Van identification and pose calculation method

Abstract

The application discloses a van identification and pose calculation method, which comprises the following steps: 1, an initial point cloud is acquired by using a loading robot TOF camera, and is converted to a world coordinate system through coordinate transformation; 2, local point clouds meeting a normal vector threshold are screened out through pretreatment and surface normal vector calculation; 3, bottom surface fitting is performed on the local point clouds; 4, local point clouds in a rear door bottom edge area are screened out; 5, linear fitting is completed by selecting feature points at a set interval; 6, boundary points of left and right frames of the rear door are determined, two lower corner points of the rear door are obtained by projecting the boundary points on the fitted straight line, and a midpoint of the rear door bottom edge is calculated; 7, an axis of the van is calculated, a lower midpoint of the upper frame of the rear door is determined, and finally complete van pose information including a van deflection angle, a coordinate of the midpoint of the rear door bottom edge, a van width and a van height is calculated. The application can accurately and quickly realize van pose calculation under a complex environment, so that the automatic loading task of the loading robot is realized.

Description

Technical Field

[0001] This invention relates to the field of van identification, detection, and pose calculation. Specifically, it is a method that enables a loading robot to automatically identify the van compartment and accurately calculate the van compartment pose under different working environments, thereby guiding the loading robot to load the van accurately. Background Technology

[0002] With the increasing development of automated factories, the application scope of loading robots is becoming wider and wider, and the penetration rate of unmanned operations is also constantly increasing. Loading robots are widely deployed in scenarios such as repetitive loading tasks, high-intensity loading work, and harsh working environments. In order to achieve a high degree of automation for loading robots, enabling them to automatically identify vans and accurately calculate the position and posture (pose) of the van, thereby automatically completing the loading operation, is crucial to improving the automation level of loading robots. Therefore, developing a van detection and pose calculation method with both sufficient accuracy and good real-time performance has become a key problem that urgently needs to be solved.

[0003] Traditional methods for recognizing and calculating the pose of vans mainly include the following: one is the recognition and pose estimation method based on the visual features of 2D images; another method is to acquire vehicle point clouds using LiDAR and match the point clouds with point cloud templates to complete the task; yet another method is to affix calibration plates, QR codes, reflective strips, and other manual markings to the rear door of the van, and calculate the vehicle pose by recognizing the corner points / coordinates of the markings and combining them with the camera calibration results.

[0004] The aforementioned traditional methods for identifying vans all have certain limitations. The first method is highly dependent on lighting conditions and is easily affected by environmental factors such as light, shadow, rain, snow, and occlusion, resulting in poor robustness. The second method, while processing point cloud data from LiDAR, suffers from poor real-time performance due to the massive amount of point cloud data and poor adaptability to vans of different sizes and structures. The third method heavily relies on manual labeling, making it unsuitable for unmarked scenarios; labels are easily worn, obscured, or soiled, leading to recognition failure; and customized labels are required for different vehicle models, resulting in poor universality and failing to meet the requirements of unmanned and automated loading operations. Summary of the Invention

[0005] The present invention aims to address the shortcomings of the existing technology by proposing a method for identifying and calculating the pose of a van, in order to achieve high-precision and high-speed van identification and pose detection, thereby realizing the goal of fully automated loading operations of loading robots.

[0006] To achieve the above-mentioned objectives, the present invention adopts the following technical solution: The present invention provides a method for recognizing and calculating the pose of a van, which is characterized by its application in cargo loading operations and is performed according to the following steps: Step 1: Establish the camera coordinate system The world coordinate system of loading robots ; Step 2: In the world coordinate system of the loading robot Obtain the original 3D point cloud of the carriage. The system then performs preprocessing to obtain a preprocessed 3D point cloud containing features of the carriage floor. ; Step 3, Calculation The surface normal vector, and based on the surface normal vector Axis component thresholding is used to extract the point cloud of the carriage floor. and to After performing a second pass-through filter, the filtered point cloud is obtained. Thus through Fitting the bottom surface of the carriage ; Step 4, find Mid-to-fit plane The distance satisfies the threshold and The point with the smallest coordinates is denoted as the rear endpoint of the base. ; Find Mid-to-fit plane The distance satisfies the threshold and The point with the smallest coordinates is denoted as the left endpoint of the base. ; Find Mid-to-fit plane The distance satisfies the threshold and The point with the largest coordinates is denoted as the right endpoint of the base. ; Step 5: Define and initialize the current selection count n=1; define and initialize the length offset for the nth selection. Define the maximum number of reselections as: N ; Step 6, based on ,exist The nth time the region point cloud is selected according to the requirements and from Filter out to the fitting plane Point clouds whose distances satisfy a threshold ; Step 7, according to Point cloud After sorting in ascending order, along Divide into equally spaced segments, and take the middle part of each segment. If the number of points in the extracted point set meets a preset threshold m, then after performing a line fitting on the point set, the nth fitted line is obtained. and will To the fitting plane After projection, the straight line after the nth projection is obtained. ; If the number of points in the extracted point set does not meet the threshold m, then... Assign to ,Will Assign to Then, return to step 6 and execute sequentially until... Until then; where b is the iteration step size; when If the carriage is not identified, it is determined that no carriage has been identified. Step 8: Place the left end point of the bottom surface and the right end point of the bottom surface Towards the straight line respectively Projection, and in Two projection points are retained in the middle. Local point cloud at the left end of the preset spatial range in the negative axis direction and the local point cloud on the right end Correspondingly, these serve as local point clouds for the left and right side frames of the side door; On the fitted line Find the boundary point of the left side of the compartment door. and the boundary point of the right frame and calculate and Midpoint between and the width of the carriage ; By fitting the plane and midpoint Find the axis of the carriage. This allows for the calculation of the carriage's deflection angle. ; Step 9: Based on the midpoint Fitting plane Determine the midpoint of the lower edge of the upper frame of the carriage. Then, the height of the carriage can be calculated. .

[0007] The method for calculating the position and orientation of a freight car body according to the present invention is characterized in that step 1 includes: Taking a fixed point in the middle of the loading robot as the origin Taking the conveying direction of the loading robot as The positive direction of the axis is defined as the direction perpendicular to the ground and pointing downwards. The positive direction of the axis is determined by the right-hand rule. The positive direction of the axis is used to establish the world coordinate system of the loading robot. ; Taking the light source point of the TOF camera as the origin The front of the TOF camera is The positive direction of the axis, the direction perpendicularly downwards from the TOF camera is... The positive direction of the axis is determined by the right-hand rule. The positive direction of the axis is used to establish the camera coordinate system. .

[0008] Furthermore, step 2 includes: Step 2.1: Obtain the camera coordinate system using a TOF camera. Original 3D point cloud of the lower carriage After coordinate transformation, the world coordinate system of the loading robot is obtained. Original 3D point cloud of the lower carriage ; Step 2.2: Remove using a pass-through filter Invalid point cloud, while preserving the world coordinate system. In The range of axis coordinates is ~ The point cloud inside is used to obtain a filtered 3D point cloud containing the features of the carriage floor. ,in, World coordinate system Down of Minimum value of the axis coordinates, World coordinate system Down of The maximum value of the axis coordinates. for Axial extreme range [ , The adaptive threshold set within; Step 2.3: Utilize rapid sampling pairs Sampling is performed to obtain a downsampled 3D point cloud containing bottom surface features. and to After performing statistical filtering to remove outliers, the result is... .

[0009] Furthermore, step 3 includes: Step 3.1, Calculation The surface normal vector, and based on the surface normal vector Threshold range for axis components, retaining point clouds that conform to the bottom surface features within the threshold range. ; Step 3.2, in Preserving the world coordinate system of Located on the axis within the preset range ~ Point clouds inside ,exist China retains Located on the axis within the interval ~ The point cloud within the carriage is used to obtain a filtered 3D point cloud containing local bottom surface features of the carriage. ; Step 3.3: Use weighted least squares method to... Perform planar fitting to obtain the bottom surface of the carriage. .

[0010] Furthermore, step 6 includes: Step 6.1, if the rear end point of the bottom surface of If the coordinates are less than or equal to 0, then the point will be... of Coordinates updated to points of The coordinates are kept constant, thus obtaining the left corner point of the bottom surface. And record the points. of The axis coordinates are ; then point of Coordinates updated to points of coordinates and The sum of these coordinates, with the other coordinates remaining unchanged, yields the right corner point of the bottom plane. , record points of Axis coordinates are marked as ; If the rear end point of the bottom surface of If the coordinates are greater than 0, then the point will be... of Coordinates updated to points of The coordinates are kept constant, thus obtaining the right corner point of the bottom surface. , record points of The axis coordinates are ; then point of Coordinates updated to points of coordinates and The sum of these coordinates, with the other coordinates remaining unchanged, yields the left corner point of the bottom surface. , record points of Coordinates are ; Step 6.2, in China retains coordinate range ~ Point clouds inside and from Select the base surface with the distance fitting Point clouds that meet the threshold .

[0011] Furthermore, the search in step 8 and The steps include: from Extract The left border point with the largest coordinates ; from Extract The rightmost bounding box point with the smallest coordinates ; Will and Towards the straight line respectively After projection, the left endpoint of the fitted line is obtained accordingly. and right endpoint .

[0012] Furthermore, in step 8, the deflection angle of the carriage is calculated. and the width of the carriage The steps include: On the fitted bottom surface Find the midpoint above And perpendicular to equation of the straight line , make the straight line Projected onto a plane The projected line is obtained. Calculate the projected line and The included angle is the deflection angle of the carriage. ; calculate and The distance between them is the allowable width of the carriage. .

[0013] Furthermore, step 9 includes: Construct a passage through the midpoint And perpendicular to the fitting plane straight line ; With midpoint Using the reference point, from China retains its line Local point cloud within a preset spatial range along the negative axis This refers to the partial dot cloud of the upper frame of the door; Will according to Sort the coordinates in descending order and select the line from them. Distance satisfies the threshold and The largest point And calculate the midpoint and points distance That is, the height of the carriage. .

[0014] The present invention provides an electronic device, including a memory and a processor, characterized in that the memory is used to store a program supporting the processor in executing the method, and the processor is configured to execute the program stored in the memory.

[0015] The present invention provides a computer-readable storage medium on which a computer program is stored, characterized in that the computer program is executed by a processor to perform the steps of the method.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention uses a TOF camera to acquire and process 3D point cloud data. Compared with traditional 2D image recognition methods, it can avoid the impact of lighting changes on the recognition results to a certain extent, improve the robustness of van recognition and pose calculation, and can stably complete recognition and pose calculation under normal lighting conditions, thus alleviating the problem of light sensitivity of existing 2D vision methods.

[0017] 2. This invention does not rely on fixed vehicle model templates, but achieves recognition and pose calculation solely through the geometric features of the vehicle body itself. It breaks away from the limitation of traditional template matching methods that are only applicable to a single vehicle model. It can adapt to a variety of common-sized and common-structure vans, and has a certain degree of versatility. There is no need to adjust the algorithm parameters separately for each vehicle model, which improves the ease of use.

[0018] 3. This invention filters effective point clouds with target features step by step through normal vector constraints, distance thresholds, etc., and removes some redundant points, invalid points and environmental interference points, thereby reducing the amount of point cloud data involved in the calculation, reducing the system's computational overhead, and improving the real-time performance of truck pose calculation to a certain extent. It can basically meet the real-time requirements of the loading robot's routine operations.

[0019] 4. This invention performs plane fitting on the bottom surface of the carriage by fitting local point clouds and further filters the point cloud of the bottom edge region of the rear door by combining distance thresholds. It focuses on key areas of the carriage, reduces redundant calculations of the global point cloud, further optimizes the algorithm's running efficiency, and reduces the impact of non-target structures on the fitting results.

[0020] 5. This invention calculates the deflection angle based on the axis of the bottom surface of the truck bed, and determines the width and height of the truck bed by combining the upper and lower midpoints of the rear door. It constructs a complete pose calculation logic, which can obtain complete pose information such as the deflection angle of the truck, the coordinates of the midpoint of the rear door, the width and height of the truck bed. This solves the problem of incomplete pose information in traditional methods and can provide a basic positioning reference for automated loading of trucks by loading robots. Attached Figure Description

[0021] Figure 1 This is a flowchart of the van recognition and pose calculation process of the present invention; Figure 2 This is a schematic diagram of the coordinate transformation of the present invention; Figure 3 This is a schematic diagram illustrating the actual application of the test scenario of this invention. Detailed Implementation

[0022] In this implementation example, a method for recognizing and calculating the pose of a van is applied to the loading operation of a loading robot, such as... Figure 1 As shown, the procedure is as follows: Step 1: Take a fixed point in the middle of the loading robot as the origin. Taking the conveying direction of the loading robot as The positive direction of the axis is defined as the direction perpendicular to the ground and pointing downwards. The positive direction of the axis is determined by the right-hand rule. The positive direction of the axis is used to establish the world coordinate system of the loading robot. .

[0023] Taking the light source point of the TOF camera as the origin The front of the TOF camera is The positive direction of the axis, the direction perpendicularly downwards from the TOF camera is... The positive direction of the axis is determined by the right-hand rule. The positive direction of the axis is used to establish the camera coordinate system. Coordinate relationships are as follows: Figure 2 As shown.

[0024] Step 2: Obtain the camera coordinate system using a TOF camera. Original 3D point cloud of the lower carriage After coordinate transformation, the world coordinate system of the loading robot is obtained. Original 3D point cloud of the lower carriage In this embodiment, the coordinate transformation is calculated based on the robot's own values ​​and the camera's mounting position, yielding the transformation matrix; the original 3D point cloud... like Figure 3 As shown in part (a) of the document.

[0025] Step 3: Remove using a pass-through filter Invalid point cloud, while preserving the world coordinate system. In The range of axis coordinates is ~ The point cloud inside is used to obtain a filtered 3D point cloud containing the features of the carriage floor. ,in, World coordinate system Down of Minimum value of the axis coordinates, World coordinate system Down of The maximum value of the axis coordinates. for Axial extreme range [ , The adaptive threshold set within; The direction is along the height of the carriage, so you can pass first. Directional pass-through filtering removes local point clouds above the bottom surface of the carriage, reducing the number of point clouds and improving computational efficiency.

[0026] Step 4: Utilize fast sampling pairs Sampling is performed to obtain a downsampled 3D point cloud containing bottom surface features. and to After performing statistical filtering to remove outliers, a point cloud is obtained. Point cloud like Figure 3 As shown in part (b) of the document.

[0027] Step 5, Calculation The surface normal vector, and based on the surface normal vector Threshold range for axis components, retaining point clouds that conform to the bottom surface features within the threshold range. Because the bottom, sides and front of the carriage are in The components of the axes differ significantly, therefore they can be identified through the normal vector. Threshold constraints on axis components are used to extract point clouds from the bottom surface of the carriage.

[0028] Step 6, in Preserving the world coordinate system of Located on the axis within the preset range ~ Point clouds inside ,exist China retains Located on the axis within the interval ~ The point cloud within the carriage is used to obtain a filtered 3D point cloud containing local bottom surface features of the carriage. ; The preset interval is based on the point cloud. of The range of coordinates is selected according to a certain proportion. Similarly, for the preset interval above, try to select the point cloud in the middle of the bottom surface of the carriage to improve the accuracy of fitting the bottom surface, such as... Figure 3 As shown in section (c) of the document.

[0029] Step 7: Apply weighted least squares method to the point cloud. Perform plane fitting to obtain the equation of the carriage bottom surface. .

[0030] Step 8: Find Mid-to-fit plane The distance satisfies the threshold and The point with the smallest coordinates is denoted as the rear endpoint of the base. ; Find Mid-to-fit plane The distance satisfies the threshold and The point with the smallest coordinates is denoted as the left endpoint of the base. ; Find Mid-to-fit plane The distance satisfies the threshold and The point with the largest coordinates is denoted as the right endpoint of the base. .

[0031] Step 9: Define and initialize the current selection count n=1; define and initialize the length offset for the nth selection. Define the maximum number of reselections as: N .

[0032] Step 10, if the bottom end point of If the coordinates are less than or equal to 0, then the point will be... of Coordinates updated to points of The coordinates are kept constant, thus obtaining the left corner point of the bottom surface. And record the points. of The axis coordinates are ; then point of Coordinates updated to points of coordinates and The sum of these coordinates, with the other coordinates remaining unchanged, yields the right corner point of the bottom plane. , record points of Axis coordinates are marked as .

[0033] If the rear end point of the bottom surface of If the coordinates are greater than 0, then the point will be... of Coordinates updated to points of The coordinates are kept constant, thus obtaining the right corner point of the bottom surface. , record points of The axis coordinates are ; then point of Coordinates updated to points of coordinates and The sum of these coordinates, with the other coordinates remaining unchanged, yields the left corner point of the bottom surface. , record points of Coordinates are .

[0034] Step 11, in China retains coordinate range ~ Point clouds inside and from Select the base surface with the distance fitting Point clouds that meet the threshold This will yield a point cloud that fits the back end of the bottom surface, without having to process the entire bottom surface point cloud, thus reducing computational overhead.

[0035] Step 12, according to Point cloud After sorting in ascending order, along Divide into equally spaced segments, and take the middle part of each segment. If the number of points in the extracted point set meets a preset threshold m, then after performing a line fitting on the point set, the nth fitted line is obtained. and will To the fitting plane After projection, the straight line after the nth projection is obtained. .

[0036] If the number of points in the extracted point set does not meet the threshold m, then... Assign to ,Will Assign to Then, return to step 10 and execute sequentially until... Until then; where b is the iteration step size; when If the time is not specified, it is determined that no carriage has been identified.

[0037] Step 13: Place the left end point of the bottom surface and the right end point of the bottom surface Towards the straight line respectively Projection, and in Two projection points are retained in the middle. Local point cloud at the left end of the preset spatial range in the negative axis direction and the local point cloud on the right end The corresponding point clouds are used as the left and right side frames of the door.

[0038] From point cloud Extract The left border point with the largest coordinates From point clouds Extract The rightmost bounding box point with the smallest coordinates ; point and points Towards a straight line After projection, the left endpoint of the fitted line is obtained. and right endpoint .

[0039] Step 14, Calculation and Midpoint between ; on the fitted bottom surface Find the midpoint above And perpendicular to equation of the straight line , make the straight line Projected onto a plane The projected line is obtained. Calculate the projected line and The included angle is the deflection angle of the carriage. ; calculate and The distance between them is the allowable width of the carriage. .

[0040] Step 15: Construct a structure passing through the midpoint And perpendicular to the fitted bottom surface straight line ; With midpoint Using the reference point, from China retains its line Local point cloud within a preset spatial range along the negative axis This refers to the partial dot cloud of the upper frame of the door; Will according to Sort the coordinates in descending order and select the line from them. Distance satisfies the threshold and The largest point And calculate the midpoint and points distance That is, the height of the carriage. .

[0041] In this embodiment, an electronic device includes a memory and a processor. The memory stores a program that supports the processor in executing the above-described method, and the processor is configured to execute the program stored in the memory.

[0042] In this embodiment, a computer-readable storage medium stores a computer program, which is executed by a processor to perform the steps of the above method.

[0043] The testing process for the van recognition and pose calculation method is as follows: exist Figure 3 The demonstrated process first focuses on a single frame of point cloud captured by the camera. Figure 3 Part (a) of the diagram undergoes preprocessing, including removing redundant and invalid point clouds, performing point cloud downsampling, and removing outliers to effectively reduce the number of point clouds, thereby obtaining the processed point cloud as shown in Figure (a). Figure 3 Part (b) of the text.

[0044] Then, calculate the normal vectors of these point clouds, based on Threshold selection of axis normal vector components for point cloud at the bottom of the carriage Furthermore, these point clouds are selected regionally, preserving local point clouds within the carriages. ,like Figure 3 As shown in section (c).

[0045] Subsequently, planar fitting was performed on these point clouds. The endpoints of the bottom surface were then located. , the left endpoint of the bottom surface Right end point of the ground By using the coordinate relationship of the three points mentioned above, the point cloud can be analyzed. Regional selection is performed to identify local point clouds at the rear of the carriage floor, and point clouds whose distances to the fitted floor surface meet a threshold are selected. These point clouds are then processed... Sort in ascending order along the axis, and select at intervals. Find the point with the smallest coordinates, fit a straight line to this series of points, and project it onto the fitted plane.

[0046] Then, find the boundary points of the left and right sides of the carriage on the fitted line, calculate the midpoint of the bottom edge of the rear door, determine the carriage's axis, and calculate the carriage's deflection angle and width. Based on the fitted plane and the midpoint of the bottom edge of the rear door, find the midpoint of the bottom edge of the top frame, thus determining the carriage's height. The pose calculation is visualized as follows: Figure 3 As shown in part (d) of the document.

[0047] This test fully demonstrates the reliability of the proposed van recognition and pose calculation method. Even in complex and varied working environments, the loading robot can achieve accurate van recognition and pose calculation using this method, and its performance is not limited by lighting conditions, maintaining excellent performance even at night or in low-light environments. Furthermore, the method has relatively low computational requirements, enabling high-speed operation on mobile devices with limited computing resources, ensuring the loading robot can complete loading tasks quickly and smoothly. Experimental results show that the method fully meets the needs of the loading robot in van recognition and pose calculation, and has never failed under preset conditions, further verifying its excellent stability.

Claims

1. A method for recognizing and calculating the pose of a van, characterized in that, It is applied to cargo loading operations and is carried out according to the following steps: Step 1: Establish the camera coordinate system The world coordinate system of loading robots ; Step 2: In the world coordinate system of the loading robot Obtain the original 3D point cloud of the carriage. The system then performs preprocessing to obtain a preprocessed 3D point cloud containing features of the carriage floor. ; Step 3, Calculation The surface normal vector, and based on the surface normal vector Axis component thresholding is used to extract the point cloud of the carriage floor. and to After performing a second pass-through filter, the filtered point cloud is obtained. Thus through Fitting the bottom surface of the carriage ; Step 4, find Mid-to-fit plane The distance satisfies the threshold and The point with the smallest coordinates is denoted as the rear endpoint of the base. ; Find Mid-to-fit plane The distance satisfies the threshold and The point with the smallest coordinates is denoted as the left endpoint of the base. ; Find Mid-to-fit plane The distance satisfies the threshold and The point with the largest coordinates is denoted as the right endpoint of the base. ; Step 5: Define and initialize the current selection count n=1; define and initialize the length offset for the nth selection. ; Define the maximum number of reselections as N ; Step 6, based on ,exist The nth selection of the region point cloud is performed according to the requirements. and from Filter out to the fitting plane Point clouds whose distances satisfy a threshold ; Step 7, according to Point cloud After sorting in ascending order, along Divide into equally spaced segments, and take the middle part of each segment. If the number of points in the extracted point set meets a preset threshold m, then after performing a line fitting on the point set, the nth fitted line is obtained. and will To the fitting plane After projection, the straight line after the nth projection is obtained. ; If the number of points in the extracted point set does not meet the threshold m, then... Assign to ,Will Assign to Then, return to step 6 and execute sequentially until... Until then; where b is the iteration step size; when If the carriage is not identified, it is determined that no carriage has been identified. Step 8: Place the left end point of the bottom surface and the right end point of the bottom surface Towards the straight line respectively Projection, and in Two projection points are retained in the middle. Local point cloud at the left end of the preset spatial range in the negative axis direction and the local point cloud on the right end Correspondingly, these serve as local point clouds for the left and right side frames of the side door; On the fitted line Find the boundary point of the left side of the compartment door. and the boundary point of the right frame and calculate and Midpoint between and the width of the carriage ; By fitting the plane and midpoint Find the axis of the carriage. This allows for the calculation of the carriage's deflection angle. ; Step 9: Based on the midpoint Fitting plane Determine the midpoint of the lower edge of the upper frame of the carriage. Then, the height of the carriage can be calculated. .

2. The method for calculating the position and orientation of a freight car body according to claim 1, characterized in that, Step 1 includes: Taking a fixed point in the middle of the loading robot as the origin Taking the conveying direction of the loading robot as The positive direction of the axis is defined as the direction perpendicular to the ground and pointing downwards. The positive direction of the axis is determined by the right-hand rule. The positive direction of the axis is used to establish the world coordinate system of the loading robot. ; Taking the light source point of the TOF camera as the origin The front of the TOF camera is The positive direction of the axis, the direction perpendicularly downwards from the TOF camera is... The positive direction of the axis is determined by the right-hand rule. The positive direction of the axis is used to establish the camera coordinate system. .

3. The method for calculating the position and orientation of a freight car body according to claim 1, characterized in that, Step 2 includes: Step 2.1: Obtain the camera coordinate system using a TOF camera. Original 3D point cloud of the lower carriage After coordinate transformation, the world coordinate system of the loading robot is obtained. Original 3D point cloud of the lower carriage ; Step 2.2: Remove using a pass-through filter Invalid point cloud, while preserving the world coordinate system. In The range of axis coordinates is ~ The point cloud inside is used to obtain a filtered 3D point cloud containing the features of the carriage floor. ,in, World coordinate system Down of Minimum value of the axis coordinates, World coordinate system Down of The maximum value of the axis coordinates. for Axial extreme range [ , The adaptive threshold set within; Step 2.3: Utilize rapid sampling pairs Sampling is performed to obtain a downsampled 3D point cloud containing bottom surface features. and to After performing statistical filtering to remove outliers, the result is... .

4. The method for calculating the position and orientation of a freight car body according to claim 1, characterized in that, Step 3 includes: Step 3.1, Calculation The surface normal vector, and based on the surface normal vector Threshold range for axis components, retaining point clouds that conform to the bottom surface features within the threshold range. ; Step 3.2, in Preserving the world coordinate system of Located on the axis within the preset range ~ Point clouds inside ,exist China retains Located on the axis within the interval ~ The point cloud within the carriage is used to obtain a filtered 3D point cloud containing local bottom surface features of the carriage. ; Step 3.3: Use weighted least squares method to... Perform planar fitting to obtain the bottom surface of the carriage. .

5. The method for calculating the position and orientation of a freight car body according to claim 1, characterized in that, Step 6 includes: Step 6.1, if the rear end point of the bottom surface of If the coordinates are less than or equal to 0, then the point will be... of Coordinates updated to points of The coordinates remain unchanged, thus obtaining the left corner point of the bottom surface. And record the points. of The axis coordinates are ; then point of Coordinates updated to points of coordinates and The sum of these coordinates, with the other coordinates remaining unchanged, yields the right corner point of the bottom plane. , record points of Axis coordinates are marked as ; If the rear end point of the bottom surface of If the coordinates are greater than 0, then the point will be... of Coordinates updated to points of The coordinates are kept constant, thus obtaining the right corner point of the bottom surface. , record points of The axis coordinates are ; then point of Coordinates updated to points of coordinates and The sum of these coordinates, with the other coordinates remaining unchanged, yields the left corner point of the bottom surface. , record points of Coordinates are ; Step 6.2, in China retains coordinate range ~ Point clouds inside and from Select the distance-fitted bottom surface Point clouds that meet the threshold .

6. The method for calculating the position and orientation of a freight car body according to claim 1, characterized in that, Finding in step 8 and The steps include: from Extract The left border point with the largest coordinates ; from Extract The rightmost bounding box point with the smallest coordinates ; Will and Towards the straight line respectively After projection, the left endpoint of the fitted line is obtained accordingly. and right endpoint .

7. The method for calculating the position and orientation of a freight car body according to claim 1, characterized in that, Calculate the deflection angle of the carriage in step 8. and the width of the carriage The steps include: On the fitted bottom surface Find the midpoint above And perpendicular to equation of the straight line , make the straight line Projected onto a plane The projected line is obtained. Calculate the projected line and The included angle is the deflection angle of the carriage. ; calculate and The distance between them is the allowable width of the carriage. .

8. The method for calculating the position and orientation of a freight car body according to claim 1, characterized in that, Step 9 includes: Construct a passage through the midpoint And perpendicular to the fitting plane straight line ; With midpoint Using the reference point, from China retains its line Local point cloud within a preset spatial range along the negative axis This refers to the partial dot cloud of the upper frame of the door; Will according to Sort the coordinates in descending order and select the line from them. Distance satisfies the threshold and The largest point And calculate the midpoint and points distance That is, the height of the carriage. .

9. An electronic device, comprising a memory and a processor, characterized in that, The memory is used to store a program that supports the processor in executing the method of any one of claims 1-8, the processor being configured to execute the program stored in the memory.

10. A computer-readable storage medium storing a computer program thereon, characterized in that, The computer program is executed by the processor to perform the steps of any of the methods described in claims 1-8.

Images