Object position detection method and device, robot, readable storage medium and program product
By obtaining the prior position and point cloud data of the target object, clustering and calculating geometric properties, and grouping to determine the target clustering center, the problem of low positioning accuracy of traditional object detection is solved, and efficient and accurate object position detection is achieved.
Patent Information
- Application Number
- CN202411079552.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-06
- Publication Date
- 2024-11-15
AI Technical Summary
In traditional object detection and positioning technology, there is a problem of low accuracy when identifying the position of an object through graphic codes, especially when the position of the object changes, it is difficult to accurately indicate the actual position of the object.
By obtaining the prior location and point cloud data of the target object, clustering to obtain the candidate cluster center, calculate the geometric properties of the candidate cluster center, group and determine the target cluster center, thereby determining the location of the target object and reducing the search range. Improve the efficiency and accuracy of position detection.
It significantly reduces the calculation amount, improves the efficiency and accuracy of position detection, can clearly and intuitively reflect geometric shape characteristics, and quickly and efficiently determine the position of the target object.
Smart Images

Figure CN118968004A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computer technology, and in particular to an object position detection method, device, robot, computer-readable storage medium and computer program product. Background Art
[0002] With the development of robotics, object detection and positioning technology has emerged. During operation, the robot collects information about the surrounding environment through sensors, and then detects the positional relationship between the surrounding objects and itself based on the environmental information, providing important data support for the robot's positioning and path planning.
[0003] In traditional technology, a graphic code with a unique code is attached to objects in the environment, and the robot determines the location and attribute information of the object by scanning the graphic code. When the position of the object changes, the position information identified by the graphic code cannot indicate the actual position of the object, resulting in low position detection accuracy. Summary of the invention
[0004] Based on this, it is necessary to provide an object position detection method, device, robot, computer-readable storage medium and computer program product that can improve the accuracy of position detection in response to the above technical problems.
[0005] In a first aspect, the present application provides an object position detection method, comprising:
[0006] Obtain the prior position corresponding to the target object, and obtain the target point cloud data for the prior position collected by the robot;
[0007] Cluster the target point cloud data to obtain multiple candidate cluster centers corresponding to the target object;
[0008] Grouping the candidate cluster centers to obtain multiple candidate cluster center combinations;
[0009] Calculate the geometric properties between each candidate cluster center included in the candidate cluster center combination;
[0010] Based on the geometric attributes corresponding to each candidate cluster center combination, a target cluster center combination corresponding to the target object is determined in each candidate cluster center combination;
[0011] The target position of the target object with respect to the robot is determined based on the target cluster center combination.
[0012] In a second aspect, the present application further provides an object position detection device, comprising:
[0013] A data acquisition module is used to obtain a priori position corresponding to the target object and obtain target point cloud data for the priori position collected by the robot;
[0014] Point cloud clustering module, used to cluster the target point cloud data and obtain multiple candidate cluster centers corresponding to the target object;
[0015] A center combination module is used to group the candidate cluster centers to obtain multiple candidate cluster center combinations;
[0016] An attribute calculation module is used to calculate the geometric attributes between each candidate cluster center included in the candidate cluster center combination;
[0017] A combination determination module, used to determine a target cluster center combination corresponding to the target object in each candidate cluster center combination based on the geometric attributes corresponding to each candidate cluster center combination;
[0018] The position determination module is used to determine the target position of the target object with respect to the robot based on the target cluster center combination.
[0019] In a third aspect, the present application further provides a robot, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned object position detection method when executing the computer program.
[0020] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above-mentioned object position detection method when executed by a processor.
[0021] In a fifth aspect, the present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned object position detection method when executed by a processor.
[0022] The object position detection method, device, robot, computer-readable storage medium and computer program product obtain the prior position corresponding to the target object, obtain the target point cloud data collected by the robot for the prior position, cluster the target point cloud data, and obtain multiple candidate cluster centers corresponding to the target object. Then, each candidate cluster center is grouped to obtain multiple candidate cluster center combinations, and the geometric properties between each candidate cluster center contained in each candidate cluster center combination are calculated. Based on the geometric properties corresponding to each candidate cluster center combination, the target cluster center combination corresponding to the target object is determined in each candidate cluster center combination. Finally, the target position of the target object for the robot is determined based on the target cluster center combination. In this way, each candidate cluster center is grouped to obtain multiple candidate cluster center combinations to be identified, and it is determined whether each candidate cluster center combination is the target cluster center combination corresponding to the target object. Compared with directly finding the cluster center corresponding to the target object from all candidate cluster center combinations, the search range can be greatly reduced, the amount of calculation can be significantly reduced, and the efficiency of position detection can be improved. Furthermore, since the geometric attributes corresponding to the candidate cluster center combination can clearly and intuitively reflect the characteristics of the geometric shapes formed by each candidate cluster center in the combination, the target cluster center combination corresponding to the target object can be quickly and efficiently determined from each candidate cluster center combination based on the geometric attributes, which can reduce the complexity of position detection, improve the efficiency of position detection, and improve the accuracy of position detection. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.
[0024] FIG1 is a diagram of an application environment of an object position detection method according to an embodiment;
[0025] FIG2 is a schematic diagram of a flow chart of an object position detection method in one embodiment;
[0026] FIG3 is a schematic flow chart of steps for collecting target point cloud data in one embodiment;
[0027] FIG4 is a schematic diagram of geometric properties in one embodiment;
[0028] FIG5 is a schematic flow chart of an object position detection method in another embodiment;
[0029] FIG6 is a block diagram of a device for detecting an object position in one embodiment;
[0030] FIG. 7 is a diagram showing the internal structure of a robot in one embodiment. DETAILED DESCRIPTION
[0031] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0032] The object position detection method provided in the embodiment of the present application can be applied to the application environment shown in Figure 1. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. The terminal 102 obtains the prior position corresponding to the target object and obtains the target point cloud data for the prior position collected by the robot. The terminal 102 clusters the target point cloud data to obtain multiple candidate cluster centers corresponding to the target object. The terminal 102 groups each candidate cluster center to obtain multiple candidate cluster center combinations. The terminal 102 calculates the geometric properties between each candidate cluster center contained in the candidate cluster center combination. The terminal 102 determines the target cluster center combination corresponding to the target object in each candidate cluster center combination based on the geometric properties corresponding to each candidate cluster center combination. The terminal 102 determines the target position of the target object for the robot based on the target cluster center combination. Among them, the terminal 102 can be, but is not limited to, various robots, personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car devices, projection devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. Head-mounted devices can be virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc. Robots can be various industrial robots that need to move autonomously (such as handling robots, stacking robots, spraying robots, etc.), service robots (such as cleaning robots, delivery robots, mowing robots, etc.) or special robots (firefighting robots, underwater robots, security robots, etc.). The server 104 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services.
[0033] In an exemplary embodiment, as shown in FIG2 , a method for detecting the position of an object is provided, and the method is described below by taking a robot as an example. The method for detecting the position of an object includes the following steps 202 to 212. Among them:
[0034] Step 202, obtaining a priori position corresponding to the target object, and obtaining target point cloud data for the priori position collected by the robot.
[0035] Among them, the target object refers to the object that needs to be detected in the robot's operating area. For example, in the intelligent warehousing scenario, the handling robot needs to move a movable load-bearing platform (such as a shelf or pallet) from one location to another. At this time, the target object is the movable load-bearing platform. Before lifting the movable load-bearing platform, it is necessary to detect the center position of the movable load-bearing platform, and then control the robot to drive to the bottom of the movable load-bearing platform and lift and carry the movable load-bearing platform to ensure stability during the handling process; when the robot performs obstacle avoidance tasks, it needs to monitor and identify obstacles in real time. At this time, the target object is the obstacle in the working environment; and so on.
[0036] The priori position refers to the approximate position of a known target object. For example, the priori position corresponding to the target object can be obtained from the position information of the target object stored in a priori map.
[0037] The target point cloud data refers to the environmental point cloud data collected by the robot through sensors. In the actual implementation process, the target point cloud data can be two-dimensional laser point cloud data collected by a lidar sensor.
[0038] Exemplarily, when the robot performs a task for a target object, it first obtains the prior position corresponding to the target object. Then, it obtains the target point cloud data for the prior position. Specifically, according to the prior position corresponding to the target object, the robot posture is adjusted so that the target object enters the sensor sensing range of the robot, and the initial point cloud data for the prior position is collected by the sensor, and then the point cloud data outside the prior position in the initial point cloud data is filtered to obtain the target point cloud data for the prior position.
[0039] Step 204 , clustering the target point cloud data to obtain a plurality of candidate cluster centers corresponding to the target object.
[0040] The candidate cluster center refers to the cluster center corresponding to the point cloud cluster extracted from the target point cloud data and corresponding to the positioning reference part of the target object. The positioning reference part refers to the key part of the target object used to assist in positioning the target object. For example, when the target object is a table, the positioning reference part can be the table leg; when the target object is a shelf, the positioning reference part can be the shelf leg; and so on.
[0041] Exemplarily, the robot clusters the target point cloud data based on the distance between each data point in the target point cloud data to obtain a plurality of initial point cloud clusters. Based on the size of each initial point cloud cluster, each initial point cloud cluster is filtered to obtain a plurality of candidate point cloud clusters. The cluster center corresponding to each candidate point cloud cluster is used as the candidate cluster center corresponding to the target object. In the actual implementation process, each initial point cloud cluster can be filtered according to at least one of the characteristics of the initial point cloud cluster, such as size, shape and density, to obtain a plurality of candidate point cloud clusters. Comprehensively filtering irrelevant point cloud data based on multiple factors can reduce the interference caused by environmental factors, improve the accuracy of position detection, and at the same time greatly reduce the amount of data calculation, save computer resources, and improve the efficiency of position detection.
[0042] Step 206: group the candidate cluster centers to obtain multiple candidate cluster center combinations.
[0043] The candidate cluster center combination refers to a set containing multiple candidate cluster centers.
[0044] Exemplarily, the robot determines the minimum number of cluster centers required for locating the target object based on the structural features of the target object. The candidate cluster centers are grouped based on the minimum number of cluster centers corresponding to the target object to obtain a plurality of candidate cluster center combinations. For example, when the target object is a rectangular shelf with four shelf legs, it is only necessary to determine the positions of three of the shelf legs to determine the position of the rectangular shelf. Therefore, the minimum number of cluster centers corresponding to the rectangular shelf is 3. Three different candidate cluster centers are selected from each candidate cluster center each time to form a candidate cluster center combination. All possible combinations are traversed to obtain a plurality of candidate cluster center combinations.
[0045] Step 208: Calculate geometric properties between candidate cluster centers included in the candidate cluster center combination.
[0046] The geometric attributes refer to the geometric properties of the geometric shapes formed by the candidate cluster centers, including at least one of the properties of connecting lines, angles, faces, etc.
[0047] Exemplarily, for any candidate cluster center combination, the robot calculates the geometric properties between each candidate cluster center in the candidate cluster center combination, and the geometric properties may include at least one of the characteristics such as the length of the connecting line between the candidate cluster centers, the angle between the connecting lines, and the plane where each candidate cluster center is located.
[0048] Step 210: determining a target cluster center combination corresponding to the target object from among the candidate cluster center combinations based on the geometric attributes corresponding to the candidate cluster center combinations.
[0049] The target cluster center combination refers to a set of cluster centers corresponding to the positioning reference parts of the target object.
[0050] Exemplarily, the robot compares the geometric properties corresponding to the candidate cluster center combination with the standard geometric properties corresponding to the target object, and determines the successfully matched candidate cluster center combination as the target cluster center combination corresponding to the target object.
[0051] In some embodiments, the robot calculates the geometric shape type and geometric size corresponding to the candidate cluster center combination based on the geometric attributes corresponding to the candidate cluster center combination. The candidate cluster center combination with the same geometric shape type and geometric size as the target object is determined as the target cluster center combination corresponding to the target object. The geometric size refers to the size of the geometric shape. For example, when the geometric shape is a square, the size of the geometric shape can be the side length of the square; when the geometric shape is a rectangle, the geometric size can be the length and width of the rectangle; and so on.
[0052] Step 212: Determine the target position of the target object with respect to the robot based on the target cluster center combination.
[0053] Exemplarily, the robot calculates the target position of the target object for the robot according to the positions of each cluster center in the target cluster center combination and the structural features of the target object, wherein the structural features include the geometric shape type. The geometric shape type is a rectangle, a regular pentagon, a circle, etc. The operation task for the target object is performed according to the target position of the target object.
[0054] In some embodiments, the target cluster center combination can be input into the center position prediction model to obtain the target position of the target object for the robot. The center position prediction model is obtained by supervised training based on the cluster center combination corresponding to multiple object samples and the known target position of each object sample for the robot.
[0055] In some embodiments, the positioning reference part corresponding to the target object may hinder the robot's travel. For example, when the target object is a shelf and the positioning reference part is the shelf leg supporting the shelf, the shelf leg is an obstacle that hinders the robot's travel. Therefore, when the positioning reference part corresponding to the target object hinders the robot's travel, the positions of all the positioning reference parts corresponding to the target object are calculated based on the target cluster center combination corresponding to the target object. Based on the positions of all the positioning reference parts corresponding to the target object and the target position, obstacle avoidance travel is performed for the target object, thereby executing the operation task for the target object.
[0056] In the above object position detection method, by obtaining the prior position corresponding to the target object, obtaining the target point cloud data for the prior position collected by the robot, clustering the target point cloud data, and obtaining multiple candidate cluster centers corresponding to the target object. Then, each candidate cluster center is grouped to obtain multiple candidate cluster center combinations, and the geometric properties between each candidate cluster center contained in each candidate cluster center combination are calculated. Based on the geometric properties corresponding to each candidate cluster center combination, the target cluster center combination corresponding to the target object is determined in each candidate cluster center combination. Finally, the target position of the target object for the robot is determined based on the target cluster center combination. In this way, each candidate cluster center is grouped to obtain multiple candidate cluster center combinations to be identified, and it is determined whether each candidate cluster center combination is the target cluster center combination corresponding to the target object. Compared with directly finding the cluster center corresponding to the target object from all candidate cluster center combinations, it can greatly reduce the search range and significantly reduce the amount of calculation, thereby improving the efficiency of position detection. Furthermore, since the geometric attributes corresponding to the candidate cluster center combination can clearly and intuitively reflect the characteristics of the geometric shapes formed by each candidate cluster center in the combination, the target cluster center combination corresponding to the target object can be quickly and efficiently determined from each candidate cluster center combination based on the geometric attributes, which can reduce the complexity of position detection, improve the efficiency of position detection, and improve the accuracy of position detection.
[0057] In an exemplary embodiment, as shown in FIG3 , obtaining target point cloud data for a priori positions collected by a robot includes steps 302 to 306 . Among them:
[0058] Step 302: Acquire initial point cloud data for a priori positions collected by the robot.
[0059] Step 304 , determining the relative position of the target object with respect to the robot based on the robot posture and the prior position corresponding to the robot.
[0060] Step 306 , based on the relative position, the initial point cloud data is filtered to obtain target point cloud data for the priori position.
[0061] The initial point cloud data refers to the initial environmental point cloud data collected by the robot through sensors. The robot posture refers to the specific position and orientation of the robot in space when collecting the initial point cloud data.
[0062] Exemplarily, the robot acquires initial point cloud data of a priori position of a target object collected by a sensor. Based on the robot posture corresponding to the robot when the initial point cloud data is collected, the priori position corresponding to the target object is converted into a relative position for the robot. Then, based on the relative position and the size information corresponding to the target object, the point cloud filtering range corresponding to the target object is determined. In the actual implementation process, the point cloud filtering range can be a range determined based on the relative position and slightly larger than the size of the target object. Data points in the initial point cloud data that are outside the point cloud filtering range are filtered to obtain target point cloud data for the priori position.
[0063] In the above embodiment, according to the robot posture and the prior position of the target object, a part of the data points in the initial point cloud data are filtered, and only the data points around the prior position are retained. This can effectively reduce the interference caused by irrelevant point cloud data and improve the accuracy of position detection. At the same time, it can greatly reduce the amount of data calculation, save computer resources, and improve computing efficiency.
[0064] In an exemplary embodiment, clustering the target point cloud data to obtain multiple candidate cluster centers corresponding to the target object includes:
[0065] Determining a reference cluster size based on the size of the positioning reference part of the target object;
[0066] The target point cloud data is clustered to obtain multiple initial point cloud clusters; each initial point cloud cluster is filtered based on a reference cluster size to obtain multiple candidate point cloud clusters; the cluster centers corresponding to the candidate point cloud clusters are used as candidate cluster centers corresponding to the target objects.
[0067] The positioning reference part refers to a key part of the target object used to assist in positioning the target object.
[0068] Exemplarily, the robot obtains the size of the positioning reference part of the target object, enlarges the size of the positioning reference part according to a certain ratio, and obtains a reference cluster size. For example, the size of the positioning reference part can be enlarged by 1.5 times to obtain a reference cluster size. The target point cloud data is clustered to obtain multiple initial point cloud clusters. In each initial point cloud data, the initial point cloud clusters with too large size are filtered according to the reference cluster size to obtain multiple candidate point cloud clusters. Specifically, the initial point cloud clusters whose minimum circumscribed rectangle length is greater than the reference cluster size can be filtered. Then, the cluster center corresponding to each candidate point cloud cluster is used as the candidate cluster center corresponding to the target object.
[0069] In the above embodiment, filtering irrelevant point cloud data according to the size of the positioning reference part can reduce the interference caused by environmental factors and improve the accuracy of position detection. At the same time, it can greatly reduce the amount of data calculation, save computer resources, and improve the efficiency of object position detection.
[0070] In an exemplary embodiment, calculating the geometric properties between the candidate cluster centers included in the candidate cluster center combination includes:
[0071] For any candidate cluster center combination, the length of the connection between each candidate cluster center in the candidate cluster center combination is calculated, and the connection angle between each connection line is calculated to obtain the geometric properties between each candidate cluster center included in the candidate cluster center combination.
[0072] Exemplarily, for any candidate cluster center combination, the robot calculates the length of the lines between each candidate cluster center in the candidate cluster center combination, and calculates the angle between each line, and uses the length of the lines between each candidate cluster center and the angle between each line as the geometric properties between each candidate cluster center. For example, when the candidate cluster center combination includes three candidate cluster centers, the geometric properties between each candidate cluster center can be as shown in FIG4, the candidate cluster center combination includes three candidate cluster centers A, B and C, line segment AC, line segment AB and line segment BC are the lines between each candidate cluster center, and the three angles of triangle ABC are the angles between each line.
[0073] In the above embodiment, for each candidate cluster center combination, the line lengths of the lines connecting each candidate cluster center and the angles between each line are used as geometric attributes between each candidate cluster center. The line lengths and angles reflect the key set characteristics of the candidate cluster center combination from different angles, and provide comprehensive and powerful data support for the subsequent determination of the target cluster center combination corresponding to the target object based on geometric attributes. Using the line lengths and angles as geometric attributes can reduce the complexity of position detection, improve the efficiency of position detection, and ensure the accuracy of position detection.
[0074] In an exemplary embodiment, the object position detection method further includes:
[0075] When the number of candidate cluster centers is greater than or equal to the quantity threshold, the step of grouping the candidate cluster centers to obtain a combination of multiple candidate cluster centers is executed; when the number of candidate cluster centers is less than the quantity threshold, the supplementary point cloud data collected by the robot for the prior position is obtained, the target point cloud data and the supplementary point cloud data are fused to obtain updated point cloud data, the updated point cloud data is used as the target point cloud data, and the target point cloud data is returned to be clustered to obtain a step of multiple candidate cluster centers corresponding to the target object.
[0076] The quantity threshold refers to the minimum number of cluster centers required to locate the target object.
[0077] Exemplarily, when the number of candidate cluster centers obtained by clustering the target point cloud data is greater than or equal to the quantity threshold, the robot enters the step of grouping the candidate cluster centers to obtain a combination of multiple candidate cluster centers. When the number of candidate cluster centers is less than the quantity threshold, the supplementary point cloud data collected by the robot for the priori position is obtained. In the actual implementation process, the robot can be equipped with multiple sensors, each of which is used to collect environmental point cloud data in different directions. When the number of candidate cluster centers is less than the quantity threshold, the environmental point cloud data collected by the sensor in another direction can be obtained as supplementary point cloud data. Then, the target point cloud data and the supplementary point cloud data are spliced to obtain updated point cloud data. The updated point cloud data is used as the target point cloud data, and the step of clustering the target point cloud data to obtain multiple candidate cluster centers corresponding to the target object is returned.
[0078] In some embodiments, for a robot with only one sensor, the robot can collect supplementary point cloud data by rotating or traveling to another position. Specifically, the robot can obtain the geometric dimensions corresponding to the target object, and determine the shortest adjustment path of the robot based on the geometric dimensions corresponding to the target object. In this way, the shortest adjustment path is determined according to the geometric dimensions of the target object, and the path length of the robot is shortened, thereby reducing the cost of position detection and improving the efficiency of position detection. After controlling the robot to move to another position along the shortest adjustment path, the environmental point cloud data for the prior position is collected again. The robot can also rotate a preset angle to collect environmental point cloud data for the prior position again. Then, the environmental point cloud data collected again is projected to the coordinate system corresponding to the target point cloud data, that is, the environmental point cloud data is transformed to the coordinate system before the movement to obtain supplementary point cloud data. Updated point cloud data is formed based on the supplementary point cloud data and the target point cloud data.
[0079] In the above embodiment, when the number of candidate cluster centers obtained by clustering the target point cloud data is less than the quantity threshold, it indicates that the target point cloud data contains too little information about the target object, and it is difficult to determine the position of the target object based on the target point cloud data. Therefore, the environmental point cloud data collected by sensors in other directions are obtained as supplementary point cloud data. The environmental point cloud data collected by sensors in different directions are integrated to obtain updated point cloud data containing more comprehensive environmental information, and then the object position detection is performed again based on the updated point cloud data, which can effectively improve the efficiency and success rate of position detection.
[0080] In an exemplary embodiment, the geometric attributes include a line length and a line angle; based on the geometric attributes corresponding to each candidate cluster center combination, determining a target cluster center combination corresponding to the target object from each candidate cluster center combination includes:
[0081] Obtain a reference length set and a reference angle corresponding to the target object; take a candidate cluster center combination whose line length successfully matches the reference length set and whose line angle successfully matches the reference angle as an intermediate cluster center combination; and determine a target cluster center combination corresponding to the target object from among the intermediate cluster center combinations.
[0082] The reference length set refers to a set of known standard side lengths corresponding to the target object. The reference angle refers to a known standard angle corresponding to the target object. For example, when the target object is a rectangular shelf, and the four corners of the rectangular shelf are supported by four shelf legs, the reference length set includes the length and width of the shelf determined based on the distance between the shelf legs, and the reference angle is 90°.
[0083] Exemplarily, the robot obtains a reference length set and a reference angle corresponding to the target object. The length of each line in each candidate cluster center combination is compared with each standard side length in the reference length set. From the candidate cluster center combinations whose line lengths successfully match the reference length set and whose line angles between successfully matched lines also successfully match the reference angle, the target cluster center combination corresponding to the target object is determined. Specifically, when the lengths of each line in the candidate cluster center combination can find matching side lengths in the reference length set, the angles between successfully matched lines are compared with the reference angles between the corresponding side lengths of the successfully matched lines in the reference length set. When the angles between the lines also successfully match the reference angles, the candidate cluster center combination is determined as the intermediate cluster center combination corresponding to the target object. Wherein, successful matching means that the difference between the two is less than a preset threshold. When there is only one intermediate cluster center combination, the intermediate cluster center combination is determined as the target cluster center combination. When multiple intermediate cluster center combinations appear, the difference between the length of the connection in each intermediate cluster center combination and the corresponding reference length, as well as the difference between the angle between the successfully matched connection lines and the corresponding reference angle, are fused to obtain the comprehensive difference values corresponding to each intermediate cluster center combination. The intermediate cluster center combination with the smallest comprehensive difference value is used as the target cluster center combination corresponding to the target object.
[0084] In the above embodiment, by comparing the lengths of the lines between the candidate cluster centers and the angles between the lines, the target candidate cluster center combination corresponding to the target object can be quickly and accurately determined, which reduces the computational complexity of position detection and improves the efficiency and accuracy of position detection.
[0085] In an exemplary embodiment, determining a target position of a target object for a robot based on a combination of target cluster centers includes:
[0086] Based on the geometric shape type corresponding to the target object and each target cluster center in the target cluster center combination, the center position corresponding to the target object is calculated; based on the robot posture and center position corresponding to the robot, the target position of the target object with respect to the robot is determined.
[0087] Exemplarily, the robot obtains the geometric shape type corresponding to the target object. The geometric shape type refers to the shape formed by each positioning reference part corresponding to the target object, for example, it can be a rectangle, trapezoid, hexagon, etc. The center position refers to the geometric center of the shape formed by each positioning reference part corresponding to the target object. Based on the target cluster center combination included in each target cluster center and the geometric shape type corresponding to the target object, the center position corresponding to the target object is calculated. For example, as shown in FIG4 , the target cluster center combination includes three candidate cluster centers A, B, and C. When the geometric shape type corresponding to the target object is a rectangle, the midpoint D of the hypotenuse AC is used as the center position of the target object. Then, the robot posture corresponding to the current robot and the robot posture corresponding to the history when the robot collects the target point cloud data are obtained. Based on the posture change between the current corresponding robot posture and the historical robot posture, the center position corresponding to the target object is adjusted to obtain the target position of the target object for the current robot posture of the robot. Based on the target position corresponding to the target object, the robot performs the operation task for the target object.
[0088] In the actual implementation process, the robot obtains the odometer data from the time when the target point cloud data is collected to the current time, and determines the robot posture corresponding to the current time based on the historical robot posture and odometer data when the robot collects the target point cloud data. The odometer data can be collected by a wheel odometer (ODOM, Odometry), which is the data that estimates the position and direction changes of the robot relative to the initial position during the movement by measuring the moving parts of the robot (such as wheels).
[0089] In the above embodiment, according to the geometric shape type corresponding to the target object and the target cluster center combination, the center position corresponding to the target object can be quickly and accurately determined, thereby improving the efficiency of position detection. Then, the center position corresponding to the target object is adjusted according to the change of the robot's posture, and the target position of the target object for the robot's posture at the current moment is obtained, thereby eliminating the error of the target object's position relative to the robot caused by the robot's movement, thereby improving the precision and accuracy of position detection.
[0090] In a specific embodiment, the object position detection method proposed in the present application can be applied to a handling robot in an intelligent warehousing system. The handling robot calculates the shelf position by accurately identifying the position of the shelf legs of the shelf to be lifted, and performs the task of lifting and handling the shelf based on the shelf position. As shown in FIG5 , the object position detection method includes the following steps:
[0091] 1. Data Collection
[0092] The handling robot obtains the environmental point cloud data for the shelf to be lifted as the initial point cloud data. It obtains the odometer data and the prior position of the shelf to be lifted. It obtains the initial robot posture corresponding to the robot when collecting the environmental point cloud data.
[0093] 2. Filter data points
[0094] Based on the initial robot posture and the prior position of the shelf to be lifted, the handling robot filters a portion of the data points in the initial point cloud data to obtain the target point cloud data. The target point cloud data only contains the data points around the prior position of the shelf to be lifted.
[0095] 3. Filter point cloud clusters
[0096] The handling robot clusters the target point cloud data according to the distance between each data point in the target point cloud data to obtain multiple initial point cloud clusters. The size threshold is determined according to the size of the shelf legs of the shelf to be lifted, and the initial point cloud clusters whose minimum circumscribed rectangle length is greater than the size threshold are filtered to obtain multiple candidate point cloud clusters, and the candidate cluster centers corresponding to each candidate point cloud cluster are calculated.
[0097] 4. Detect shelf position
[0098] When the number of candidate cluster centers obtained by screening is less than or equal to two, the environmental point cloud data collected by another sensor is obtained as the supplementary point cloud data. The target point cloud data and the supplementary point cloud data are fused to obtain the updated point cloud data, and the updated point cloud data is used as the target point cloud data, and the step of filtering the point cloud is returned to execute.
[0099] When the number of candidate cluster centers screened is greater than two, the candidate cluster centers are combined to obtain multiple candidate cluster center combinations. Each candidate cluster center combination contains three different candidate cluster centers. For any candidate cluster center combination, calculate the length of the line between each candidate cluster center and the angle between the lines. Among each candidate cluster center combination, select the candidate cluster center combination with the line length closest to the shelf side length of the shelf to be lifted and the angle closest to 90° as the target cluster center combination. Calculate the center position of the shelf to be lifted based on each cluster center in the target cluster center combination. Then, according to the odometer data and the initial robot posture, calculate the current robot posture corresponding to the handling robot at the current moment. Based on the posture change between the current robot posture and the initial robot posture, adjust the center position corresponding to the shelf to be lifted to obtain the target position of the shelf to be lifted for the current robot posture.
[0100] In the actual implementation process, since sensor data is often interfered by various noises, the noise covariance matrix can be introduced through the extended Kalman filter (EKF, Extended Kalman Filter), the target position is filtered, and the filtered target position is output, which can suppress the interference of noise on the object position detection to a certain extent. Based on the target position corresponding to the shelf to be lifted, the handling robot drives to the center position of the shelf to be lifted and lifts and transports the shelf to be lifted. Specifically, the handling robot can calculate the positions of all shelf legs corresponding to the shelf to be lifted based on the target cluster center combination, and avoid obstacles based on the positions of each shelf leg and the target position. After driving to the center position under the shelf to be lifted, it lifts and transports the shelf to be lifted.
[0101] In the above embodiment, a portion of the point cloud data is filtered according to the prior position of the shelf, and position detection is performed based on the filtered point cloud data, which can save computing resources and improve computing efficiency. Clustering the filtered target point cloud data to obtain multiple initial point cloud clusters, and filtering larger initial point cloud clusters according to the size of the shelf legs can further save computing resources, improve computing efficiency, and improve the accuracy of position detection. The candidate cluster centers corresponding to the filtered candidate point cloud clusters are combined, and the target cluster center combination corresponding to the shelf to be lifted is determined based on the length of the connection line between each candidate cluster center in each candidate cluster center combination and the angle between the connection lines. The center position of the shelf to be lifted is determined based on the target cluster center combination, which can reduce the complexity of position detection, improve the efficiency of position detection, and improve the accuracy of position detection.
[0102] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.
[0103] Based on the same inventive concept, the embodiment of the present application also provides an object position detection device for implementing the object position detection method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more object position detection device embodiments provided below can refer to the limitations of the object position detection method above, and will not be repeated here.
[0104] In an exemplary embodiment, as shown in FIG6 , an object position detection device is provided, comprising: a data acquisition module 602, a point cloud clustering module 604, a center combination module 606, an attribute calculation module 608, a combination determination module 610 and a position determination module 612, wherein:
[0105] The data acquisition module 602 is used to acquire the priori position corresponding to the target object and acquire the target point cloud data corresponding to the priori position collected by the robot.
[0106] The point cloud clustering module 604 is used to cluster the target point cloud data to obtain multiple candidate cluster centers corresponding to the target object.
[0107] The center combination module 606 is used to group the candidate cluster centers to obtain multiple candidate cluster center combinations.
[0108] The attribute calculation module 608 is used to calculate the geometric attributes between the candidate cluster centers included in the candidate cluster center combination.
[0109] The combination determination module 610 is used to determine a target cluster center combination corresponding to the target object from among the candidate cluster center combinations based on the geometric attributes corresponding to the candidate cluster center combinations.
[0110] The position determination module 612 is used to determine the target position of the target object with respect to the robot based on the target cluster center combination.
[0111] In one embodiment, the data acquisition module 602 is further used to:
[0112] Acquire initial point cloud data for a priori position collected by the robot; determine the relative position of the target object to the robot based on the robot posture and priori position corresponding to the robot; filter the initial point cloud data based on the relative position to obtain target point cloud data for the priori position.
[0113] In one embodiment, the point cloud clustering module 604 is further configured to:
[0114] Based on the size of the positioning reference part of the target object, the reference cluster size is determined; the target point cloud data is clustered to obtain multiple initial point cloud clusters; each initial point cloud cluster is filtered based on the reference cluster size to obtain multiple candidate point cloud clusters; the cluster center corresponding to the candidate point cloud cluster is used as the candidate cluster center corresponding to the target object.
[0115] In one embodiment, the attribute calculation module 608 is further used to:
[0116] For any candidate cluster center combination, the length of the connection between each candidate cluster center in the candidate cluster center combination is calculated, and the connection angle between each connection line is calculated to obtain the geometric properties between each candidate cluster center included in the candidate cluster center combination.
[0117] In one embodiment, the data acquisition module 602 is further used to:
[0118] When the number of candidate cluster centers is greater than or equal to the quantity threshold, the step of grouping the candidate cluster centers to obtain a combination of multiple candidate cluster centers is executed; when the number of candidate cluster centers is less than the quantity threshold, the supplementary point cloud data collected by the robot for the prior position is obtained, the target point cloud data and the supplementary point cloud data are fused to obtain updated point cloud data, the updated point cloud data is used as the target point cloud data, and the target point cloud data is returned to be clustered to obtain a step of multiple candidate cluster centers corresponding to the target object.
[0119] In one embodiment, the geometric attributes include line length and line angle; based on the geometric attributes corresponding to each candidate cluster center combination, the attribute calculation module 608 is further used to:
[0120] Obtain a reference length set and a reference angle corresponding to the target object; take a candidate cluster center combination whose line length successfully matches the reference length set and whose line angle successfully matches the reference angle as an intermediate cluster center combination; and determine a target cluster center combination corresponding to the target object from among the intermediate cluster center combinations.
[0121] In one embodiment, the location determination module 612 is further configured to:
[0122] Based on the geometric shape type corresponding to the target object and each target cluster center in the target cluster center combination, the center position corresponding to the target object is calculated; based on the robot posture and center position corresponding to the robot, the target position of the target object with respect to the robot is determined.
[0123] Each module in the above-mentioned object position detection device can be implemented in whole or in part by software, hardware or a combination thereof. Each module can be embedded in or independent of the processor in the robot in the form of hardware, or can be stored in the memory in the robot in the form of software, so that the processor can call and execute the corresponding operations of each module.
[0124] In an exemplary embodiment, a robot is provided, and its internal structure diagram can be shown in FIG7. The robot includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. The processor, the memory, and the input / output interface are connected via a system bus, and the communication interface, the display unit, and the input device are connected to the system bus via the input / output interface. The processor of the robot is used to provide computing and control capabilities. The memory of the robot includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the robot is used to exchange information between the processor and an external device. The communication interface of the robot is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, an object position detection method is implemented. The display unit of the robot is used to form a visually visible picture, which can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the robot can be a touch layer covering the display screen, or a button, trackball or touchpad set on the robot shell, or an external keyboard, touchpad or mouse.
[0125] Those skilled in the art will understand that the structure shown in FIG. 7 is merely a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the robot to which the scheme of the present application is applied. The specific robot may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.
[0126] In one embodiment, a robot is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.
[0127] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.
[0128] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.
[0129] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.
[0130] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.
[0131] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.
Claims
1. A method for detecting an object position, characterized in that: The method comprises: Acquire a priori position corresponding to the target object, and acquire target point cloud data collected by the robot for the priori position; Clustering the target point cloud data to obtain a plurality of candidate cluster centers corresponding to the target object; Grouping the candidate cluster centers to obtain multiple candidate cluster center combinations; Calculating geometric properties between the candidate cluster centers included in the candidate cluster center combination; Based on the geometric attributes respectively corresponding to the candidate cluster center combinations, determining a target cluster center combination corresponding to the target object from among the candidate cluster center combinations; The target position of the target object with respect to the robot is determined based on the target cluster center combination.
2. The method according to claim 1, characterized in that The step of obtaining target point cloud data collected by the robot for the priori position includes: Acquire initial point cloud data for the priori position collected by the robot; Determining a relative position of the target object with respect to the robot based on a robot pose corresponding to the robot and the prior position; Based on the relative position, the initial point cloud data is filtered to obtain target point cloud data for the priori position collected by the robot.
3. The method according to claim 1, characterized in that The step of clustering the target point cloud data to obtain a plurality of candidate cluster centers corresponding to the target object includes: Determining a reference cluster size based on a size of a positioning reference part of the target object; Clustering the target point cloud data to obtain a plurality of initial point cloud clusters; Filtering each of the initial point cloud clusters based on the reference cluster size to obtain a plurality of candidate point cloud clusters; The cluster center corresponding to the candidate point cloud cluster is used as the candidate cluster center corresponding to the target object.
4. The method according to claim 1, characterized in that: The calculating of geometric properties between the candidate cluster centers included in the candidate cluster center combination includes: For any one of the candidate cluster center combinations, the lengths of the lines between each of the candidate cluster centers in the candidate cluster center combination are calculated, and the line angles between the lines are calculated to obtain the geometric properties between the candidate cluster centers included in the candidate cluster center combination.
5. The method according to claim 1, characterized in that The method further comprises: When the number of the candidate cluster centers is greater than or equal to the number threshold, the step of grouping the candidate cluster centers to obtain a plurality of candidate cluster center combinations is executed; When the number of the candidate cluster centers is less than the number threshold, the supplementary point cloud data collected by the robot for the prior position is obtained, the target point cloud data and the supplementary point cloud data are fused to obtain updated point cloud data, the updated point cloud data is used as the target point cloud data, and the step of clustering the target point cloud data to obtain multiple candidate cluster centers corresponding to the target object is returned.
6. The method according to claim 1, characterized in that The geometric attributes include a line length and a line angle; and determining a target cluster center combination corresponding to the target object from among the candidate cluster center combinations based on the geometric attributes corresponding to the candidate cluster center combinations, includes: Obtaining a reference length set and a reference angle corresponding to the target object; The candidate cluster center combination whose line length successfully matches the reference length set and whose line angle successfully matches the reference angle is used as the intermediate cluster center combination; A target cluster center combination corresponding to the target object is determined from each of the intermediate cluster center combinations.
7. The method according to claim 1, characterized in that The determining the target position of the target object with respect to the robot based on the target cluster center combination includes: Calculating the center position corresponding to the target object based on the geometric shape type corresponding to the target object and each target cluster center in the target cluster center combination; Based on the robot posture corresponding to the robot and the center position, a target position of the target object with respect to the robot is determined.
8. An object position detection device, characterized in that: The device comprises: A data acquisition module is used to acquire a priori position corresponding to the target object and acquire target point cloud data collected by the robot for the priori position; A point cloud clustering module, used to cluster the target point cloud data to obtain a plurality of candidate cluster centers corresponding to the target object; A center combination module, used for grouping each of the candidate cluster centers to obtain multiple candidate cluster center combinations; An attribute calculation module, used for calculating the geometric attributes between the candidate cluster centers included in the candidate cluster center combination; A combination determination module, configured to determine a target cluster center combination corresponding to the target object from among the candidate cluster center combinations based on the geometric attributes corresponding to the candidate cluster center combinations; A position determination module is used to determine the target position of the target object with respect to the robot based on the target cluster center combination.
9. A robot comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.
Citation Information
Cited By
Method, device and equipment for recognizing trailer posture by robot, storage medium and program product
CN120147679A
Method, device, equipment, storage medium and program product for robot to identify trailer posture
CN120147679B
Air conditioner, control method of air conditioner and computer equipment
CN121474668A
Object position detection method and apparatus, robot, readable storage medium, and program product
WO2026032234A1
