Data processing method and device, and computer device

Abstract

The application relates to a data processing method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: obtaining structure data of a target object; performing feature extraction on the structure data to obtain feature data of the target object; comparing the feature data with historical feature data to determine the similarity between the target object and each historical object; in the case that there is a historical object with a similarity greater than a first threshold, generating reference resource exchange information of the target object according to historical resource exchange information of the historical object; the resource exchange information is used for representing the number of exchangeable resources of an object; in the case that there is no historical object with a similarity greater than the first threshold, inputting the feature data of the target object into a resource prediction model to obtain the reference resource exchange information of the target object. The method can improve the accuracy and efficiency of resource exchange information generation.

Description

Technical Field

[0001] This application relates to the field of big data technology, and in particular to a data processing method, apparatus and computer equipment. Background Technology

[0002] With the rapid development of digital design and manufacturing technologies, the storage and management of engineering drawings has shifted from paper-based archives to digital formats. Various CAD systems (computer-aided design systems) can accurately describe the geometry, dimensions, and technical requirements of components. Traditionally, when faced with new part requirements, companies typically rely on engineers manually reviewing historical drawings, comparing part structures, assessing process similarities, and estimating required time and material consumption based on personal experience to generate a cost quote. This approach is highly dependent on engineers' subjective judgment and personal experience, leading to inconsistent estimates and difficulty in ensuring cost consistency and stability. Furthermore, manual drawing comparison is inefficient and slow, failing to meet the business demands for rapid cost quotations. Summary of the Invention

[0003] Therefore, it is necessary to provide a data processing method, apparatus, computer equipment, computer-readable storage medium, and computer program product to address the aforementioned technical problems.

[0004] Firstly, this application provides a data processing method. The method includes:

[0005] Obtain the structure data of the target object;

[0006] Feature extraction is performed on the structural data to obtain the feature data of the target object;

[0007] The feature data is compared with historical feature data to determine the similarity between the target object and each historical object;

[0008] In the case of historical objects with a similarity greater than a first threshold, reference resource exchange information of the target object is generated based on the historical resource exchange information of the historical objects; the resource exchange information is used to characterize the quantity of exchangeable resources of the object.

[0009] In the absence of historical objects with a similarity greater than the first threshold, the feature data of the target object is input into the resource prediction model to obtain the reference resource exchange information of the target object.

[0010] Secondly, this application also provides a data processing apparatus. The apparatus includes:

[0011] The data acquisition module is used to acquire the structural data of the target object;

[0012] The feature extraction module is used to extract features from the structural data to obtain the feature data of the target object;

[0013] The feature comparison module is used to compare the feature data with historical feature data to determine the similarity between the target object and each historical object;

[0014] The information generation module is used to generate reference resource exchange information of the target object based on the historical resource exchange information of the historical object when there is a historical object with a similarity greater than a first threshold; the resource exchange information is used to characterize the quantity of exchangeable resources of the object.

[0015] The information generation module is also used to input the feature data of the target object into the resource prediction model to obtain the reference resource exchange information of the target object when there is no historical object with a similarity greater than the first threshold.

[0016] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the data processing method as described in any one of the embodiments of this disclosure.

[0017] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the data processing method as described in any one of the embodiments of this disclosure.

[0018] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the data processing method as described in any of the embodiments of this disclosure.

[0019] The aforementioned data processing methods, apparatus, computer equipment, storage media, and computer program products acquire structural data of the target object and extract its features, transforming them into feature data usable for analysis. The extracted feature data is compared with data in a historical feature database to calculate the similarity between the current object and each historical object. Two cases are handled based on the similarity: if a historical object with a similarity greater than a preset threshold exists, its resource exchange information is directly used to generate reference resource exchange information for the target object; if all similarities are below the threshold, the feature data of the current object is input into a pre-trained resource prediction model, which calculates and outputs reference resource exchange information. By prioritizing the reuse of highly similar historical data, the reliance on the prediction model is reduced, enabling rapid result generation in common scenarios and reducing computational overhead. Furthermore, results based on real historical information have practical reference value. On the other hand, for new objects lacking highly similar historical data, model prediction provides feasible reference information, ensuring the applicability of the solution and avoiding processing failures due to data gaps. Prioritizing the reuse of highly similar historical data reduces reliance on complex calculations, further reducing processing latency and resulting in faster overall business response and more efficient resource utilization. Attached Figure Description

[0020] Figure 1 This is a diagram illustrating the application environment of a data processing method in one embodiment.

[0021] Figure 2 This is a flowchart illustrating a data processing method in one embodiment;

[0022] Figure 3 This is a flowchart illustrating the data processing implementation in one embodiment;

[0023] Figure 4 This is a structural block diagram of a data processing device in one embodiment;

[0024] Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0026] The data processing method provided in this application embodiment can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or placed in the cloud or on other network servers. Terminal 102 acquires the structural data of the target object, for example, by capturing images through a camera or obtaining 3D structural information through sensors, and sends this structural data to server 104 via the network. After receiving the structural data, server 104 extracts features from the structural data to obtain the feature data of the target object. Server 104 compares the extracted feature data with historical feature data in the data storage system to determine the similarity between the target object and each historical object. If there is a historical object with a similarity greater than a first threshold, server 104 generates reference resource exchange information for the target object based on the historical resource exchange information of that historical object. If there is no historical object with a similarity greater than the first threshold, server 104 inputs the feature data of the target object into a resource prediction model and obtains the reference resource exchange information for the target object through model calculation. Finally, server 104 returns the generated reference resource exchange information to terminal 102, which can then display or further process the information. The terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle systems. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0027] In one embodiment, such as Figure 2 As shown, a data processing method is provided, including the following steps:

[0028] Step S200: Obtain the structural data of the target object.

[0029] The target object can include parts to be processed, product components to be quoted, or engineering drawings to be evaluated. Structural data describes the geometric shape, topological relationships, and technical attributes of the target object, and can be in formats such as 2D CAD drawings, 3D model files, or point cloud data. Specifically, user-submitted structural data can be received through a file upload interface. Methods for acquiring structural data also include collecting 3D point cloud data of physical parts using optical scanning equipment, and then converting it into a usable structural data format using reverse engineering software.

[0030] Step S202: Extract features from the structural data to obtain feature data of the target object.

[0031] The feature data can include information from multiple dimensions such as structural features, dimensional features, text features, and layer features, used to comprehensively characterize the technical characteristics and process requirements of the target object. The feature extraction process can include multi-level analysis of the structural data to extract key parameters from the original geometric data that can be used for similarity comparison and cost estimation. Structural feature extraction can include the identification of primitive features and geometric features. Dimensional feature extraction can include linear features, angular features, and tolerance features. Text feature extraction can include two levels: character features and identifier features. Layer feature extraction can include layer composition features and color distribution features.

[0032] In one exemplary embodiment, feature extraction can be achieved by reading primitive information, traversing all entity objects to statistically analyze the quantity distribution of basic primitives such as lines, arcs, circles, and polylines, and simultaneously calculating geometric parameters such as the total length, total area, and bounding box dimensions of the drawing. Linear dimensions, angular dimensions, and tolerance annotations in the drawing are identified, and annotation values ​​and their associated geometric elements are extracted. All text entities are traversed, identifying identification information such as part numbers, material grades, and version numbers, and statistically analyzing the character type distribution. The layer organization structure is analyzed, and the number of layers, the number of primitives contained in each layer, and the color scheme are statistically analyzed. For 3D model files, boundary representation information is extracted, and quality attributes such as volume, surface area, and centroid position are calculated. Machining features such as holes, slots, and bosses in the feature tree are identified. The extracted feature data are organized according to a preset data structure to form a feature description file of the target object, providing standardized input for subsequent similarity comparison and cost prediction.

[0033] Step S204: Compare the feature data with historical feature data to determine the similarity between the target object and each historical object.

[0034] Historical feature data can be stored in a database and may include multi-dimensional feature information of historical objects with which resource exchanges have been completed. Determining similarity can quantify the degree of similarity between the target object and historical objects, providing a basis for subsequent information generation. Specifically, feature vectors of the target object can be generated based on the extracted feature data. These feature vectors map multi-dimensional feature information to a unified numerical space, facilitating similarity calculation. The generation of feature vectors can employ a weighted normalization method, assigning corresponding weight coefficients to different types of features to reflect the importance of each feature dimension in similarity assessment. For example, the number of primitives and geometric parameters in structural features can reflect the basic morphological complexity of the object; tolerance information in dimensional features can reflect processing accuracy requirements; and material identifiers in text features can directly correlate with resource consumption levels. These features can be configured with different weight values ​​according to the actual application scenario during vector generation.

[0035] In another exemplary embodiment, similarity judgment can be performed on each feature data to obtain the similarity corresponding to each feature data, and the overall similarity between the target object and each historical object can be determined based on the similarity corresponding to each feature data. Specifically, structural feature similarity, size feature similarity, text feature similarity, and layer feature similarity can be calculated separately, and the similarities of each dimension can be integrated into a comprehensive similarity according to a preset fusion strategy. Structural feature similarity can be obtained by comparing the Euclidean distance or cosine similarity between the primitive quantity distribution vector and the geometric parameter vector; size feature similarity can be determined by comparing the quantity and numerical distribution of linear dimensions, angular dimensions, and tolerance annotations; text feature similarity can be calculated by using string matching algorithms or semantic similarity models to calculate the consistency of identification information; layer feature similarity can be obtained by comparing the degree of difference between layer organization structure and color configuration scheme. The similarities of each dimension can be weighted and summed according to preset weights to obtain the comprehensive matching score between the target object and historical objects, i.e., the similarity. The weight configuration can be dynamically adjusted according to the application scenario. For example, in scenarios that focus on processing costs, the weight of size features and material identification can be increased, and in scenarios that focus on design specifications, the weight of layer features and text features can be increased.

[0036] Step S206: If there are historical objects with a similarity greater than the first threshold, generate reference resource exchange information for the target object based on the historical resource exchange information of the historical objects.

[0037] The first threshold can be set as a high similarity standard to filter historical objects highly similar to the target object. When the similarity exceeds the first threshold, it indicates the existence of historical objects with similar technical characteristics to the target object, whose resource exchange information can be directly used as a reference. Resource exchange information can include specific data such as processing time, material consumption, equipment occupancy time, labor costs, and outsourcing costs, as well as price information such as supplier quotations, contract prices, and actual settlement amounts. Specifically, the historical object with the highest similarity can be selected, and its historical resource exchange information can be directly used as the reference resource exchange information for the target object; alternatively, multiple historical objects with similarity exceeding the first threshold can be selected, and a weighted average can be calculated based on the similarity between each historical object and the target object to generate comprehensive reference resource exchange information. The weights of the weighted average can be allocated according to the similarity level, with higher-similar historical objects receiving greater weights to fully reflect the impact of similarity on the reliability of the reference information. Furthermore, adjustments can be made based on the timeliness of the historical resource exchange information, assigning higher weights to more recent historical data and appropriately reducing the weights to older historical data to reflect the impact of market price fluctuations and technological advancements.

[0038] In an exemplary embodiment, when multiple historical objects with similarity greater than a first threshold exist, a multi-dimensional information filtering mechanism can be established. Historical objects are sorted from highest to lowest similarity, and a predetermined number of these are selected as candidate reference objects. Consistency verification is performed on the historical resource exchange information of each candidate reference object, identifying and removing abnormal data that significantly deviates from the normal range. Consistency verification can employ statistical analysis methods to calculate the mean and standard deviation of resource exchange information for each dimension, classifying data exceeding a predetermined multiple of the standard deviation as outliers. For historical resource exchange information that passes consistency verification, a comprehensive reference value can be calculated based on a dual weighting of similarity and timeliness. The timeliness weight can be determined based on the time node when the historical object completed the resource exchange, using an exponential decay function or a piecewise linear function for quantification; more recent historical data receives a higher timeliness weight. Finally, the weighted historical resource exchange information of each candidate reference object is aggregated to generate reference resource exchange information for the target object. Aggregation methods can include weighted arithmetic mean, weighted geometric mean, or adaptive fusion based on confidence level.

[0039] Step S208: In the absence of any historical object with a similarity greater than the first threshold, the feature data of the target object is input into the resource prediction model to obtain the reference resource exchange information of the target object.

[0040] Resource prediction models, built upon machine learning algorithms, are used to directly predict resource exchange information based on the technical characteristics of a target object when highly similar historical references are lacking. This model establishes a mapping relationship between feature data and resource consumption by learning from a large number of historical objects and their resource exchange information, thereby estimating the resource requirements of new objects. Specifically, resource prediction models can employ various algorithmic architectures such as regression models, neural networks, or ensemble learning. The input layer receives the feature data of the target object, and the output layer generates predicted values ​​for various resource exchange information.

[0041] In one exemplary embodiment, during the model training phase, multi-dimensional feature data of historical objects can be used as training samples, and corresponding historical resource exchange information can be used as label data. Supervised learning is then used to optimize model parameters. Feature data can be standardized before being fed into the model to eliminate the impact of differences in the dimensions of different features on model training. During model training, cross-validation can be used to evaluate the model's generalization ability, and hyperparameter configurations can be adjusted using techniques such as grid search or Bayesian optimization to obtain optimal prediction performance. For the trained model, performance monitoring and iterative updates can be performed periodically. When newly accumulated historical data reaches a certain scale, incremental learning or full retraining of the model is triggered to maintain the stability of prediction accuracy.

[0042] During the model inference phase, the feature data of the target object can be preprocessed according to the model input format and then input into the trained resource prediction model. The model outputs prediction results for various resource exchange information through forward computation. The prediction results can include point estimates and confidence intervals. The point estimates serve as the main content of the reference resource exchange information, while the confidence intervals characterize the degree of uncertainty of the prediction results, providing a risk reference for subsequent decision-making. When the confidence interval width of the prediction results exceeds a preset range, a manual review mechanism can be triggered, prompting the user to carefully evaluate the prediction results.

[0043] In another exemplary embodiment, the resource prediction model can also be configured with an anomaly detection mechanism. When the distribution of the input feature data deviates significantly from the training samples, a corresponding warning message is output, indicating that the target object may belong to a special case not fully covered by the model. The model can also output feature importance analysis results, identifying the feature dimensions that contribute the most to the current prediction result, helping users understand the basis for the prediction and identify potential data quality issues.

[0044] In the above data processing method, the structural data of the target object is acquired and its features are extracted, transforming it into feature data that can be used for analysis. The extracted feature data is compared with data in a historical feature database to calculate the similarity between the current object and each historical object. Two cases are handled based on the similarity: if there is a historical object with a similarity greater than a preset threshold, the resource exchange information of that historical object is directly used to generate reference resource exchange information for the target object; if all similarities are below the threshold, the feature data of the current object is input into a pre-trained resource prediction model, which calculates and outputs reference resource exchange information. By prioritizing the reuse of highly similar historical data, the reliance on the prediction model is reduced, enabling rapid result generation in common scenarios and reducing computational overhead. Furthermore, results based on real historical information have practical reference value. On the other hand, for new objects lacking highly similar historical data, feasible reference information is provided through model prediction, ensuring the applicability of the solution and avoiding processing failures due to data gaps. By prioritizing the reuse of highly similar historical data, the reliance on complex calculations is reduced, further lowering processing latency and making the overall business response faster and resource utilization more efficient.

[0045] In one embodiment, the step of inputting the feature data of the target object into a resource prediction model to obtain reference resource exchange information of the target object when there are no historical objects with a similarity greater than a first threshold includes:

[0046] In the absence of any historical objects with a similarity greater than the first threshold, but with a similarity greater than the second threshold, the feature data of the target object is input into the resource prediction model to obtain the reference resource exchange information of the target object.

[0047] In the absence of historical objects with a similarity greater than the second threshold, the feature data and material identifier of the target object are obtained;

[0048] Based on the feature data of the target object, the complexity of the target object is determined;

[0049] Reference resource exchange information for the target object is generated based on the complexity and material identifier of the target object.

[0050] In an exemplary embodiment, when there are no historical objects with a similarity greater than a first threshold but a storage similarity greater than a second threshold, the feature data of the target object can be input into a resource prediction model. The model then infers reference resource exchange information for the target object. The resource prediction model can be built based on machine learning algorithms, trained using the mapping relationship between the structural features, size features, text features, and layer features of a large number of historical objects and their corresponding actual resource consumption data. The model input layer receives multi-dimensional feature vectors of the target object, which undergo nonlinear transformation and feature combination in the hidden layer, generating a prediction result for resource exchange information at the output layer. The resource exchange information may include quantitative indicators such as estimated processing time, material costs, equipment occupancy time, and energy consumption, as well as guiding information such as recommended processing routes, required tool lists, and quality control points. The second threshold can be set lower than the first threshold to distinguish between highly similar objects and moderately similar objects. For moderately similar objects, the resource prediction model can make reasonable inferences and corrections based on their historical data, providing reference information with a certain degree of reliability.

[0051] In one exemplary embodiment, if no similarity exceeds a second threshold, the target object can be considered to have low or zero similarity with all historical objects, making it impossible to directly rely on the resource data of historical objects for prediction. In this case, a resource estimation method based on rules and empirical formulas can be used. The resource estimation method may include estimating material costs and processing costs, and summing the material costs and processing costs to obtain reference resource exchange information.

[0052] In one exemplary embodiment, material cost estimation may include estimating material costs based on the volume, density, and material identification of the target object. Specifically, material cost can be obtained using the following formula: Material Cost = Volume * Density * Material Unit Price. Processing cost estimation may include: reading process capability data from a configuration file; calculating a batch adjustment factor based on the production quantity; obtaining the process cost rate and complexity factor; calculating a time adjustment factor based on process capability and part precision; calculating labor hours = complexity * complexity factor * time adjustment factor; process cost = labor hours * hourly rate * batch factor; processing cost = feature complexity * process complexity factor * time adjustment factor * cost rate * batch factor; summing all costs gives the total processing cost. The time adjustment factor may include process precision divided by the part precision requirement; the higher the precision requirement, the larger the time factor, etc. The process complexity coefficient can be determined according to the process type. For example, the complexity coefficient of EDM and wire EDM is 2; the complexity of grinding is 1.8; the complexity of CNC machining is 1.5; the complexity of turning is 0.8; the complexity of milling and drilling is 0.6; and the complexity of other processes is 1.

[0053] In an exemplary embodiment, the total complexity = (basic complexity + shape complexity + contour complexity + size complexity) * material complexity; where, basic complexity = number of faces * 0.5 + number of edges * 0.3 + number of vertices * 0.2; shape complexity can be obtained based on sphericity: (1 - sphericity) * 100, where sphericity is an index used to measure the similarity of an object's shape to an ideal sphere, with a value range between [0,1], and the closer the value is to 1, the closer the shape is to a sphere; sphericity = (π^(1 / 3) × (6 × volume)^(2 / 3)) / surface area, and can also be obtained based on elongation: (elongation - 1) * 5 0, where elongation can be used to measure the ratio of an object's length to its width and height, indicating the object's slenderness; elongation = maximum size / minimum size; contour complexity can be measured by convexity, contour complexity = (1 - convexity) * 100, convexity is the ratio of the object's volume to its convex hull volume, with a value range between [0,1]; convexity = object volume / convex hull volume; dimensional complexity = (dimensional variance / ((maximum model size - minimum model size)² / 4)) × 50; material complexity can be predefined in a preset way, for example, difficult-to-process materials are 1.5, easy-to-process materials are 0.8, and ordinary steel is 1.0.

[0054] In this embodiment, a multi-level threshold judgment mechanism is established to achieve intelligent collaboration between similarity matching and resource prediction. When highly similar historical objects exist, their resource data is directly reused to ensure prediction accuracy; when only moderately similar objects exist, intelligent inference is performed using a resource prediction model to balance efficiency and accuracy; when similar references are lacking, a rule-based empirical calculation process is initiated to ensure basic prediction capabilities in scenarios without historical data. This hierarchical processing strategy not only fully utilizes the value of historical data assets but also establishes a standardized resource estimation system through complex quantification models and process parameter libraries. This enables the generation of resource exchange information to have complete coverage from data-driven to knowledge-driven approaches, significantly improving the adaptability and reliability of resource prediction under different data conditions.

[0055] In one embodiment, inputting the feature data of the target object into a resource prediction model to obtain reference resource exchange information of the target object includes:

[0056] Obtain the material identifier of the target object.

[0057] Material identification can include material grade or material code identified from character features, used to determine the physical performance parameters and processing characteristics of the target object. Material identification can be obtained by parsing the material labeling field in the text features, matching it against a standard material thesaurus or a company-defined material coding system, and extracting unique material category information.

[0058] The feature data and material identifier of the target object are input into the resource prediction model to obtain the first resource exchange information of the target object.

[0059] If the first resource exchange information is valid, the first resource exchange information is determined as the reference resource exchange information.

[0060] If the first resource exchange information is invalid, historical resource exchange information of multiple historical objects with a similarity greater than a second threshold to the target object and the similarity between each historical object and the target object are obtained. Based on the similarity and the corresponding historical resource exchange information, second resource exchange information is obtained. Based on the second resource exchange information and the material identifier, reference resource exchange information of the target object is generated. Wherein, invalid first resource exchange information includes at least one of the following: the resource prediction model does not exist, the amount of historical data input to the resource prediction model does not reach a preset number, or the resource prediction model malfunctions during operation.

[0061] In one exemplary embodiment, a valid model prediction may include the prediction model operating normally and returning a valid result from a larger-than-a-certain amount of similar data. An invalid prediction may include a non-existent model, insufficient data (if the amount of historical similar data exceeds a certain threshold, failure will be returned), or errors occurring during the calculation process.

[0062] In one exemplary embodiment, when the first resource exchange information is invalid, the average similarity between the current part and the learning data can be calculated, and the weight can be calculated based on the similarity (the higher the similarity, the greater the weight of the learning data). The weighted average is used to calculate the optimized cost = average cost of learning data * similarity weight + basic cost * (1 - similarity weight), etc.

[0063] In this embodiment, by introducing material identifiers as a key input parameter of the resource prediction model, joint modeling of material properties and geometric features is achieved, enabling the prediction results to accurately reflect the impact of different materials on processing technology and cost. The validity verification mechanism for the first resource exchange information ensures the reliability of the model output. When the model cannot provide effective predictions due to insufficient training data, abnormal operation, or lack of deployment, the system automatically switches to a weighted inference mode based on similar historical objects. This mode aggregates resource exchange information from multiple moderately similar objects and performs fusion calculations based on similarity weights. This avoids the bias that may be caused by a single similar object and fully utilizes the collective wisdom in historical data. Combined with material identifiers for final correction, it ensures that valuable resource exchange information can still be provided even in scenarios without model support.

[0064] In one embodiment, comparing the feature data with historical feature data to determine the similarity between the target object and each historical object includes:

[0065] Based on the feature data, a feature vector of the target object is generated.

[0066] The similarity between the target object and each historical object is determined by comparing the feature vector of the target object with the historical feature vectors.

[0067] Feature vectors can be intermediate data forms that numerically represent and structurally organize the multi-dimensional feature information of the target object, facilitating similarity measurement within a unified mathematical space. The generation of feature vectors can include operations such as normalization, weight allocation, and dimension mapping of the original feature data. Normalization can eliminate the influence of differences in the dimensions of different feature dimensions; for example, it can uniformly scale heterogeneous data such as the number of primitives, size values, and text length to the same numerical range. Weight allocation can be dynamically adjusted according to the importance of each feature dimension in similarity assessment, highlighting the influence of key features on similarity through a preset weight coefficient matrix or an adaptive weight mechanism learned from historical data. Dimension mapping can employ techniques such as principal component analysis, autoencoders, or feature hashing to compress the high-dimensional sparse feature space into a low-dimensional dense vector representation, reducing the complexity of subsequent calculations while retaining key information.

[0068] In an exemplary embodiment, the generation of feature vectors may further include: performing logarithmic transformation on the statistical data of primitives in structural features to compress numerical ranges with large differences in magnitude, and converting the quantities of various primitives such as lines, arcs, circles, and polylines into relative proportion vectors; performing interval encoding on linear and angular dimensions in dimensional features, discretizing continuous values ​​into preset level identifiers, and extracting precision level features from tolerance annotations; converting character types and identifier information into numerical vectors using word embedding or one-hot encoding methods for text features, wherein key identifiers such as part numbers and material grades can be normalized to generate hash features; and statistically analyzing the proportion of layers and color distribution entropy values ​​for layer features to form feature sub-vectors describing the organizational structure of the drawing. The above feature sub-vectors are combined into a complete feature vector according to preset splicing rules, and the vector dimension can be dynamically expanded or truncated according to the richness of the actual feature extraction to ensure the consistency of the vector length.

[0069] In another exemplary embodiment, historical feature vectors can be stored using a distributed database or a vector database, supporting efficient similarity retrieval operations. Feature vector comparison can employ an approximate nearest neighbor search algorithm to quickly locate the set of historical feature vectors closest to the target vector in a high-dimensional vector space.

[0070] In this embodiment, the comparability quantification between different feature types is achieved by transforming multi-dimensional feature data into a unified feature vector representation. The feature vector generation process comprehensively considers the numerical characteristics and physical meaning of each feature dimension, and employs targeted encoding strategies to ensure the integrity and accuracy of information expression.

[0071] In one embodiment, the feature data includes structural features; the structural features include at least one of the following: primitive features and geometric features;

[0072] The acquisition of the primitive features includes: traversing the structural data, counting the number of each primitive in the target object, and obtaining the primitive features.

[0073] The primitive features can include identifying and counting the basic graphic elements contained in the structural data to form a quantitative index describing the complexity of the object's composition. The statistics of primitive features can cover various CAD primitive types, such as line segments, arcs, circles, ellipses, polylines, spline curves, polygons, text annotations, dimension annotations, fill patterns, and block references. During the traversal process, a recursive algorithm can be used to process nested block reference structures, expanding the primitives within the block before including them in the statistics, thus avoiding information loss caused by block encapsulation.

[0074] The acquisition of the geometric features includes: obtaining the geometric data of the target object from the structural data to obtain the geometric features, wherein the geometric features include at least one of the following: total length, total area, and bounding box size.

[0075] Geometric features can include quantitative parameters extracted from structural data to describe the spatial morphology and scalability of an object. Geometric data can be acquired by parsing the graphic database of CAD files, accessing the geometric definitions of primitives, and performing corresponding geometric calculations. Total length can be calculated by summing the lengths of all linear primitives (including line segments, arcs, polylines, etc.), reflecting the linear material consumption scale of the object; total area can be calculated by accumulating the area of ​​all closed regions and infill patterns, used to estimate the requirements for sheet metal or surface treatment; bounding box dimensions can determine the maximum span of the object in the X, Y, and Z directions, characterizing the object's physical space occupation and transportation and storage requirements. For three-dimensional structural data, further advanced geometric features such as volume, surface area, and centroid location can be extracted to more comprehensively describe the spatial characteristics of the object.

[0076] In another exemplary embodiment, the combined features may further include shape complexity, which can be used to quantify the irregularity and construction fineness of the geometric contour of a target object. Shape complexity can be obtained by analyzing the proportional relationship between the bounding box size and the total area or volume in the geometric features. For example, calculating the ratio of the object's surface area to its volume; a larger ratio indicates richer surface details and a more complex structure. Alternatively, it can be obtained by statistically analyzing the number and type diversity of the primitives constituting the object; the more types and the greater the number of primitives, the higher the shape complexity. For structural data containing curved surfaces, parameters such as the frequency of curvature changes and the number of surface patches can be further introduced to comprehensively evaluate shape complexity. Shape complexity features can effectively distinguish between simple standard parts and complex custom parts, providing important reference for subsequent data classification, process difficulty assessment, and manufacturing cost estimation.

[0077] In this embodiment, by systematically identifying and statistically analyzing the primitives in the structural data, primitive features describing the constituent elements of the target object are obtained. Simultaneously, geometric features reflecting the spatial scale of the object, such as total length, total area, and bounding box dimensions, are extracted from the geometric data, achieving a dual characterization of the target object's physical composition and spatial form. Primitive features focus on the analysis of the object's compositional complexity, while geometric features focus on the object's space occupancy and material scale assessment. These two aspects complement each other, forming a complete structural feature description system. This multi-dimensional structural feature extraction method not only preserves key information from the original design data but also transforms it into calculable and comparable quantitative indicators, laying a data foundation for subsequent similarity matching and resource prediction.

[0078] In one embodiment, the feature data includes dimensional features; the dimensional features include at least one of the following features: linear features, angular features, and tolerance features;

[0079] The acquisition of the linear features includes traversing the structural data, identifying each linear dimension annotation in the target object, and obtaining the linear features.

[0080] Linear features can include the systematic identification and extraction of linear dimensions from all annotations within the target object, forming a feature set describing the object's length scale distribution. Linear dimension annotation identification can be achieved by parsing dimension entities in CAD files, extracting their annotation values, locations, and associated geometric elements. During the traversal, horizontal, vertical, and inclined dimensions can be distinguished, and their quantity and numerical distribution can be statistically analyzed.

[0081] The acquisition of the angular features includes traversing the structural data, identifying the angular dimension annotations in the target object, and obtaining the angular features;

[0082] The angular features can include identifying and extracting all angular dimensions labeled in the target object, forming a feature set describing the distribution of angular relationships within the object. Angular dimension label identification can be achieved by parsing angular dimension entities in CAD files, extracting their angle values, vertex positions, and associated geometric elements on both sides. During the traversal, different angle types such as acute, right, obtuse, and straight angles can be distinguished, and the proportion and numerical distribution range of each type of angle can be statistically analyzed. Furthermore, angle combinations with special technological significance, such as symmetrical angles and complementary angles, can be identified.

[0083] The acquisition of the tolerance features includes traversing the structural data, identifying the tolerance markings of the target object, and obtaining the tolerance features; the tolerance markings are used to characterize the machining accuracy of the target object.

[0084] Tolerance features can include identifying and extracting tolerance information from all annotations on the target object, forming a feature set describing the object's machining accuracy requirements. Tolerance annotation identification can be achieved by parsing tolerance annotation entities in CAD files, extracting information such as tolerance type, tolerance value, datum system, and associated dimensions. Tolerance types can include various categories such as dimensional tolerances, geometric tolerances, and surface roughness. Dimensional tolerances can be further divided into unidirectional tolerances, bidirectional tolerances, and symmetrical tolerances. Geometric tolerances can cover straightness, flatness, roundness, cylindricity, line profile, surface profile, parallelism, perpendicularity, angularity, coaxiality, symmetry, position, and runout. During the traversal, the quantity distribution, accuracy grade distribution, and tolerance band width range of various tolerances can be statistically analyzed to identify the high-precision requirements of key dimensions and the strict constraints of geometrical tolerances.

[0085] In this embodiment, by traversing structural data to identify various dimensional annotations, linear features, angular features, and tolerance features describing the size specifications of the target object are obtained, thus achieving a systematic extraction of the target object's dimensional information. Linear features reflect the object's length specifications in various directions, angular features describe the object's angular relationship requirements, and tolerance features characterize the machining accuracy level. Together, these three constitute a complete dimensional feature system, providing crucial dimensional information for subsequent similarity matching and resource prediction.

[0086] In one embodiment, the feature data includes text features; the text features include at least one of the following features: character features and identifier features;

[0087] The acquisition of character features includes: traversing the structured data, identifying text units in the target object, acquiring the characters of each text unit, and obtaining character features.

[0088] Character features can include character-level recognition and statistics of all text content in the target object, forming a basic feature set describing the text information of the object. Text unit recognition can be achieved by parsing text entities in CAD files, extracting attribute information such as character content, font style, rotation angle, and insertion position. During the traversal, single-line and multi-line text can be distinguished, and their quantity distribution can be statistically analyzed. Special character types, such as numbers, letters, Chinese characters, symbols, and their mixed patterns, can also be identified, and the frequency and proportion of each type of character can be statistically analyzed. Character features can include quantitative indicators such as total text quantity, average text length, character type distribution, and number of font types, reflecting the information annotation density and text complexity of the target object.

[0089] The acquisition of the identification features includes: identifying the identification data of the target object from the character features to obtain the identification features; wherein the identification data includes at least one of the following: part identification, material identification, and version identification.

[0090] The identification features can include extracting key identification information with specific semantic meanings from character features, forming a high-level feature set describing the identity attributes and technical specifications of the target object. Part identification can be extracted through pre-defined numbering rule pattern matching or regular expressions, such as identifying part numbers, drawing numbers, or material codes that conform to the enterprise coding system. These identifiers typically have fixed format features, such as containing specific prefixes, segmented structures, or check digits. Material identification can be matched based on a standard thesaurus of material grades to extract key information reflecting the material properties and physical performance of the target object. Version identification can be obtained by tracking the revision history of drawings to extract information such as version number, revision date, and change order number, used to distinguish different design iteration states of the same object and ensure version consistency during data matching.

[0091] In this embodiment, character features are obtained by traversing structural data to identify text units, and identification features such as part identifiers, material identifiers, and version identifiers are extracted from them, realizing multi-level parsing of the target object's text information. Character features provide basic statistical information of the text content, while identification features further mine key semantic information in the text. The combination of the two can not only describe the information annotation features of the target object, but also accurately identify the object's identity attributes and technical specifications, providing reliable data support for text content-based similarity matching and classification retrieval.

[0092] In one embodiment, the feature data includes layer features; the layer features include at least one of the following features: layer composition features and color distribution features;

[0093] The acquisition of the layer composition features includes: identifying the layers in the structural data, counting the number of layers and the number of primitives in each layer, and obtaining the layer composition features.

[0094] Layer composition features can include parsing and statistically analyzing the layer system defined in the structural data to obtain a set of structured features describing the hierarchical organization of the target object's data. Layer identification can be achieved by parsing the layer table information in the CAD file to obtain metadata such as layer name, layer status, and layer attributes, while simultaneously traversing all graphic element entities to record their respective layer affiliations. The number of layers can reflect the complexity of the target object's data organization; a higher number of layers indicates a more detailed structural division and richer information hierarchy. The distribution of graphic elements within each layer can reveal the proportion of the object's composition in different functional modules or geometric regions, such as distinguishing the scale of graphic elements carried by contour layers, annotation layers, auxiliary line layers, and fill layers.

[0095] The acquisition of color distribution features includes: acquiring the colors corresponding to each layer, counting the frequency of use of each color, and obtaining color distribution features.

[0096] Color distribution characteristics can include analyzing and statistically analyzing the color attributes of each layer and graphic entity to form a feature set describing the visual coding patterns of the target object. Colors can be obtained by parsing the color index table in the CAD file or directly reading the color attribute values ​​of the graphic elements, supporting unified processing of both true color and indexed color modes. The frequency of use statistics for each color can identify the distribution of the main color tone and the matching patterns of auxiliary colors in the target object, reflecting the visual differentiation strategies for different geometric elements or functional areas in the design specifications. Color distribution characteristics can include quantitative indicators such as the number of color types, the proportion of main colors, and the uniformity of color gradation distribution. Further analysis can be conducted on the association patterns between specific colors and graphic element types, such as the conventional coding habits of red being commonly used for warning labels and blue for dimensioning. For drawings using a layer color inheritance mechanism, the hierarchical structure characteristics of color distribution can be analyzed in conjunction with layer composition characteristics.

[0097] In this embodiment, by identifying layer information in structural data and statistically analyzing layer composition and color distribution, a systematic extraction of the data organization method and visual presentation characteristics of the target object is achieved. Layer composition features reflect the structured, layered design concept of the drawing; different layers typically correspond to different functional modules or professional divisions of labor. The number of layers and the distribution of elements in each layer can reveal the design complexity and professional collaboration characteristics of the object. Color distribution features characterize the visual coding standards of the drawing; different colors are often used to distinguish object types, status attributes, or process requirements. Statistics on color usage frequency can identify the ratio of dominant and auxiliary color systems. Layer features, structural features, dimensional features, and text features complement each other, jointly forming a multi-dimensional object description system, providing additional discriminative dimensions for matching and retrieval based on data organization patterns and visual similarity.

[0098] In one embodiment, the data processing method provided in this application can be applied to an intelligent quotation system in the machining industry, such as... Figure 3 As shown, the specific steps may include the following:

[0099] Step S300: Obtain Part Data. Receive the target part data uploaded by the user for which a quotation is required. The user can upload the part's engineering file, etc., through the user interface or application interface. The engineering file can include two-dimensional DXF format drawings or three-dimensional STP or STEP format models. Simultaneously, the user can also set key parameters for this quotation on the interface, such as material type and production quantity.

[0100] Step S302: Parsing and Preprocessing Data. Upon receiving a file, its type can be automatically determined based on the file extension to call different processing engines. If it is a 2D drawing, it can interpret DXF format data; if it is 3D data, it can parse STP or STEP format data. A unique task identifier is also created for this task.

[0101] Step S304: Extract features from the 2D drawing. Traverse all graphic entities in the drawing, counting the number of various graphic elements such as lines, circles, arcs, and polylines. Calculate the overall geometric features of the drawing, including the total length of all lines, the total area of ​​closed regions, and the width, height, and area of ​​the smallest bounding rectangle of the entire graphic, thus determining the shape complexity of the drawing. Identify dimensioning information in the drawing, distinguishing between linear dimensions, angular dimensions, radial dimensions, and diameter dimensions, and count the number of each type of dimension and the number of tolerance annotations it contains. Extract text content from the drawing, identify characters, and further parse out possible key identifiers such as part numbers, material information, and version numbers. Analyze the layer composition of the drawing, counting the total number of layers, the number of graphic elements on each layer, and the color distribution used by each layer.

[0102] Step S306: Extract 3D model features. Calculate the model's basic geometric properties, such as volume and surface area, and calculate a compactness index reflecting the model's compactness based on volume and surface area. Next, obtain the model's length, width, and height dimensions in 3D space, and count the number of faces, edges, and vertices constituting the model's topology. Also calculate shape descriptors such as sphericity, rectangularity, elongation, and flatness to improve the precision of describing the model's shape. Analyze the model's inertia matrix to extract its moments of inertia along the three principal axes. The model's convex hull volume can also be calculated to derive convexity and solidity features reflecting the model's degree of concavity and convexity. Finally, systematically analyze the model's mirror symmetry in three directions and the mean and standard deviation of surface curvature, thus completing the comprehensive extraction of eight categories of features from the 3D model.

[0103] Step S308: Construct and normalize feature vectors. For 2D drawings, the system constructs multi-dimensional vectors based on primitive distribution, color distribution, and size scale; for 3D models, the system combines the aforementioned eight categories of features into a high-dimensional feature vector. These feature vectors are then normalized, for example through scale normalization, distance normalization, or logarithmic transformation, to uniformly map all feature values ​​to the range of 0 to 1, ensuring fairness in subsequent calculations. Simultaneously, the system calculates a unique feature hash value based on the normalized feature vectors, used for extremely fast and accurate part comparison.

[0104] Step S310, Similarity Matching and Path Decision. The system traverses the database and calculates the comprehensive similarity score between the current part and each historical part based on the preset weights of different feature categories. The similarity judgment is divided into three levels according to the user-defined threshold standards: when the similarity to a historical part is higher than the first threshold, the system determines it as high similarity; when the similarity is between the first and second thresholds, it is determined as medium similarity; when the similarity is lower than the second threshold, it is determined as no similar historical parts. The system automatically enters different cost calculation paths based on the similarity.

[0105] Step S312: Generating a quote for highly similar parts. Historical quote data for this part can be directly accessed from the database. This involves obtaining its historical process route, processing time for each step, material costs, and total cost. Based on the specific material requirements and production quantity of the current target part, these historical cost data are fine-tuned. For example, material costs are adjusted according to differences in unit material prices, and fixed costs are adjusted according to quantity differences, ultimately generating a reference quote for the current part.

[0106] Step S314: Machine learning-based cost optimization for parts with moderate similarity. Collect cost data, manufacturing process data, and similarity scores between similar historical parts and the current part. Use this data, along with the feature vector of the current part, as input to a pre-trained machine learning model. This model can learn the complex nonlinear relationship between features and cost, and, combined with the actual costs of similar parts, predict and optimize the manufacturing cost of the current part. Add the optimized manufacturing cost predicted by the model to the material cost calculated based on the current part's volume and material density to obtain the final total cost quote.

[0107] Step S316: Standard time quote for parts with no similarity. The blank weight of the part is calculated by multiplying the volume extracted from the 3D model or the bounding box volume calculated from the 2D drawing by the user-specified material density, and then multiplying by the material unit price to obtain the material cost. For processing costs, the system calculates the estimated processing cost based on the part's complexity score, process coefficient, material coefficient, and quantity coefficient, multiplied by a preset time rate. The material cost and processing cost are added together to obtain the initial quote for the new part.

[0108] Step S318: Result Output and Data Storage. The generated cost quotation results are summarized. Regardless of the calculation path used, the results clearly display material costs, processing costs, total costs, and quotation information, along with suggested process routes and estimated working hours. These results are intuitively presented to the user through the user interface and simultaneously saved to the database as new historical data. At the same time, the feature vectors, similarity information, and complete calculation process logs used in this calculation can be recorded for future traceability and to provide new training samples for the continuous optimization of the machine learning model.

[0109] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0110] Based on the same inventive concept, this application also provides a data processing apparatus for implementing the data processing method described above. The solution provided by this apparatus is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more data processing apparatus embodiments provided below can be found in the limitations of the data processing method described above, and will not be repeated here.

[0111] In one embodiment, such as Figure 4 As shown, a data processing device 400 is provided, including: a data acquisition module 401, a feature extraction module 403, a feature comparison module 405, and an information generation module 407, wherein:

[0112] The data acquisition module is used to acquire the structural data of the target object;

[0113] The feature extraction module is used to extract features from the structural data to obtain the feature data of the target object;

[0114] The feature comparison module is used to compare the feature data with historical feature data to determine the similarity between the target object and each historical object;

[0115] The information generation module is used to generate reference resource exchange information of the target object based on the historical resource exchange information of the historical object when there is a historical object with a similarity greater than a first threshold; the resource exchange information is used to characterize the quantity of exchangeable resources of the object.

[0116] The information generation module is also used to input the feature data of the target object into the resource prediction model to obtain the reference resource exchange information of the target object when there is no historical object with a similarity greater than the first threshold.

[0117] In one embodiment, the information generation module is further configured to:

[0118] In the absence of any historical objects with a similarity greater than the first threshold, but with a similarity greater than the second threshold, the feature data of the target object is input into the resource prediction model to obtain the reference resource exchange information of the target object.

[0119] In the absence of historical objects with a similarity greater than the second threshold, the feature data and material identifier of the target object are obtained;

[0120] Based on the feature data of the target object, the complexity of the target object is determined;

[0121] Reference resource exchange information for the target object is generated based on the complexity and material identifier of the target object.

[0122] In one embodiment, the information generation module is further configured to:

[0123] Obtain the material identifier of the target object;

[0124] The feature data and material identifier of the target object are input into the resource prediction model to obtain the first resource exchange information of the target object;

[0125] If the first resource exchange information is valid, the first resource exchange information is determined as the reference resource exchange information;

[0126] If the first resource exchange information is invalid, historical resource exchange information of multiple historical objects with a similarity greater than a second threshold to the target object and the similarity between each historical object and the target object are obtained. Based on the similarity and the corresponding historical resource exchange information, second resource exchange information is obtained. Based on the second resource exchange information and the material identifier, reference resource exchange information of the target object is generated. Wherein, invalid first resource exchange information includes at least one of the following: the resource prediction model does not exist, the amount of historical data input to the resource prediction model does not reach a preset number, or the resource prediction model malfunctions during operation.

[0127] In one embodiment, the feature comparison module is further configured to:

[0128] Based on the feature data, generate the feature vector of the target object;

[0129] The similarity between the target object and each historical object is determined by comparing the feature vector of the target object with the historical feature vectors.

[0130] In one embodiment, the feature data includes structural features; the structural features include at least one of the following: primitive features and geometric features; the feature extraction module is further configured to:

[0131] Traverse the structured data, count the number of each primitive in the target object, and obtain primitive features;

[0132] Geometric data of the target object is obtained from the structural data to obtain geometric features, wherein the geometric features include at least one of total length, total area, and bounding box size.

[0133] In one embodiment, the feature data includes dimensional features; the dimensional features include at least one of the following features: linear features, angular features, and tolerance features; the feature extraction module is further configured to:

[0134] By traversing the structural data, each linear dimension annotation in the target object is identified to obtain linear features;

[0135] By traversing the structural data, the angular dimension annotations in the target object are identified to obtain angular features;

[0136] The structural data is traversed to identify the tolerance markings of the target object, thereby obtaining the tolerance features; the tolerance markings are used to characterize the machining accuracy of the target object.

[0137] In one embodiment, the feature data includes text features; the text features include at least one of the following features: character features and identifier features, and are further used for:

[0138] Traverse the structured data, identify the text units in the target object, obtain the characters of each text unit, and obtain character features;

[0139] Identify the identification data of the target object from the character features to obtain identification features; wherein the identification data includes at least one of the following: part identification, material identification, and version identification.

[0140] In one embodiment, the feature data includes layer features; the layer features include at least one of the following features: layer composition features and color distribution features; the feature extraction module is further configured to:

[0141] Identify the layers in the structural data, count the number of layers and the number of primitives in each layer, and obtain the layer composition features;

[0142] Obtain the colors corresponding to each layer, count the frequency of use of each color, and obtain the color distribution characteristics.

[0143] Each module in the aforementioned data processing device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0144] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5 As shown, this computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores target object data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a data processing method.

[0145] Those skilled in the art will understand that Figure 5The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0146] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0147] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0148] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0149] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A data processing method, characterized in that, The method includes: Obtain the structure data of the target object; Feature extraction is performed on the structural data to obtain the feature data of the target object; The feature data is compared with historical feature data to determine the similarity between the target object and each historical object; In the case of historical objects with a similarity greater than a first threshold, reference resource exchange information of the target object is generated based on the historical resource exchange information of the historical objects; the resource exchange information is used to characterize the quantity of exchangeable resources of the object. In the absence of historical objects with a similarity greater than the first threshold, the feature data of the target object is input into the resource prediction model to obtain the reference resource exchange information of the target object.

2. The method according to claim 1, characterized in that, In the absence of historical objects with a similarity greater than a first threshold, the feature data of the target object is input into the resource prediction model to obtain reference resource exchange information of the target object, including: In the absence of any historical objects with a similarity greater than the first threshold, but with a similarity greater than the second threshold, the feature data of the target object is input into the resource prediction model to obtain the reference resource exchange information of the target object. In the absence of historical objects with a similarity greater than the second threshold, the feature data and material identifier of the target object are obtained; Based on the feature data of the target object, the complexity of the target object is determined; Reference resource exchange information for the target object is generated based on the complexity and material identifier of the target object.

3. The method according to claim 2, characterized in that, The step of inputting the feature data of the target object into the resource prediction model to obtain the reference resource exchange information of the target object includes: Obtain the material identifier of the target object; The feature data and material identifier of the target object are input into the resource prediction model to obtain the first resource exchange information of the target object; If the first resource exchange information is valid, the first resource exchange information is determined as the reference resource exchange information; If the first resource exchange information is invalid, historical resource exchange information of multiple historical objects with a similarity greater than a second threshold to the target object and the similarity between each historical object and the target object are obtained. Based on the similarity and the corresponding historical resource exchange information, second resource exchange information is obtained. Based on the second resource exchange information and the material identifier, reference resource exchange information of the target object is generated. Wherein, invalid first resource exchange information includes at least one of the following: the resource prediction model does not exist, the amount of historical data input to the resource prediction model does not reach a preset number, or the resource prediction model malfunctions during operation.

4. The method according to claim 1, characterized in that, The step of comparing the feature data with historical feature data to determine the similarity between the target object and each historical object includes: Based on the feature data, generate the feature vector of the target object; The similarity between the target object and each historical object is determined by comparing the feature vector of the target object with the historical feature vectors.

5. The method according to claim 4, characterized in that, The feature data includes structural features; the structural features include at least one of the following features: primitive features and geometric features; The acquisition of the primitive features includes: traversing the structural data, counting the number of each primitive in the target object, and obtaining the primitive features; The acquisition of the geometric features includes: obtaining the geometric data of the target object from the structural data to obtain the geometric features, wherein the geometric features include at least one of the following: total length, total area, and bounding box size.

6. The method according to claim 4, characterized in that, The feature data includes dimensional features; the dimensional features include at least one of the following features: linear features, angular features, and tolerance features; The acquisition of the linear features includes traversing the structural data, identifying each linear dimension annotation in the target object, and obtaining the linear features. The acquisition of the angular features includes traversing the structural data, identifying the angular dimension annotations in the target object, and obtaining the angular features; The acquisition of the tolerance features includes traversing the structural data, identifying the tolerance markings of the target object, and obtaining the tolerance features; the tolerance markings are used to characterize the machining accuracy of the target object.

7. The method according to claim 4, characterized in that, The feature data includes text features; the text features include at least one of the following features: character features and identifier features; The acquisition of character features includes: traversing the structured data, identifying text units in the target object, acquiring the characters of each text unit, and obtaining character features; The acquisition of the identification features includes: identifying the identification data of the target object from the character features to obtain the identification features; wherein the identification data includes at least one of the following: part identification, material identification, and version identification.

8. The method according to claim 4, characterized in that, The feature data includes layer features; the layer features include at least one of the following features: layer composition features and color distribution features; The acquisition of the layer composition features includes: identifying the layers in the structural data, counting the number of layers and the number of primitives in each layer, and obtaining the layer composition features; The acquisition of color distribution features includes: acquiring the colors corresponding to each layer, counting the frequency of use of each color, and obtaining color distribution features.

9. A data processing apparatus, characterized in that, The device includes: The data acquisition module is used to acquire the structural data of the target object; The feature extraction module is used to extract features from the structural data to obtain the feature data of the target object; The feature comparison module is used to compare the feature data with historical feature data to determine the similarity between the target object and each historical object; The information generation module is used to generate reference resource exchange information of the target object based on the historical resource exchange information of the historical object when there is a historical object with a similarity greater than a first threshold; the resource exchange information is used to characterize the quantity of exchangeable resources of the object. The information generation module is also used to input the feature data of the target object into the resource prediction model to obtain the reference resource exchange information of the target object when there is no historical object with a similarity greater than the first threshold.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 8.

Images