A method for segmenting carriage point clouds based on geometric features and iterative optimization

Abstract

This invention discloses a truck body point cloud segmentation method based on geometric features and iterative optimization. The method includes: acquiring a 3D point cloud of a truck; determining the length, width, and height of the vehicle using a directed bounding box and a direction correction index, aligning the 3D point cloud of the truck along the corresponding coordinate axes; projecting the 3D point cloud of the truck onto a plane to extract the projected contour point cloud, dividing the contour point cloud into left and right sides and performing line fitting on each side, obtaining a set of 2D interior points corresponding to the left and right side lines through distance filtering; mapping the 2D interior points back to the 3D point cloud of the truck to obtain the left and right mapped point sets, calculating the local normalized gradient energy of the points and filtering salient points, combining the salient point set with a priority index to determine the initial left and right candidate segmentation points for the truck head and body; determining the guiding direction and generating guiding points based on the current initial left and right candidate segmentation points, obtaining optimized segmentation points through index verification, and completing the point cloud segmentation of the truck body. This invention can accurately segment the truck body point cloud.

Description

Technical Field

[0001] This invention relates to the fields of computer vision and 3D point cloud processing, and in particular to a method for segmenting carriage point clouds based on geometric features and iterative optimization. Background Technology

[0002] With the increasing demand for intelligent transportation, logistics monitoring, and vehicle supervision, higher requirements are being placed on the inspection of truck compartments. Point cloud data, as an information carrier that can accurately reflect the three-dimensional geometric features of objects, has been widely used in the field of vehicle inspection and modeling.

[0003] Currently, there are many existing methods for segmenting truck cargo compartments, including clustering-based point cloud segmentation methods and deep learning-based point cloud segmentation methods. Clustering-based methods use Euclidean clustering and other clustering techniques to segment the cab and cargo compartment based on the clustering results. However, the clustering results are sensitive to parameters, and the cab and cargo compartment point clouds are easily misclassified when they are connected. Deep learning-based point cloud segmentation methods use deep learning networks to perform semantic segmentation of vehicle point clouds, but they rely on a large amount of labeled data for training, resulting in insufficient generalization and high computational cost, making them unsuitable for real-time applications. Furthermore, during truck cargo compartment segmentation, there may be uncontrollable factors such as significant differences in vehicle structure, point cloud noise interference, and uncertain vehicle point cloud poses, all of which can affect the segmentation results. Summary of the Invention

[0004] In view of the above-mentioned defects and uncontrollable factors in the existing technology, the present invention provides a carriage point cloud segmentation method based on geometric features and iterative optimization to achieve accurate segmentation of carriage point clouds.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: In a first aspect, the present invention provides a method for segmenting carriage point clouds based on geometric features and iterative optimization, the method comprising: S1. Obtain the 3D point cloud of the truck; S2. Based on the directed bounding box of the truck's 3D point cloud, extract three pairwise orthogonal feature vectors and their corresponding side lengths from the directed bounding box. Combine the direction correction index to determine the length, width, and height directions of the vehicle. Map the vehicle's length, width, and height directions to the X, Y, and Z coordinate axes of the 3D Cartesian coordinate system, respectively. Align the truck's 3D point cloud along the corresponding coordinate axes through rigid body rotation. S3. Project the 3D point cloud of the truck onto the XY plane, extract the projection contour point cloud through the projection extreme points, divide the contour point cloud into left and right sides based on the global centroid, perform line fitting on the point clouds on both sides respectively, and then filter the two-dimensional interior point sets corresponding to the left and right side lines by the distance from the points to the fitted lines. S4. Map the two-dimensional interior points back to the three-dimensional point cloud of the truck to obtain the left and right mapped point sets. Calculate the local normalized gradient energy of the points in the mapped point set and screen out the significant points. Combine the significant point set with the priority index to determine the initial left and right candidate segmentation points for the truck head and the truck body. S5. Determine the guiding direction and generate guiding points based on the current initial left and right candidate segmentation points. Calculate the local forward guiding normal angle and local curvature of the guiding points. Obtain the optimized segmentation points through index verification. Use the optimized segmentation points to complete the point cloud segmentation of the truck body.

[0006] Following the above technical solution, step S1 further includes: The 3D point cloud of the truck is subjected to denoising and downsampling processing, including statistical filtering and voxel filtering.

[0007] Based on the above technical solution, the length, width, and height of the vehicle are determined as follows: The 3D point cloud computing of the truck generates a directed bounding box, from which three pairwise orthogonal feature vectors are extracted. and corresponding side length ,in ; Based on the relationship between the side lengths, the longest side is... Corresponding direction Determine the length direction of the vehicle, corresponding to the X-axis; then determine the second largest side length. Corresponding direction As a candidate direction for vehicle height, corresponding to the Z-axis; the minimum side length Corresponding direction The candidate direction for vehicle width corresponds to the Y-axis; The 3D point cloud of the truck is divided along the direction. and direction Divided into several sections; Define direction correction index ;in, Indicates along direction No. n Average height of the section Indicates along direction No. n Number of point sets within the segment Indicates along direction No. n All points within the section, Point The Z-coordinate value, Indicates along direction The number of segments divided; like Then determine the direction. Determine the direction based on the vehicle's height. For the width of the vehicle; conversely, for the direction. Determine the direction in the width direction of the vehicle. In the direction of vehicle height.

[0008] Following the above technical solution, the 3D point cloud of the truck is divided along the direction... and direction Divided into several sections, including: Define along direction No. n The point set within the segment is: ; ; in, Indicates along direction No. n The point set within the segment, Representing the 3D point cloud of a truck The point in the middle, For along direction Section interval, Point In direction Projected coordinates on; Indicates along direction The number of effective differences after removing zero differences. Indicates along direction After removing the zero difference, the first j The X coordinates of each valid point Indicates along direction After removing the zero difference, the first j The Y coordinate values ​​of each valid point; This determines the direction. The interval between the divided sections Number of segments and the point set within each section .

[0009] Following the above technical solution, step S3 includes: The 3D point cloud of the truck is projected onto the XY plane to obtain a 2D projected point cloud; The two-dimensional projection point cloud is divided into several continuous intervals along the length of the vehicle. The minimum and maximum extreme points of the Y coordinate in each interval are extracted as the outer edge points of that interval. The projection contour point cloud is formed by the outer edge points of all intervals. Determine the global centroid of the projected contour point cloud, and divide the projected contour point cloud into left and right sides according to the global centroid corresponding to the Y-axis; The Theil–Sen estimation method is used to fit straight lines to the point clouds on the left and right sides respectively, and the distances from the point clouds on the left and right sides to their respective fitted lines are calculated. Points whose distance is less than a preset threshold are taken as the two-dimensional interior points of the corresponding fitted lines, thus obtaining the two-dimensional interior point sets corresponding to the fitted lines on the left and right sides.

[0010] Following the above technical solution, step S4 includes: Mapping the two-dimensional interior points back to the three-dimensional point cloud of the truck yields mapping candidate points. The local average point spacing of each mapping candidate point is calculated, and the mapping candidate point with the smallest local average point spacing is selected as the final mapping point of the two-dimensional interior point, thus obtaining the left and right mapping point sets. For each internal point in the left and right mapping point sets, calculate the point... i Locally normalized gradient energy centered at: ; in, Represents the set of mapped points as points i The local normalized gradient energy centered on the center, and These represent the points in the mapping point set sorted by their X-coordinate values ​​from smallest to largest. i -1 and the first i +1 point's X coordinate value, and These represent the points in the mapping point set sorted by their X-coordinate values ​​from smallest to largest. i -1 and the first i +1 point Z coordinate value, This indicates the number of points in the corresponding mapped point set; Using locally normalized gradient energy Filter the set of significant gradient points and define the set of significant gradient points. for: ; in, k Indicates by The th in descending order k One point, T Indicates the threshold for gradient energy contribution. Represents the aligned 3D point cloud of the truck. The point in the middle; Aligned 3D point cloud of truck Calculate the center of the interval in the X direction , serving as a reference point for initial candidate segmentation points; Define the initial candidate split point priority index : ; in, Set of significant gradient points midpoint The X coordinate; Prioritize the initial candidate split points Sort the data in descending order and take the first point as the initial candidate split point for the corresponding side.

[0011] Following the above technical solution, step S5 includes: Determine the current left and right initial candidate split points Find the centroid of the neighborhood point set and obtain the X coordinate value. ; Obtain the aligned 3D point cloud of the truck Minimum value in the X direction With the maximum value ,like If the value is positive, the guidance direction is towards the positive X-axis, defined as +1; otherwise, it is towards the negative X-axis, defined as +1. 1. The guiding symbol is: ; Generate guide points based on guide symbols: ;in, Indicates the first i The current guiding point. express The next guide point along the guiding direction, To guide the step size, It is a unit vector; Determine the current guide point Local normal vector and next guiding point The angle between the local normal vectors is used as the angle between the local forward guiding normal vectors, and the current guiding point is determined. The local curvature; Verify whether the included angle of the local forward guiding normal and the local curvature meet the preset indicators. If not, continue guiding; if they meet, determine the current guiding point. To optimize the segmentation points, the optimal left and right segmentation points are obtained. and ; Then optimize the left and right segmentation points. and Perform left and right synchronization verification: ,in This is expressed as the allowable error threshold in the X direction; if the left and right synchronization verification passes, the left and right optimized split points will be... and As the final left and right optimization segmentation point; By utilizing the final optimized left and right segmentation points, a segmentation plane is constructed along the length of the vehicle to complete the point cloud segmentation of the truck body.

[0012] Following the above technical solution, verify whether the included angle of the local forward guiding normal and the local curvature both meet the preset indicators, including: Set the threshold for the included angle of the local forward guidance normal. and local curvature threshold If the angle between the local forward guiding normals And local curvature If the angle between the local forward guiding normal and the local curvature meet the preset index, then it is determined that the local forward guiding normal angle and the local curvature meet the preset index. If the left and right synchronization verification fails, then determine the current guiding direction: if the current guiding direction is the positive X-axis direction, then optimize the left and right split points. and Points with smaller X-coordinate values ​​continue iterative guidance along the guiding direction until both left and right sides pass simultaneous verification; if the current guiding direction is the negative X-axis direction, then the left and right optimization points are selected. and Points with large X-coordinate values ​​continue to be iteratively guided along the guiding direction until the left and right sides are simultaneously verified.

[0013] In a second aspect, the present invention provides a computer device / apparatus / system, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in the first aspect.

[0014] Thirdly, the present invention provides a computer-readable storage medium having a computer program / instructions stored thereon, which, when executed by a processor, implement the steps of the method described in the first aspect.

[0015] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects: This invention, by combining directed bounding boxes and orientation correction indices, can adaptively determine the length, width, and height of a vehicle and align the point cloud along the corresponding axes, thereby eliminating the influence of vehicle posture differences on the segmentation results and improving the adaptability and stability of the method. By combining projection contour extraction, significant gradient point screening, initial candidate segmentation point localization, and guided optimization strategies, it can accurately obtain the segmentation feature points of the truck front and the truck body, completing the accurate segmentation of the truck body point cloud. In addition, this method does not rely on a single geometric feature, but combines geometric features with iterative optimization strategies for truck body segmentation, thus having a certain degree of universality for different truck models. Attached Figure Description

[0016] Figure 1 This is a flowchart of a carriage point cloud segmentation method based on geometric features and iterative optimization according to an embodiment of the present invention; Figure 2 This is a comparison image of the truck point cloud before and after alignment according to an embodiment of the present invention; Figure 3 This is a projected outline view of a truck according to an embodiment of the present invention; Figure 4 This is a diagram showing the result of dividing the cargo compartment of a freight car according to an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. All other embodiments obtained by those skilled in the art based on the embodiments provided by this invention without inventive effort are within the scope of protection of this invention.

[0018] Obviously, the accompanying drawings described below are merely some examples or embodiments of the present invention. Those skilled in the art can apply the present invention to other similar scenarios based on these drawings without any inventive effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this invention, modifications to design, manufacturing, or production based on the technical content disclosed in this invention are merely conventional technical means and should not be construed as insufficient disclosure of the present invention.

[0019] In this invention, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this invention may be combined with other embodiments without conflict.

[0020] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "a," "an," "an," "the," and similar words used in this invention do not indicate quantity limitation and may indicate singular or plural. The terms "comprising," "including," "having," and any variations thereof used in this invention are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms "connected," "linked," "coupled," and similar words used in this invention are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "A plurality" used in this invention refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships may exist; for example, "A and / or B" can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following objects have an "or" relationship. The terms "first," "second," and "third" used in this invention are merely to distinguish similar objects and do not represent a specific ordering of the objects.

[0021] This invention provides a method for segmenting truck point clouds based on geometric features and iterative optimization. The method includes: denoising and downsampling the 3D point cloud of the truck; determining the length, width, and height directions of the vehicle using a directed bounding box (OBB) and a direction correction index, and then mapping these directions to the X-axis of a 3D Cartesian coordinate system. 、 Y 、 The Z-axis is used to align the truck's 3D point cloud along the corresponding axis through rigid body rotation. The truck's 3D point cloud is projected onto the XY plane, and the projected contour point cloud is obtained through the projection extrema. The contour point cloud is then grouped into left and right groups, and the Theil-Sen estimation method is used to obtain the left and right side lines and the set of two-dimensional interior points of the projected contour. Based on the left and right mapping point sets formed by mapping the two-dimensional interior points back to the truck's 3D point cloud, the local normalized gradient energy is calculated and salient points are selected. Initial candidate segmentation points are determined based on the salient point set and priority index. Guide points are generated based on the current initial candidate segmentation points and the determined guide direction. Optimized segmentation points are obtained by calculating the local forward guide normal angle and local curvature, and these optimized segmentation points are used to complete the truck body point cloud segmentation. This invention can accurately segment the truck body point cloud.

[0022] like Figure 1 As shown in the figure, the carriage point cloud segmentation method based on geometric features and iterative optimization according to an embodiment of the present invention includes the following steps: S1. Denoise and downsample the 3D point cloud of the truck; S2. After determining the length, width, and height of the vehicle using the Directed Bounding Box (OBB) and incorporating orientation correction indices, map these dimensions to the X-axis of the three-dimensional Cartesian coordinate system. 、 Y 、 The Z-axis is used to align the truck's 3D point cloud along the corresponding axis through rigid body rotation. S3. Project the 3D point cloud of the truck onto the XY plane and obtain the projected contour point cloud through the projection extreme points. Then, group the contour point cloud into left and right sides and use the Theil-Sen estimation method to obtain the left and right side lines and the two-dimensional interior point set of the projected contour. S4. Calculate the local normalized gradient energy and screen salient points based on the left and right mapping point sets formed by mapping the two-dimensional interior points back to the point cloud of the freight car, and determine the initial candidate segmentation points based on the set of salient points and priority index. S5. After generating guide points based on the current initial candidate segmentation points and the determined guide direction, optimize segmentation points are obtained by calculating the included angle of the local forward guide normal and the local curvature, and then the optimized segmentation points are used to complete the point cloud segmentation of the truck body.

[0023] In some embodiments, the specific method of step S1 is as follows: The acquired 3D point cloud data of the truck is denoted as: ;in, Represents the coordinates of three-dimensional points in a point cloud. The total number of points.

[0024] Subsequently, statistical filtering and voxel filtering are applied sequentially to the point cloud data to achieve noise removal and downsampling, resulting in a preprocessed point cloud. .

[0025] In some embodiments, the specific method of step S2 is as follows: Based on the obtained 3D point cloud of the truck Calculate the directed bounding box (OBB) and extract three pairwise orthogonal feature vectors from the OBB. and its corresponding side length ,in ; Based on the relationship between the side lengths, the direction corresponding to the longest side is... The direction is determined to be along the length of the vehicle. Since the width and height of the truck are similar, the direction corresponding to the second largest side length is chosen. The direction corresponding to the minimum side length is selected as a candidate direction for vehicle height. As a candidate direction for vehicle width.

[0026] Define along direction The nThe point set within the segment is: ; ; in, Indicates along direction No. n The point set within the segment, Point cloud Three-dimensional points in For along direction Section interval, Point In direction Projected coordinates on Indicates along direction The number of effective differences after removing zero differences. Indicates along direction After removing the zero difference, the first j The X coordinates of each valid point Indicates along direction After removing the zero difference, the first j The Y coordinate values ​​of each valid point.

[0027] It should be noted that, in this embodiment, "along the direction" refers to... Difference refers to calculating the coordinate differences between adjacent points within a point cloud and removing those with a difference of 0. For example, along a direction... Difference is the process of calculating the difference between the X coordinates of adjacent points and then removing the X coordinates with a difference of 0. This is equivalent to sorting all the X coordinates and then removing duplicates, keeping only one of each X coordinate value.

[0028] Define the direction correction index: ; in, Indicates along direction No. n Average height of the section Indicates along direction No. n Number of point sets within the segment Point The Z-coordinate value, Indicates along direction The number of segments divided.

[0029] Based on the geometric characteristics of trucks, there is usually a significant height abrupt change at the junction of the cab and the cargo box along the length of the vehicle. If it is established, then the direction is determined. Determine direction based on vehicle height. As for the width direction of the vehicle; if the condition is not met, that is Then exchange and Definition, determining direction Determine the direction as the width direction of the vehicle. As the vehicle's height direction.

[0030] like Figure 2 As shown, after determining the vehicle length, width, and height directions corresponding to the feature vectors extracted by OBB using the direction correction index, the vehicle length, width, and height directions are respectively mapped to X... 、 Y 、 The Z-axis is used, and the truck point cloud is aligned along the corresponding axis by rigid body rotation, thus obtaining the aligned truck point cloud. .

[0031] In some embodiments, the specific method of step S3 is as follows: 3D point cloud of the truck Projecting onto the XY plane yields a two-dimensional projected point cloud set: ; in, This represents a set of two-dimensional projected point clouds.

[0032] Then, the projected point cloud is divided into several continuous intervals along the vehicle's length direction, i.e., the X direction. In each interval... Calculate separately: ; get , and take the corresponding point. , As the outer edge point of this interval. Figure 3 As shown, the projected contour point cloud is obtained by using the set of outer edge points of all intervals: ; in, Representing the 3D point cloud of a truck The average distance between adjacent points along the length direction. This indicates the number of effective differences after removing zero differences. This indicates a range index.

[0033] Subsequently, based on the global centroid The projected contour point cloud is mapped according to the coordinates of its relative centroid. Divided into left and right sides: ; Finally, Theil–Sen estimation is used to fit a straight line to the point sets on both sides, and the fitted line is expressed in standard form: ; Therefore, we can further calculate the distances from the points in the point sets on both the left and right sides to the line. This is used to filter out the sets of interior points corresponding to the left and right lines. Taking the set of interior points corresponding to the left line as an example, we calculate the points in the left point set. to the straight line Distance: ; when When, then the point is considered Belongs to a straight line The interior points of the line to the left can be obtained by considering the interior points of the line. Similarly, the set of interior points of the line on the right can be obtained. .

[0034] In some embodiments, step S4 is specifically implemented as follows: To ensure that the interior points of the straight line model are mapped back to the point cloud of the freight car. The mapping point lies on the local dense plane, and the two-dimensional interior points are calculated respectively. These corresponding mapping candidate points Local average point spacing: ; The point with the smallest local average point spacing is selected as the mapping point of this interior point. Wherein, Candidate points for mapping The local neighborhood point set (which does not contain other candidate points). This represents the number of local neighborhood points. It is a point within the neighborhood.

[0035] Therefore, the mapping point set of the left and right lines can be obtained: ; in, Indicates the left and right sides of the projected outline. These represent truck point clouds. The set of mapping points on the left and right sides of the middle.

[0036] For each internal point in the set of mapped points ( i =2,3,... n -1), calculated using points i Locally normalized gradient energy centered at: ; in, Represents the set of mapped points as points i The local normalized gradient energy centered on the center, and These represent the points in the mapping point set sorted by their X-coordinate values ​​from smallest to largest. i-1 and the first i +1 point's X coordinate value, and These represent the points in the mapping point set sorted by their X-coordinate values ​​from smallest to largest. i -1 and the first i +1 point Z coordinate value, This represents the number of points in the corresponding mapping point set.

[0037] By using locally normalized gradient energy Calculate the set of significant gradient points, and define the set of significant gradient points as: ; in, k Indicates by The th in descending order k One point, T This represents the gradient energy contribution threshold, which is 0.9 in this embodiment.

[0038] 3D point cloud of truck of X Intervals in the direction Calculate the center of the interval : ; Among them, the As a reference point for the initial candidate split points, used to constrain the selection of the initial candidate split points, the priority index of the initial candidate split points is defined as follows: ; in, This indicates the priority index of the initial candidate split points. Set of significant gradient points midpoint The X coordinate.

[0039] according to Sort the data in descending order and take the first point as the initial candidate split point for the sidebar. This gives us two initial candidate split points for the left and right sides. .

[0040] In some embodiments, the specific method of step S5 is as follows: Let the initial candidate split points be set. The neighborhood point set is Its center of mass is: ; in, The centroid of the neighborhood set of the initial candidate split point is represented by X coordinate, Indicates the initial candidate segmentation points on the left and right sides. This represents the number of neighborhood points of the initial candidate split point.

[0041] Based on the truck point cloud obtained above Minimum value in the X direction Maximum value ,like If the value is positive, the guidance direction is towards the positive X-axis, defined as +1; otherwise, it is guided towards the negative X-axis, defined as +1. 1. Therefore, the guiding symbol can be written as: ; Therefore, the guiding point can be calculated: ; among which three-dimensional points Indicates the first i A guiding point. express The next point along the guiding direction, To guide the step size, i.e., the truck point cloud The average distance between adjacent points along the length direction It is a unit vector.

[0042] Next, the neighborhood point set of the guiding point... and Perform comprehensive feature analysis: calculate the current local normal vector. With forward-guided neighborhood normal The included angle is the included angle of the local forward guiding normal. and local curvature .in, , They represent , The neighborhood point set, For point set After covariance analysis, the eigenvalues ​​of the covariance matrix are obtained and .

[0043] Based on the above verification criteria, whether the following are met: If the conditions are met, the current guiding point is determined to be the optimized segmentation point. Finally, perform left and right synchronization verification: Therefore, the optimal split points on both sides can be obtained. If the left and right synchronization verification fails, and the guiding direction is the positive (or negative) X-axis direction, then the guided optimization split points on both sides are compared. The guiding point with the larger (or smaller) current X-value continues to be used as the optimized split point on that side, and the guided optimization split point on the other side continues to be iteratively calculated along the guiding direction until the verification conditions are met. and These are respectively represented as the local forward guiding normal angle threshold and the local curvature threshold. Desirable , The acceptable values ​​are [0.01, 0.05]. Representing vectors exist X Projected length in the direction, The permissible error threshold in the X direction can be expressed as... .

[0044] In this embodiment, The value is 75°. The value is 0.05; Values .like Figure 4 As shown, after constructing a segmentation plane along the length of the vehicle using optimized segmentation points, the points in the point cloud located on both sides of the segmentation plane are divided into the front point set and the cargo box point set, respectively, thus completing the segmentation of the truck cargo box point cloud.

[0045] It should be noted that constructing a segmentation plane along the length of the vehicle using optimized segmentation points specifically involves: using the final determined left and right optimized segmentation points to determine a segmentation plane perpendicular to the XOY plane, thereby completing the point cloud segmentation of the truck body.

[0046] Furthermore, the present invention also provides a computer device / apparatus / system, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-described method.

[0047] The present invention also provides a computer-readable storage medium having a computer program / instructions stored thereon, which, when executed by a processor, implement the steps of the above-described method.

[0048] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, systems, and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or block diagrams.

[0049] These computer program instructions may also be stored in a computer-readable storage medium capable of directing a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more blocks in a block diagram.

[0050] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable data processing apparatus to produce a computer-implemented process, thereby providing steps for implementing the functions specified in one or more flowcharts and / or one or more blocks in a block diagram.

[0051] It should be noted that, depending on the implementation needs, the various steps / components described in this invention can be broken down into more steps / components, or two or more steps / components or parts of the operation of steps / components can be combined into new steps / components to achieve the purpose of this invention.

[0052] Those skilled in the art will readily understand that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for segmenting carriage point clouds based on geometric features and iterative optimization, characterized in that, The method includes: S1. Obtain the 3D point cloud of the truck; S2. Based on the directed bounding box of the truck's 3D point cloud, extract three pairwise orthogonal feature vectors and their corresponding side lengths from the directed bounding box. Combine the direction correction index to determine the length, width, and height directions of the vehicle. Map the vehicle's length, width, and height directions to the X, Y, and Z coordinate axes of the 3D Cartesian coordinate system, respectively. Align the truck's 3D point cloud along the corresponding coordinate axes through rigid body rotation. S3. Project the 3D point cloud of the truck onto the XY plane, extract the projection contour point cloud through the projection extreme points, divide the contour point cloud into left and right sides based on the global centroid, perform line fitting on the point clouds on both sides respectively, and then filter the two-dimensional interior point sets corresponding to the left and right side lines by the distance from the points to the fitted lines. S4. Map the two-dimensional interior points back to the three-dimensional point cloud of the truck to obtain the left and right mapped point sets. Calculate the local normalized gradient energy of the points in the mapped point set and screen out the significant points. Combine the significant point set with the priority index to determine the initial left and right candidate segmentation points for the truck head and the truck body. S5. Determine the guiding direction and generate guiding points based on the current initial left and right candidate segmentation points. Calculate the local forward guiding normal angle and local curvature of the guiding points. Obtain the optimized segmentation points through index verification. Use the optimized segmentation points to complete the point cloud segmentation of the truck body.

2. The method for segmenting carriage point clouds based on geometric features and iterative optimization according to claim 1, characterized in that, Step S1 also includes: The 3D point cloud of the truck is subjected to denoising and downsampling processing, including statistical filtering and voxel filtering.

3. The method for segmenting carriage point clouds based on geometric features and iterative optimization according to claim 1, characterized in that, Determine the length, width, and height of the vehicle, specifically as follows: The 3D point cloud computing of the truck generates a directed bounding box, from which three pairwise orthogonal feature vectors are extracted. and corresponding side length ,in ; Based on the relationship between the side lengths, the longest side is... Corresponding direction Determine the length direction of the vehicle, corresponding to the X-axis; then determine the second largest side length. Corresponding direction As a candidate direction for vehicle height, corresponding to the Z-axis; the minimum side length Corresponding direction The candidate direction for vehicle width corresponds to the Y-axis; The 3D point cloud of the truck is divided along the direction. and direction Divided into several sections; Define direction correction index ;in, Indicates along direction No. n Average height of the section Indicates along direction No. n Number of point sets within the segment Indicates along direction No. n All points within the section, Point The Z-coordinate value, Indicates along direction The number of segments divided; like Then determine the direction. Determine the direction based on the vehicle's height. For the width of the vehicle; conversely, for the direction. Determine the direction in the width direction of the vehicle. In the direction of vehicle height.

4. The carriage point cloud segmentation method based on geometric features and iterative optimization according to claim 3, characterized in that, The 3D point cloud of the truck is divided along the direction. and direction Divided into several sections, including: Define along direction No. n The point set within the segment is: ； ； in, Indicates along direction No. n The point set within the segment, Representing the 3D point cloud of a truck The point in the middle, For along direction Section interval, Point In direction Projected coordinates on; Indicates along direction The number of effective differences after removing zero differences. Indicates along direction After removing the zero difference, the first j The X coordinates of each valid point Indicates along direction After removing the zero difference, the first j The Y coordinate values ​​of each valid point; This determines the direction. The interval between the divided sections Number of segments and the point set within each section .

5. The method for segmenting carriage point clouds based on geometric features and iterative optimization according to claim 1, characterized in that, Step S3 includes: The 3D point cloud of the truck is projected onto the XY plane to obtain a 2D projected point cloud; The two-dimensional projection point cloud is divided into several continuous intervals along the length of the vehicle. The minimum and maximum extreme points of the Y coordinate in each interval are extracted as the outer edge points of that interval. The projection contour point cloud is formed by the outer edge points of all intervals. Determine the global centroid of the projected contour point cloud, and divide the projected contour point cloud into left and right sides according to the global centroid corresponding to the Y-axis; The Theil–Sen estimation method is used to fit straight lines to the point clouds on the left and right sides respectively, and the distances from the point clouds on the left and right sides to their respective fitted lines are calculated. Points whose distance is less than a preset threshold are taken as the two-dimensional interior points of the corresponding fitted lines, thus obtaining the two-dimensional interior point sets corresponding to the fitted lines on the left and right sides.

6. The method for segmenting carriage point clouds based on geometric features and iterative optimization according to claim 1, characterized in that, Step S4 includes: Mapping the two-dimensional interior points back to the three-dimensional point cloud of the truck yields mapping candidate points. The local average point spacing of each mapping candidate point is calculated, and the mapping candidate point with the smallest local average point spacing is selected as the final mapping point of the two-dimensional interior point, thus obtaining the left and right mapping point sets. For each internal point in the left and right mapping point sets, calculate the point... i Locally normalized gradient energy centered at: ； in, Represents the mapping point set as points i The local normalized gradient energy centered on the center and These represent the points in the mapping point set sorted by their X-coordinate values ​​from smallest to largest. i -1 and the first i +1 point's X coordinate value, and These represent the points in the mapping point set sorted by their X-coordinate values ​​from smallest to largest. i -1 and the first i +1 point Z coordinate value, This indicates the number of points in the corresponding mapped point set; Using locally normalized gradient energy Filter the set of significant gradient points and define the set of significant gradient points. for: ； in, k Indicates by The th in descending order k One point, T This represents the threshold for gradient energy contribution. Represents the aligned 3D point cloud of the truck. The point in the middle; Aligned 3D point cloud of truck Calculate the center of the interval in the X direction , serving as a reference point for initial candidate segmentation points; Define the initial candidate split point priority index : ； in, Set of significant gradient points midpoint The X coordinate; Prioritize the initial candidate split points Sort the data in descending order and take the first point as the initial candidate split point for the corresponding side.

7. The method for segmenting carriage point clouds based on geometric features and iterative optimization according to claim 1, characterized in that, Step S5 includes: Determine the current left and right initial candidate split points Find the centroid of the neighborhood point set and obtain the X coordinate value. ; Obtain the aligned 3D point cloud of the truck Minimum value in the X direction With the maximum value ,like If the value is positive, the guidance direction is towards the positive X-axis, defined as +1; otherwise, it is towards the negative X-axis, defined as +1.

1. The guiding symbol is: ; Generate guide points based on guide symbols: ;in, Indicates the first i The current guiding point. express The next guide point along the guiding direction, To guide the step size, It is a unit vector; Determine the current guide point Local normal vector and next guiding point The angle between the local normal vectors is used as the angle between the local forward guiding normal vectors, and the current guiding point is determined. Local curvature; Verify whether the included angle of the local forward guiding normal and the local curvature meet the preset indicators. If not, continue guiding; if they meet, determine the current guiding point. To optimize the segmentation points, the optimal left and right segmentation points are obtained. and ; Then optimize the left and right segmentation points. and Perform left and right synchronization verification: ,in This is expressed as the allowable error threshold in the X direction; if the left and right synchronization verification passes, the left and right optimized split points will be... and As the final left and right optimization segmentation point; By utilizing the final optimized left and right segmentation points, a segmentation plane is constructed along the length of the vehicle to complete the point cloud segmentation of the truck body.

8. The method for segmenting carriage point clouds based on geometric features and iterative optimization according to claim 7, characterized in that, Verify whether the included angle of the local forward guiding normal and the local curvature both meet the preset indicators, including: Set the threshold for the included angle of the local forward guidance normal. and local curvature threshold If the angle between the local forward guiding normals And local curvature If the angle between the local forward guiding normal and the local curvature meet the preset index, then it is determined that the local forward guiding normal angle and the local curvature meet the preset index. If the left and right synchronization verification fails, then determine the current guiding direction: if the current guiding direction is the positive X-axis direction, then optimize the left and right split points. and Points with smaller X-coordinate values ​​continue iterative guidance along the guiding direction until both left and right sides pass simultaneous verification; if the current guiding direction is the negative X-axis direction, then the left and right optimization points are selected. and Points with large X-coordinate values ​​continue to be iteratively guided along the guiding direction until the left and right sides are simultaneously verified.

9. A computer device / equipment / system, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 8.

Images