Intelligent computing cloud platform carries out automobile drawing size marking automatic detection method and device through computing power

Abstract

The application provides a kind of intelligent computing cloud platform and the automatic detection method and device of automobile drawing size marking by computing power, it is related to the technical field of wisdom calculation cloud platform and computing power infrastructure, this method includes: based on at least one of first standard parameter, second standard parameter and third standard parameter, construct sample data set, and sample data set is divided into training set and verification set;Intelligent computing cloud platform utilizes computing power, based on training set, initial model is trained N times, and final model is obtained;Final model is deployed in the intelligent computing cloud platform, to receive the size marking detection of batch automobile drawing.The intelligent computing cloud platform can accurately locate and identify the size marking anomaly in the automobile drawing by computing power for automobile drawing size marking automatic detection, reduces the understanding difference in the process of manual automobile drawing size marking review, improves the detection efficiency and accuracy.

Description

Technical Field

[0001] This invention relates to the field of intelligent computing cloud platform and computing power infrastructure technology, specifically to an automatic detection method and device for automotive drawing dimension annotation using computing power on an intelligent computing cloud platform. Background Technology

[0002] With the rapid development of artificial intelligence technology, the "intelligent computing cloud platform" has emerged.

[0003] "Intelligent computing cloud platform" refers to a cloud computing platform that provides intelligent computing services based on the hardware and software resources of an intelligent computing center.

[0004] "Intelligent computing" refers to computing driven by intelligent computing power (GPU, FPGA, ASIC, etc.), and is a general term for computing theories, methods, architectures, and technical capabilities characterized by data intensity and model-based nature. Its goal is to obtain better results by utilizing specific computing resources and appropriate computing methods according to the needs of the task.

[0005] A "Smart Computing Center" refers to a facility that utilizes large-scale heterogeneous computing resources, including general-purpose computing power (CPU) and intelligent computing power (GPU, FPGA, ASIC, etc.), to provide the necessary computing power, data, and algorithms for artificial intelligence applications (such as the development, training, and inference of deep learning models). A Smart Computing Center encompasses facilities, hardware, and software, and can provide full-stack capabilities from underlying computing power to top-level application enablement.

[0006] "Computing power" is the core of "intelligent computing cloud platform". It is the ability of computer equipment or computing / data center to process information. It is the ability of computer hardware and software to work together to perform a certain computing requirement. It is the computing power to achieve the target output by processing information data. It is a new type of productivity that integrates information computing power, network carrying capacity and data storage capacity. It mainly provides services to society through computing power infrastructure.

[0007] In the production, inspection, and acceptance stages of automobiles, the dimensions on the drawings are essential. The lines and numbers on the drawings are the visual representation of design intent, processing requirements, and assembly standards. Therefore, the standardization and accuracy of the dimensions on automotive drawings directly determine whether the production end can accurately reproduce the automotive design scheme. However, with the industry's digital transformation, facing the archiving needs of automotive companies that often involve tens of thousands of drawings, manual review requires checking the lines and numbers on each drawing one by one, which is time-consuming and labor-intensive, severely slowing down the digital archiving process. Furthermore, different reviewers have different understandings of the dimensions on automotive drawings, making it impossible to form standardized inspection conclusions.

[0008] While some automated detection methods exist in existing technologies, none are specifically designed for the automated detection of dimensions on automotive drawings. Furthermore, it's crucial to emphasize that current automated detection methods are typically based on a simple hardware architecture of a single workstation (CPU + entry-level GPU), generally using traditional image processing libraries like OpenCV (such as Canny / Hough operators). This is vastly different from the application scenarios of automated detection methods in intelligent computing centers. Intelligent computing centers require the integration of kilo-level GPU clusters (such as NVIDIA A100 / H100) and high-speed RDMA networks to achieve EFLOPS-level computing power. This architecture demands 400Gbps inter-node communication in the computing network, 200Gbps bandwidth in the storage network, and overall latency controlled to the sub-microsecond level.

[0009] It is evident that since the emergence of intelligent computing centers, how to use computing power to automatically detect the dimension annotations of automotive drawings in order to improve detection efficiency and accuracy has become an urgent problem to be solved. Summary of the Invention

[0010] This invention provides a method and apparatus for automatically detecting automotive drawing dimension annotations using computing power on an intelligent computing cloud platform, in order to solve the problem of how to improve detection efficiency and accuracy by automatically detecting automotive drawing dimension annotations using computing power since the emergence of intelligent computing centers.

[0011] To solve the above problems, the present invention is implemented as follows:

[0012] In a first aspect, embodiments of the present invention provide a method for automatically detecting the dimension annotations of automotive drawings using a smart computing cloud platform, the method comprising:

[0013] Step S1: Based on at least one of the first standard parameter, the second standard parameter, and the third standard parameter, construct a sample dataset and divide the sample dataset into a training set and a validation set. The first standard parameter is the parameter corresponding to the line segment mark in the dimension annotation of the car drawing under the standard style. The second standard parameter is the parameter corresponding to the number mark in the dimension annotation of the car drawing under the standard style. The third standard parameter is the parameter corresponding to the positional relationship between the line segment mark and the number mark under the standard style.

[0014] Step S2: The intelligent computing cloud platform uses computing power to perform N iterations of training on the initial model based on the training set. The initial model after the Mth iteration is determined as the final model. The loss value of the initial model after the Mth iteration in the validation set is less than the loss value of the initial model after the Lth iteration in the validation set. N is a positive integer, M is a positive integer less than or equal to N, L is a positive integer less than or equal to N, and L is not equal to M. The Lth iteration is any one of the N iterations except the Mth iteration.

[0015] Step S3: Deploy the final model on the intelligent computing cloud platform to perform batch dimension annotation detection on the received automotive drawings.

[0016] In one embodiment, step S1 includes:

[0017] Step S11: Inject the first abnormal feature into the first standard parameter to obtain the first sample dataset. The first sample dataset includes multiple abnormal line segment markers. Each abnormal line segment marker carries a first label. The first label is used to indicate the abnormal type of the abnormal line segment marker. The abnormal type of the abnormal line segment marker includes at least one of the following: missing line segment, misaligned line segment, overlapping line segment, twisted line segment, missing arrow, and reversed arrow direction.

[0018] Step S12: Inject the second abnormal feature into the second standard parameter to obtain the second sample dataset. The second sample dataset includes multiple abnormal digit markers. Each abnormal digit marker carries a second label. The second label is used to indicate the abnormal type of the abnormal digit marker. The abnormal type of the abnormal digit marker includes at least one of digit missing, digit reversed, and digit overlapping.

[0019] Step S13: Inject the third abnormal feature into the third standard parameter to obtain the third sample dataset. The third sample dataset includes multiple abnormal positional relationships. Each abnormal positional relationship carries a third label. The third label is used to indicate the abnormal type of the abnormal positional relationship. The abnormal type of the abnormal positional relationship includes at least one of the following: overlapping of line segment markers and number markers, abnormal relative position of line segment markers and number markers, and abnormal lead line between line segment markers and number markers.

[0020] Step S14: Construct a sample dataset based on the positive sample set and the negative sample set. The negative sample set includes the first sample dataset, the second sample dataset, and the third sample dataset. The positive sample set includes the standard style corresponding to the line segment markers generated according to the first standard parameters, the standard style corresponding to the number markers generated according to the second standard parameters, and the standard style corresponding to the positional relationship generated according to the third standard parameters.

[0021] In one embodiment, step S14 includes:

[0022] Step 141: Perform a first preset processing on each sample data in the positive sample set and each sample data in the negative sample set to obtain the sample dataset. The first preset processing includes at least one of the following: adding scanning noise, blurring, grayscale deviation processing, adjusting line thickness, changing font style, image tilting, and overlaying background elements.

[0023] In one embodiment, step S2 includes:

[0024] Step S21: Select a lightweight deep learning detection network as the initial model. The initial model includes a detection head and a classification head. The detection head is used to output the position information of the dimension annotations in the car drawing. The classification head is used to identify the abnormal information corresponding to the line segment marks and / or number marks in the dimension annotations.

[0025] Step S22: The intelligent computing cloud platform uses computing power and adopts a distributed data parallel framework to perform a second preset processing on a preset number of sample data randomly extracted from the training set to obtain processed sample data. The second preset processing includes at least one of multi-scale enhancement, rotational perturbation and angle regularization.

[0026] Step S23: After inputting the processed sample data into the initial model for feature extraction, the detection head outputs predicted location information, and the classification head outputs predicted anomaly information.

[0027] Step S24: Calculate the total loss value of the initial model based on the first loss value and the second loss value. The first loss value represents the error between the predicted location information and the real location information corresponding to the size label in the processed sample data. The second loss value represents the error between the predicted anomaly information and the real anomaly information corresponding to the size label in the processed sample data. The first loss value is calculated using CIoU Loss, and the second loss value is calculated using Focal Loss.

[0028] Step S25: Based on the total loss value, backpropagation is performed using the chain rule to calculate the gradient of the model parameters corresponding to the initial model, and the model parameters are iterated N times according to the gradient and the learning rate to obtain the initial model after N iterations of training, where N is a preset value. The model parameters include convolution kernel weights, bias terms, and regression parameters of the detection head. In each iteration, the learning rate is reduced according to the cosine annealing algorithm.

[0029] Step S26: If the loss value of the initial model after the Mth iteration of training in the validation set is less than the loss value of the initial model after the Lth iteration of training in the validation set, the initial model after the Mth iteration of training is determined as the final model.

[0030] In one embodiment, step S3 includes:

[0031] Step S31: After converting the final model into an inference format adapted to the intelligent computing cloud platform, deploy it on the GPU inference node of the intelligent computing cloud platform, and encapsulate the final model into a callable API interface. The API interface supports receiving batch automotive drawings and outputting the detection results of the dimension annotation detection.

[0032] In one embodiment, after step S3, the method further includes:

[0033] Step S4: Generate a review report based on the test results. The review report includes location information and anomaly information corresponding to the dimensions of the vehicle drawings, references to standard clauses, and suggested correction schemes. The references to standard clauses are used to indicate the standard clauses corresponding to the anomaly information, and the suggested correction schemes are used to indicate the correction suggestions for the anomaly information.

[0034] Step S5: Based on the review report, the test results are reviewed. If it is determined that there are abnormal results in the test results, the dimension annotations in the car drawings corresponding to the abnormal results are used as new sample data.

[0035] Secondly, embodiments of the present invention provide an automatic detection device for automotive drawing dimension annotation using computing power on an intelligent computing cloud platform, the device comprising:

[0036] A construction module is used to construct a sample dataset based on at least one of a first standard parameter, a second standard parameter, and a third standard parameter, and to divide the sample dataset into a training set and a validation set. The first standard parameter is the parameter corresponding to the line segment mark in the dimension annotation of the car drawing under the standard style. The second standard parameter is the parameter corresponding to the number mark in the dimension annotation of the car drawing under the standard style. The third standard parameter is the parameter corresponding to the positional relationship between the line segment mark and the number mark under the standard style.

[0037] The training module is used by the intelligent computing cloud platform to perform N iterations of training on the initial model based on the training set using computing power. The initial model after the Mth iteration of training is determined as the final model. The loss value of the initial model after the Mth iteration of training in the validation set is less than the loss value of the initial model after the Lth iteration of training in the validation set. N is a positive integer, M is a positive integer less than or equal to N, L is a positive integer less than or equal to N, and L is not equal to M. The Lth iteration of training is any one of the N iterations of training except for the Mth iteration of training.

[0038] The detection module is used to deploy the final model on the intelligent computing cloud platform to perform batch dimension annotation detection on the received automotive drawings.

[0039] Thirdly, embodiments of the present invention provide an electronic device, including: a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the method described in the first aspect.

[0040] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0041] Fifthly, embodiments of the present invention provide a computer program product including computer instructions that, when executed by a processor, implement the steps of the method described in the first aspect.

[0042] In this embodiment of the invention, a sample dataset is constructed based on at least one of a first standard parameter, a second standard parameter, and a third standard parameter. This sample dataset is then divided into a training set and a validation set to address the problem of sample scarcity and lay the foundation for the model to learn complete features. The intelligent computing cloud platform utilizes computing power to accelerate the iterative process of the initial model through data parallelism, improving training efficiency. The final model is deployed on the intelligent computing cloud platform, and the platform dynamically schedules computing power to achieve parallel detection of batch drawings. Thus, based on the features trained in the early stages, the model can accurately locate and identify anomalies in dimension annotations in automotive drawings, reducing misunderstandings during manual review of automotive drawing dimension annotations and improving detection efficiency and accuracy. Attached Figure Description

[0043] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 This is a flowchart of an automatic detection method for automotive drawing dimension annotation using computing power provided by an embodiment of the present invention;

[0045] Figure 2 This is one of the standard styles corresponding to the dimension annotations on automotive drawings provided in this embodiment of the invention;

[0046] Figure 3 This is the second standard style corresponding to the dimension annotation of automobile drawings provided in the embodiments of the present invention;

[0047] Figure 4 This is one of the abnormal types of abnormal line segment marking provided in the embodiments of the present invention;

[0048] Figure 5 This is the second type of anomaly in the abnormal line segment marking provided in this embodiment of the invention;

[0049] Figure 6 This is the third type of anomaly in the abnormal line segment marking provided in this embodiment of the invention;

[0050] Figure 7 This is the fourth type of anomaly marking for abnormal line segments provided in this embodiment of the invention;

[0051] Figure 8 This is the fifth type of anomaly marking for abnormal line segments provided in this embodiment of the invention;

[0052] Figure 9 This is one of the abnormal types of abnormal number markers provided in the embodiments of the present invention;

[0053] Figure 10 This is the second type of anomaly in the abnormal number markers provided in this embodiment of the invention;

[0054] Figure 11 This is the third type of anomaly in the abnormal number markers provided in this embodiment of the invention;

[0055] Figure 12 This is one of the abnormal types of abnormal positional relationships provided in the embodiments of the present invention;

[0056] Figure 13 This is the second type of abnormal positional relationship provided in the embodiments of the present invention;

[0057] Figure 14 This is the third type of abnormal positional relationship provided in the embodiments of the present invention;

[0058] Figure 15 This is a structural diagram of an automatic detection device for marking automotive drawing dimensions using computing power, provided by an embodiment of the present invention;

[0059] Figure 16 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below 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.

[0061] The technical terms involved in this invention will be briefly explained below.

[0062] The “computing power” mentioned in this invention refers to: the ability of computer equipment or computing / data center to process information; the ability of computer hardware and software to work together to perform a certain computing requirement; the computing power to achieve the target result output by processing information data; and a new type of productivity that integrates information computing power, network carrying capacity, and data storage capacity, mainly providing services to society through computing power infrastructure.

[0063] The "computational power" (CP) described in this invention refers to the ability of a data center server to process data and output results. It is a comprehensive indicator of a data center's computing power, encompassing general computing power, supercomputing power, and intelligent computing power. The commonly used unit of measurement is floating-point operations per second (FLOPS, 1 EFLOPS = 10^18 FLOPS), with higher values ​​indicating stronger overall computing power. It is estimated that 1 EFLOPS is approximately the computing power output of 5 Tianhe-2A supercomputers, 500,000 mainstream server CPUs, or 2 million mainstream laptops. The calculation formula is: CP = CP 通用 +CP 智能 +CP 超级 .

[0064] The "Network Power" (NP) mentioned in this invention refers to the performance of data transmission capability of computing facilities, which includes comprehensive capabilities such as network architecture, network bandwidth, transmission latency, intelligent management and scheduling, and involves network transmission within and between data centers. It is a comprehensive indicator for measuring network transmission scheduling capability.

[0065] The "Storage Power" (SP) described in this invention refers to the comprehensive capabilities of a data center in four aspects: data storage capacity, performance, security and reliability, and green and low-carbon operation. It is a comprehensive indicator for measuring the data storage capacity of a data center, including external storage devices such as storage arrays and internal storage devices within servers. The commonly used unit of measurement for storage capacity is exabytes (EB, 1EB = 2^60 bytes), the commonly used unit of measurement for performance is the number of read / write operations per second (IOPS / TB), and the disaster recovery ratio is an important indicator of security and reliability.

[0066] The "computing infrastructure" mentioned in this invention refers to a new type of information infrastructure that integrates information computing power, network carrying capacity, and data storage capacity, enabling centralized computing, storage, transmission, and application of information.

[0067] The "new information infrastructure" mentioned in this invention refers to network infrastructure such as 5G networks, fiber optic broadband networks, backbone networks, international communication networks, and satellite internet; computing infrastructure such as data centers, general computing centers, intelligent computing centers, and supercomputing centers; and new technology facilities such as artificial intelligence, blockchain, and quantum computing.

[0068] The “computing power” mentioned in this invention includes: general computing power, intelligent computing power, and supercomputing power.

[0069] The "general computing power" mentioned in this invention refers to the computing power provided by servers based on CPU (Central Processing Unit) chips, which is used to support basic general computing such as cloud computing and edge computing.

[0070] The "intelligent computing power" mentioned in this invention refers to: a computing platform deployed on a large scale based on dedicated chips such as GPU (Graphics Processing Unit), FPGA (Field Programmable Gate Array), and ASIC (Application Specific Integrated Circuit) for various artificial intelligence innovative applications, such as natural language processing and machine vision.

[0071] The “supercomputing power” mentioned in this invention refers to the computing power provided by high-performance computing clusters such as supercomputers. It utilizes the centralized computing resources of multiple computer systems working in parallel and uses a dedicated operating system to handle extremely complex or data-intensive problems. It is mainly used for computing in cutting-edge scientific fields, such as planetary simulation, drug molecule design, and gene analysis.

[0072] The "intelligent computing" mentioned in this invention refers to computing driven by intelligent computing power (GPU, FPGA, ASIC, etc.), and is a general term for computing theories, methods, architectures and technical capabilities with characteristics such as data intensity and model-based computing. Its goal is to obtain better results by utilizing specific computing resources and appropriate computing methods according to the needs of the task.

[0073] The "intelligent computing cloud platform" mentioned in this invention refers to a cloud computing platform that provides intelligent computing services based on the hardware and software resources of an intelligent computing center.

[0074] The "intelligent computing center" described in this invention refers to a facility that, through the use of large-scale heterogeneous computing resources, including general-purpose computing power (CPU) and intelligent computing power (GPU, FPGA, ASIC, etc.), primarily provides the necessary computing power, data, and algorithms for artificial intelligence applications (such as the development, training, and inference of deep learning models). The intelligent computing center encompasses facilities, hardware, and software, and can provide full-stack capabilities from underlying computing power to top-level application enablement.

[0075] The "intelligent computing center" mentioned in this invention includes, but is not limited to, "smart computing center".

[0076] The "intelligent computing center" mentioned in this invention, also known as an artificial intelligence computing center, is a type of computing infrastructure that provides computing power services, data services, and algorithm services required for artificial intelligence applications, based on artificial intelligence theory and adopting an artificial intelligence computing architecture.

[0077] The "computing center" mentioned in this invention refers to a facility that is mainly composed of infrastructure such as wind, thermal, hydro, and electricity, and IT hardware and software equipment, and has computing power, carrying capacity, and storage capacity, including general data centers, intelligent computing centers, supercomputing centers, etc.

[0078] The "supercomputing center" mentioned in this invention refers to a supercomputing data center, which is a data center based on supercomputers or large-scale computing clusters. It can provide large-scale computing, storage and network services and is widely used in aerospace, defense, oil exploration, climate modeling and genome sequencing and other application scenarios.

[0079] The “computing resources” mentioned in this invention refer to the technologies and facilities required for the development of the digital society that have the ability to compute, transmit, store and apply information, including but not limited to computing resources such as CPUs and GPUs, network resources such as switches and routers, storage resources such as storage arrays and distributed storage, security resources such as firewalls and intrusion detection systems, and supporting and guaranteeing resources such as wind, fire, water and electricity.

[0080] The "computing power service" mentioned in this invention refers to a service that provides users with on-demand computing power through computing resources, such as computing tasks, model inference, and data processing.

[0081] The "model" mentioned in this invention includes, but is not limited to, "large language model" and "multimodal large model".

[0082] The "large language model" (LLM) described in this invention refers to a language model with a large parameter scale, designed to understand and generate human language. It is trained on a large amount of text data and can perform a wide range of tasks, including text summarization, translation, and sentiment analysis.

[0083] The “Multimodal Large Models” mentioned in this invention refer to models that combine multimodal information such as text, images, videos, and audio for training, including but not limited to multimodal large language models.

[0084] The "automotive drawings" mentioned in this invention refer to engineering and technical drawings used throughout the entire process of automotive R&D, production, and testing to convey the design intent, processing requirements, and assembly standards of automobiles and their components. Automotive drawings can include: assembly drawings of the complete vehicle and chassis, assembly drawings, and component drawings. Assembly drawings of the complete vehicle and chassis show the assembly relationships of components in the complete vehicle or chassis, marking the relative positions and connection methods of key components. Assembly drawings (such as engine assembly, transmission assembly, and body frame assembly drawings) mark the assembly dimensions of each sub-component within the assembly. Component drawings (such as drawings of individual parts like bolts, bearings, gears, and body sheet metal parts) mark the detailed dimensions, tolerances, and processing requirements of the parts. Since the dimensional annotations on automotive component drawings are dense and directly affect processing quality, the automatic dimensional annotation detection of this invention primarily targets automotive component drawings.

[0085] The "dimension annotation" described in this invention is the core object for automatic detection of automotive drawings. Dimension annotation includes line segment markers and numerical markers. The standard corresponding to the automatic detection of dimension annotations in automotive drawings is: the national standard GB / T4458.4-2003 Mechanical Drawing Part 4: Dimensioning, or the equivalent international standard ISO 129-1 Technical drawing — Indication of dimensions and tolerances — Part 1: General principles.

[0086] With the digital transformation of the automotive industry, facing a large number of historical drawings, traditional paper drawings are typically scanned to obtain image files (formats such as JPG, PNG, TIFF, etc.) for review and archiving. However, since these historical drawings are not CAD source files, it is impossible to extract line segments and annotation information from the geometric layer, rendering traditional CAD review tools unusable. Therefore, this invention provides a method and apparatus for automatically detecting automotive drawing dimension annotations using a smart computing cloud platform, addressing the challenge of improving detection efficiency and accuracy since the emergence of smart computing centers.

[0087] It is important to emphasize that this application is based on a distributed architecture built on an intelligent computing cloud platform, integrating computing power (e.g., a kilocalorie-level GPU cluster and a high-speed network). Through the deep integration of computing power, algorithms, and data by the intelligent computing cloud platform, a "super intelligent inspection system" specifically tailored for the inspection of automotive drawing dimension annotations has been constructed. Its core advantages are reflected in the breakthroughs in all dimensions of technical architecture, processing power, inspection accuracy, and scenario adaptability.

[0088] The intelligent computing cloud platform, through the collaboration of a cluster of 1000-level NVIDIA A100 / H100 GPUs and a high-speed RDMA network, creates an EFLOPS-level (100 quadrillion calculations per second) computing power pool. The communication rate between computing nodes exceeds 400Gbps, the storage network bandwidth reaches 200Gbps, and the overall latency is compressed to sub-microsecond levels. This distributed architecture completely breaks the hardware constraints of traditional single-machine (CPU + entry-level GPU): traditional methods are limited by the memory and computing power of a single GPU, and can only process up to 4K resolution drawings, with a single drawing processing time exceeding 1 second (parallel processing capability of only 1-2 drawings per second); while the intelligent computing center can easily parse 16K ultra-high-definition drawings (accuracy down to 0.01px) and achieve batch processing of 200+ drawings per second through the parallel computing capabilities of GPU tensor cores, adapting to the mass production and batch inspection needs of the automotive industry.

[0089] The intelligent computing cloud platform utilizes a multimodal fusion model combining computer vision (CV), geometric topology analysis, and a physical rule engine, along with the feature association capabilities of graph neural networks (GNNs), to achieve holographic detection of complex assembly drawings.

[0090] For example, the computer vision module is responsible for accurately extracting pixel-level features (such as line edges and digit contours) of dimensions from images; the geometric topology analysis module constructs a spatial relationship map between parts (such as the topological relationship of an assembly with 2000+ feature points), avoiding related dimension errors caused by isolated detection; and the physical rule engine verifies whether the annotations meet manufacturing feasibility based on geometric tolerance standards (such as GB / T 1182) (such as whether the cumulative tolerance exceeds the assembly threshold). This collaborative mechanism improves detection accuracy, which is completely unattainable by single-machine methods relying solely on traditional OpenCV operators (such as Canny edge detection and Hough line detection).

[0091] Furthermore, the intelligent computing cloud platform possesses dynamic evolution capabilities, completely overcoming the static limitations of traditional methods: through an online learning system, it can automatically identify newly emerging annotation styles (such as designer handwriting annotations or enterprise-specific simplified annotation rules), eliminating the need for manual model retraining; by leveraging generative adversarial networks to generate adversarial examples for extreme scenarios such as uneven lighting and scanning noise, the model's robustness to complex environments is improved by more than 30%; and the real-time feedback mechanism directly connects the detection results with product lifecycle management and manufacturing execution systems. For example, when the cumulative tolerance of 872 positioning holes in a new energy vehicle battery pack exceeds 0.2mm, the system automatically generates a heatmap to mark the risk area and triggers a design change process, achieving a paradigm shift from passive verification after drawings are completed to proactive verification during the design process.

[0092] Please see Figure 1 , Figure 1 This is a flowchart of a method for automatically detecting automotive drawing dimension annotations using computing power on an intelligent computing cloud platform, as provided in an embodiment of the present invention. Figure 1 As shown, it includes the following steps:

[0093] Step S1: Based on at least one of the first standard parameter, the second standard parameter, and the third standard parameter, construct a sample dataset and divide the sample dataset into a training set and a validation set. The first standard parameter is the parameter corresponding to the line segment mark in the dimension annotation of the car drawing under the standard style. The second standard parameter is the parameter corresponding to the number mark in the dimension annotation of the car drawing under the standard style. The third standard parameter is the parameter corresponding to the positional relationship between the line segment mark and the number mark under the standard style.

[0094] In this step, the standard styles corresponding to the three elements—line segment markings, number markings, and positional relationships—can be clearly defined according to GB / T 4458.4-2003 or ISO 129-1. For example, line segment markings and number markings can be styled according to… Figure 2 ( Figure 2(Using the number 20 as an example) Write the numbers in the direction shown in the diagram, and try to avoid marking numbers within the 30° range shown in the diagram; if this cannot be avoided, follow the... Figure 3 ( Figure 3 The dimensions are labeled in the form of the number 16 (for example). Then, quantifiable indicators are extracted from these styles, such as the line width range of line segments, arrow integrity, the letter height ratio and direction angle of numbers, the spacing threshold of positional relationships, and parallelism error, which are used as the first standard parameter, the second standard parameter, and the third standard parameter, respectively, to provide a quantitative basis for constructing the sample data in the sample dataset.

[0095] The first standard parameter extracted based on the standard style of line segment marking can include: dimension line width, dimension line spacing, dimension boundary line extension length, arrow length to line width ratio, arrow integrity, and line segment straightness. For example, in constructing the sample dataset, sample data that does not meet any of the following quantitative indicators corresponding to the first standard parameter can be defined as a counterexample sample corresponding to the line segment marking:

[0096] 1. Dimension line width ≤ 0.25mm ≤ 1mm;

[0097] 2. Spacing between dimension lines ≥ 7mm;

[0098] 3. 2mm ≤ the length of the end of the dimension line ≤ 5mm;

[0099] 4. 3d ≤ arrow length ≤ 5d (d is the dimension line width);

[0100] 5. Double-headed arrow;

[0101] 6. The curvature of the dimension lines / boundaries shall be ≤0.5mm / m.

[0102] Conversely, the positive sample corresponding to the line segment label is the line segment label.

[0103] Sample data that does not meet any of the following first criterion parameters in the quantitative indicators can be defined as counterexamples corresponding to the line segment markers:

[0104] The extracted second standard parameters, based on the standard style of digit markers, may include: character height, character width to character height ratio, digit orientation angle, and digit clarity (pixel level). For example, in constructing the sample dataset, sample data that does not meet any of the following second standard parameters corresponding to the quantification indicators can be defined as counterexamples for the digit markers:

[0105] 1. 3.5mm ≤ character height ≤ 10mm (according to the drawing scale, such as 1:100, then use 5mm);

[0106] 2. The ratio of character width to character height is 0.707±5%;

[0107] 3. The numbers are horizontal: 0° (letters facing up);

[0108] 4. The numbers are oriented vertically at a 90° angle (with the top of the letter facing left).

[0109] 5. The numbers should be tilted: the angle between the tilted number and the dimension line should be ≤30° (avoid inversion);

[0110] 6. The edges of the numbers are not blurry (blur ≤ 2 pixels) and there is no overlap (overlap rate with other numbers / lines ≤ 5%).

[0111] Conversely, the positive sample corresponds to the numerical label.

[0112] The third standard parameter extracted based on the standard pattern of positional relationships may include: spacing between the numeral and dimension line, parallelism between the numeral and dimension line, leader length, number of leader bends, and overlap rate between the numeral and dimension boundary line. For example, in constructing the sample dataset, sample data that does not meet any of the following quantitative indicators corresponding to the third standard parameter can be defined as a counterexample sample corresponding to the positional relationship:

[0113] 1. The distance between numbers and dimension lines should be ≥1mm;

[0114] 2. When tilted, the angle between the number and the dimension line should be ≤3°;

[0115] 3. 5mm ≤ lead wire length ≤ 30mm;

[0116] 4. Number of lead wire bends ≤ 1;

[0117] 5. The overlap rate between numbers and size boundaries (the ratio of the overlapping area between the edge of a number and the boundary to the total area of ​​the number) is ≤5%.

[0118] Conversely, the positive sample corresponds to the positional relationship.

[0119] After constructing a sample dataset including positive and negative samples based on at least one of the first, second, and third standard parameters, the sample data can be divided into a training set for initial model learning and a validation set for evaluating the model's generalization ability during iteration in a 7:3 ratio. Stratified sampling can be used to ensure that the proportion of positive and negative samples and the distribution of error types are consistent between the two sets of samples.

[0120] Furthermore, the intelligent computing cloud platform can call the graphics engine in parallel across multiple nodes to synchronously draw sample data with a style close to real drawings, systematically generating 100,000 samples covering all types of errors. This avoids the scarcity of negative examples, improves the diversity of samples during subsequent model training, and allows the model to learn the characteristics of standard-defined negative examples, thereby improving the comprehensiveness and accuracy of model detection.

[0121] Step S2: The intelligent computing cloud platform uses computing power to perform N iterations of training on the initial model based on the training set. The initial model after the Mth iteration is determined as the final model. The loss value of the initial model after the Mth iteration in the validation set is less than the loss value of the initial model after the Lth iteration in the validation set. N is a positive integer, M is a positive integer less than or equal to N, L is a positive integer less than or equal to N, and L is not equal to M. The Lth iteration is any one of the N iterations except the Mth iteration.

[0122] In this step, a lightweight deep learning detection network (such as YOLOv8n) can be selected as the initial model. Data parallelism is achieved through a smart computing cloud platform using computing power (e.g., multi-GPU clusters, using a distributed training framework such as PyTorchDistributed) to avoid insufficient computing power on a single node. Every K iterations (e.g., K = 5), the performance of the initial model after the Kth iteration is evaluated using a validation set, until a preset N iterations (e.g., N = 100) are completed, at which point iterations stop. Then, the loss values ​​of the N initial models trained after N iterations are compared in the validation set. The initial model trained after the Mth iteration is determined as the final model. The loss value of the initial model trained after the Mth iteration in the validation set is less than the loss value of the initial model trained after the Lth iteration in the validation set; in other words, the initial model trained after the Mth iteration has the smallest loss value and its training effect is optimal. In this way, the initial model trained in the Mth iteration fully learns all types of error features (such as missing arrows, reversed numbers, etc.), and ultimately achieves a minimum loss value on the validation set. This improves the accuracy of detecting dimension annotations on automotive drawings and lays an efficient and accurate model foundation for subsequent batch detection. See the following description for details:

[0123] In one embodiment, step S2 includes:

[0124] Step S21: Select a lightweight deep learning detection network as the initial model. The initial model includes a detection head and a classification head. The detection head is used to output the position information of the dimension annotations in the car drawing. The classification head is used to identify the abnormal information corresponding to the line segment marks and / or number marks in the dimension annotations.

[0125] Step S22: The intelligent computing cloud platform uses computing power and adopts a distributed data parallel framework to perform a second preset processing on a preset number of sample data randomly extracted from the training set to obtain processed sample data. The second preset processing includes at least one of multi-scale enhancement, rotational perturbation and angle regularization.

[0126] Step S23: After inputting the processed sample data into the initial model for feature extraction, the detection head outputs predicted location information, and the classification head outputs predicted anomaly information.

[0127] Step S24: Calculate the total loss value of the initial model based on the first loss value and the second loss value. The first loss value represents the error between the predicted location information and the real location information corresponding to the size label in the processed sample data. The second loss value represents the error between the predicted anomaly information and the real anomaly information corresponding to the size label in the processed sample data. The first loss value is calculated using CIoU Loss, and the second loss value is calculated using Focal Loss.

[0128] Step S25: Based on the total loss value, backpropagation is performed using the chain rule to calculate the gradient of the model parameters corresponding to the initial model, and the model parameters are iterated N times according to the gradient and the learning rate to obtain the initial model after N iterations of training, where N is a preset value. The model parameters include convolution kernel weights, bias terms, and regression parameters of the detection head. In each iteration, the learning rate is reduced according to the cosine annealing algorithm.

[0129] Step S26: If the loss value of the initial model after the Mth iteration of training in the validation set is less than the loss value of the initial model after the Lth iteration of training in the validation set, the initial model after the Mth iteration of training is determined as the final model.

[0130] In this embodiment, lightweight deep learning detection networks such as YOLOv8n or PP-YOLOE-tiny are selected as the initial model, balancing accuracy and computational efficiency, and adapting to the training and subsequent deployment requirements of cloud platforms. The detection head in the initial model is responsible for locating the specific positions of dimension annotations on the drawing, while the classification head is responsible for identifying anomalies in line segment and number markers within the annotations. This dual-branch model structure clearly defines the goals of subsequent model training. Then, a predetermined number of samples are randomly selected from the training set, and a second set of pre-defined processing techniques, such as multi-scale enhancement, rotational perturbation, and angle regularization, are applied to generate diverse samples adapted to the model input. This optimizes the original sample data and provides a foundation for the model to learn complex scene features. After the second-pre-defined sample data is input into the initial model, the feature extraction network extracts key features of the annotations (such as line segment edges and number contours). The detection head then outputs the predicted position information of the annotation boxes, and the classification head outputs the predicted results of anomaly types (such as the probability of reversed numbers), completing the model inference process.

[0131] To quantify the difference between the predicted results and the true labels, the loss is further calculated based on the inference output: CIoU Loss is used to calculate the first loss value to measure the deviation between the predicted and true locations, including position, scale, and angle errors; Focal Loss is used to calculate the second loss value to measure the difference between the predicted and true anomalies, focusing on optimizing hard example identification; and the two are weighted and summed to obtain the total loss, thus quantifying the model's prediction accuracy. Then, parameter optimization is initiated based on the total loss: the gradients of model parameters (e.g., convolutional kernel weights, bias terms, detector regression parameters, etc.) can be calculated through backpropagation using the chain rule, combined with a cosine annealing strategy to dynamically adjust the learning rate, which decays with the cosine annealing algorithm in each iteration to avoid parameter oscillation. This allows for multiple iterations of model parameter updates, gradually bringing the model's prediction results closer to the true labels. The model updated in each iteration can then be applied to the validation set for inference, and the validation set loss value can be calculated. If the loss value of the initial model after the Mth iteration of training is less than the loss value of the initial model after the Lth iteration of training in the same validation set, then the initial model after the Mth iteration of training is the optimal model among the N initial models trained in the same validation set. Therefore, the initial model after the Mth iteration of training can be determined as the final model, training is completed, and a model usable for actual detection is obtained. This ultimately achieves iterative evolution from the initial model to a high-precision final model.

[0132] Step S3: Deploy the final model on the intelligent computing cloud platform to perform batch dimension annotation detection on the received automotive drawings.

[0133] In this step, the final model is deployed on an intelligent computing cloud platform, enabling image-level specification inspection without CAD dependency. This allows for batch dimension annotation inspection of received automotive drawings (e.g., scanned images or PDF files), determining compliance with drafting standards, and automatically identifying non-standard dimension annotations. It boasts high accuracy, high versatility, and high robustness, showing broad application prospects in industrial design review, engineering archiving verification, and automated drafting quality control, completely eliminating reliance on vector layer data. Simultaneously, it avoids discrepancies in the understanding of dimension annotations on automotive drawings among different reviewers, improving inspection efficiency and accuracy.

[0134] Specifically, in one embodiment, step S3 includes:

[0135] Step S31: After converting the final model into an inference format adapted to the intelligent computing cloud platform, deploy it on the GPU inference node of the intelligent computing cloud platform, and encapsulate the final model into a callable API interface. The API interface supports receiving batch automotive drawings and outputting the detection results of the dimension annotation detection.

[0136] In this embodiment, to achieve efficient inference environment adaptation of the final model to the cloud platform, model format conversion and service encapsulation can be performed. Model format conversion converts the model parameters saved during the training phase into a cloud platform-friendly inference format (such as ONNX or TensorRT format), reduces computational load through INT8 quantization compression, and ensures that accuracy loss is controlled within 1% to improve inference speed. Service encapsulation integrates the converted model into the cloud platform's inference framework, encapsulating it as an API interface that can be called via HTTP or HTTPS, and explicitly defining the input of the interface as batch automotive drawing files and the output as structured detection results.

[0137] Furthermore, the intelligent computing cloud platform can dynamically allocate GPU inference nodes based on the batch size of received automotive drawings (e.g., 10,000 drawings) using cluster management tools (such as Kubernetes). For example, one NVIDIA T4 / V100 node can be allocated for every 500 drawings, and each node can deploy multiple model instances. These model instances can be parallel replicas based on the same final model, isolated by Docker containers to achieve high-concurrency processing. The API interface can receive batches of automotive drawings uploaded by users, supporting formats including scanned images (JPG / PNG / TIFF, etc.) and PDF image pages. It automatically extracts image layers from the automotive drawings and then uniformly converts them to grayscale images to reduce the number of channels, lower computational load, and standardize the resolution (e.g., 300 dpi to ensure clear annotation details).

[0138] For example, given the large size of automotive drawings (e.g., A3 drawings with 4961×7016 pixels) and dense annotations, the drawings can be cropped into 640×640 pixel blocks. Simultaneously, the coordinate offset of each sub-image relative to the original image is obtained, and minor enhancements (such as contrast adjustment and noise reduction filtering) are applied to eliminate the effects of scanning noise or uneven lighting. Then, the images are normalized using the mean and variance from model training. The preprocessed sub-images are then batch-input into model instances on each GPU node, and inference is performed in parallel: the detection head outputs the predicted coordinates of the bounding boxes (x1, y1, x2, y2 relative to the sub-image); while the classification head outputs the probability distribution of error types, such as a missing arrow probability of 0.95 and a reversed digit probability of 0.03. The category with the highest confidence is then selected as the final anomaly type. If the confidence threshold is set to 0.5, a manual review mechanism can be triggered if the probability distribution falls below the threshold, improving the accuracy of the detection results.

[0139] In one embodiment, after step S3, the method further includes:

[0140] Step S4: Generate a review report based on the test results. The review report includes location information and anomaly information corresponding to the dimensions of the vehicle drawings, references to standard clauses, and suggested correction schemes. The references to standard clauses are used to indicate the standard clauses corresponding to the anomaly information, and the suggested correction schemes are used to indicate the correction suggestions for the anomaly information.

[0141] Step S5: Based on the review report, the test results are reviewed. If it is determined that there are abnormal results in the test results, the dimension annotations in the car drawings corresponding to the abnormal results are used as new sample data.

[0142] In this embodiment, a review report containing complete information is generated based on batch inspection results, transforming the raw inspection data output by the model into a decision-making basis that auditors can directly use. The review report clearly identifies the specific location and type of the anomaly in the automotive drawings, ensuring accurate positioning. For each anomaly type, the standard clause reference automatically associates it with the corresponding drafting standard clause. For example, a missing arrow can be associated with clause 5.2.2 of GB / T 4458.4-2003, which states that the terminal of a dimension line should be a double-headed arrow, explaining why the annotation is in violation and enhancing the authority and traceability of the inspection results. Suggested corrections provide specific modification guidance based on standard requirements. For example, for a missing arrow, it is suggested to add a right-hand arrow; for reversed numbers, it is suggested to rotate the numbers 180° to ensure the top of the text is facing upwards, allowing auditors to make corrections directly without consulting the standard. This meets the visualization needs of the review process and facilitates integration with the enterprise's drawing management system, enabling rapid implementation of inspection results.

[0143] Furthermore, manual review is used to address potential misjudgments and missed detections in the model, and newly discovered anomalies are converted into samples to continuously improve model performance. The specific process is as follows:

[0144] First, professional drafters manually verify the generated review report to confirm whether the model has made any misjudgments, such as classifying compliant annotations as abnormal; and to confirm whether there are any missed detections, such as the model failing to identify genuine anomalies. If an anomaly is found during the review, the automotive drawing fragment containing the anomaly is cropped into a new sample, its true location and anomaly type are labeled, and it is added to the original sample set as new sample data.

[0145] After cleaning, the newly added samples can be used for subsequent model retraining to specifically optimize the model's ability to identify edge cases. This method of supplementing samples through human feedback compensates for the model's limitations in complex scenarios, enabling the cloud platform to continuously evolve with real-world applications and ultimately achieve long-term improvements in detection accuracy.

[0146] In this embodiment of the invention, a sample dataset is constructed based on at least one of a first standard parameter, a second standard parameter, and a third standard parameter. This sample dataset is then divided into a training set and a validation set to address the problem of sample scarcity and lay the foundation for the model to learn complete features. The intelligent computing cloud platform utilizes computing power to accelerate the iterative process of the initial model through data parallelism, improving training efficiency. The final model is deployed on the intelligent computing cloud platform, and the platform dynamically schedules computing power to achieve parallel detection of batch drawings. Thus, based on the features trained in the early stages, the model can accurately locate and identify anomalies in dimension annotations in automotive drawings, reducing misunderstandings during manual review of automotive drawing dimension annotations and improving detection efficiency and accuracy.

[0147] In one embodiment, step S1 includes:

[0148] Step S11: Inject the first abnormal feature into the first standard parameter to obtain the first sample dataset. The first sample dataset includes multiple abnormal line segment markers. Each abnormal line segment marker carries a first label. The first label is used to indicate the abnormal type of the abnormal line segment marker. The abnormal type of the abnormal line segment marker includes at least one of the following: missing line segment, misaligned line segment, overlapping line segment, twisted line segment, missing arrow, and reversed arrow direction.

[0149] Step S12: Inject the second abnormal feature into the second standard parameter to obtain the second sample dataset. The second sample dataset includes multiple abnormal digit markers. Each abnormal digit marker carries a second label. The second label is used to indicate the abnormal type of the abnormal digit marker. The abnormal type of the abnormal digit marker includes at least one of digit missing, digit reversed, and digit overlapping.

[0150] Step S13: Inject the third abnormal feature into the third standard parameter to obtain the third sample dataset. The third sample dataset includes multiple abnormal positional relationships. Each abnormal positional relationship carries a third label. The third label is used to indicate the abnormal type of the abnormal positional relationship. The abnormal type of the abnormal positional relationship includes at least one of the following: overlapping of line segment markers and number markers, abnormal relative position of line segment markers and number markers, and abnormal lead line between line segment markers and number markers.

[0151] Step S14: Construct a sample dataset based on the positive sample set and the negative sample set. The negative sample set includes the first sample dataset, the second sample dataset, and the third sample dataset. The positive sample set includes the standard style corresponding to the line segment markers generated according to the first standard parameters, the standard style corresponding to the number markers generated according to the second standard parameters, and the standard style corresponding to the positional relationship generated according to the third standard parameters.

[0152] In this embodiment, after extracting the first standard parameters based on the standard style of line segment marking, first abnormal features are injected to generate non-standard negative samples (i.e., samples that do not meet the quantitative indicators corresponding to the first standard parameters, i.e., abnormal line segment markings). Simultaneously, standard positive samples are generated based on the first standard parameters, together forming the first sample dataset. The number of negative samples is greater than the number of positive samples, covering all possible violation scenarios for line segment marking. For example, the abnormal type of abnormal line segment marking may include: missing line segments (e.g.,...). Figure 4 As shown, Figure 4 (Taking the numbers 1234 as an example), line segment misalignment (such as...) Figure 5 As shown, Figure 5 (Taking the numbers 1234 as an example), overlapping line segments (such as...) Figure 6 As shown, Figure 6 (Examples include numbers 1234 and 1235) and missing arrows (e.g.) Figure 7 As shown, Figure 7 (Taking the numbers 1234 as an example), the arrow direction is reversed (e.g.) Figure 8 As shown, Figure 8 The algorithm uses at least one of the following: numerical markers (e.g., 1234) and line segment distortion. A first label is added to each abnormal line segment marker to clearly indicate the abnormal type, serving as a supervision signal during model training. This facilitates the model's learning of the standard and abnormal features of line segment markers in automotive drawings, improving the model's accuracy in identifying abnormal line segment markers.

[0153] After extracting the second standard parameters from the standard pattern of the digit markers, second abnormal features are injected to generate non-standard negative samples (i.e., samples that do not meet the quantitative indicators corresponding to the second standard parameters, i.e., abnormal digit markers). Simultaneously, standard positive samples are generated based on the second standard parameters, together forming the second sample dataset. The number of negative samples is greater than the number of positive samples, covering all possible violation scenarios for digit markers. For example, the abnormal types of abnormal digit markers may include: missing digits (e.g.,...). Figure 9 As shown, Figure 9 (Taking the numbers 1234 as an example) and reversing the numbers (e.g.) Figure 10 As shown, Figure 10 Taking the numbers 1234 as an example), overlapping numbers (such as...) Figure 11 As shown, Figure 11 (Taking two numerical markers, 1234, as an example, at least one of them is used.) A second label is added to each abnormal numerical marker to clearly indicate the abnormal type corresponding to each abnormal numerical marker, serving as a supervision signal during model training. This facilitates the model learning the standard and abnormal features of numerical markers in automotive drawings during training, thereby improving the accuracy of the model in identifying abnormal numerical markers.

[0154] After extracting the third standard parameter based on the standard pattern of the positional relationship between line segment markers and number markers, third abnormal features are injected to generate non-standard negative samples (i.e., samples that do not meet the quantitative indicators corresponding to the third standard parameters, i.e., abnormal positional relationships). Simultaneously, standard positive samples are generated based on the third standard parameters, together forming the third sample dataset. The number of negative samples is greater than the number of positive samples, covering all possible violation scenarios of positional relationships. For example, the abnormal types of abnormal positional relationships can include: overlap between line segment markers and number markers (e.g., ...). Figure 12 As shown, Figure 12 (Taking the number markers 1234 as an example), the relative positions of the line segment markers and the number markers are abnormal (e.g.) Figure 13 As shown, Figure 13 (Taking the number markings 1234 as an example), abnormal lead lines between line segment markings and number markings (such as...) Figure 14 As shown, Figure 14 At least one of the following (taking the numerical labels 1234 as an example). A third label is added to each abnormal positional relationship to clearly indicate the abnormal type corresponding to each abnormal positional relationship, serving as a supervision signal during model training. This facilitates the model learning the standard and abnormal features of the positional relationships between line segment labels and numerical labels in automotive drawings during training, improving the accuracy of the model in identifying abnormal positional relationships.

[0155] After generating the first, second, and third sample datasets, positive and negative samples can be mixed in a preset ratio (e.g., 3:1) to ensure a balanced distribution of each category in the sample set. The final sample dataset contains both references to standard styles and cases covering various anomaly types, solving the problem of scarce and incomplete error samples in real automotive drawings. This ensures that subsequent model training can learn multiple possible annotation violation features, laying a data foundation for improving detection accuracy.

[0156] In one embodiment, step S14 includes:

[0157] Step 141: Perform a first preset processing on each sample data in the positive sample set and each sample data in the negative sample set to obtain the sample dataset. The first preset processing includes at least one of the following: adding scanning noise, blurring, grayscale deviation processing, adjusting line thickness, changing font style, image tilting, and overlaying background elements.

[0158] In this embodiment, the first preset processing includes a variety of operations that simulate the characteristics of real drawings. Each processing is designed for the non-ideal states that may occur in the actual circulation of automotive drawings (such as scanning, copying, and archiving). In this way, by simulating various interference factors of real automotive drawings, the generated samples are made closer to the actual application scenario, and the model is prevented from failing in real detection. Considering that real drawing scans often produce noise due to paper texture and scanner precision, the following measures are taken: Adding scanning noise to inject randomly distributed granular noise or spots (such as Gaussian noise or salt-and-pepper noise) into the sample image simulates the rough texture of scanned paper drawings; using Gaussian blur or motion blur simulates blurry annotations caused by inaccurate focusing, paper wear (such as creases), or low-resolution scanning; adjusting the grayscale distribution of the sample (such as local bright / dark areas) simulates the differences in brightness caused by uneven lighting during scanning; randomly fine-tuning the width of line segment markers within the standard line width range (0.25mm - 1mm) simulates the changes in line thickness caused by differences in CAD software pen width settings or scanning scaling; using a slightly modified version of the Song typeface to simulate the different drawing habits or font library differences of different designers; randomly rotating the sample image by ±15° simulates the skewed placement of the drawing during scanning; and adding simple part outlines, auxiliary lines, or annotation text (not dimensioning content) around the annotations simulates the complex background where annotations and part graphics coexist in real automotive drawings.

[0159] Through the above processing, the sample data automatically synthesized based on standards is transformed from ideally pure labeled data into scenario-based samples containing real interference. The final sample dataset can more comprehensively cover the interference that may be encountered in actual detection. This ensures that the model not only learns the compliance rules of the annotation during training, but also adapts to interference factors in real-world scenarios, thereby improving the robustness of subsequent detections and avoiding detection failures caused by the disconnect between samples and real-world scenarios.

[0160] Please see Figure 15 , Figure 15 This is a structural diagram of an automatic detection device for annotating automotive drawing dimensions using computing power, provided by an embodiment of the present invention. Figure 15 As shown, the intelligent computing cloud platform utilizes computing power to automatically detect the dimension annotation of automotive drawings using an automatic detection device 1500, which includes:

[0161] The construction module 1501 is used to construct a sample dataset based on at least one of a first standard parameter, a second standard parameter, and a third standard parameter, and to divide the sample dataset into a training set and a validation set. The first standard parameter is the parameter corresponding to the line segment mark in the dimension annotation of the car drawing under the standard style. The second standard parameter is the parameter corresponding to the number mark in the dimension annotation of the car drawing under the standard style. The third standard parameter is the parameter corresponding to the positional relationship between the line segment mark and the number mark under the standard style.

[0162] Training module 1502 is used by the intelligent computing cloud platform to perform N iterations of training on the initial model based on the training set using computing power, and to determine the initial model after the Mth iteration of training as the final model. The loss value of the initial model after the Mth iteration of training in the validation set is less than the loss value of the initial model after the Lth iteration of training in the validation set. N is a positive integer, M is a positive integer less than or equal to N, L is a positive integer less than or equal to N, and L is not equal to M. The Lth iteration of training is any one of the N iterations of training except for the Mth iteration of training.

[0163] The detection module 1503 is used to deploy the final model on the intelligent computing cloud platform to perform batch dimension annotation detection on the received automotive drawings.

[0164] In one embodiment, the construction module 1501 is specifically used for:

[0165] Injecting a first abnormal feature into the first standard parameter yields a first sample dataset. The first sample dataset includes multiple abnormal line segment markers, each carrying a first label. The first label indicates the abnormal type of the abnormal line segment marker. The abnormal type of the abnormal line segment marker includes at least one of the following: missing line segment, misaligned line segment, overlapping line segment, twisted line segment, missing arrow, and reversed arrow direction.

[0166] A second abnormal feature is injected into the second standard parameter to obtain a second sample dataset. The second sample dataset includes multiple abnormal digit markers, each of which carries a second label. The second label is used to indicate the abnormal type of the abnormal digit marker. The abnormal type of the abnormal digit marker includes at least one of digit missing, digit reversed, and digit overlapping.

[0167] A third abnormal feature is injected into the third standard parameter to obtain a third sample dataset. The third sample dataset includes multiple abnormal positional relationships. Each abnormal positional relationship carries a third label. The third label is used to indicate the abnormal type of the abnormal positional relationship. The abnormal type of the abnormal positional relationship includes at least one of the following: overlapping of line segment markers and number markers, abnormal relative position of line segment markers and number markers, and abnormal lead line between line segment markers and number markers.

[0168] A sample dataset is constructed based on the positive example set and the negative example set. The negative example set includes the first sample dataset, the second sample dataset, and the third sample dataset. The positive example set includes the standard style corresponding to the line segment markers generated according to the first standard parameter, the standard style corresponding to the number markers generated according to the second standard parameter, and the standard style corresponding to the positional relationship generated according to the third standard parameter.

[0169] In one embodiment, constructing the sample dataset based on the positive example set and the negative example set includes:

[0170] The sample dataset is obtained by performing a first preset processing on each sample data in the positive sample set and each sample data in the negative sample set. The first preset processing includes at least one of the following: adding scanning noise, blurring, grayscale deviation processing, adjusting line thickness, changing font style, image tilting, and overlaying background elements.

[0171] In one embodiment, the training module 1502 is specifically used for:

[0172] A lightweight deep learning detection network is selected as the initial model. The initial model includes a detection head and a classification head. The detection head is used to output the position information of the dimension annotations in the automotive drawings. The classification head is used to identify the abnormal information corresponding to the line segment marks and / or number marks in the dimension annotations.

[0173] The intelligent computing cloud platform utilizes computing power and adopts a distributed data parallel framework to perform a second preset processing on a preset number of sample data randomly extracted from the training set to obtain processed sample data. The second preset processing includes at least one of multi-scale enhancement, rotational perturbation, and angle regularization.

[0174] After the processed sample data is input into the initial model for feature extraction, the predicted location information is output through the detection head, and the predicted anomaly information is output through the classification head.

[0175] The total loss value of the initial model is calculated based on the first loss value and the second loss value. The first loss value represents the error between the predicted location information and the true location information corresponding to the size label in the processed sample data. The second loss value represents the error between the predicted anomaly information and the true anomaly information corresponding to the size label in the processed sample data. The first loss value is calculated using CIoU Loss, and the second loss value is calculated using FocalLoss.

[0176] Based on the total loss value, the gradient of the model parameters corresponding to the initial model is calculated by backpropagation using the chain rule, and the model parameters are iterated N times according to the gradient and the learning rate to obtain the initial model after N iterations of training, where N is a preset value. The model parameters include convolution kernel weights, bias terms, and regression parameters of the detection head. In each iteration, the learning rate is reduced according to the cosine annealing algorithm.

[0177] If the loss value of the initial model after the Mth iteration of training is less than the loss value of the initial model after the Lth iteration of training in the same validation set, the initial model after the Mth iteration of training is determined as the final model.

[0178] In one embodiment, the detection module 1503 is specifically used for:

[0179] Step S31: After converting the final model into an inference format adapted to the intelligent computing cloud platform, deploy it on the GPU inference node of the intelligent computing cloud platform, and encapsulate the final model into a callable API interface. The API interface supports receiving batch automotive drawings and outputting the detection results of the dimension annotation detection.

[0180] In one embodiment, the device further includes:

[0181] The generation module is used to generate a review report based on the detection results. The review report includes location information and anomaly information corresponding to the dimensions of the vehicle drawing, reference to standard clauses, and suggested correction schemes. The reference to standard clauses is used to indicate the standard clauses corresponding to the anomaly information, and the suggested correction schemes are used to indicate the correction suggestions for the anomaly information.

[0182] The review module is used to review the test results based on the review report. If it is determined that there are abnormal results in the test results, the dimension annotations in the car drawings corresponding to the abnormal results are used as new sample data.

[0183] The intelligent computing cloud platform automatic detection device for automotive drawing dimension annotation provided in this embodiment of the invention can realize the various processes of the above-mentioned intelligent computing cloud platform automatic detection method for automotive drawing dimension annotation through computing power. The technical features are one-to-one and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0184] It should be noted that the intelligent computing cloud platform in this embodiment of the invention can be a device, or it can be a component, integrated circuit, or chip in an electronic device.

[0185] This invention also provides an electronic device, see [link to relevant documentation]. Figure 16 , Figure 16 This is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. The electronic device includes a memory 1601, a processor 1602, and a program or instructions stored in the memory 1601 that run on the processor. When the program or instructions are executed by the processor 1602, they can achieve the following: Figure 1 The corresponding intelligent computing cloud platform achieves the same beneficial effects by performing any step in the embodiment of the automatic detection method for dimension annotation of automotive drawings through computing power, and will not be elaborated here.

[0186] The processor 1602 can be a CPU, ASIC, FPGA or GPU.

[0187] Those skilled in the art will understand that all or part of the steps of the above-described intelligent computing cloud platform for automatically detecting the dimension annotation of automotive drawings using computing power can be accomplished by hardware related to program instructions, and the program can be stored in a readable medium.

[0188] This invention also provides a readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described functions. Figure 1 The corresponding intelligent computing cloud platform can perform any step in the automatic detection method embodiment for automotive drawing dimension annotation using computing power, and can achieve the same technical effect. To avoid repetition, it will not be described again here. The storage medium mentioned is such as read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, etc.

[0189] The present invention also provides a computer program product, including computer instructions that, when executed by a processor, implement the above-described... Figure 1 The corresponding intelligent computing cloud platform performs each process of the automatic detection method for dimension annotation of automotive drawings through computing power, and can achieve the same technical effect. To avoid repetition, it will not be described in detail here.

[0190] In the embodiments of this invention, the terms "first," "second," etc., are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products, or devices. Additionally, the use of "and / or" in this invention indicates at least one of the connected objects, such as A and / or B and / or C, representing seven possibilities: A alone, B alone, C alone, both A and B present, both B and C present, both A and C present, and A, B, and C present.

[0191] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0192] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or second terminal device, etc.) to execute the methods of the various embodiments of the present invention.

[0193] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other modifications under the guidance of the present invention without departing from the spirit and scope of the claims, and all of these modifications are within the protection scope of the present invention.

Claims

1. A method for automatically detecting dimensional annotations on automotive drawings using a smart computing cloud platform, characterized in that: The method includes: Step S1: Based on at least one of the first standard parameter, the second standard parameter, and the third standard parameter, construct a sample dataset and divide the sample dataset into a training set and a validation set. The first standard parameter is the parameter corresponding to the line segment mark in the dimension annotation of the car drawing under the standard style. The second standard parameter is the parameter corresponding to the number mark in the dimension annotation of the car drawing under the standard style. The third standard parameter is the parameter corresponding to the positional relationship between the line segment mark and the number mark under the standard style. Step S2: The intelligent computing cloud platform uses computing power to perform N iterations of training on the initial model based on the training set. The initial model after the Mth iteration is determined as the final model. The loss value of the initial model after the Mth iteration in the validation set is less than the loss value of the initial model after the Lth iteration in the validation set. N is a positive integer, M is a positive integer less than or equal to N, L is a positive integer less than or equal to N, and L is not equal to M. The Lth iteration is any one of the N iterations except the Mth iteration. Step S3: Deploy the final model on the intelligent computing cloud platform to perform batch dimension annotation detection on the received automotive drawings; Step S1 includes: Step S11: Inject the first abnormal feature into the first standard parameter to obtain the first sample dataset. The first sample dataset includes multiple abnormal line segment markers. Each abnormal line segment marker carries a first label. The first label is used to indicate the abnormal type of the abnormal line segment marker. The abnormal type of the abnormal line segment marker includes at least one of the following: missing line segment, misaligned line segment, overlapping line segment, twisted line segment, missing arrow, and reversed arrow direction. Step S12: Inject the second abnormal feature into the second standard parameter to obtain the second sample dataset. The second sample dataset includes multiple abnormal digit markers. Each abnormal digit marker carries a second label. The second label is used to indicate the abnormal type of the abnormal digit marker. The abnormal type of the abnormal digit marker includes at least one of digit missing, digit reversed, and digit overlapping. Step S13: Inject the third abnormal feature into the third standard parameter to obtain the third sample dataset. The third sample dataset includes multiple abnormal positional relationships. Each abnormal positional relationship carries a third label. The third label is used to indicate the abnormal type of the abnormal positional relationship. The abnormal type of the abnormal positional relationship includes at least one of the following: overlapping of line segment markers and number markers, abnormal relative position of line segment markers and number markers, and abnormal lead line between line segment markers and number markers. Step S14: Construct a sample dataset based on the positive sample set and the negative sample set. The negative sample set includes the first sample dataset, the second sample dataset, and the third sample dataset. The positive sample set includes the standard style corresponding to the line segment markers generated according to the first standard parameters, the standard style corresponding to the number markers generated according to the second standard parameters, and the standard style corresponding to the positional relationship generated according to the third standard parameters.

2. The method as described in claim 1, characterized in that, Step S14 includes: Step 141: Perform a first preset processing on each sample data in the positive sample set and each sample data in the negative sample set to obtain the sample dataset. The first preset processing includes at least one of adjusting line thickness, changing font style, image tilting, and overlaying background elements.

3. The method as described in claim 1, characterized in that, Step S2 includes: Step S21: Select a lightweight deep learning detection network as the initial model. The initial model includes a detection head and a classification head. The detection head is used to output the position information of the dimension annotations in the car drawing. The classification head is used to identify the abnormal information corresponding to the line segment marks and / or number marks in the dimension annotations. Step S22: The intelligent computing cloud platform uses computing power and adopts a distributed data parallel framework to perform a second preset processing on a preset number of sample data randomly extracted from the training set to obtain processed sample data. The second preset processing includes at least one of multi-scale enhancement, rotational perturbation and angle regularization. Step S23: After inputting the processed sample data into the initial model for feature extraction, the detection head outputs predicted location information, and the classification head outputs predicted anomaly information. Step S24: Calculate the total loss value of the initial model based on the first loss value and the second loss value. The first loss value represents the error between the predicted location information and the real location information corresponding to the size label in the processed sample data. The second loss value represents the error between the predicted anomaly information and the real anomaly information corresponding to the size label in the processed sample data. The first loss value is calculated using CIoU Loss, and the second loss value is calculated using Focal Loss. Step S25: Based on the total loss value, backpropagation is performed using the chain rule to calculate the gradient of the model parameters corresponding to the initial model, and the model parameters are iterated N times according to the gradient and the learning rate to obtain the initial model after N iterations of training, where N is a preset value. The model parameters include convolution kernel weights, bias terms, and regression parameters of the detection head. In each iteration, the learning rate is reduced according to the cosine annealing algorithm. Step S26: If the loss value of the initial model after the Mth iteration of training in the validation set is less than the loss value of the initial model after the Lth iteration of training in the validation set, the initial model after the Mth iteration of training is determined as the final model.

4. The method according to any one of claims 1 to 3, characterized in that, Step S3 includes: Step S31: After converting the final model into an inference format adapted to the intelligent computing cloud platform, deploy it on the GPU inference node of the intelligent computing cloud platform, and encapsulate the final model into a callable API interface. The API interface supports receiving batch automotive drawings and outputting the detection results of the dimension annotation detection.

5. The method as described in claim 4, characterized in that, After step S3, the method further includes: Step S4: Generate a review report based on the test results. The review report includes location information and anomaly information corresponding to the dimensions of the vehicle drawings, references to standard clauses, and suggested correction schemes. The references to standard clauses are used to indicate the standard clauses corresponding to the anomaly information, and the suggested correction schemes are used to indicate the correction suggestions for the anomaly information. Step S5: Based on the review report, the test results are reviewed. If it is determined that there are abnormal results in the test results, the dimension annotations in the car drawings corresponding to the abnormal results are used as new sample data.

6. An automatic detection device for dimension annotation of automotive drawings using intelligent computing cloud platform, characterized in that, The device includes: A construction module is used to construct a sample dataset based on at least one of a first standard parameter, a second standard parameter, and a third standard parameter, and to divide the sample dataset into a training set and a validation set. The first standard parameter is the parameter corresponding to the line segment mark in the dimension annotation of the car drawing under the standard style. The second standard parameter is the parameter corresponding to the number mark in the dimension annotation of the car drawing under the standard style. The third standard parameter is the parameter corresponding to the positional relationship between the line segment mark and the number mark under the standard style. The training module is used by the intelligent computing cloud platform to perform N iterations of training on the initial model based on the training set using computing power. The initial model after the Mth iteration of training is determined as the final model. The loss value of the initial model after the Mth iteration of training in the validation set is less than the loss value of the initial model after the Lth iteration of training in the validation set. N is a positive integer, M is a positive integer less than or equal to N, L is a positive integer less than or equal to N, and L is not equal to M. The Lth iteration of training is any one of the N iterations of training except for the Mth iteration of training. The detection module is used to deploy the final model on the intelligent computing cloud platform to perform batch dimension annotation detection on the received automotive drawings; The building module is specifically used for: Injecting a first abnormal feature into the first standard parameter yields a first sample dataset. The first sample dataset includes multiple abnormal line segment markers, each carrying a first label. The first label indicates the abnormal type of the abnormal line segment marker. The abnormal type of the abnormal line segment marker includes at least one of the following: missing line segment, misaligned line segment, overlapping line segment, twisted line segment, missing arrow, and reversed arrow direction. A second abnormal feature is injected into the second standard parameter to obtain a second sample dataset. The second sample dataset includes multiple abnormal digit markers, each of which carries a second label. The second label is used to indicate the abnormal type of the abnormal digit marker. The abnormal type of the abnormal digit marker includes at least one of digit missing, digit reversed, and digit overlapping. A third abnormal feature is injected into the third standard parameter to obtain a third sample dataset. The third sample dataset includes multiple abnormal positional relationships. Each abnormal positional relationship carries a third label. The third label is used to indicate the abnormal type of the abnormal positional relationship. The abnormal type of the abnormal positional relationship includes at least one of the following: overlapping of line segment markers and number markers, abnormal relative position of line segment markers and number markers, and abnormal lead line between line segment markers and number markers. A sample dataset is constructed based on the positive example sample set and the negative example sample set. The negative example sample set includes the first sample dataset, the second sample dataset, and the third sample dataset. The positive example sample set includes the standard style corresponding to the line segment markers generated according to the first standard parameter, the standard style corresponding to the number markers generated according to the second standard parameter, and the standard style corresponding to the positional relationship generated according to the third standard parameter.

7. An electronic device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method as described in any one of claims 1 to 5.

9. A computer program product, characterized in that, Includes computer instructions that, when executed by a processor, implement the steps of the method as described in any one of claims 1 to 5.

Images