Target pose calculation method and device for oil-electric hybrid chassis mechanical arm composite robot on non-paved road surface and storage medium

Abstract

The invention discloses a target pose calculation method and device for an oil-electric hybrid chassis mechanical arm composite robot on a non-paved road surface and a storage medium, and belongs to the technical field of machine vision systems.The method comprises the following steps that hand-eye calibration is conducted on a shot image to obtain a conversion matrix of a camera at the tail end of a mechanical arm; obtaining a target object template point cloud; obtaining a vibration point cloud of a target object; fPFH features are calculated respectively, two sets of feature vectors are obtained, and the two sets of feature vectors are registered to obtain a coarse registration conversion matrix; calculating to obtain a precise registration conversion matrix; performing rigid body transformation on the vibration point cloud of the target object by using the fine registration transformation matrix, calculating the number of matching points, and then calculating to obtain an adaptive threshold value; and iteration is conducted on the rigid body transformation matrix of the target object template point cloud, transformation poses are transformed through the transformation matrix, and the poses of the target under the mechanical arm base are obtained. The method is small in hardware size, low in cost and low in computing power requirement, and the target object pose estimation speed and precision in the shaking state can be improved.

Description

Technical Field

[0001] The present invention relates to a target posture calculation method, device and storage medium for a hybrid chassis manipulator composite robot for unpaved roads, belonging to the technical field of machine vision systems. Background Art

[0002] Currently, on unpaved roads, when a hybrid chassis is idling and generating power, the engine typically maintains an inefficient speed range. This leads to uneven combustion pressure within the cylinders, which can cause second-order vibrations. Furthermore, there is a lag in the dynamic matching of engine speed regulation with power generation demand, which can cause transient shocks. These vibrations are transmitted to the chassis, causing jitter in the sensor mounted at the end of the robotic arm. Camera stabilization technology currently uses either mechanical or electronic image stabilization. Mechanical camera stabilization primarily compensates for offset by physically driving the image sensor. This hardware is expensive and bulky, and its effectiveness at compensating for high-frequency jitter is limited. Electronic image stabilization primarily uses multi-sensor fusion for motion compensation, which requires high computing power. Neither of these two methods can guarantee the speed and accuracy of object pose estimation under jitter conditions. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to overcome the defects of the existing technology and provide a method, device and storage medium for calculating the target pose of a hybrid chassis manipulator arm composite robot on non-paved roads. The hardware has a small size, low cost, and low computing power requirements, and can improve the speed and accuracy of target object pose estimation in a shaking state.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0005] In a first aspect, the present invention provides a method for calculating the target pose of a hybrid chassis manipulator-arm composite robot for unpaved roads, comprising the following steps:

[0006] Control the end of the robotic arm to move to the specified position, use the structured light camera on the end of the robotic arm to capture the calibration plate to obtain a captured image, and perform hand-eye calibration on the captured image to obtain the camera's transformation matrix under the robotic arm;

[0007] Scan the target object using structured light in a stationary state to obtain the initial point cloud data of the target object;

[0008] The noise points in the initial point cloud data are removed through the model filtering algorithm, and then the target object template point cloud is obtained through voxel grid downsampling and point cloud decentralization;

[0009] When the structured light device is in vibration, the point cloud of the target object is collected, and the vibration point cloud of the target object is obtained by statistical outlier removal and voxel grid downsampling;

[0010] The FPFH features are calculated for the acquired target template point cloud and the target vibration point cloud respectively, and two sets of feature vectors are obtained. The SAC-IA algorithm is used to align the two sets of feature vectors to obtain a coarse registration transformation matrix.

[0011] Use the coarse registration transformation matrix as the initial transformation matrix, and use the ICP algorithm to calculate the target template point cloud and the target vibration point cloud to obtain the fine registration transformation matrix;

[0012] The target object vibration point cloud is transformed into a rigid body using a precise registration transformation matrix, and the number of matching points between the target object template point cloud and the target object vibration point cloud is calculated. The adaptive threshold is then calculated based on the number of matching points.

[0013] Based on the adaptive threshold, the rigid body transformation matrix of the target template point cloud is iterated to obtain the matrix transformation results under each posture. The matrix transformation result with the largest number of matching point pairs is used as the transformed pose of the target and the template. The transformed pose is then transformed through the conversion matrix to obtain the pose of the target under the base of the robotic arm.

[0014] The hand-eye calibration of the captured image to obtain the conversion matrix of the camera at the end of the robotic arm includes:

[0015] Get the transformation matrix of the calibration plate under the camera in each image ,in 、 and are the rotation matrices of the calibration plate under the camera in the 1st, 2nd and nth images respectively, 、 and are the displacement vectors of the calibration plate under the camera in the first, second and nth images respectively;

[0016] Get the transformation matrix of the corresponding robotic arm end under the base ,in 、 and are the rotation matrices of the end of the robotic arm under the base in the first, second and mth images respectively, 、 and are the displacement vectors of the end of the robotic arm under the base in the first, second, and mth images respectively;

[0017] The transformation matrix of the camera under the end of the robotic arm is obtained by calibrating the transformation matrix of the calibration plate under the camera and the transformation matrix of the end of the robotic arm under the base. ,in is the rotation matrix of the camera under the end of the robotic arm, is the displacement vector of the camera under the end of the robotic arm.

[0018] The value range of m is 15~25.

[0019] The FPFH features are calculated for the obtained target template point cloud and target vibration point cloud respectively, and the obtained feature vectors include:

[0020] The 3D geometric relationships between points in the target object template point cloud and between points in the target object vibration point cloud are converted into 33*1-dimensional vectors by calculating FPFH features.

[0021] The adaptive threshold value obtained by calculating the number of matching points includes:

[0022] ;

[0023] in is the adaptive threshold, is the first ICP matching point, is the attenuation rate, is the target point cloud density.

[0024] The iterating of the rigid body transformation matrix of the target object template point cloud includes:

[0025] Perform rigid body transformation matrix on the target object template point cloud Transformation, where is the identity matrix, and are the rotation matrix and displacement vector in the unit matrix, is the rigid body transformation matrix The linear increase or decrease of the translation vector in the adaptive threshold On the basis of Perform linear increase and decrease to achieve iteration.

[0026] The matrix transformation result with the largest number of matching point pairs as the transformation pose of the target object and the template includes:

[0027] Through iteration Get the corresponding matching points ,in 、 、 and are the matching points corresponding to the 1st, 2nd, 3rd and tth iterations respectively, and the matrix corresponding to the maximum matching point number is recorded as , that is, the transformed pose of the target object and the template, where and are the rotation matrix and displacement vector in the matrix corresponding to the maximum number of matching points.

[0028] The transformed pose is then transformed by a transformation matrix to obtain the pose of the target under the base of the robotic arm, which includes:

[0029] The position of the target under the base of the robotic arm obtained when the camera is shaken for:

[0030] = ;

[0031] in, and They are the rotation matrix and displacement vector of the target object under the base of the robotic arm, is the position of the end of the robotic arm under the base, and They are the rotation matrix and displacement vector of the position of the end of the robotic arm under the base, is the fine registration transformation matrix, and are the rotation matrix and displacement vector in the precise registration transformation matrix respectively.

[0032] In a second aspect, the present invention provides a target pose calculation device for a hybrid chassis manipulator arm composite robot for unpaved roads, comprising:

[0033] The hand-eye calibration module is used to control the movement of the end of the robotic arm to the specified position. The structured light camera on the end of the robotic arm is used to capture the calibration plate to obtain an image. The hand-eye calibration is performed on the captured image to obtain the transformation matrix of the camera under the end of the robotic arm.

[0034] An initial point cloud data acquisition module is used to scan a target object using structured light in a stationary state to obtain initial point cloud data of the target object;

[0035] The target object template point cloud acquisition module is used to remove noise points in the initial point cloud data through a model filtering algorithm, and then obtain the target object template point cloud through voxel grid downsampling and point cloud decentralization;

[0036] The target object vibration point cloud acquisition module is used to collect the point cloud of the target object when the structured light device is in vibration, and obtain the target object vibration point cloud by statistical outlier removal and voxel grid downsampling;

[0037] The coarse registration conversion matrix acquisition module calculates the FPFH features for the acquired target template point cloud and the target vibration point cloud, obtains two sets of feature vectors, and uses the SAC-IA algorithm to align the two sets of feature vectors to obtain the coarse registration conversion matrix;

[0038] The fine registration transformation matrix calculation module is used to use the coarse registration transformation matrix as the initial transformation matrix, and use the ICP algorithm to calculate the fine registration transformation matrix for the target template point cloud and the target vibration point cloud;

[0039] The adaptive threshold calculation module is used to perform rigid body transformation on the target object vibration point cloud using the precise registration transformation matrix, calculate the number of matching points between the target object template point cloud and the target object vibration point cloud, and then calculate the adaptive threshold based on the number of matching points;

[0040] The pose acquisition module is used to iterate the rigid body transformation matrix of the target template point cloud based on the adaptive threshold, obtain the matrix transformation results under each posture, and use the matrix transformation result with the largest number of matching point pairs as the transformed pose of the target and the template. The transformed pose is then transformed through the conversion matrix to obtain the pose of the target under the base of the robotic arm.

[0041] In a third aspect, the present invention provides a computer-readable storage medium having a computer program / instruction stored thereon. When the computer program / instruction is executed by a processor, the target pose calculation method of the hybrid chassis manipulator arm composite robot on a non-paved road is implemented.

[0042] Beneficial effects of the present invention: The present invention provides a method, device and storage medium for calculating the target posture of a hybrid chassis manipulator composite robot on non-paved roads. It only requires hand-eye calibration first, and then template point cloud production. After that, the posture of the target object can be directly judged under vibration. The process is simple and easy to operate, meeting the use requirements in actual engineering. This method solves the shortcomings of the huge size and complex control system of the mechanical anti-shake device, while avoiding the use of multi-sensor fusion and reducing the computing power requirements of the equipment. In summary, the present invention has the advantages of small hardware size, low cost, and low computing power requirements. At the same time, the target posture calculation method itself can improve the speed and accuracy of target posture estimation under vibration. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a flow chart of a method for calculating the target pose of a hybrid chassis manipulator-arm composite robot for unpaved roads according to the present invention;

[0044] Figure 2 This is a structural diagram of the hybrid chassis manipulator arm composite robot for non-paved roads of the present invention;

[0045] Figure 3 This is the template point cloud image of the present invention without vibration;

[0046] Figure 4 This is a point cloud image of the target object under vibration conditions of the present invention;

[0047] Figure 5The point cloud images of all target objects that meet the matching point threshold in the iterative process of the present invention;

[0048] Figure 6 The target point cloud set and pose graph that meet the threshold of the present invention;

[0049] The reference numerals in the figure are as follows: 1-hybrid chassis; 2-robotic arm base; 3-structured light camera; 4-robotic arm end flange; 5-robotic arm grasping actuator; 6-daily standard object; 7-unpaved road surface. DETAILED DESCRIPTION

[0050] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention.

[0051] Example 1

[0052] like Figure 1 As shown, the present invention discloses a method for calculating the target pose of a hybrid chassis manipulator composite robot on an unpaved road, comprising the following steps:

[0053] Step 1: Control the end of the robotic arm to move to the specified position, use the structured light camera on the end of the robotic arm to shoot the calibration plate to obtain a captured image, and perform hand-eye calibration on the captured image to obtain the transformation matrix of the camera under the end of the robotic arm.

[0054] Step 2: Scan the target object using structured light in a stationary state to obtain initial point cloud data of the target object.

[0055] Step three: Use the model filtering algorithm to remove noise points in the initial point cloud data, and then obtain the target object template point cloud through voxel grid downsampling and point cloud decentralization.

[0056] Step 4: When the structured light device is in vibration, the point cloud of the target object is collected, and the vibration point cloud of the target object is obtained by statistical outlier removal and voxel grid downsampling.

[0057] Step 5: Calculate the FPFH features for the acquired target template point cloud and target vibration point cloud respectively, obtain two sets of feature vectors, and use the SAC-IA algorithm to align the two sets of feature vectors to obtain a coarse registration transformation matrix.

[0058] Step 6: Use the coarse registration transformation matrix as the initial transformation matrix, and use the ICP algorithm to calculate the target template point cloud and the target vibration point cloud to obtain the fine registration transformation matrix.

[0059] Step 7: Use the precise registration transformation matrix to perform rigid body transformation on the target object vibration point cloud, calculate the number of matching points between the target object template point cloud and the target object vibration point cloud, and then calculate the adaptive threshold based on the number of matching points.

[0060] In step eight, the rigid body transformation matrix of the target template point cloud is iterated based on the adaptive threshold to obtain the matrix transformation results under each posture. The matrix transformation result with the largest number of matching point pairs is used as the transformed posture of the target and the template. The transformed posture is then transformed using the conversion matrix obtained in step one to obtain the posture of the target under the base of the robotic arm.

[0061] The present invention only requires hand-eye calibration first, and then template point cloud production, after which the target object's posture can be directly judged under vibration. The process is simple and easy to operate, meeting the needs of use in actual engineering. This method solves the shortcomings of the mechanical anti-shake device, which is large in size and complex in control system, while avoiding the use of multi-sensor fusion and reducing the computing power requirements of the equipment. In summary, the present invention has the advantages of small hardware size, low cost, and low computing power requirements. At the same time, the target posture calculation method itself can improve the speed and accuracy of target object posture estimation under vibration.

[0062] Example 2

[0063] like Figure 2 As shown in FIG, the present invention discloses a hybrid chassis and manipulator arm composite robot structure for non-paved roads, which is arranged on non-paved roads 7 and includes a hybrid chassis 1, a manipulator arm base 2, a structured light camera 3, a manipulator arm end flange 4, a manipulator arm grasping actuator 5 and a standard object 6. Figure 1 As shown, the present invention discloses a target pose calculation method for a hybrid chassis manipulator-arm composite robot for non-paved roads, based on the above-mentioned hybrid chassis manipulator-arm composite robot structure for non-paved roads, comprising the following steps:

[0064] Step 1: Use structured light to shoot about 20 images of the calibration plate fixed in place, and obtain the transformation matrix of the calibration plate under the camera in each image. , and the corresponding transformation matrix of the end of the robot under the base Calibrate to obtain the pose conversion relationship between the camera and the end of the robotic arm Among them 、 and are the rotation matrices of the calibration plate under the camera in the 1st, 2nd and 20th images respectively, 、 and are the displacement vectors of the calibration plate under the camera in the 1st, 2nd and 20th images respectively.

[0065] Step 2: Scan the target using structured light in a stationary state to obtain the initial point cloud data of the target .

[0066] Step 3: Use model filtering algorithm to filter Noise points are removed and then the point cloud is obtained by downsampling the voxel grid. , Decentralized acquisition of target object template point cloud files , the template point cloud of the target object without vibration is as follows Figure 3 As shown, (a) is the front view of the template point cloud, and (b) is the side view of the template point cloud.

[0067] Step 4: Collect the point cloud of the target object while the structured light device is vibrating, and obtain the target object vibration point cloud by statistical outlier removal and voxel grid downsampling , the point cloud of the target object under vibration is as follows Figure 4 As shown in the figure, (a) is the front view of the target point cloud under vibration, and (b) is the side view of the target point cloud.

[0068] Step 5: Use FPFH to calculate the features of the target template point cloud and the target vibration point cloud obtained in steps 2 and 3 to compress the 3D geometric relationship into a 33*1 dimensional vector , use the SAC-IA algorithm to compare the two sets of eigenvectors Find the nearest neighbor for registration to obtain the coarse registration transformation matrix ,in is the rotation matrix of the camera under the end of the robotic arm, is the displacement vector of the camera under the end of the robotic arm.

[0069] Step 6: Obtain the coarse registration transformation matrix from step 5 As a precondition, the ICP algorithm is used for precise registration to obtain the transformation matrix .

[0070] Step 7: Use the transformation matrix obtained in step 6 Transform the target object vibration point cloud and calculate the number of matching point pairs between the target object template point cloud and the target object vibration point cloud ,by Calculate the adaptive threshold based on ,in is the first ICP matching point, is the attenuation rate, is the target point cloud density:

[0071] .

[0072] Step 8: Perform the template point cloud Transformation, is the identity matrix, and when the threshold is met On the basis of Perform linear increase and decrease, is the rigid body transformation matrix The linear increase or decrease of the translation vector in Iterate, and during the iteration process, all target point clouds that meet the matching point threshold are as follows: Figure 5 As shown, the target point cloud set and pose that meet the threshold are as follows Figure 6 As shown, (a) is the front view of the target point cloud set that meets the threshold, (b) is the side view of the target point cloud set that meets the threshold, and (c) is the pose matrix obtained in the experiment. The transformation matrix with the largest number of matching point pairs is recorded as the final matrix , and then the position of the target object in the robot base coordinate system under vibration is obtained by transformation .

[0073] Specifically, the transformation matrix of the camera at the end of the robotic arm is: ,

[0074] Through iteration Corresponding matching points ,in 、 、 and are the matching points corresponding to the 1st, 2nd, 3rd and tth iterations respectively, and the matrix corresponding to the maximum matching point number is recorded as .

[0075] The transformation matrix of the target object vibration point cloud under the template point cloud is:

[0076] ;

[0077] The position of the end of the robotic arm under the base:

[0078] ;

[0079] The pose of the target object under the base of the robotic arm obtained when the camera is shaken is:

[0080] = .

[0081] Example 3

[0082] This embodiment provides a target pose calculation device for a hybrid chassis manipulator-arm composite robot for unpaved roads, comprising:

[0083] The hand-eye calibration module is used to control the movement of the end of the robotic arm to the specified position. The structured light camera on the end of the robotic arm is used to capture the calibration plate to obtain an image. The hand-eye calibration is performed on the captured image to obtain the transformation matrix of the camera under the end of the robotic arm.

[0084] An initial point cloud data acquisition module is used to scan a target object using structured light in a stationary state to obtain initial point cloud data of the target object;

[0085] The target object template point cloud acquisition module is used to remove noise points in the initial point cloud data through a model filtering algorithm, and then obtain the target object template point cloud through voxel grid downsampling and point cloud decentralization;

[0086] The target object vibration point cloud acquisition module is used to collect the point cloud of the target object when the structured light device is in vibration, and obtain the target object vibration point cloud by statistical outlier removal and voxel grid downsampling;

[0087] The coarse registration conversion matrix acquisition module calculates the FPFH features for the acquired target template point cloud and the target vibration point cloud, obtains two sets of feature vectors, and uses the SAC-IA algorithm to align the two sets of feature vectors to obtain the coarse registration conversion matrix;

[0088] The fine registration transformation matrix calculation module is used to use the coarse registration transformation matrix as the initial transformation matrix, and use the ICP algorithm to calculate the fine registration transformation matrix for the target template point cloud and the target vibration point cloud;

[0089] The adaptive threshold calculation module is used to perform rigid body transformation on the target object vibration point cloud using the precise registration transformation matrix, calculate the number of matching points between the target object template point cloud and the target object vibration point cloud, and then calculate the adaptive threshold based on the number of matching points;

[0090] The pose acquisition module is used to iterate the rigid body transformation matrix of the target template point cloud based on the adaptive threshold, obtain the matrix transformation results under each posture, and use the matrix transformation result with the largest number of matching point pairs as the transformed pose of the target and the template. The transformed pose is then transformed through the conversion matrix to obtain the pose of the target under the base of the robotic arm.

[0091] Example 4

[0092] This embodiment provides a computer-readable storage medium having a computer program / instruction stored thereon. When the computer program / instruction is executed by a processor, the target pose calculation method of the hybrid chassis manipulator arm composite robot on an unpaved road surface in Example 1 or Example 2 is implemented.

[0093] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0094] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0095] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0096] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for calculating the target pose of a hybrid robot with a chassis and manipulator arm for unpaved roads, characterized by: The following steps are involved: Control the end of the robotic arm to move to the specified position, use the structured light camera on the end of the robotic arm to capture the calibration plate to obtain a captured image, and perform hand-eye calibration on the captured image to obtain the camera's transformation matrix under the robotic arm; Scan the target object using structured light in a stationary state to obtain initial point cloud data of the target object; The noise points in the initial point cloud data are removed through the model filtering algorithm, and then the target object template point cloud is obtained through voxel grid downsampling and point cloud decentralization; When the structured light device is in vibration, the point cloud of the target object is collected, and the vibration point cloud of the target object is obtained by statistical outlier removal and voxel grid downsampling; The FPFH features are calculated for the acquired target template point cloud and the target vibration point cloud respectively, and two sets of feature vectors are obtained. The SAC-IA algorithm is used to align the two sets of feature vectors to obtain a coarse registration transformation matrix. Use the coarse registration transformation matrix as the initial transformation matrix, and use the ICP algorithm to calculate the target template point cloud and the target vibration point cloud to obtain the fine registration transformation matrix; The target object vibration point cloud is transformed into a rigid body using a precise registration transformation matrix, and the number of matching points between the target object template point cloud and the target object vibration point cloud is calculated. The adaptive threshold is then calculated based on the number of matching points. Based on the adaptive threshold, the rigid body transformation matrix of the target template point cloud is iterated to obtain the matrix transformation results under each posture. The matrix transformation result with the largest number of matching point pairs is used as the transformed pose of the target and the template. The transformed pose is then transformed through the conversion matrix to obtain the pose of the target under the base of the robotic arm.

2. The target pose calculation method for a hybrid chassis manipulator-arm composite robot for unpaved roads according to claim 1 is characterized by: The hand-eye calibration of the captured image to obtain the conversion matrix of the camera at the end of the robotic arm includes: Get the transformation matrix of the calibration plate under the camera in each image ,in 、 and are the rotation matrices of the calibration plate under the camera in the 1st, 2nd and nth images respectively, 、 and are the displacement vectors of the calibration plate under the camera in the first, second and nth images respectively; Get the transformation matrix of the corresponding robotic arm end under the base ,in 、 and are the rotation matrices of the end of the robotic arm under the base in the first, second and mth images respectively, 、 and are the displacement vectors of the end of the robotic arm under the base in the first, second, and mth images respectively; The transformation matrix of the camera under the end of the robotic arm is obtained by calibrating the transformation matrix of the calibration plate under the camera and the transformation matrix of the end of the robotic arm under the base. ,in is the rotation matrix of the camera under the end of the robotic arm, is the displacement vector of the camera under the end of the robotic arm.

3. The target pose calculation method for a hybrid chassis manipulator-arm composite robot for unpaved roads according to claim 2 is characterized by: The value range of m is 15~25.

4. The target pose calculation method for a hybrid chassis manipulator-arm composite robot for unpaved roads according to claim 1 is characterized by: The FPFH features are calculated for the obtained target template point cloud and target vibration point cloud respectively, and the obtained feature vectors include: The 3D geometric relationships between points in the target object template point cloud and between points in the target object vibration point cloud are converted into 33*1-dimensional vectors by calculating FPFH features.

5. The target pose calculation method for a hybrid chassis manipulator-arm composite robot for non-paved roads according to claim 1 is characterized by: The adaptive threshold value obtained by calculating the number of matching points includes: ； in is the adaptive threshold, is the first ICP matching point, is the attenuation rate, is the target point cloud density.

6. The target pose calculation method for a hybrid chassis manipulator-arm composite robot for unpaved roads according to claim 5 is characterized by: The iterating of the rigid body transformation matrix of the target object template point cloud includes: Perform rigid body transformation matrix on the target object template point cloud Transformation, where is the identity matrix, and are the rotation matrix and displacement vector in the unit matrix, is the rigid body transformation matrix The linear increase or decrease of the translation vector in the adaptive threshold On the basis of Perform linear increase and decrease to achieve iteration.

7. The target pose calculation method for a hybrid chassis manipulator-arm composite robot for unpaved roads according to claim 6 is characterized by: The matrix transformation result with the largest number of matching point pairs as the transformation pose of the target object and the template includes: Through iteration Get the corresponding matching points ,in 、 、 and are the matching points corresponding to the 1st, 2nd, 3rd and tth iterations respectively, and the matrix corresponding to the maximum matching point number is recorded as , that is, the transformed pose of the target object and the template, where and are the rotation matrix and displacement vector in the matrix corresponding to the maximum number of matching points.

8. The target pose calculation method for a hybrid chassis manipulator-arm composite robot for non-paved roads according to claim 7 is characterized by: The transformed pose is then transformed by a transformation matrix to obtain the pose of the target under the base of the robotic arm, which includes: The position of the target under the base of the robotic arm obtained when the camera is shaken for: = ； in, and They are the rotation matrix and displacement vector of the target object under the base of the robotic arm, is the position of the end of the robotic arm under the base, and They are the rotation matrix and displacement vector of the position of the end of the robotic arm under the base, is the fine registration transformation matrix, and are the rotation matrix and displacement vector in the precise registration transformation matrix respectively.

9. A target position calculation device for a hybrid-electric chassis manipulator-arm composite robot for unpaved roads, characterized by: include: The hand-eye calibration module is used to control the movement of the end of the robotic arm to the specified position. The structured light camera on the end of the robotic arm is used to capture the calibration plate to obtain an image. The hand-eye calibration is performed on the captured image to obtain the transformation matrix of the camera under the end of the robotic arm. An initial point cloud data acquisition module is used to scan a target object using structured light in a stationary state to obtain initial point cloud data of the target object; The target object template point cloud acquisition module is used to remove noise points in the initial point cloud data through a model filtering algorithm, and then obtain the target object template point cloud through voxel grid downsampling and point cloud decentralization; The target object vibration point cloud acquisition module is used to collect the point cloud of the target object when the structured light device is in vibration, and obtain the target object vibration point cloud by statistical outlier removal and voxel grid downsampling; The coarse registration conversion matrix acquisition module calculates the FPFH features for the acquired target template point cloud and the target vibration point cloud, obtains two sets of feature vectors, and uses the SAC-IA algorithm to align the two sets of feature vectors to obtain the coarse registration conversion matrix; The fine registration transformation matrix calculation module is used to use the coarse registration transformation matrix as the initial transformation matrix, and use the ICP algorithm to calculate the fine registration transformation matrix for the target template point cloud and the target vibration point cloud; The adaptive threshold calculation module is used to perform rigid body transformation on the target object vibration point cloud using the precise registration transformation matrix, calculate the number of matching points between the target object template point cloud and the target object vibration point cloud, and then calculate the adaptive threshold based on the number of matching points; The pose acquisition module is used to iterate the rigid body transformation matrix of the target template point cloud based on the adaptive threshold, obtain the matrix transformation results under each posture, and use the matrix transformation result with the largest number of matching point pairs as the transformed pose of the target and the template. The transformed pose is then transformed through the conversion matrix to obtain the pose of the target under the base of the robotic arm.

10. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instruction is executed by a processor, the target pose calculation method of the hybrid chassis manipulator arm composite robot for non-paved roads as described in any one of claims 1-8 is implemented.

Images