Object positioning method and device, nonvolatile storage medium and computer device

Abstract

This invention discloses an object localization method, apparatus, non-volatile storage medium, and computer device. The method includes: acquiring an initial point cloud image based on a depth camera on a target vehicle; transforming the initial point cloud image from the coordinate system corresponding to the depth camera to the coordinate system corresponding to the target vehicle to obtain a target point cloud image, wherein the coordinate system corresponding to the depth camera has its origin at the center of the depth camera; identifying point cloud images in the target point cloud image that do not exceed a preset object height to determine a first point cloud image; identifying and removing ground point cloud images from the first point cloud image based on a ground detection algorithm to obtain an object point cloud image; and determining the object's position in the coordinate system corresponding to the target vehicle based on the object point cloud image. This invention solves the technical problem of difficulty in accurately reaching the material picking station position every time during material tray localization due to complex and variable scenarios, resulting in a certain degree of deviation.

Description

Technical Field

[0001] This invention relates to the field of computer vision, and more specifically, to an object localization method, apparatus, non-volatile storage medium, and computer device. Background Technology

[0002] With the rapid iteration and maturation of autonomous navigation technology, forklift-type AGVs have been increasingly widely used. They have successfully overcome application bottlenecks in complex industrial environments and have achieved large-scale deployment in multiple fields such as automobile manufacturing, food and pharmaceuticals, e-commerce warehousing, and chemical logistics. For example, in automotive parts production workshops, forklift-type AGVs can autonomously and precisely deliver raw material pallets to the production line according to the production rhythm; in e-commerce warehousing centers, they can work with warehouse management systems to achieve cross-warehouse pallet transfers and inventory replenishment, effectively improving warehouse space utilization and cargo turnover efficiency.

[0003] In autonomous pallet transfer operations, forklift-type AGVs first need to navigate to a designated station location. At this station, a pallet of material to be transferred is typically placed. Upon arrival, the AGV needs to accurately align its forks into the square through-holes in the pallet before it can stably lift and transfer the material. After reaching the designated picking station, the AGV generally directly invokes a pre-tuned forklift-pallet picking motion. This operation relies on the AGV consistently reaching the same picking position as during the teaching demonstration. However, during AGV operation, real-world scenarios are often complex and dynamically changing. Therefore, ensuring the AGV accurately reaches the picking station location every time is challenging, and errors can occur in multiple directions. When these errors exist, the AGV's forks cannot directly lift the pallet according to the taught motion and are more likely to collide with the pallet during the process.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This invention provides an object positioning method, apparatus, non-volatile storage medium, and computer device to at least solve the technical problem that it is difficult to accurately reach the material picking station position every time when positioning a material tray due to the complexity and variability of the scene, resulting in a certain degree of deviation.

[0006] According to one aspect of the present invention, an object localization method is provided, comprising: acquiring an initial point cloud image based on a depth camera on a target vehicle; transforming the initial point cloud image from a coordinate system corresponding to the depth camera to a coordinate system corresponding to the target vehicle to obtain a target point cloud image, wherein the coordinate system corresponding to the depth camera has the center of the depth camera as the origin; identifying point cloud images in the target point cloud image that do not exceed a preset object height to determine a first point cloud image; identifying and removing ground point cloud images in the first point cloud image based on a ground detection algorithm to obtain an object point cloud image; and determining the position of the object in the coordinate system corresponding to the target vehicle based on the object point cloud image.

[0007] Optionally, the initial point cloud image is switched from the coordinate system corresponding to the depth camera to the coordinate system corresponding to the target vehicle to obtain the target point cloud image, including: determining a transformation matrix based on the coordinate system corresponding to the depth camera and the coordinate system corresponding to the target vehicle; obtaining the first coordinate of the initial point cloud image in the coordinate system corresponding to the depth camera; determining the second coordinate of the initial point cloud image in the coordinate system corresponding to the target vehicle based on the first coordinate and the transformation matrix; and determining the target point cloud image based on the second coordinate and the initial point cloud image.

[0008] Optionally, identifying the portion of the target point cloud image that does not exceed a preset object height to determine the first point cloud image includes: constructing an initial voxel grid structure based on a preset voxel grid size, wherein the initial voxel grid structure includes multiple voxel units; assigning any point in the target point cloud image to multiple voxel units in the initial voxel grid structure to obtain the first voxel grid structure; removing voxel units in the first voxel grid structure that do not contain points to obtain a second voxel grid structure; traversing the voxel units in the second voxel grid structure and selecting points that meet preset conditions as target points corresponding to voxel units in the second voxel grid structure; determining the filtered target point cloud image based on the target points corresponding to voxel units in the second voxel grid structure; and identifying the portion of the filtered target point cloud image that does not exceed a preset object height to determine the first point cloud image.

[0009] Optionally, identifying the portion of the target point cloud image that does not exceed the preset object height and determining the first point cloud image includes: acquiring the three-dimensional coordinates corresponding to each of multiple points in the target point cloud image; determining a z-axis coordinate threshold based on the preset object height; and filtering points whose z-axis coordinates are lower than or equal to the z-axis coordinate threshold based on the three-dimensional coordinates corresponding to each of the multiple points in the target point cloud image, thereby determining the first point cloud image.

[0010] Optionally, based on a ground detection algorithm, identifying and removing ground point cloud images from the first point cloud image to obtain an object point cloud image includes: randomly selecting multiple point sets in the first point cloud image based on the ground detection algorithm; fitting multiple point sets to obtain planar models corresponding to each point set; detecting the matching degree between points in the first point cloud image and the planar models corresponding to each point set, and determining the matching degree corresponding to each point set, wherein the matching degree represents the number of points in the first point cloud image whose distance to the planar model is lower than a preset distance threshold; selecting planar models whose matching degree meets preset matching conditions as ground point cloud images; and removing ground point cloud images from the first point cloud image to obtain the object point cloud image.

[0011] Optionally, determining the position of the object in the coordinate system corresponding to the target vehicle based on the object point cloud image includes: determining the center position of the object based on the object point cloud image; determining the principal direction of the object point cloud image based on the principal component analysis algorithm; determining the pose of the object relative to the target vehicle based on the principal direction; and determining the position of the object in the coordinate system corresponding to the target vehicle based on the center position of the object and the pose of the object relative to the target vehicle.

[0012] According to another aspect of the present invention, an object positioning device is also provided, comprising: an acquisition module for acquiring an initial point cloud image based on a depth camera on a target vehicle; a conversion module for converting the initial point cloud image from a coordinate system corresponding to the depth camera to a coordinate system corresponding to the target vehicle to obtain a target point cloud image, wherein the coordinate system corresponding to the depth camera has the center of the depth camera as its origin; a first recognition module for recognizing point cloud images in the target point cloud image that do not exceed a preset object height to determine a first point cloud image; a second recognition module for recognizing and removing ground point cloud images in the first point cloud image based on a ground detection algorithm to obtain an object point cloud image; and a determination module for determining the position of the object in the coordinate system corresponding to the target vehicle based on the object point cloud image.

[0013] According to another aspect of the present invention, a non-volatile storage medium is also provided, the non-volatile storage medium including a stored program, wherein, when the program is running, it controls the device where the non-volatile storage medium is located to execute any of the above-described object positioning methods.

[0014] According to another aspect of the present invention, a computer device is also provided, the computer device including a processor, the processor being configured to run a program, wherein the program, when running, executes any of the above-described object positioning methods.

[0015] According to another aspect of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements any of the above-described object positioning methods.

[0016] In this embodiment of the invention, an object localization method is employed. An initial point cloud image is acquired using a depth camera on the target vehicle. This initial point cloud image is then transformed from the coordinate system corresponding to the depth camera to the coordinate system corresponding to the target vehicle, resulting in a target point cloud image. The coordinate system corresponding to the depth camera has its origin at the center of the depth camera. Point cloud images within the target point cloud image that do not exceed a preset object height are identified, thus determining a first point cloud image. Based on a ground detection algorithm, ground point cloud images are identified and removed from the first point cloud image to obtain an object point cloud image. Based on the object point cloud image, the position of the object in the coordinate system corresponding to the target vehicle is determined, achieving the goal of accurately reaching the material picking station. This reduces the positioning error of the material tray and solves the technical problem of difficulty in accurately reaching the material picking station position each time due to complex and variable scenarios, resulting in a certain degree of deviation. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0018] Figure 1 A hardware structure block diagram of a computer terminal for implementing an object localization method is shown.

[0019] Figure 2 This is a flowchart illustrating the object positioning method provided according to an embodiment of the present invention;

[0020] Figure 3 This is a structural block diagram of an object positioning device provided according to an embodiment of the present invention. Detailed Implementation

[0021] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0022] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0023] According to an embodiment of the present invention, a method embodiment for object positioning is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0024] The method embodiment provided in Embodiment 1 of this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1 A hardware block diagram of a computer terminal for implementing an object localization method is shown. Figure 1 As shown, the computer terminal 10 may include one or more processors (shown as 102a, 102b, ..., 102n in the figure) (the processor may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing data. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0025] It should be noted that the aforementioned one or more processors and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10. As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0026] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the object positioning method in this embodiment of the invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby implementing the object positioning method of the aforementioned application. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0027] The display can be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10.

[0028] Figure 2 This is a flowchart illustrating the object positioning method provided in an embodiment of the present invention, as shown below. Figure 2 As shown, the method includes the following steps:

[0029] Step S202: Acquire an initial point cloud image based on the depth camera on the target vehicle.

[0030] In this step, the depth camera can be precisely mounted on the front end of the forklift-type AGV to ensure its field of view covers the pallet area. The camera's angle and position may need to be fine-tuned according to the actual application scenario to ensure optimal image quality. Furthermore, the transformation matrix of the depth camera relative to the AGV's body coordinate system is calculated using a camera hand-eye calibration method.

[0031] Furthermore, once the forklift AGV arrives at the designated picking station, its control system triggers the depth camera to begin capturing point cloud images. The "initial point cloud image" refers to the raw point cloud data directly output by the depth camera, containing depth information about the environment surrounding the picking station. These data points are described in the depth camera's coordinate system, with the origin located at the camera's optical center. To ensure high-quality point cloud images, the depth camera captures images when the forklift AGV is stationary. Capturing images while the forklift AGV is moving may result in blurred or erroneous depth information due to the relative motion between the camera and the target object.

[0032] After acquiring the initial point cloud image, the data preprocessing stage begins immediately. Preprocessing may include operations such as denoising and downsampling to reduce computational load and improve computational accuracy. For example, voxel filtering can be used to reduce point cloud density, while the SOR (Statistical Outlier Removal) algorithm can remove outliers from the point cloud.

[0033] Step S204: Transform the initial point cloud image from the coordinate system corresponding to the depth camera to the coordinate system corresponding to the target vehicle to obtain the target point cloud image, wherein the coordinate system corresponding to the depth camera takes the center of the depth camera as the origin.

[0034] In this step, a transformation matrix needs to be determined based on the coordinate system corresponding to the depth camera and the coordinate system corresponding to the target vehicle (e.g., a forklift AGV). The transformation matrix can be obtained through camera hand-eye calibration and can accurately describe the pose relationship between the depth camera coordinate system and the target vehicle coordinate system. The depth camera coordinate system has its origin at the center of the depth camera, while the target vehicle coordinate system may have its origin at a specific point on the vehicle (such as the center of the chassis). Therefore, the transformation matrix needs to consider both translation and rotation transformations.

[0035] Furthermore, point cloud data of the initial point cloud image in the depth camera coordinate system is acquired. This data exists in the form of three-dimensional coordinates (x, y, z), where x and y represent the position in the horizontal and vertical directions of the depth camera, and z represents depth information, i.e., distance from the depth camera. Coordinate transformation is performed based on the determined transformation matrix and the extracted three-dimensional coordinates of the initial point cloud image. By applying the transformation matrix, the new coordinates of each point cloud data in the target vehicle coordinate system, i.e., the second coordinates, are calculated. This process involves matrix multiplication, multiplying the coordinate vector of each point cloud by the transformation matrix to obtain the corresponding coordinates in the target vehicle coordinate system.

[0036] Finally, based on the calculated second coordinates and the original initial point cloud image, the point cloud image is reconstructed. The newly generated point cloud image, i.e., the target point cloud image, contains the position information of all point cloud data in the target vehicle coordinate system.

[0037] Step S206: Identify the point cloud image of the target point cloud image that does not exceed the preset object height, and determine the first point cloud image.

[0038] In this step, after acquiring the target point cloud image, the point cloud image that does not exceed the preset object height is identified and determined. This is achieved through a distance filtering algorithm, which separates the point cloud data of the ground and the object above the material tray, retaining only the point cloud data of the material tray, thus obtaining the first point cloud image as the input for subsequent algorithm processing.

[0039] First, a preset object height threshold needs to be set. This threshold is based on the actual physical dimensions of the material tray and the installation position and angle of the depth camera. The threshold should be high enough to cover the entire height of the material tray, but not so high that it includes the point cloud of objects above the material tray. For example, if the average height of the material tray is 1.2 meters, the installation height of the depth camera is 1.5 meters, and the downward angle towards the material tray is 45 degrees, then the preset object height threshold can be set to 1.3 meters.

[0040] Height detection is performed on each point cloud data in the target point cloud image. Since the target point cloud image has been transformed to the coordinate system corresponding to the target vehicle (AGV) using a transformation matrix, the three-dimensional coordinates (x, y, z) of each point cloud data can be directly analyzed in the AGV coordinate system. By comparing the z-axis coordinate value of each point cloud data with a preset object height threshold, point cloud data with z-axis coordinate values ​​below the threshold are filtered out.

[0041] Furthermore, in practice, this process can be achieved by traversing every point in the target point cloud image and comparing whether its z-axis coordinate value is lower than a preset height threshold. Points below the threshold are retained in the first point cloud image; points above the threshold are removed from the image, thus obtaining a first point cloud image containing only the material tray point cloud data.

[0042] Step S208: Based on the ground detection algorithm, identify and remove the ground point cloud image in the first point cloud image to obtain the object point cloud image.

[0043] First, based on the characteristics of the current scene, the parameters of the ground detection algorithm are initialized, such as the number of iterations, the minimum number of points in the fitted plane, and the maximum residual value. In the first point cloud image, the RANSAC algorithm is used to randomly select a certain number of points as initial samples. The RANSAC algorithm is used to estimate the parameters of the mathematical model from point cloud data containing outliers, which are points that deviate significantly from the true model. In this invention, outliers mainly refer to non-ground points above and around the material tray. Based on the selected point set, mathematical methods such as least squares are used for fitting to obtain a planar model.

[0044] Furthermore, the distances between all points in the first point cloud image and the planar model are calculated, and the number of points whose distances are below a preset distance threshold is counted. This number is used as the matching degree of the planar model. The preset distance threshold needs to be determined in conjunction with the flatness of the scene's ground and the accuracy requirements of the algorithm to ensure that ground points and non-ground points can be effectively distinguished.

[0045] Repeat the above process iteratively, generating a new planar model in each iteration. Among all generated planar models, the one with the highest matching degree is selected as the representative of the ground point cloud image. Once the optimal planar model is determined, all points whose matching degree with this model exceeds a preset matching condition are considered ground point cloud images and removed from the first point cloud image. The obtained object point cloud image is then preliminarily verified to ensure no residual ground point cloud data remains. If necessary, voxel filtering, statistical outlier removal (SOR), and other algorithms can be applied again to optimize the object point cloud image, further improving the accuracy and reliability of object localization.

[0046] Step S210: Based on the object point cloud image, determine the position of the object in the coordinate system corresponding to the target vehicle.

[0047] First, the centroid (center point) coordinates of the object are calculated based on all points in the object's point cloud image. This typically involves calculating the mean of the 3D coordinates of all point cloud data to estimate the object's geometric center. Further, principal component analysis (PCA) is used to determine the principal orientations of the object's point cloud image. This is done by calculating the covariance matrix of the point cloud data and then performing eigenvalue decomposition, yielding a set of eigenvalues ​​and corresponding eigenvectors. The order of the eigenvalues ​​reflects the variance of the point cloud data in different directions, and the eigenvectors represent these directions. Generally, the eigenvector corresponding to the largest eigenvalue represents the longest axis of the object, the eigenvector corresponding to the second largest eigenvalue represents the second longest axis, and the eigenvector corresponding to the smallest eigenvalue is perpendicular to the object's surface, representing the object's height.

[0048] Based on the principal component analysis results described above, the object's attitude relative to the target vehicle can be estimated, i.e., the object's offset and rotation angle. Combining the calculated center point of the object with the estimated attitude, the object's specific position in the target vehicle's coordinate system can be determined. This typically includes the coordinates of the object's center point (relative to the origin of the AGV coordinate system) and the object's attitude (rotation angle relative to the AGV coordinate system). For forklift-type AGVs, this means that they can adjust their movement trajectory based on this positional information to ensure that the forks can accurately align and lift the pallet. Even if there is a positional deviation when the AGV arrives at the picking station each time, it can be effectively corrected through this precise positioning process, thereby achieving high-precision pallet picking and transfer.

[0049] By following the steps above, the technical effect of reducing the positioning error of the material tray can be achieved, thereby solving the technical problem that it is difficult to accurately reach the position of the material picking station every time when the material tray is positioned due to the complexity and variability of the scene, resulting in a certain deviation.

[0050] As an optional embodiment, the following steps can be used to switch the initial point cloud image from the coordinate system corresponding to the depth camera to the coordinate system corresponding to the target vehicle to obtain the target point cloud image, including: determining a transformation matrix based on the coordinate system corresponding to the depth camera and the coordinate system corresponding to the target vehicle; obtaining the first coordinates of the initial point cloud image in the coordinate system corresponding to the depth camera; determining the second coordinates of the initial point cloud image in the coordinate system corresponding to the target vehicle based on the first coordinates and the transformation matrix; and determining the target point cloud image based on the second coordinates and the initial point cloud image.

[0051] First, the transformation matrix of the depth camera relative to the target vehicle coordinate system needs to be obtained using a camera hand-eye calibration method. This step involves capturing points that are visible to both the depth camera and the target vehicle coordinate system as the target vehicle moves, establishing the relationship, and then calculating the transformation matrix. The transformation matrix contains rotation and translation information to describe the spatial relationship between the depth camera coordinate system and the target vehicle coordinate system.

[0052] Furthermore, point cloud data is extracted from the raw images captured by the depth camera. Each pixel corresponds to a point in the depth image, and the three-dimensional coordinates of this point in the depth camera coordinate system are the first coordinates. Using the obtained transformation matrix, the first coordinates of each point in the extracted point cloud image are transformed, converting the coordinates of each point from the depth camera coordinate system to the target vehicle coordinate system, thus obtaining the second coordinates of the point in the target vehicle coordinate system.

[0053] Finally, after coordinate transformation, each point has the correct coordinates in the target vehicle coordinate system. At this point, the second coordinates of all points are matched with their relevant information (such as color, intensity, etc.) in the depth camera image, and reorganized into a new point cloud image, which is the target point cloud image.

[0054] As an optional embodiment, the following steps can be used to identify point cloud images in the target point cloud image that do not exceed the height of a preset object, and determine the first point cloud image, including: constructing an initial voxel grid structure based on a preset voxel grid size, wherein the initial voxel grid structure includes multiple voxel units; assigning any point in the target point cloud image to multiple voxel units in the initial voxel grid structure to obtain the first voxel grid structure; removing voxel units in the first voxel grid structure that do not contain points to obtain a second voxel grid structure; traversing the voxel units in the second voxel grid structure and selecting points that meet preset conditions as target points corresponding to voxel units in the second voxel grid structure; determining the filtered target point cloud image based on the target points corresponding to voxel units in the second voxel grid structure; and identifying point cloud images in the filtered target point cloud image that do not exceed the height of a preset object to determine the first point cloud image.

[0055] First, a voxel grid size needs to be defined, which depends on the expected size of the object and the required precision. For example, if the pallet is large, the voxel grid can be set relatively large to reduce computation; conversely, if high-precision positioning is required, the voxel grid should be set smaller. The initial voxel grid structure covers the entire spatial range of the point cloud image, where each voxel unit is a three-dimensional cube. Each point in the target point cloud image is traversed, and each point is assigned to a corresponding voxel unit based on its coordinates. In this way, each voxel unit may contain zero or more points, distributing the dense point cloud data across different voxel units.

[0056] Specifically, in the first voxel mesh structure, some voxel units may not be assigned any points due to the characteristics of the point cloud image. These empty voxel units do not contribute to subsequent object recognition and can therefore be removed to optimize storage and computational efficiency, resulting in a more compact second voxel mesh structure. For each non-empty voxel unit, the points within it need to be further filtered to ensure they meet specific conditions. For example, a point count threshold can be set within a voxel unit to remove those voxel units with too few points, thus eliminating potential noise or irrelevant points.

[0057] Through the above filtering process, the points within each voxel unit are further confirmed, and the resulting point cloud is called the "filtered target point cloud image." In the filtered target point cloud image, the set of points that does not exceed the height of a preset object (here, the material tray) is identified and filtered. The preset object height is an empirical value, usually set based on the design height of the material tray plus a small safety margin. By removing points higher than the predetermined height, the point cloud image of the material tray can be isolated, thus determining the first point cloud image.

[0058] As an optional embodiment, the following steps can be used to identify the point cloud image in the target point cloud image that does not exceed the preset object height, and determine the first point cloud image, including: obtaining the three-dimensional coordinates corresponding to each of the multiple points in the target point cloud image; determining the z-axis coordinate threshold based on the preset object height; and filtering points whose z-axis coordinates are lower than or do not exceed the z-axis coordinate threshold based on the three-dimensional coordinates corresponding to each of the multiple points in the target point cloud image, and determining the first point cloud image.

[0059] From the target point cloud image, the 3D coordinate information of each point is extracted, including values ​​along the x, y, and z axes. In the point cloud image acquired by the depth camera, each pixel has a corresponding 3D coordinate. Based on the depth camera's viewpoint and the expected height of the target object (such as a material tray), a preset object height threshold is set. This threshold is determined by adding a certain safety margin to the actual height of the material tray.

[0060] In a 3D coordinate system, the z-axis represents the vertical height. Based on a preset object height, a z-axis coordinate threshold is further determined. This threshold is used to filter out point cloud data within the height range of the material tray in the target point cloud image. For each point in the target point cloud image, its z-axis value in 3D coordinates is checked to see if it is lower than or equal to the z-axis coordinate threshold. This filtering process can effectively remove point cloud data that may exist in the image, such as sky, buildings, trees, or other objects higher than the preset object height.

[0061] Furthermore, after filtering the point cloud data, all point cloud data that meet the criteria (i.e., points with z-axis coordinates lower than or equal to a z-axis coordinate threshold) are collected to form a new point cloud image, called the first point cloud image. This image is then further optimized, such as through voxel filtering to improve point cloud clarity, or by using other point cloud processing algorithms to enhance image quality.

[0062] As an optional embodiment, this can be achieved through the following steps: based on a ground detection algorithm, identifying and removing ground point cloud images from a first point cloud image to obtain an object point cloud image, including: randomly selecting multiple point sets in the first point cloud image based on the ground detection algorithm; fitting multiple point sets to obtain planar models corresponding to each of the multiple point sets; detecting the matching degree between points in the first point cloud image and the planar models corresponding to each of the multiple point sets, and determining the matching degree corresponding to each of the multiple point sets, wherein the matching degree represents the number of points in the first point cloud image whose distance to the planar model is lower than a preset distance threshold; selecting planar models whose matching degree meets preset matching conditions as ground point cloud images; and removing ground point cloud images from the first point cloud image to obtain the object point cloud image.

[0063] First, a certain number of points are randomly selected from the first point cloud image. The selection of the point set should take into account the distribution of the point cloud image, ensuring that the point set can cover different regions of the entire image. The RANSAC (Random Sample Consensus) algorithm can be used as the basic framework for the ground detection algorithm. For each randomly selected point set, a suitable fitting method is used to determine a planar model. Planar fitting methods include least squares or Hough transform. Least squares finds the optimal model parameters by minimizing the sum of squared errors, and this method is suitable for cases with low noise. Hough transform is more suitable for processing noisy data. It can find lines or planes in datasets that are not completely or partially occluded, thereby identifying the ground.

[0064] Specifically, after fitting planar models, the matching degree between points in the point cloud images and these models is evaluated. The matching degree is evaluated based on the distance from the point to the plane; that is, for each point in the first point cloud image, the distance to the fitted planar model for each point set is calculated. If the distance between a point and a planar model is below a preset distance threshold, then this point is considered a matching point for that planar model. The preset distance threshold is chosen considering the accuracy of the point cloud and the actual ground conditions, typically at the centimeter level. The matching degree of all fitted planar models is statistically analyzed, i.e., the number of matching points for each model is calculated. A model with a high matching degree indicates that more points in the point cloud image match it, and this model is more likely to represent the true planar state of the ground. By comparing the matching degrees of multiple models, the planar model that meets the preset matching conditions is selected as the representative of the ground point cloud image. The preset matching conditions can be the absolute number of matching points or a relative proportion based on the total number of points in the point cloud.

[0065] Finally, after determining the optimal ground plane model, points belonging to the ground in the first point cloud image are removed, leaving only the object point cloud image. By comparing the distance of each point in the point cloud to the optimal ground model, any point whose distance exceeds a preset distance threshold is considered not to belong to the ground and is retained as part of the object point cloud image. Filtering out the ground point cloud image from the first point cloud image yields the clean object point cloud image. This set of point cloud images contains no ground information, only the object's point cloud data.

[0066] As an optional embodiment, the following steps can be used to determine the position of the object in the coordinate system corresponding to the target vehicle based on the object point cloud image, including: determining the center position of the object based on the object point cloud image; determining the principal direction of the object point cloud image based on the principal component analysis algorithm; determining the pose of the object relative to the target vehicle based on the principal direction; and determining the position of the object in the coordinate system corresponding to the target vehicle based on the center position of the object and the pose of the object relative to the target vehicle.

[0067] Specifically, the centroid of the object is calculated from the point cloud image to determine the center position of the object in three-dimensional space, using the following formula:

[0068] ;

[0069] Where N is the number of points in the point cloud. These are the coordinates of the object's center position.

[0070] Furthermore, based on Principal Component Analysis (PCA), the principal orientation of the object's point cloud image is determined. First, the covariance matrix of the object's point cloud image is constructed. Then, the eigenvalues ​​and eigenvectors of the covariance matrix are calculated. The eigenvector with the largest eigenvalue points to the principal orientation of the data, which is the principal orientation of the object's point cloud image. Based on this principal orientation, the object's pose relative to the target vehicle is determined.

[0071] In the principal directions of an object's point cloud image, the eigenvector corresponding to the largest eigenvalue indicates the object's width, the eigenvector corresponding to the second largest eigenvalue indicates the object's height, and the eigenvector corresponding to the smallest eigenvalue indicates the object's depth (i.e., roughly perpendicular to the point cloud surface). By analyzing the relationship between these three directions and their corresponding axes in the target vehicle's coordinate system, the object's rotation angle relative to the target vehicle can be estimated, thus determining its attitude.

[0072] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0073] Through the above description of the embodiments, those skilled in the art can clearly understand that the object positioning method according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0074] According to embodiments of the present invention, an object positioning device for implementing the above-described object positioning method is also provided. Figure 3 This is a structural block diagram of an object positioning device provided according to an embodiment of the present invention, such as... Figure 3 As shown, the object positioning device includes: an acquisition module 302, a conversion module 304, a first identification module 306, a second identification module 308, and a determination module 310. The object positioning device will be described below.

[0075] The acquisition module 302 is used to acquire an initial point cloud image based on the depth camera on the target vehicle;

[0076] The conversion module 304, connected to the acquisition module 302, is used to convert the initial point cloud image from the coordinate system corresponding to the depth camera to the coordinate system corresponding to the target vehicle to obtain the target point cloud image, wherein the coordinate system corresponding to the depth camera takes the center of the depth camera as the origin;

[0077] The first identification module 306, connected to the conversion module 304, is used to identify the point cloud image of the target point cloud image that does not exceed the preset object height, and to determine the first point cloud image;

[0078] The second recognition module 308 is connected to the first recognition module 306 and is used to identify and remove the ground point cloud image in the first point cloud image based on the ground detection algorithm to obtain the object point cloud image.

[0079] The determination module 310, connected to the second recognition module 308, is used to determine the position of the object in the coordinate system corresponding to the target vehicle based on the object point cloud image.

[0080] It should be noted that the acquisition module 302, conversion module 304, first identification module 306, second identification module 308, and determination module 310 mentioned above correspond to steps S202 to S210 in the embodiments. Multiple modules and their corresponding steps implement the same instances and application scenarios, but are not limited to the content disclosed in the above embodiments. It should also be noted that the above modules, as part of the device, can run on the computer terminal 10 provided in the embodiments.

[0081] Embodiments of the present invention may provide a computer device. Optionally, in this embodiment, the computer device may be located in at least one of a plurality of network devices in a computer network. The computer device includes a memory and a processor.

[0082] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the object positioning method and apparatus in this embodiment of the invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the object positioning method described above. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to a computer terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0083] The processor can access the information and application program stored in the memory via the transmission device to perform the following steps: acquiring an initial point cloud image based on the depth camera on the target vehicle; transforming the initial point cloud image from the coordinate system corresponding to the depth camera to the coordinate system corresponding to the target vehicle to obtain a target point cloud image, wherein the coordinate system corresponding to the depth camera has the center of the depth camera as the origin; identifying the point cloud image in the target point cloud image that does not exceed a preset object height to determine the first point cloud image; identifying and removing the ground point cloud image in the first point cloud image based on a ground detection algorithm to obtain an object point cloud image; and determining the position of the object in the coordinate system corresponding to the target vehicle based on the object point cloud image.

[0084] Optionally, the processor may also execute program code for the following steps: switching the initial point cloud image from the coordinate system corresponding to the depth camera to the coordinate system corresponding to the target vehicle to obtain the target point cloud image, including: determining a transformation matrix based on the coordinate system corresponding to the depth camera and the coordinate system corresponding to the target vehicle; obtaining the first coordinate of the initial point cloud image in the coordinate system corresponding to the depth camera; determining the second coordinate of the initial point cloud image in the coordinate system corresponding to the target vehicle based on the first coordinate and the transformation matrix; and determining the target point cloud image based on the second coordinate and the initial point cloud image.

[0085] Optionally, the processor may also execute program code for the following steps: identifying point cloud images in the target point cloud image that do not exceed a preset object height, and determining a first point cloud image, including: constructing an initial voxel grid structure based on a preset voxel grid size, wherein the initial voxel grid structure includes multiple voxel units; assigning any point in the target point cloud image to multiple voxel units in the initial voxel grid structure to obtain the first voxel grid structure; removing voxel units in the first voxel grid structure that do not contain points to obtain a second voxel grid structure; traversing the voxel units in the second voxel grid structure, selecting points that meet preset conditions as target points corresponding to voxel units in the second voxel grid structure; determining a filtered target point cloud image based on the target points corresponding to voxel units in the second voxel grid structure; identifying point cloud images in the filtered target point cloud image that do not exceed a preset object height, and determining the first point cloud image.

[0086] Optionally, the processor may also execute program code for the following steps: identifying point cloud images in the target point cloud image that do not exceed a preset object height, and determining a first point cloud image, including: acquiring the three-dimensional coordinates corresponding to each of multiple points in the target point cloud image; determining a z-axis coordinate threshold based on the preset object height; and filtering points whose z-axis coordinates are lower than or equal to the z-axis coordinate threshold based on the three-dimensional coordinates corresponding to each of multiple points in the target point cloud image, and determining the first point cloud image.

[0087] Optionally, the processor may also execute program code for the following steps: based on a ground detection algorithm, identifying and removing ground point cloud images from a first point cloud image to obtain an object point cloud image, including: randomly selecting multiple point sets in the first point cloud image based on a ground detection algorithm; fitting multiple point sets to obtain planar models corresponding to each of the multiple point sets; detecting the matching degree between points in the first point cloud image and the planar models corresponding to each of the multiple point sets, and determining the matching degree corresponding to each of the multiple point sets, wherein the matching degree represents the number of points in the first point cloud image whose distance to the planar model is lower than a preset distance threshold; selecting planar models whose matching degree meets preset matching conditions as ground point cloud images; and removing ground point cloud images from the first point cloud image to obtain an object point cloud image.

[0088] Optionally, the processor may also execute program code for the following steps: determining the position of an object in the coordinate system corresponding to the target vehicle based on the object point cloud image, including: determining the center position of the object based on the object point cloud image; determining the principal direction of the object point cloud image based on the principal component analysis algorithm; determining the pose of the object relative to the target vehicle based on the principal direction; and determining the position of the object in the coordinate system corresponding to the target vehicle based on the center position of the object and the pose of the object relative to the target vehicle.

[0089] This invention provides a method for object localization. An initial point cloud image is acquired using a depth camera on a target vehicle. This initial point cloud image is then transformed from the coordinate system corresponding to the depth camera to the coordinate system corresponding to the target vehicle, resulting in a target point cloud image. The coordinate system corresponding to the depth camera has its origin at the center of the depth camera. Point cloud images within the target point cloud image that do not exceed a preset object height are identified, forming a first point cloud image. Ground point cloud images within the first point cloud image are identified and removed using a ground detection algorithm, resulting in an object point cloud image. Based on the object point cloud image, the object's position in the coordinate system corresponding to the target vehicle is determined, achieving accurate positioning at the material picking station. This reduces the positioning error of the material tray and solves the technical problem of difficulty in accurately reaching the material picking station position each time due to complex and variable scenarios, resulting in a certain degree of deviation.

[0090] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a non-volatile storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0091] Embodiments of the present invention also provide a non-volatile storage medium. Optionally, in this embodiment, the non-volatile storage medium can be used to store the program code executed by the object positioning method provided in the above embodiments.

[0092] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0093] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: acquiring an initial point cloud image based on a depth camera on the target vehicle; transforming the initial point cloud image from the coordinate system corresponding to the depth camera to the coordinate system corresponding to the target vehicle to obtain a target point cloud image, wherein the coordinate system corresponding to the depth camera has the center of the depth camera as its origin; identifying point cloud images in the target point cloud image that do not exceed a preset object height to determine a first point cloud image; identifying and removing ground point cloud images in the first point cloud image based on a ground detection algorithm to obtain an object point cloud image; and determining the position of the object in the coordinate system corresponding to the target vehicle based on the object point cloud image.

[0094] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: switching the initial point cloud image from the coordinate system corresponding to the depth camera to the coordinate system corresponding to the target vehicle to obtain the target point cloud image, including: determining a transformation matrix based on the coordinate system corresponding to the depth camera and the coordinate system corresponding to the target vehicle; obtaining the first coordinates of the initial point cloud image in the coordinate system corresponding to the depth camera; determining the second coordinates of the initial point cloud image in the coordinate system corresponding to the target vehicle based on the first coordinates and the transformation matrix; and determining the target point cloud image based on the second coordinates and the initial point cloud image.

[0095] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: identifying point cloud images in a target point cloud image that do not exceed a preset object height, and determining a first point cloud image, including: constructing an initial voxel grid structure based on a preset voxel grid size, wherein the initial voxel grid structure includes multiple voxel units; assigning any point in the target point cloud image to multiple voxel units in the initial voxel grid structure to obtain a first voxel grid structure; removing voxel units in the first voxel grid structure that do not contain points to obtain a second voxel grid structure; traversing the voxel units in the second voxel grid structure, selecting points that meet preset conditions as target points corresponding to voxel units in the second voxel grid structure; determining a filtered target point cloud image based on the target points corresponding to voxel units in the second voxel grid structure; identifying point cloud images in the filtered target point cloud image that do not exceed a preset object height, and determining a first point cloud image.

[0096] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: identifying point cloud images in the target point cloud image that do not exceed a preset object height, and determining a first point cloud image, including: acquiring the three-dimensional coordinates corresponding to each of multiple points in the target point cloud image; determining a z-axis coordinate threshold based on the preset object height; and filtering points whose z-axis coordinates are lower than or equal to the z-axis coordinate threshold based on the three-dimensional coordinates corresponding to each of the multiple points in the target point cloud image, and determining the first point cloud image.

[0097] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: based on a ground detection algorithm, identifying and removing ground point cloud images from a first point cloud image to obtain an object point cloud image, including: randomly selecting multiple point sets in the first point cloud image based on the ground detection algorithm; fitting multiple point sets to obtain planar models corresponding to each of the multiple point sets; detecting the matching degree between points in the first point cloud image and the planar models corresponding to each of the multiple point sets, and determining the matching degree corresponding to each of the multiple point sets, wherein the matching degree characterizes the number of points in the first point cloud image whose distance to the planar model is lower than a preset distance threshold; selecting planar models whose matching degree meets preset matching conditions as ground point cloud images; and removing ground point cloud images from the first point cloud image to obtain an object point cloud image.

[0098] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: determining the position of an object in the coordinate system corresponding to the target vehicle based on the object point cloud image, including: determining the center position of the object based on the object point cloud image; determining the principal direction of the object point cloud image based on the principal component analysis algorithm; determining the pose of the object relative to the target vehicle based on the principal direction; and determining the position of the object in the coordinate system corresponding to the target vehicle based on the center position of the object and the pose of the object relative to the target vehicle.

[0099] Embodiments of the present invention also provide a computer program product, including a computer program. Optionally, in this embodiment, when the computer program is executed by a processor, it can perform the following: acquiring an initial point cloud image based on a depth camera on a target vehicle; transforming the initial point cloud image from the coordinate system corresponding to the depth camera to the coordinate system corresponding to the target vehicle to obtain a target point cloud image, wherein the coordinate system corresponding to the depth camera has the center of the depth camera as its origin; identifying point cloud images in the target point cloud image that do not exceed a preset object height to determine a first point cloud image; identifying and removing ground point cloud images in the first point cloud image based on a ground detection algorithm to obtain an object point cloud image; and determining the position of the object in the coordinate system corresponding to the target vehicle based on the object point cloud image.

[0100] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0101] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0102] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0103] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0104] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0105] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a non-volatile storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0106] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for locating an object, characterized in that, include: Acquire initial point cloud images using the depth camera on the target vehicle; The initial point cloud image is transformed from the coordinate system corresponding to the depth camera to the coordinate system corresponding to the target vehicle to obtain the target point cloud image, wherein the coordinate system corresponding to the depth camera has the center of the depth camera as the origin; Identify the portion of the target point cloud image that does not exceed the preset object height, and determine the first point cloud image; Based on the ground detection algorithm, the ground point cloud image in the first point cloud image is identified and removed to obtain the object point cloud image; Based on the point cloud image of the object, the position of the object in the coordinate system corresponding to the target vehicle is determined.

2. The method according to claim 1, characterized in that, The step of switching the initial point cloud image from the coordinate system corresponding to the depth camera to the coordinate system corresponding to the target vehicle to obtain the target point cloud image includes: Based on the coordinate system corresponding to the depth camera and the coordinate system corresponding to the target vehicle, the transformation matrix is ​​determined; Obtain the first coordinates of the initial point cloud image in the coordinate system corresponding to the depth camera; Based on the first coordinates and the transformation matrix, the second coordinates of the initial point cloud image in the coordinate system corresponding to the target vehicle are determined; The target point cloud image is determined based on the second coordinates and the initial point cloud image.

3. The method according to claim 1, characterized in that, The step of identifying the portion of the target point cloud image that does not exceed the preset object height, and determining the first point cloud image, includes: Based on a preset voxel grid size, an initial voxel grid structure is constructed, wherein the initial voxel grid structure includes multiple voxel units; Any point in the target point cloud image is assigned to multiple voxel units of the initial voxel mesh structure to obtain the first voxel mesh structure. Remove the voxel cells that do not contain points from the first voxel mesh structure to obtain the second voxel mesh structure; Traverse the voxel cells in the second voxel mesh structure, and select points that meet the preset conditions as the target points corresponding to the voxel cells in the second voxel mesh structure; Based on the target points corresponding to the voxel units in the second voxel mesh structure, the filtered target point cloud image is determined. Identify the portion of the filtered target point cloud image that does not exceed the preset object height, and determine the first point cloud image.

4. The method according to claim 1, characterized in that, The step of identifying the portion of the target point cloud image that does not exceed the preset object height, and determining the first point cloud image, includes: Obtain the three-dimensional coordinates of each of the multiple points in the target point cloud image; Based on the preset object height, determine the z-axis coordinate threshold; Based on the three-dimensional coordinates of multiple points in the target point cloud image, points with z-axis coordinates lower than or equal to the z-axis coordinate threshold are selected to determine the first point cloud image.

5. The method according to claim 1, characterized in that, The method based on ground detection algorithm, which identifies and removes ground point cloud images from the first point cloud image to obtain an object point cloud image, includes: Based on the ground detection algorithm, multiple point sets are randomly selected in the first point cloud image; By fitting the multiple point sets, a planar model corresponding to each of the multiple point sets is obtained; The matching degree between points in the first point cloud image and the corresponding planar models of the plurality of point sets is detected, and the matching degree of the plurality of point sets is determined. The matching degree represents the number of points in the first point cloud image whose distance from the planar model is less than a preset distance threshold. A planar model whose matching degree meets the preset matching conditions is selected as the ground point cloud image; The ground point cloud image is removed from the first point cloud image to obtain the object point cloud image.

6. The method according to any one of claims 1 to 5, characterized in that, Determining the position of the object in the coordinate system corresponding to the target vehicle based on the object point cloud image includes: Based on the point cloud image of the object, determine the center position of the object; Based on the principal component analysis algorithm, the main orientation of the object point cloud image is determined; Based on the main direction, the attitude of the object relative to the target vehicle is determined; Based on the center position of the object and the orientation of the object relative to the target vehicle, the position of the object in the coordinate system corresponding to the target vehicle is determined.

7. An object positioning device, characterized in that, include: The acquisition module is used to acquire an initial point cloud image based on the depth camera on the target vehicle; The conversion module is used to convert the initial point cloud image from the coordinate system corresponding to the depth camera to the coordinate system corresponding to the target vehicle to obtain the target point cloud image, wherein the coordinate system corresponding to the depth camera takes the center of the depth camera as the origin; The first recognition module is used to recognize the point cloud image of the target point cloud image that does not exceed the preset object height, and to determine the first point cloud image; The second recognition module is used to identify and remove the ground point cloud image in the first point cloud image based on the ground detection algorithm to obtain the object point cloud image. The determination module is used to determine the position of the object in the coordinate system corresponding to the target vehicle based on the object point cloud image.

8. A non-volatile storage medium, characterized in that, The non-volatile storage medium includes a stored program, wherein, when the program is executed, it controls the device containing the non-volatile storage medium to perform the object positioning method according to any one of claims 1 to 6.

9. A computer device, characterized in that, include: Memory and processor The memory stores computer programs; The processor is configured to execute a computer program stored in the memory, wherein when the computer program is executed, the processor performs the object positioning method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the object positioning method according to any one of claims 1 to 6.

Images