pitch angle, roll angle and speed fluctuation value of the UAV; S402, subtract the product of the porosity of the N car body slices and the standard density of the cargo from the standard density of the cargo to determine the initial bulk density of the N car body slices. S403, determine the cargo porosity correction coefficient of the target carriage based on the pitch angle, the roll angle, the speed fluctuation value, and preset pitch angle weight, roll angle weight, and speed fluctuation value weight; S404, determine the corrected bulk density of the N car body slices based on the product of the initial bulk density of the cargo in the N car body slices and the cargo porosity correction coefficient; S405, determine the actual bulk density of the goods in the N car body slices based on the corrected bulk density of the goods in the N car body slices, the standard density of the goods, and the preset cargo density weight.

7. The method for detecting off-center loading of a carriage based on an unmanned aerial vehicle (UAV) according to claim 5, characterized in that, Step S500 includes: S501, for N car body slices, sum the products of the actual bulk density of the goods in the nth car body slice, the actual bulk volume of the goods in the nth car body slice, and the center of gravity coordinates of the goods in the nth car body slice, and divide by the sum of the products of the actual bulk density of the goods in the nth car body slice and the actual bulk volume of the goods in the nth car body slice to determine the initial center of gravity coordinates of the goods in the target car body; S502, Based on the initial center of gravity coordinates of the cargo, the X-axis tilt angle, and the Y-axis tilt angle, determine the X-axis tilt amount and the Y-axis tilt amount of the target carriage; S503, the initial center of gravity coordinates of the cargo are subtracted from the X-axis tilt and the Y-axis tilt to determine the actual center of gravity coordinates of the cargo in the target carriage.

8. The method for detecting off-center loading of a carriage based on an unmanned aerial vehicle (UAV) according to claim 5, characterized in that, Following step S600, the following is also included: S701, Obtain the current speed of the target carriage; S702, based on the ideal center of gravity range, determine the maximum permissible lateral offset, maximum permissible longitudinal offset, maximum permissible vertical offset, and ideal cargo center of gravity coordinates of the target car body; S703, based on the actual center of gravity coordinates of the cargo and the ideal center of gravity coordinates of the cargo, determine the cargo center of gravity X-axis offset, cargo center of gravity Y-axis offset, and cargo center of gravity Z-axis offset of the target carriage. S704, based on the cargo center of gravity X-axis offset, cargo center of gravity Y-axis offset, cargo center of gravity Z-axis offset, running speed, maximum permissible lateral offset of the center of gravity, maximum permissible longitudinal offset of the center of gravity, maximum permissible vertical offset of the center of gravity, and preset running speed threshold, Y-axis tilt angle threshold, offset weight, running speed weight, and Y-axis tilt angle weight, determine the cargo off-center loading risk coefficient of the target car; S705, when the cargo off-center loading risk coefficient is greater than or equal to the preset off-center loading risk coefficient threshold, a risk warning is immediately triggered, and a status adjustment suggestion is output simultaneously.

9. A vehicle overload detection system based on unmanned aerial vehicles (UAVs), characterized in that, The UAV-based vehicle overload detection system includes a ground processing terminal and a UAV. The ground processing terminal and the UAV communicate with each other, and the ground processing terminal and the UAV work together to implement the UAV-based vehicle overload detection method according to any one of claims 1 to 4.

10. A vehicle carriage off-center load detection system based on unmanned aerial vehicles (UAVs), characterized in that, The UAV-based carriage overload detection system includes a ground processing terminal and a UAV. The ground processing terminal and the UAV communicate with each other, and the ground processing terminal and the UAV work together to implement the UAV-based carriage overload detection method according to any one of claims 5 to 8.

Images