A generalized contour data processing method, object recognition method and device

Abstract

The application provides a generalized contour data processing method, an object cognition method and equipment. It is applied to the technical field of robot perception. The method comprises the following steps: performing feature learning according to generalized contour data of a target object to obtain generalized contour coding; the generalized contour data and the generalized contour coding both comprise generalized contour features of a plurality of points, the number of points in the generalized contour coding is less than the number of points in the generalized contour data, and the number of generalized contour features in the generalized contour coding is greater than the number of generalized contour features in the generalized contour data; and the generalized contour coding is decoded to obtain the generalized contour data of the target object. The application significantly improves the efficiency and quality of generalized contour data processing, and realizes efficient generalized contour construction driven by a small amount of data.

Description

Technical Field

[0001] This invention relates to the field of robot perception technology, specifically to a generalized contour data processing method, an object recognition method, and a device. Background Technology

[0002] Object recognition is a crucial foundation for robots to achieve intelligent operation. Currently, both vision and touch can provide some object recognition capabilities. Vision can capture the geometric contours of objects when visible, but it cannot provide force-related information (such as inertia, stiffness, and friction). Touch can provide force-related information, but corresponding modeling methods are currently lacking. Acquiring tactile data relies on direct interaction with objects, which may lead to damage to the robot, sensors, and the target object. Data acquisition is also costly; therefore, efficient data utilization is paramount. Summary of the Invention

[0003] In view of this, the present invention provides a method for processing generalized contour data based on generalized contour coding, comprising: performing feature learning based on the generalized contour data of a target object to obtain a generalized contour code, wherein both the generalized contour data and the generalized contour code include generalized contour features of several points, and the number of points in the generalized contour code is less than the number of points in the generalized contour data, and the number of generalized contour features in the generalized contour code is greater than the number of generalized contour features in the generalized contour data; and decoding the generalized contour code to obtain the generalized contour data of the target object.

[0004] Optionally, feature learning is performed based on the generalized contour data of the target object to obtain a generalized contour code, including: sampling the farthest point of the generalized contour data to obtain a preset number of first representative points, the preset number being less than the number of points in the generalized contour data; based on each first representative point, filtering points in the generalized contour data within the first representative point and multiple preset neighborhoods to obtain neighborhood points of each first representative point; fusing the generalized contour features of the neighborhood points of each first representative point to obtain a preset number of fused neighborhood features of the first representative points; performing dimensionality transformation on the fused neighborhood features of each first representative point using a multilayer perceptron to obtain a preset number of target learning features of the first representative points; and concatenating the generalized contour features and target learning features of the preset number of first representative points to obtain the generalized contour code.

[0005] Optionally, the preset neighborhood is a spherical region defined by the neighborhood radius.

[0006] A second aspect of the present invention provides a method for object cognition based on robot tactile perception, comprising: acquiring generalized contour data and tactile data of a target object; performing feature learning on the generalized contour data using any generalized contour data processing method based on generalized contour coding to obtain a generalized contour code; performing feature learning on the tactile data using a tactile coding model to obtain a tactile code; extracting noise from the generalized contour code and tactile data with added random noise using a generalized contour diffusion model to obtain noise, and removing noise from the generalized contour code with added random noise based on the noise to obtain a target generalized contour code; and decoding the target generalized contour code to obtain the true generalized contour data of the target object.

[0007] Optionally, before using the generalized contour diffusion model to extract noise from the generalized contour coding and tactile data with added random noise, the method further includes: generating random noise and time steps with the same shape as the generalized contour coding based on the generalized contour coding; adding random noise to the generalized contour coding to obtain the generalized contour coding with added random noise.

[0008] Optionally, noise extraction is performed on the generalized contour coding and tactile data with added random noise using a generalized contour diffusion model to obtain noise, including: concatenating the features of the generalized contour coding and tactile data with added random noise to obtain fused features; and iteratively extracting noise from the fused features based on a time step using a generalized contour diffusion model to obtain noise.

[0009] Optionally, acquiring the generalized contour data of the target object includes: acquiring the point cloud of the target object using a vision acquisition device; sampling based on the point cloud to obtain a second representative point; acquiring displacement information of the target object caused by applying a force at the second representative point; obtaining the force-displacement ratio information based on the applied force and the corresponding displacement information; and obtaining the generalized contour data of the target object based on the point cloud and the force-displacement ratio information.

[0010] Optionally, the force applied to the target object at the second representative point is applied in the normal direction at the second representative point.

[0011] Optionally, acquiring tactile data of the target object includes: controlling the robot end effector to move to the target position based on a preset position and preset direction; controlling the robot end effector to grasp the target object with a preset load; adjusting the position and load of the robot end effector and collecting tactile data of the target object during the adjustment process.

[0012] Optionally, the load on the robot's end effector can be adjusted, including: increasing the preset load until it reaches a preset maximum load, and then stopping the increase of the preset load; or, acquiring the position of the target object during the increase of the preset load; determining whether the position of the target object has changed; if the position of the target object has not changed, then continuing to increase the preset load; if the position of the target object has changed and the displacement reaches a preset displacement, then stopping the increase of the preset load.

[0013] Optionally, the position of the robot end effector can be adjusted, including: keeping the preset load constant and adjusting the position and orientation of the robot end effector.

[0014] A third aspect of the present invention provides an object recognition device based on robot tactile perception, the device comprising: a processor and a memory connected to the processor; wherein the memory stores instructions executable by the processor, the instructions being executed by the processor to cause the processor to perform the aforementioned object recognition method based on robot tactile perception.

[0015] The generalized contour data processing method based on generalized contour coding of the present invention generates a generalized contour code with fewer points but richer generalized contour features by performing feature learning on the generalized contour data of the target object; then, by decoding this code, the generalized contour data of the target object is obtained. This embodiment requires only a small amount of original generalized contour data to generate a high-information-density code through feature learning and compression enhancement, reducing the amount of data while retaining core contour features, and quickly decoding and reconstructing generalized contour data consistent with the real generalized contour, significantly improving the efficiency and quality of generalized contour data processing, and realizing efficient generalized contour construction driven by a small amount of data.

[0016] This invention presents a robot tactile object cognition method. It uses a small amount of generalized contour and tactile data of a target object as basic input. A generalized contour encoding model is used to learn and compress the contour data, retaining core features and learning new fusion features, significantly reducing the data volume. Then, a tactile encoding model is used to extract and compress the tactile data, simultaneously reducing the scale of tactile information. A generalized contour diffusion model is then used to learn potential biases in the data by adding noise and removing them, enhancing the model's adaptability to complex distributions. Simultaneously, the generalized contour diffusion model integrates generalized contour encoding and tactile encoding, co-mapping geometric structure and contact state information to the generalized contour space. Finally, a generalized contour encoder decodes the data to obtain the true generalized contour, reconstructing a high-precision generalized contour that includes the complete shape of the object and retains contact details. Furthermore, it reconstructs a complete generalized contour consistent with the true generalized contour from a small amount of original data, achieving efficient acquisition of a large number of generalized contours driven by a small amount of input data. This provides a cost-effective generalized contour input solution for robot tactile cognition, improving efficiency. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a generalized contour data processing method based on generalized contour coding in an embodiment of the present invention;

[0019] Figure 2 This is a structural diagram of the generalized contour coding model in an embodiment of the present invention;

[0020] Figure 3 This is a flowchart of an object recognition method based on robot touch according to an embodiment of the present invention. Detailed Implementation

[0021] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0023] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can also refer to the internal connection of two components; and they can refer to a wireless connection or a wired connection. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0024] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0025] like Figure 1 As shown, this embodiment of the invention provides a method for processing generalized contour data based on generalized contour coding. This method is executed by an electronic device such as a computer or server, and specifically includes:

[0026] S11. Based on the generalized contour data of the target object, feature learning is performed to obtain the generalized contour code. Both the generalized contour data and the generalized contour code include generalized contour features of several points. The number of points in the generalized contour code is less than the number of points in the generalized contour data, and the number of generalized contour features in the generalized contour code is greater than the number of generalized contour features in the generalized contour data.

[0027] In this embodiment, the target object is any object whose contour data needs to be acquired (such as a person or object). Point cloud data of the target object, or force response data at point cloud coordinates, is acquired using a force response testing device to obtain generalized contour data, which is used to characterize the geometric contour morphology, mechanical response distribution characteristics, and structural spatial properties of the target object. Feature learning is then performed on the generalized contour data of the target object to ultimately generate a generalized contour code with fewer points but richer generalized contour features.

[0028] S12, decode the generalized contour code to obtain the generalized contour data of the target object.

[0029] Generalized contour data refers to the spatial coordinates (such as three-dimensional coordinates (x, y, z)) of several points on the surface of a target object, used to accurately describe the geometry of the target object, as well as the force response data at the point cloud coordinates.

[0030] The contour data processing method based on generalized contour coding provided in this invention is essentially a contour data processing algorithm implemented using a generalized contour coding model. Besides using common unsupervised dimensionality reduction techniques such as autoencoders or variational autoencoders, the generalized contour coding model can also employ an encoder-decoder architecture, specifically as follows... Figure 2 As shown, the generalized contour encoder 101 includes multiple downsampling modules, such as... Figure 2 The downsampling modules 1-m in the model are used to learn features from the generalized contour data of the target object to obtain the generalized contour code. The generalized contour decoder 102 includes multiple upsampling modules, such as... Figure 2The upsampling modules 1-m are used to decode the generalized contour encoding to obtain the generalized contour data of the target object. The downsampling modules need to maintain the meaning of the generalized contour features at the architectural level, while the upsampling modules do not need to maintain the meaning of the generalized contour features at the architectural level and can be implemented using a multilayer perceptron.

[0031] In this embodiment, the generalized contour encoding model can be trained in advance using generalized contour training data. During training, the generalized contour training data is used as input and the generalized contour data is used as the target. Key contour features are extracted and compressed into generalized contour codes by multiple downsampling modules of the generalized contour encoder, and then decoded and restored into generalized contours by multiple upsampling modules of the generalized contour decoder. The training process uses the generalized contour reconstruction error as the loss function, and optimizes the encoder and decoder parameters through backpropagation, so that the model can efficiently reconstruct the target generalized contour data from the generalized contour data.

[0032] like Figure 2 As shown, the steps for processing generalized contour data using a trained generalized contour coding model are as follows:

[0033] The input is generalized contour data of the target object, in the format [B,C,N], where B is the batch size of the input data, C is the number of generalized contour features (including features from the generalized contour data and contact data), and N is the number of points. Then, the downsampling module in the generalized contour encoder transforms the generalized contour data, resulting in the data format [B,C+D_1,N_1], where C is the number of generalized contour features (including features from the generalized contour data and contact data), D_1 is the number of features learned by the model, and N_1 is the number of points after downsampling, which is less than N. Then, as needed, more downsampling modules are added to continue the transformation process, resulting in data formats [B,C+D_2,N_2], [B,C+D_3,N_3], ..., [B,C+D_m,N_m]. Overall, this continuously increases the number of learned features and decreases the number of points. Finally, multiple upsampling modules in the generalized contour decoder are used to transform the data in the format [B,C+D_m,N_m] to the data in the format [B,C+D_m-1,N_m-1], ..., and finally to the data in the format [B, C, N], which is the generalized contour data of the target object.

[0034] This embodiment generates a generalized contour code with fewer points but richer generalized contour features by performing feature learning on the generalized contour data of the target object. This code is then decoded to obtain the generalized contour data of the target object. This embodiment requires only a small amount of original generalized contour data to generate a high-information-density code through feature learning and compression enhancement, reducing the amount of data while preserving core contour features. It also quickly decodes and reconstructs generalized contour data consistent with the real generalized contour, significantly improving the efficiency and quality of generalized contour data processing and achieving efficient generalized contour construction driven by a small amount of data.

[0035] In some optional embodiments of this example, step S11 involves feature learning based on the generalized contour data of the target object to obtain a generalized contour code, specifically including:

[0036] S111, sample the farthest point of the generalized contour data to obtain a preset number of first representative points, the preset number being less than the number of points in the generalized contour data.

[0037] In practice, each downsampling module performs feature learning on the generalized contour data. Taking one downsampling module as an example, based on the preset number N' points required for sampling, the module uses the farthest sampling point to find N' points from the generalized contour data according to the spatial coordinates of the points in the input shape [B,C+D,N]. This yields the preset number N' of the first representative points, where N' < N, and D represents the new features learned by each downsampling module. If we take the first downsampling module as an example, then the input shape is [B,C,N], i.e., D=0.

[0038] S112, based on each first representative point, filter the points in the generalized contour data within the first representative point and multiple preset neighborhoods to obtain the neighborhood points of each first representative point.

[0039] Given k neighborhoods, for each of the n' first representative points, we need to find the points within the k neighborhoods of the first representative point, and finally obtain the neighborhood points of each first representative point.

[0040] S113, perform feature fusion on the generalized contour features of the neighborhood points of each first representative point to obtain a preset number of fused neighborhood features of the first representative points.

[0041] The feature fusion method can be to take the average, maximum or minimum value of the generalized contour features of the neighborhood points corresponding to each first representative point, or it can be other complex combination grid methods. In this case, the fused neighborhood feature data format is [B,k(C+D),N'].

[0042] S114, using a multilayer perceptron to perform dimensional transformation on the fused neighborhood features of each first representative point, to obtain a preset number of target learning features for the first representative points.

[0043] The dimensionality of the fused neighborhood features of each first representative point is changed to obtain the target learning feature format [B,d',n'].

[0044] S115, the generalized contour features and target learning features in the first representative points of a preset number are concatenated to obtain the generalized contour code.

[0045] The generalized contour encoding is obtained by combining the original C generalized contour features in step S111 with the d' generalized contour features learned in step S114, with the format [B,C+d',n'].

[0046] This embodiment selects a predetermined number of key representative points (fewer than the original data) from the original generalized contour data through farthest point sampling, reducing redundant points while preserving the core contour location information. Subsequently, based on these representative points, neighborhood point selection and feature fusion are performed to integrate the generalized contour features of local regions, enhancing the ability to represent details. Then, new fusion features are learned through dimensional transformation of a multilayer perceptron, concatenating the original features with clear physical meaning with the newly added fusion features to generate a generalized contour code with higher information density. This process significantly reduces the data size (the reduction in the number of points far exceeds the increase in the number of features) while learning optimal new features, fully leveraging the optimal information compression capability of the deep learning encoder. Ultimately, the generalized contour code can still completely preserve the core contour information of the target object with less data, laying the foundation for subsequent efficient decoding and reconstruction of high-precision generalized contour data.

[0047] In some optional embodiments of this example, the preset neighborhood in step S112 is a spherical region determined by the neighborhood radius.

[0048] For each of the n' first representative points, multiple spherical regions need to be formed with the first representative point as the center and the radius of each region. That is, one first representative point corresponds to k spherical regions. Then, the original point in each spherical region is found to obtain the neighborhood point of a first representative point. The neighborhood points of the remaining first representative points are found in the same way.

[0049] This embodiment sets the preset neighborhood as a spherical region determined by the neighborhood radius, and collects the original points inside the spherical region corresponding to each first representative point as neighborhood points. This enables more flexible and comprehensive capture of local spatial features, enhances the ability to collect neighborhood information at different locations, and thus improves the richness of feature fusion and the model's representation effect on contour details.

[0050] like Figure 3 As shown, this embodiment of the invention provides a method for object cognition based on robot tactile perception. This method is executed by an electronic device such as a computer or server, and specifically includes:

[0051] S21, Obtain the generalized contour data and tactile data of the target object.

[0052] The process involves collecting generalized contour data of the target object using force response testing equipment and tactile data of the target object using a robot end effector. The generalized contour data includes point cloud data of the target object or force response data at point cloud coordinates, used to characterize the geometric contour morphology, mechanical response distribution characteristics, and structural spatial properties of the target object. The tactile data includes parameters such as the position coordinates of the end effector during adjustment, the magnitude and direction of the applied force, the displacement change at the contact point, pressure distribution characteristics, and deformation response under contact conditions, used to characterize the contact mechanical properties of the target object (e.g., stiffness, flexibility), the fine structure of the local geometric contour (e.g., curvature characteristics), and surface pressure distribution patterns. Data acquisition can be completed with only a small amount of contact information from discrete points or local areas. This step eliminates the need for high-density, full-coverage acquisition methods, significantly reducing data acquisition costs and time.

[0053] S22, using the generalized contour data processing method based on generalized contour coding described in any of the above embodiments, feature learning is performed on the generalized contour data to obtain the generalized contour coding.

[0054] That is, feature learning is performed on the generalized contour data according to step S11, that is, feature learning is achieved through the generalized contour encoder of the generalized contour coding model to obtain the generalized contour code of the target object. For details of the specific processing, please refer to the relevant description of the above embodiment, which will not be repeated here.

[0055] S23, use the tactile coding model to learn features from tactile data to obtain tactile codes.

[0056] Tactile data is learned and encoded using a tactile encoding model. The tactile encoding model can be an existing supervised encoder or a common unsupervised dimensionality reduction technique such as an autoencoder or variational autoencoder. The tactile data is represented by a vector to obtain the tactile code, which compresses the original high-dimensional and sparse tactile data into a compact low-dimensional vector representation, ultimately achieving a significant reduction in the amount of data.

[0057] S24. The generalized contour diffusion model is used to extract noise from the generalized contour code and tactile data with added random noise, and the noise is removed from the generalized contour code with added random noise based on the noise to obtain the target generalized contour code.

[0058] In this embodiment, the random noise and the extracted noise do not refer to physical noise, but rather to the random offset added to the generalized profile diffusion model. This process of adding and removing noise can help the generalized profile diffusion model learn the underlying structure and distribution of the data.

[0059] S25, decode the target generalized contour code to obtain the real generalized contour data of the target object.

[0060] By using the generalized contour encoder of the trained generalized contour encoding model to decode the generalized contour encoding of the target object, the true generalized contour of the target object can be obtained. Specifically, the "compressed information" in the low-dimensional generalized contour encoding is gradually restored to the specific coordinates of the high-dimensional generalized contour through multiple upsampling modules of the generalized contour encoder, and finally the generalized contour data that is highly consistent with the true contour of the target object in terms of geometric shape and mechanical response distribution is reconstructed.

[0061] This embodiment uses a small amount of generalized contour and tactile data of a target object as basic input. A generalized contour encoding model is used to learn and compress the contour data, retaining core key features and learning new fusion features, significantly reducing the data volume. Then, a tactile encoding model is used to extract and compress the tactile data, simultaneously reducing the scale of tactile information. A generalized contour diffusion model is then used to learn potential biases in the data by adding noise and removing it, enhancing the model's adaptability to complex distributions. Simultaneously, the generalized contour diffusion model integrates generalized contour encoding and tactile encoding, co-mapping geometric structure and contact state information to the generalized contour space. Finally, a generalized contour encoder decodes the data to obtain the true generalized contour, reconstructing a high-precision generalized contour that includes the complete shape of the object and retains contact details. Moreover, it reconstructs a complete generalized contour consistent with the true generalized contour from a small amount of original data, achieving efficient acquisition of a large number of generalized contours driven by a small amount of input data. This provides a cost-effective generalized contour input solution for robot tactile cognition, improving efficiency.

[0062] In some optional embodiments of this example, before using the generalized contour diffusion model to extract noise from the generalized contour encoding and tactile data with added random noise in step S24, the object cognition method based on robot tactile perception further includes: generating random noise and time steps with the same shape as the generalized contour encoding based on the generalized contour encoding.

[0063] Based on the shape of the generalized contour encoding, random offsets (such as coordinate offsets and feature offsets) are generated to match it. This process does not depend on actual time, but only controls the offset generation logic through time steps (number of iterations).

[0064] Adding random noise to the generalized contour coding results in a generalized contour coding with added random noise.

[0065] The generated random offsets are progressively superimposed onto the generalized contour code over time steps. For example, a small offset is added at time step t=1, and two offsets are added at t=2 (or the offset strength is increased). Each time step represents a "noise diffusion" operation, which iteratively disrupts the structure of the original code, simulating the cumulative perturbation of data in a complex environment.

[0066] This embodiment adds offsets controlled by time steps, which not only ensures that the subsequent generalized contour diffusion model focuses on core contour information (rather than local perturbations), but also enhances the reconstruction capability of the original data through inverse denoising. Ultimately, the model can decode the target generalized contour more accurately from a small number of generalized contour codes.

[0067] In some optional embodiments of this example, in step S24, a generalized contour diffusion model is used to extract noise from the generalized contour encoding and tactile data with added random noise, specifically including:

[0068] S241, the generalized contour coding with added random noise and the tactile data are concatenated to obtain the fused features.

[0069] The generalized contour encoding with added random offsets is concatenated with tactile data at the feature level to form a fused feature. This step fuses visual geometric information with tactile mechanical information, allowing the model to utilize global perturbation information of both "shape" and "contact state" simultaneously, providing a more comprehensive input for subsequent offset extraction.

[0070] S242 uses a generalized contour diffusion model based on time steps to iteratively extract noise from the fused features to obtain the noise.

[0071] Using time steps as the control unit (each time step represents one noise diffusion operation), the fused features are processed in multiple iterations. The generalized contour diffusion model progressively separates and extracts the offset components at each time step, ultimately obtaining a clean offset signal.

[0072] This embodiment uses time-step controlled offset iteration extraction. The generalized contour diffusion model simultaneously utilizes global information from both geometry and mechanics, and accurately separates the offset components in stages. The generalized contour diffusion model can more accurately recover the original unoffset generalized contour encoding from the offset generalized contour encoding, providing more realistic input for subsequent generalized contour data reconstruction, and significantly improving the robustness and accuracy of object shape perception in robot tactile cognition.

[0073] In some optional embodiments of this example, obtaining the generalized contour data of the target object in step S21 specifically includes:

[0074] S211a, uses a visual acquisition device to acquire the point cloud of the target object.

[0075] A series of spatial points of the target object are collected using a visual acquisition device (such as a 3D scanner), for example, 2048 points in the form of (x, y, z). The collected point cloud data can also be called contour data. During the acquisition, the target object can be placed in a certain way, such as freely placed on a plane or suspended by an elastic string.

[0076] S212a, based on point cloud sampling, obtains the second representative point.

[0077] A typical way to select the second representative point is to sample from the farthest point to reduce the amount of data. For details, please refer to step S11.

[0078] S213a, Obtain displacement information of the target object caused by applying force at the second representative point.

[0079] At each second representative point, a force is applied along that point, and the displacement of the target object is measured. The force can be applied along the normal direction at the second representative point. The method for determining the local normal direction is to fit a plane to points "near" the second representative point, and the normal of those points is the normal of the second representative point.

[0080] S214a, based on the applied force and the corresponding displacement information, obtains the proportional relationship information between force and displacement.

[0081] For example, when suspending a target object with an elastic string, the force and displacement should be directly proportional. In this case, this proportionality coefficient is extracted as information about the ratio of force to displacement.

[0082] For example, when a target object is freely placed on a plane, there should be no displacement (or very small displacement) when the force is less than a certain value, and a linear relationship should exist when the force is greater than a certain value. In this case, the limit of the force and the proportional coefficient of the linear relationship are extracted as information on the proportional relationship between force and displacement.

[0083] This step can also involve installing a high-density array sensor at the end of the force-applying device to apply force and measure displacement and force distribution, thereby obtaining information on the ratio of force to displacement.

[0084] S215a, based on point cloud and force-displacement ratio information, obtains the generalized contour data of the target object.

[0085] This embodiment uses a vision-based data acquisition device to collect point cloud data of the target object as basic contour data, further reducing the data volume through sampling at the farthest point. Then, force is applied at a small number of secondary representative points, and displacement is measured. The direction of the force is determined by combining local normal fitting. Mechanical features are extracted by analyzing the relationship between force and displacement. Finally, the point cloud and the proportional relationship between force and displacement are used as generalized contour data. This process requires only a small amount of raw data acquisition to obtain high-information-density generalized contour data, thus reducing data acquisition costs.

[0086] In some optional embodiments of this example, obtaining tactile data of the target object in step S21 specifically includes:

[0087] S211b controls the robot's end effector to move to the target position based on a preset position and preset direction.

[0088] Using a robotic arm or other tactile device, the device moves to a target location near the target object. The preset position and direction can be determined by first finding a general graspable position and direction based on vision using existing grasping decision algorithms, and then randomly fine-tuning it around that position. Alternatively, it can be determined manually.

[0089] S212b controls the robot's end effector to grasp the target object with a preset load.

[0090] The preset load can be a small force to grip the target object to prevent damage to its shape.

[0091] S213b adjusts the position and load of the robot's end effector and collects tactile data of the target object during the adjustment process.

[0092] The robot continuously adjusts the gripping load or the position of the end effector, and collects contact status information of the robot end effector when gripping the target object in real time during the adjustment process, such as contact pressure distribution (force value at each contact point), contact pressure change rate, contact area ratio, etc.

[0093] This embodiment first controls the robot's end effector to move to a preset position, gently gripping the object with a small load to prevent deformation. Then, by dynamically adjusting the gripping load or the device's position, tactile data of the contact during the adjustment process is collected. This process requires only a few gripping adjustments to obtain key tactile information reflecting the object's contact characteristics, reducing data acquisition complexity and providing core inputs for describing the contact state in subsequent generalized contour coding and tactile coding models. This allows the models to more accurately learn the coupling relationship between geometry and mechanics, ultimately reconstructing extended contour information that includes both shape and contact details.

[0094] In some optional embodiments of this example, the load of the robot end effector is adjusted in step S213b, specifically in two ways:

[0095] The first method:

[0096] Increase the preset load until it reaches the preset maximum load, then stop increasing the preset load.

[0097] The second method:

[0098] The position of the target object is collected during the process of increasing the preset load.

[0099] Determine whether the position of the target object has changed.

[0100] If the position of the target object does not change, continue to increase the preset load.

[0101] If the position of the target object changes and the displacement reaches the preset displacement, then the preset load will stop increasing.

[0102] Both of the above methods start with a preset load (small load) and gradually increase the load until a certain criterion is met, such as reaching the preset maximum load or the target object's position changing and reaching a certain displacement when the load is increased, at which point the increase in load can be stopped.

[0103] This embodiment achieves dual optimization of accurate tactile data acquisition and object protection through two load adjustment methods: starting with a small load and gradually increasing it avoids deformation or damage to the object caused by sudden force; and triggering a stop by monitoring position changes precisely captures the transition of contact state, ensuring that the acquired tactile data covers the contact characteristics under different loads. These two methods enhance the diversity of tactile data, including contact details under different loads, providing a more comprehensive input for subsequent model learning of the coupling relationship between geometry and mechanics.

[0104] In some optional embodiments of this example, adjusting the position of the robot's end effector in step S213b specifically includes:

[0105] Keep the preset load unchanged, and adjust the position and orientation of the robot's end effector.

[0106] The position and orientation of the robot's end effector can be adjusted randomly or fixedly.

[0107] This embodiment achieves diversified acquisition of tactile data and full coverage of contact scenarios by keeping the preset load of the grip constant and adjusting the position and orientation of the robot's end effector. Random adjustment can cover the contact state of the target object at different positions and angles, avoiding local data deviations caused by fixed positions; fixed adjustment can specifically acquire contact details at key locations (such as edges and recesses). The combination of these two methods makes the tactile data contain richer scene information, reflecting the mechanical properties of the object at different contact positions, and providing a more comprehensive input for subsequent model learning of the coupling relationship between geometry and mechanics.

[0108] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0109] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0110] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0111] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0112] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for processing generalized contour data based on generalized contour coding, characterized in that, include: Feature learning is performed on the generalized contour data of the target object to obtain a generalized contour code. Both the generalized contour data and the generalized contour code include generalized contour features of several points. The number of points in the generalized contour code is less than the number of points in the generalized contour data, and the number of generalized contour features in the generalized contour code is greater than the number of generalized contour features in the generalized contour data. Decode the generalized contour code to obtain the generalized contour data of the target object; The step of performing feature learning based on the generalized contour data of the target object to obtain the generalized contour code includes: The farthest point is sampled from the generalized contour data to obtain a preset number of first representative points, where the preset number is less than the number of points in the generalized contour data. Based on each first representative point, points in the generalized contour data that are within the first representative point and multiple preset neighborhoods are filtered to obtain the neighborhood points of each first representative point. The generalized contour features of the neighborhood points of each first representative point are fused to obtain a preset number of fused neighborhood features of the first representative points. The dimensionality transformation of the fused neighborhood features of each first representative point is performed using a multilayer perceptron to obtain a preset number of target learning features for the first representative points. The generalized contour features in the first representative points of the preset number and the target learned features are concatenated to obtain the generalized contour code; The acquisition of the generalized contour data of the target object includes: The point cloud of the target object is acquired using a visual acquisition device; Based on the point cloud, a second representative point is obtained by sampling. Obtain displacement information when a force is applied to the target object at the second representative point, causing the target object to shift; Based on the applied force and the corresponding displacement information, the proportional relationship between force and displacement is obtained; The generalized contour data of the target object is obtained based on the point cloud and the proportional relationship between force and displacement.

2. The method according to claim 1, characterized in that, The preset neighborhood is a spherical region defined by the neighborhood radius.

3. A method for object cognition based on robot tactile perception, characterized in that, include: Acquire generalized contour data and tactile data of the target object; Using the method described in any one of claims 1 or 2, feature learning is performed on the generalized contour data to obtain a generalized contour code; The tactile data is used to learn features using a tactile coding model to obtain tactile codes; The noise is extracted from the generalized contour code with added random noise and the tactile data using the generalized contour diffusion model to obtain the noise, and the noise is removed from the generalized contour code with added random noise based on the noise to obtain the target generalized contour code. Decode the target generalized contour to obtain the real generalized contour data of the target object; The acquisition of the generalized contour data of the target object includes: The point cloud of the target object is acquired using a visual acquisition device; Based on the point cloud, a second representative point is obtained by sampling. Obtain displacement information when a force is applied to the target object at the second representative point, causing the target object to shift; Based on the applied force and the corresponding displacement information, the proportional relationship between force and displacement is obtained; The generalized contour data of the target object is obtained based on the point cloud and the proportional relationship between force and displacement.

4. The method according to claim 3, characterized in that, Before using the generalized contour diffusion model to extract noise from the generalized contour encoding with added random noise and the tactile data, the method further includes: Based on the generalized contour coding, random noise and time steps with the same shape as the generalized contour coding are generated; The random noise is added to the generalized contour code to obtain the generalized contour code with added random noise.

5. The method according to claim 4, characterized in that, The noise is extracted from the generalized contour coding with added random noise and the tactile data using the generalized contour diffusion model, and includes: The generalized contour code with added random noise and the tactile data are concatenated to obtain fused features; The noise is obtained by iteratively extracting noise from the fused features using a generalized contour diffusion model based on the time step.

6. The method according to claim 3, characterized in that, The force applied to the target object at the second representative point is applied in the normal direction at the second representative point.

7. The method according to claim 3, characterized in that, Acquire tactile data of the target object, including: Based on a preset position and preset direction, control the robot's end effector to move to the target position; Control the robot's end effector to grasp the target object with a preset load; The position and load of the robot's end effector are adjusted, and tactile data of the target object are collected during the adjustment process.

8. The method according to claim 7, characterized in that, Adjusting the load on the robot's end effector includes: Increase the preset load until it reaches a preset maximum load, then stop increasing the preset load; or... The position of the target object is acquired during the process of increasing the preset load; Determine whether the position of the target object has changed; If the position of the target object does not change, the preset load will continue to increase; If the position of the target object changes and the displacement reaches the preset displacement, then the increase of the preset load will stop.

9. The method according to claim 8, characterized in that, Adjusting the position of the robot's end effector includes: While keeping the preset load constant, adjust the position and orientation of the robot end effector.

10. An object recognition device based on robotic tactile perception, characterized in that, include: A processor and a memory connected to the processor; wherein the memory stores instructions executable by the processor, the instructions being executed by the processor to cause the processor to perform the method as described in any one of claims 1-9.

Images