Intelligent computing cloud platform for automatic detection of text crossings in automotive drawing dimension lines using computing power.

Abstract

This invention provides a method and apparatus for automatically detecting dimension lines crossing text on automotive drawings using a smart computing cloud platform. It relates to the field of smart computing cloud platforms and computing infrastructure technology. The method includes: constructing a sample dataset based on positive and negative examples, and dividing the sample dataset into a training set and a validation set; the smart computing cloud platform using computing power to iteratively train an initial model N times based on the training set, and determining the initial model after the Mth iteration as the final model; deploying the final model on the smart computing cloud platform to perform batch detection of dimension lines crossing text on received automotive drawings. Based on features from thorough prior training, the model can accurately output text positions and identify the spatial relationship between dimension lines and text in automotive drawings, improving the efficiency and accuracy of detecting dimension lines crossing text on automotive drawings.

Description

Technical Field

[0001] This invention relates to the field of intelligent computing cloud platform and computing infrastructure technology, specifically to a method and device for automatically detecting text crossing dimension lines on automotive drawings 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 entire process of automotive R&D, design, parts processing, and quality inspection, industrial drawings are the core carrier for conveying design intent and clarifying technical parameters. Textual annotations, dimensions, and symbol explanations, along with dimension lines, constitute the complete expressive system of the drawings. According to drafting standards such as "Mechanical Drawing Dimensioning" (GB / T 4458.4-2003), graphic elements such as lines and boundary lines must not cross text areas. This specification directly affects the accuracy of information transmission in drawings. If "lines cover text" or "text is cut off by lines," it can easily lead to misinterpretation of dimensions and confusion of technical requirements by processing personnel, thereby causing production risks such as parts processing deviations and assembly conflicts.

[0008] However, many historical archived drawings, on-site scanned drawings, and some automatically generated digital drawings currently suffer from overlapping and misaligned lines and text after being scanned, compressed, and vectorized. Furthermore, due to their age or differences in management processes, most archived drawings no longer have traceable CAD or other engineering source files, retaining only planar image data (such as JPG, PNG, and TIFF formats), lacking the original geometric layer information (such as the overlay relationship between text and line layers). This makes it difficult to accurately determine the spatial relationship between lines and text through layer separation. Additionally, some scanned drawings suffer from additional problems such as blurred text and worn lines due to equipment precision and paper aging, further increasing the difficulty of spatial relationship identification.

[0009] While some automatic detection methods exist in existing technologies, none are designed to address the requirement of dimension lines crossing text on automotive drawings, and they differ fundamentally from the application scenarios supported by intelligent computing centers. Current automatic detection methods are typically based on a simple hardware architecture of a single workstation (CPU + entry-level GPU), generally using traditional image processing libraries such as OpenCV (e.g., Canny / Hough operators). This is vastly different from the application scenarios of automatic detection methods in intelligent computing centers. Intelligent computing centers require the integration of kilo-level GPU clusters (e.g., NVIDIA A100 / H100) and high-speed RDMA networks to achieve EFLOPS-level computing power. This architecture requires the computing network to achieve 400Gbps inter-node communication, the storage network to reach 200Gbps bandwidth, and overall latency to be controlled at the sub-microsecond level.

[0010] It is evident that since the emergence of intelligent computing centers, how to use computing power to automatically detect the intersection of dimension lines and text on automotive drawings in order to improve detection efficiency and accuracy has become an urgent problem to be solved. Summary of the Invention

[0011] This invention provides a method and apparatus for automatically detecting text crossings in automotive drawing dimensions using computing power on an intelligent computing cloud platform. This addresses the problem of how to improve detection efficiency and accuracy by automatically detecting text crossings in automotive drawing dimensions using computing power since the emergence of intelligent computing centers.

[0012] To solve the above problems, the present invention is implemented as follows:

[0013] In a first aspect, embodiments of the present invention provide a method for automatically detecting text crossings by dimension lines on automotive drawings using a smart computing cloud platform, the method comprising:

[0014] Step S1: Based on positive and negative examples, construct a sample dataset and divide the sample dataset into a training set and a validation set. The positive example example indicates that there are no intersection, overlap, or coverage anomalies between dimension lines and text in the car drawing. The negative example example indicates that there are at least one of the following anomalies between dimension lines and text in the car drawing: intersection, overlap, or coverage anomalies.

[0015] 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.

[0016] Step S3: Deploy the final model on the intelligent computing cloud platform to perform batch detection of dimension lines crossing text on the received car drawings.

[0017] In one embodiment, step S1 includes:

[0018] Step S11: Generate positive example patterns according to preset rules, the preset rules including drafting specifications and enterprise testing standards;

[0019] Step S12: Perform cross processing on the positive example styles to obtain the first negative example style set. The first negative example style set includes negative example styles where the dimension line intersects the text at any angle.

[0020] Step S13: Overlap the positive example styles to obtain a second set of negative example styles. The second set of negative example styles includes negative example styles with overlapping dimension lines, negative example styles with overlapping text, and negative example styles with overlapping dimension lines and text.

[0021] Step S14: Override the positive example style to obtain the third negative example style set, which includes negative example styles where the dimension line and text overlap.

[0022] Step S15: Construct a sample dataset based on the positive example sample set and the negative example sample set. The negative example sample set includes the first negative example pattern set, the second negative example pattern set, and the third negative example pattern set. The positive example sample set includes the positive example patterns.

[0023] In one embodiment, step S15 includes:

[0024] Step 151: Perform preset processing on each positive example pattern in the positive example sample set and each negative example pattern in the negative example sample set to obtain the sample data after preset processing. The preset processing includes at least one of adding scanning noise, blurring, grayscale deviation processing, tilting processing, and image overlay background elements. Adding scanning noise is to randomly apply optical character recognition (OCR) noise to the positive example pattern or negative example pattern.

[0025] Step 152: Set labels for each of the pre-processed sample data to obtain the sample dataset.

[0026] In one embodiment, step S2 includes:

[0027] Step S21: Select a lightweight deep learning detection network as the initial model. The initial model is used to output the position information of the text in the car drawing, as well as the detection results of the dimension lines and text in the car drawing.

[0028] Step S22: The intelligent computing cloud platform utilizes computing power and adopts a distributed data parallel framework to input a preset number of sample data randomly selected from the training set into the initial model for feature extraction. Then, the initial model outputs the predicted location information and the predicted detection results.

[0029] Step S23: 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 position information of the text in the car drawing in the sample data output by the initial model and the actual position information of the text in the car drawing in the sample data. The second loss value represents the error between the predicted detection results of the dimension lines and text in the car drawing in the sample data output by the initial model and the actual detection results of the dimension lines and text in the sample data. The first loss value is calculated using the intersection-union ratio loss (IoULoss), and the second loss value is calculated using the multi-class cross-entropy loss.

[0030] Step S24: 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, and the learning rate is reduced according to the cosine annealing algorithm in each iteration.

[0031] Step S25: If the total loss value of the initial model after the Mth iteration of training is less than the total 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.

[0032] In one embodiment, step S3 includes:

[0033] 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 batches of car drawings and outputting the detection results of the dimension lines of the car drawings crossing the text.

[0034] In one embodiment, after step S3, the method further includes:

[0035] Step S4: Generate a review report based on the detection results. The review report includes the position information of the text in the car drawing, the anomaly type of dimension lines crossing the text in the car drawing, and the confidence level corresponding to the anomaly type.

[0036] 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 abnormal patterns in the car drawings corresponding to the abnormal results are used as new sample data.

[0037] Secondly, embodiments of the present invention provide an automatic detection device for text crossing dimension lines on automotive drawings using computing power on an intelligent computing cloud platform, the device comprising:

[0038] The construction module is used to construct a sample dataset based on positive example patterns and negative example patterns, and divide the sample dataset into a training set and a validation set. The positive example patterns represent that the dimension lines and text in the car drawings have no intersection anomalies, no overlap anomalies, and no coverage anomalies. The negative example patterns represent that the dimension lines and text in the car drawings have at least one of the intersection anomalies, overlap anomalies, and coverage anomalies.

[0039] 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.

[0040] The deployment module is used to deploy the final model on the intelligent computing cloud platform to perform batch detection of dimension lines crossing text on the received automotive drawings.

[0041] 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.

[0042] 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.

[0043] 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.

[0044] In this embodiment of the invention, negative example patterns are obtained by processing positive example patterns. A sample dataset is constructed based on the positive and negative example patterns and divided into a training set and a validation set. The data synthesis method solves the problem of sample scarcity, laying 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. In this way, based on the features trained in the early stage, the model can accurately output the text position and identify the spatial relationship between dimension lines and text in automotive drawings, improving the efficiency and accuracy of detecting dimension lines crossing text in automotive drawings. Attached Figure Description

[0045] 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.

[0046] Figure 1 This is a flowchart of a method for automatically detecting text crossing dimension lines on automotive drawings using computing power, provided by an embodiment of the present invention.

[0047] Figure 2 This is one of the counterexamples provided in the embodiments of the present invention;

[0048] Figure 3 This is the second counterexample provided in the embodiments of the present invention;

[0049] Figure 4 This is the third counterexample provided in the embodiments of the present invention;

[0050] Figure 5 This is the fourth counterexample provided in the embodiments of the present invention;

[0051] Figure 6 This is the fifth counterexample provided in the embodiments of the present invention;

[0052] Figure 7 This is a structural diagram of an intelligent computing cloud platform that uses computing power to automatically detect text crossings in automotive drawing dimension lines, as provided in an embodiment of the present invention.

[0053] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0054] 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.

[0055] The technical terms involved in this invention will be briefly explained below.

[0056] 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.

[0057] 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 .

[0058] 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.

[0059] 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.

[0060] 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.

[0061] 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.

[0062] The “computing power” mentioned in this invention includes: general computing power, intelligent computing power, and supercomputing power.

[0063] 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.

[0064] 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.

[0065] 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.

[0066] 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.

[0067] 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.

[0068] 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.

[0069] The "intelligent computing center" mentioned in this invention includes, but is not limited to, "smart computing center".

[0070] 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.

[0071] 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.

[0072] 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.

[0073] 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.

[0074] 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.

[0075] The "model" mentioned in this invention includes, but is not limited to, "large language model" and "multimodal large model".

[0076] 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.

[0077] 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.

[0078] The "automotive drawings" described in this invention refer to engineering 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 are used to show the assembly relationships of components in the complete vehicle or chassis, and to indicate the relative positions and connection methods of key components. Assembly drawings (such as engine assembly, transmission assembly, and body frame assembly drawings) are used to indicate 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) are used to indicate the detailed dimensions, tolerances, and processing requirements of the parts. Textual annotations, dimension markings, and symbol explanations, together with dimension lines, constitute a complete expressive system for the drawings. According to drafting standards such as "Mechanical Drawing Dimensioning Method" (GB / T 4458.4-2003), dimension lines must not cross text areas. This specification directly relates to the accuracy of information transmission in the drawings; therefore, the automatic detection of this invention mainly targets the detection of dimension lines and text in automotive drawings.

[0079] With the deepening of digital transformation in the automotive industry, enterprises face the dual demands of reviewing tens of thousands of historical drawings and verifying massive amounts of new drawings. Currently, a large number of historical archived drawings, on-site scanned drawings, and some automatically generated digital drawings generally suffer from overlapping and misaligned lines and text after scanning, compression, and vectorization processes. Furthermore, due to their age or differences in management processes, most archived drawings no longer have traceable CAD or other engineering source files, retaining only planar image data and lacking the original geometric layer information. This makes it difficult to accurately determine the spatial relationship between lines and text through layer separation. Additionally, some scanned drawings suffer from blurred text and worn lines due to factors such as equipment precision and paper aging, further increasing the difficulty of spatial relationship recognition. Therefore, this invention provides a method and apparatus for automatically detecting line crossings in automotive drawings using computing power on an intelligent computing cloud platform. This addresses the problem of how to improve detection efficiency and accuracy by automatically detecting line crossings in automotive drawings using computing power, a challenge that has existed since the emergence of intelligent computing centers.

[0080] It is important to emphasize that this application is based on a distributed architecture built on an intelligent computing cloud platform that integrates computing power (e.g., a kilocalorie-level GPU cluster and a high-speed network) in an intelligent computing center scenario. Through the deep integration of computing power, algorithms, and data by the intelligent computing cloud platform, a "super intelligent detection system" has been built specifically for detecting text crossings in automotive drawing dimension lines. Its core advantages are reflected in the breakthroughs in all dimensions of technical architecture, processing power, detection accuracy, and scenario adaptability.

[0081] 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.

[0082] The intelligent computing cloud platform utilizes a multimodal fusion model combining computer vision (CV), geometric topology analysis, and a physical rule engine. Leveraging the feature association capabilities of graph neural networks (GNNs), it achieves holographic detection of complex assembly drawings. For example, the computer vision module focuses on core detection targets, accurately extracting pixel-level correlation features between dimension lines and text from images, capturing subtle crossing marks even in blurry or tilted drawings. The geometric topology analysis module constructs spatial relationships between text and dimension lines, avoiding misjudgments caused by isolated detection. The physical rule engine strictly adheres to the "Mechanical Drawing Dimensioning Method" (GB / T 4458.4-2003) to verify the compliance of spatial relationships between dimension lines and text, ensuring that detection results conform to drafting standards. This significantly improves the detection accuracy of dimension lines crossing text. This is something that single-machine methods relying solely on traditional OpenCV operators (such as Canny edge detection and Hough line detection) cannot achieve.

[0083] 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 dimension lines or text styles (such as designer handwriting or enterprise-specific simplified dimension line rules) without requiring manual model retraining; by leveraging generative adversarial networks to generate adversarial samples 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 a critical dimension line in a vehicle body structure drawing is detected to cross material thickness annotation text, the system can automatically generate a screenshot and coordinate information indicating the crossing location, triggering the design review process, and simultaneously synchronizing the violation data to the drawing management system. This represents a paradigm shift from passively checking after drawing completion to proactively preventing dimension lines from crossing text during the design process, comprehensively ensuring the accuracy of automotive drawing information transmission.

[0084] Please see Figure 1 , Figure 1 This is a flowchart of a method for automatically detecting text crossing dimension lines on automotive drawings using computing power, provided by an embodiment of the present invention. Figure 1 As shown, it includes the following steps:

[0085] Step S1: Based on positive and negative examples, construct a sample dataset and divide the sample dataset into a training set and a validation set. The positive example example indicates that there are no intersection, overlap, or coverage anomalies between dimension lines and text in the car drawing. The negative example example indicates that there are at least one of the following anomalies between dimension lines and text in the car drawing: intersection, overlap, or coverage anomalies.

[0086] In this step, the standard styles of dimension lines and text can be determined according to the "Mechanical Drawing Dimensioning Method" (GB / T 4458.4-2003), namely, the positive example style: the dimension line and text have no intersection abnormalities, no overlap abnormalities, and no covering abnormalities. For example, the text can be a dimension annotation (50mm), and the distance between the dimension line next to the text and the text is greater than or equal to 2mm. After determining the standard style, abnormal styles can be obtained by processing the standard style, namely, the negative example style: the dimension line and text have intersection abnormalities, overlap abnormalities, or covering abnormalities. For example, the dimension line diagonally crosses the center of the text, the dimension line partially coincides with the edge of the text, the dimension line completely obscures the key strokes of the text, or multiple lines cross multiple characters, or slanted dimension lines cross deformed text, etc.

[0087] After constructing a sample dataset including positive and negative examples based on mapping standards and enterprise detection standards, it can be divided into a training set and a validation set in a 7:3 ratio. The training set is used for the model to learn the feature differences between positive and negative example patterns; the validation set is used to evaluate the model's training performance and avoid overfitting.

[0088] In this way, by using the intelligent computing cloud platform to call the graphics engine in parallel across multiple nodes, sample data that closely resembles the style of real drawings is drawn. This systematically generates 100,000 samples covering all types of errors, avoiding the scarcity of negative examples and improving the diversity of samples during subsequent model training. This allows the model to learn the pixel distance features between dimension lines and text in compliant samples, as well as the cross-pixel features between the two in non-compliant samples, thereby improving the comprehensiveness and accuracy of model detection.

[0089] 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.

[0090] 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.

[0091] In this way, the initial model trained in the Mth iteration fully learns all types of error features (such as intersection anomalies, overlap anomalies, and coverage anomalies between dimension lines and text), and ultimately achieves a minimum loss value on the validation set. This improves the accuracy of detecting dimension lines crossing text on automotive drawings, laying an efficient and accurate model foundation for subsequent batch detection. See the following description for details:

[0092] In one embodiment, step S2 includes:

[0093] Step S21: Select a lightweight deep learning detection network as the initial model. The initial model includes a detection branch and a classification branch. The detection branch is used to output the position information of the text in the car drawing, and the classification branch is used to output the detection results of the dimension lines and text in the car drawing.

[0094] Step S22: The intelligent computing cloud platform utilizes computing power and adopts a distributed data parallel framework to input a preset number of sample data randomly selected from the training set into the initial model for feature extraction. Then, it outputs the predicted location information through the detection branch and the predicted detection result through the classification branch.

[0095] Step S23: 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 position information of the text in the car drawing in the sample data output by the detection branch and the actual position information of the text in the car drawing in the sample data. The second loss value represents the error between the predicted detection results of the dimension lines and text in the car drawing in the sample data output by the classification branch and the actual detection results of the dimension lines and text in the sample data. The first loss value is calculated using the intersection-union ratio loss (IoULoss), and the second loss value is calculated using the multi-class cross-entropy loss.

[0096] Step S24: 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, and the learning rate is reduced according to the cosine annealing algorithm in each iteration.

[0097] Step S25: If the total loss value of the initial model after the Mth iteration of training is less than the total 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.

[0098] In this embodiment, a lightweight deep learning detection network such as YOLOv8n is selected as the initial model. This ensures detection accuracy while reducing computational overhead through a simplified network structure, adapting to the training efficiency and subsequent inference deployment requirements of the intelligent computing cloud platform. The initial model includes a detection branch and a classification branch. The detection branch outputs pixel-level coordinate information of the text in the drawing, such as the rectangular coordinates of the text area as (x1, y1, x2, y2). This provides a text position benchmark for subsequent judgment of the spatial relationship between dimension lines and text, avoiding errors in association judgment due to inaccurate text positioning. By combining the text position output by the detection branch with the feature information of the dimension lines, the spatial relationship detection result between the dimension lines and text is output. The detection result can include binary conclusions, such as "crossing" or "not crossing," or more specific conclusions, such as "normal," "intersecting," "overlapping," or "covering." Based on the constructed initial model, it can adapt to the training and subsequent deployment requirements of the cloud platform, and realize text positioning and spatial relationship detection between dimension lines and text, clarifying the training objectives and laying the architectural foundation for subsequent accurate training.

[0099] After constructing the initial model, the intelligent computing cloud platform utilizes a kilo-level GPU cluster and high-speed network resources to batch process the training set samples using a distributed data parallel framework. For example, a predetermined number of samples can be randomly selected from the training set (e.g., 128 drawings per sample). After being input into the initial model, the backbone network (e.g., ResNet-50) first extracts global image features. These global features can include the stroke outlines of text, the linear features of dimension lines, and the pixel correlation features between the two. Then, the detection branch outputs the predicted position information of the text based on these features, and the classification branch outputs the prediction result of whether the dimension line crosses the text based on the features and the text position. The prediction result can be expressed as a confidence score. With the support of distributed computing power, the efficiency of feature extraction and branch inference for single batches of samples is improved, solving the computing power bottleneck of traditional single-machine processing of large-scale samples.

[0100] To comprehensively measure the difference between model predictions and true labels, and to ensure that optimization covers the three dimensions of localization, recognition, and confidence, the total loss of the initial model can be calculated based on the first and second loss values. The first loss value can be calculated using the Intersection over Union (IoU) loss to measure the error between the predicted text location information output by the detection branch and the actual text location information in the sample. This error can be obtained by calculating the IoU ratio between the predicted and actual bounding boxes. The closer the IoU value is to 1, the more accurate the localization and the smaller the loss value. This loss function can effectively optimize the localization accuracy of text positions and avoid misjudgments of the spatial relationship between the size lines and the text due to text box offset. The second loss can be calculated using multi-class cross-entropy loss to measure the error between the predicted detection results output by the classification branch and the actual detection results. The total loss value is the weighted sum of the two. By using the combined loss function of IoU loss and multi-class cross-entropy loss, the accuracy of text localization and the accuracy of relationship judgment can be optimized simultaneously.

[0101] Then, parameter optimization is initiated based on the total loss: the gradients of all model parameters, such as convolutional kernel weights, bias terms, and regression parameters, are calculated through backpropagation using the chain rule. This is combined with a dynamically adjusted learning rate using cosine annealing, which decays with each iteration to avoid parameter oscillations. This iterative update of model parameters gradually brings the model's predictions closer to the true labels. The updated model from each iteration can then be applied to the validation set for inference, and the validation set loss is calculated. If the loss value of the initial model trained in the Mth iteration is less than that of the initial model trained in the Lth iteration, the initial model trained in the Mth iteration is considered the optimal model among the N initial models trained in the Mth iteration. This initial model can then be selected as the final model, completing the training and obtaining a model suitable for actual detection. Ultimately, this achieves iterative evolution from the initial model to a high-precision final model.

[0102] Step S3: Deploy the final model on the intelligent computing cloud platform to perform batch detection of dimension lines crossing text on the received car drawings.

[0103] 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 detection of dimension lines crossing text in received automotive drawings (e.g., scanned images or PDF files), completing drafting standard compliance judgments. It can automatically detect anomalies such as dimension lines crossing text, exhibiting high accuracy, high versatility, and high robustness. It has broad application prospects in industrial design review, engineering archiving verification, and automated drafting quality control, completely eliminating reliance on vector layer data and improving detection efficiency and accuracy.

[0104] Specifically, in one embodiment, step S3 includes:

[0105] 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 batches of car drawings and outputting the detection results of the dimension lines of the car drawings crossing the text.

[0106] 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.

[0107] 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).

[0108] 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 cropped image within 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. Finally, the processed cropped images are batch-input into model instances on each GPU node to perform parallel text detection of dimension lines crossing the automotive drawings, improving detection efficiency.

[0109] In one embodiment, after step S3, the method further includes:

[0110] Step S4: Generate a review report based on the detection results. The review report includes the position information of the text in the car drawing, the anomaly type of dimension lines crossing the text in the car drawing, and the confidence level corresponding to the anomaly type.

[0111] 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 abnormal patterns in the car drawings corresponding to the abnormal results are used as new sample data.

[0112] In this embodiment, a review report containing complete information is generated based on the batch detection results, transforming the raw detection data output by the model into a decision-making basis that reviewers can directly use. The review report includes text location information, detection results, and confidence levels. The text location information can directly correspond to the bounding box coordinates output by the model, presented in the format of "top left x-coordinate, top left y-coordinate, bottom right x-coordinate, bottom right y-coordinate." Engineers can quickly locate text areas interfered with by dimension lines on drawings using these coordinates, eliminating the need to check each drawing individually. Detection results can be dimension line crossing anomaly types output by the model, such as "dimension line overlap." The confidence level can directly use the model's output confidence level value (e.g., 0.82) to indicate the reliability of the detection results; the closer the value is to 1, the more accurate the model's judgment of the anomaly. The review report can be shown in the table below:

[0113]

[0114] Then, the test results are reviewed based on the review report. If abnormal results are found in the test results, the abnormal patterns in the car drawings corresponding to the abnormal results are used as new sample data to continuously improve the model performance.

[0115] The specific process is as follows:

[0116] First, professional reviewers verify the test results one by one based on the generated review report, focusing on confirming whether there are any misjudgments in the model. For example, if the test result corresponding to "the style where the dimension line and text overlap" is judged as normal, then it is considered that there is an abnormal result in the test result. At this time, the drawing fragment containing the style where the dimension line and text overlap can be cut into a new sample, its true error type can be marked, and it can be added to the original sample set.

[0117] 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.

[0118] In this embodiment of the invention, negative example patterns are obtained by processing positive example patterns. A sample dataset is constructed based on the positive and negative example patterns and divided into a training set and a validation set. The data synthesis method solves the problem of sample scarcity, laying 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 cloud platform dynamically schedules computing power to achieve parallel detection of batch drawings. In this way, based on the features trained in the early stage, the model can accurately output the text position and identify the spatial relationship between dimension lines and text in automotive drawings, improving the efficiency and accuracy of detecting dimension lines crossing text in automotive drawings.

[0119] In one embodiment, step S1 includes:

[0120] Step S11: Generate positive example patterns according to preset rules, the preset rules including drafting specifications and enterprise testing standards;

[0121] Step S12: Perform cross processing on the positive example styles to obtain the first negative example style set. The first negative example style set includes negative example styles where the dimension line intersects the text at any angle.

[0122] Step S13: Overlap the positive example styles to obtain a second set of negative example styles. The second set of negative example styles includes negative example styles with overlapping dimension lines, negative example styles with overlapping text, and negative example styles with overlapping dimension lines and text.

[0123] Step S14: Override the positive example style to obtain the third negative example style set, which includes negative example styles where the dimension line and text overlap.

[0124] Step S15: Construct a sample dataset based on the positive example sample set and the negative example sample set. The negative example sample set includes the first negative example pattern set, the second negative example pattern set, and the third negative example pattern set. The positive example sample set includes the positive example patterns.

[0125] In this embodiment, the preset rules may include drafting standards and enterprise testing standards. Drafting standards may be national standards such as "Mechanical Drawing Dimensioning" (GB / T 4458.4-2003), ensuring compliance of the spatial relationship between dimension lines and text, such as a spacing of ≥2mm between dimension lines and text areas, and dimension lines not touching text strokes. Enterprise testing standards may incorporate the specific drafting specifications of automotive companies, such as the font and line width requirements for specific car model drawings, making the positive example style more closely aligned with actual production scenarios. Positive example styles are generated according to the preset rules, producing standard styles with no anomalies in dimension lines and text, providing a compliance reference for subsequent negative example sample generation.

[0126] After establishing a standard style, the direction of the dimension lines can be adjusted to allow them to intersect the text at any angle. For example, dimension lines can intersect text at a 45° angle, a 90° angle perpendicularly, or non-standard angles such as 15° or 30°, simulating abnormal intersection scenarios caused by drawing scan offsets or drafting errors. Figure 2 As shown, the text "50mm" is diagonally crossed by a 45° angled dimension line from the text area corresponding to "mm". After assigning corresponding labels to the various cross-anomaly counterexample styles, a first set of counterexample styles is formed, which the model uses to learn cross-anomaly features.

[0127] After constructing the standard styles, a second set of negative styles can be obtained by performing local pixel overlap processing on the positive example styles. The negative examples in the second set of negative styles focus on edge overlap between dimension lines, text, and similar elements. Specifically, dimension line overlap can be the partial overlap of the edges of two or more dimension lines. For example, the dimension line corresponding to 50mm overlaps with the dimension line corresponding to 60mm. Figure 3 As shown; text overlap can be the overlapping of the edges of different characters, for example, an overlap between 50mm and 60mm, as shown. Figure 4 As shown; overlap between dimension lines and text can be between a first dimension line and a second text, where the first dimension line corresponds to the first text, and the second text corresponds to the second dimension line. In other words, overlap between dimension lines and text can be between a dimension line in one set of dimension lines and text and a text in another set of dimension lines and text. For example, the dimension line corresponding to 50mm overlaps with the dimension line corresponding to 60mm. Figure 5 As shown, after assigning corresponding labels to various counterexample patterns of overlap anomalies, a second set of counterexample patterns is formed, which is used by the model to learn overlap anomaly features.

[0128] After constructing the standard styles, a third set of negative examples can be obtained by extensively occluding the positive examples. This third set includes negative examples where dimension lines and text overlap. Unlike negative examples of overlap anomalies, when generating negative examples of overlap anomalies, the dimension lines can completely cover key strokes of the text, such as numbers or core words, or the text can cover key segments of the dimension lines, such as arrows. In this way, the key segments of the dimension lines or text in the third set of negative examples of overlap anomalies are completely obscured, making the drawing annotations incomprehensible to the user. For example, in the text "10℃±2℃", the "±" is completely covered by the dimension line, and only "10℃" and "2℃" are visible. Figure 6 As shown, after assigning corresponding labels to various cover anomaly counterexample styles, a third counterexample style set is formed, which is used by the model to learn cover anomaly features.

[0129] After generating the first, second, and third counterexample pattern sets, positive and counterexample samples can be mixed in a preset ratio (e.g., 3:1) to ensure a balanced distribution of categories across the sample sets. The positive sample set is used by the model to learn the features of standard patterns; the counterexample sample set, which summarizes the first, second, and third counterexample pattern sets and covers the corresponding anomaly types of intersection, overlap, and coverage, is used by the model to learn anomaly features. The final constructed sample dataset includes compliance references and covers all scenarios of anomaly detection, such as dimension lines crossing text on automotive drawings. Each sample has a clear label, providing comprehensive and accurate supervised learning data for subsequent model training and improving the model's ability to identify various anomalies.

[0130] In one embodiment, step S15 includes:

[0131] Step 151: Perform preset processing on each positive example pattern in the positive example sample set and each negative example pattern in the negative example sample set to obtain the sample data after preset processing. The preset processing includes at least one of adding scanning noise, blurring, grayscale deviation processing, tilting processing, and image overlay background elements. Adding scanning noise is to randomly apply optical character recognition (OCR) noise to the positive example pattern or negative example pattern.

[0132] Step 152: Set labels for each of the pre-processed sample data to obtain the sample dataset.

[0133] In this embodiment, interference factors from real car drawings are simulated and clearly labeled to make the sample dataset more closely resemble actual detection scenarios, providing the model with accurate and robust training data. Specifically, various preset processing methods are used to simulate non-ideal states that may occur in real car drawings during scanning and archiving, targeting standard patterns in the positive sample set and abnormal patterns in the negative sample set. This ensures that the model can adapt to complex environmental interference during training. These preset processing methods include at least one of the following: adding scanning noise, blurring, grayscale deviation processing, tilting, and overlaying background elements onto the image. Through these processing methods, the sample data automatically synthesized based on standard-driven methods 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 labels during training but also adapts to interference factors in real scenarios, thereby improving the robustness of subsequent detection and avoiding detection failures caused by the disconnect between samples and actual scenarios.

[0134] For example, random OCR noise can be applied to simulate paper texture, insufficient scanner precision, or ink stains caused by paper aging during scanning, such as black spots appearing on the edge of the "m" in "50mm". This allows the model to learn to extract core features from the noise, improving the model's robustness.

[0135] Blur processing can employ Gaussian blur or motion blur to simulate inaccurate focusing during scanning, image blurring at paper creases, or detail loss caused by low-resolution scanning. For example, it can blur the edges of the "℃" symbol and blur the boundaries between dimension lines and text. This improves the model's adaptability to blurred drawings and avoids missed detections due to decreased image clarity.

[0136] Gray-scale deviation processing can adjust the gray-scale value distribution of samples (local brightening or darkening) to simulate the differences in brightness caused by uneven lighting during scanning (such as local attenuation of the scanner light source). For example, the "assembly gap" annotation area is generally dark, with a significant difference in gray-scale compared to the surrounding drawings. This solves the difficulty of feature extraction caused by uneven lighting in real drawings, allowing the model to identify targets without relying on uniform gray-scale.

[0137] Tilt processing can be achieved by randomly rotating the sample image by ±15° to simulate the skewed state when manually placing the scanned drawing. For example, the entire "120kPa±5kPa" may be tilted by 10°, and the characters may be arranged non-horizontally. This is to adapt to the tilted text and dimension lines in real drawings and reduce the misjudgment rate of position.

[0138] Image overlay background elements can be simple part outlines, grid lines, or irrelevant text overlaid around dimension lines or text to simulate the complex background with multiple elements coexisting in real automotive drawings. This allows the model to better distinguish the combination of dimension lines and text from the interfering background and avoid misjudging background lines as dimension lines.

[0139] Furthermore, each sample data point, after pre-processing, is assigned a precise label as a criterion for model training. If the sample originates from the positive example set, its label can indicate that the dimension line and text are compliant; if the sample originates from the negative example set, its label must clearly indicate the specific anomaly type, such as labeling a dimension line crossing text, and the anomaly type as a coverage anomaly. Ultimately, a high-quality sample dataset is constructed that covers all anomaly types, adapts to real-world interference, and has clear labels, laying a data foundation for the accuracy and robustness of model training.

[0140] Please see Figure 7 , Figure 7 This is a structural diagram of an intelligent computing cloud platform that uses computing power to automatically detect text crossings in automotive drawing dimension lines, as provided in an embodiment of the present invention. Figure 7 As shown, the intelligent computing cloud platform uses computing power to automatically detect text crossings in automotive drawing dimension lines. The device 700 includes:

[0141] The construction module 701 is used to construct a sample dataset based on positive example patterns and negative example patterns, and to divide the sample dataset into a training set and a validation set. The positive example patterns represent that there are no intersection anomalies, no overlap anomalies, and no coverage anomalies between dimension lines and text in the automotive drawings. The negative example patterns represent that there are at least one of the intersection anomalies, overlap anomalies, and coverage anomalies between dimension lines and text in the automotive drawings.

[0142] Training module 702 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.

[0143] The deployment module 703 is used to deploy the final model on the intelligent computing cloud platform to perform batch detection of dimension lines crossing text on the received automotive drawings.

[0144] In one embodiment, the construction module 701 is specifically used for:

[0145] Positive example patterns are generated according to preset rules, which include drafting standards and enterprise testing standards;

[0146] The positive example styles are cross-processed to obtain a first set of negative example styles, which includes negative example styles in which the dimension line intersects the text at any angle.

[0147] The positive example styles are overlapped to obtain a second set of negative example styles. The second set of negative example styles includes negative example styles with overlapping dimension lines, negative example styles with overlapping text, and negative example styles with overlapping dimension lines and text.

[0148] The positive example style is overridden to obtain a third negative example style set, which includes negative example styles in which dimension lines and text overlap.

[0149] 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 negative example style set, the second negative example style set, and the third negative example style set. The positive example sample set includes the positive example style.

[0150] In one embodiment, constructing the sample dataset based on the positive example set and the negative example set includes:

[0151] Pre-processing is performed on each positive example pattern in the positive example sample set and each negative example pattern in the negative example sample set to obtain sample data after pre-processing. The pre-processing includes at least one of adding scanning noise, blurring, grayscale deviation processing, tilting processing, and image overlaying background elements. Adding scanning noise is to randomly apply optical character recognition (OCR) noise to the positive example pattern or the negative example pattern.

[0152] Label each of the pre-processed sample data to obtain the sample dataset.

[0153] In one embodiment, the training module 702 is specifically used for:

[0154] A lightweight deep learning detection network was selected as the initial model. The initial model includes a detection branch and a classification branch. The detection branch is used to output the position information of the text in the car drawing, and the classification branch is used to output the detection results of the dimension lines and text in the car drawing.

[0155] The intelligent computing cloud platform utilizes computing power and adopts a distributed data parallel framework to input a preset number of sample data randomly selected from the training set into the initial model for feature extraction. After that, the detection branch outputs the predicted location information, and the classification branch outputs the predicted detection result.

[0156] 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 position information of the text in the car drawing in the sample data output by the detection branch and the actual position information of the text in the car drawing in the sample data. The second loss value represents the error between the predicted detection results of the dimension lines and text in the car drawing in the sample data output by the classification branch and the actual detection results of the dimension lines and text in the sample data. The first loss value is calculated using the intersection-union ratio (IoU) loss, and the second loss value is calculated using the multi-class cross-entropy loss.

[0157] 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, and the learning rate is reduced according to the cosine annealing algorithm in each iteration.

[0158] If the total loss value of the initial model after the Mth iteration of training is less than the total 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.

[0159] In one embodiment, deployment module 703 is specifically used for:

[0160] After converting the final model into an inference format adapted to the intelligent computing cloud platform, it is deployed on the GPU inference node of the intelligent computing cloud platform, and the final model is encapsulated into a callable API interface. The API interface supports receiving batches of automotive drawings and outputting the detection results of the dimension lines of the automotive drawings crossing the text.

[0161] In one embodiment, the apparatus further includes:

[0162] The generation module is used to generate a review report based on the detection results. The review report includes the position information of the text in the car drawing, the anomaly type of the dimension line crossing the text in the car drawing, and the confidence level corresponding to the anomaly type.

[0163] 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 abnormal patterns in the car drawings corresponding to the abnormal results are used as new sample data.

[0164] The intelligent computing cloud platform provided in this invention provides an automatic detection device for detecting text crossings in automotive drawing dimensions using computing power. This device can realize the various processes of the above-mentioned intelligent computing cloud platform automatic detection method for detecting text crossings in automotive drawing dimensions using 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.

[0165] 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.

[0166] This invention also provides an electronic device, see [link to relevant documentation]. Figure 8 , Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. The electronic device includes a memory 801, a processor 802, and a program or instructions stored in the memory 801 that run on the processor. When the program or instructions are executed by the processor 802, they can achieve the following: Figure 1 The corresponding intelligent computing cloud platform uses computing power to perform any step in the embodiment of the automatic detection method for car drawing dimension lines crossing text and achieves the same beneficial effect, which will not be elaborated here.

[0167] The processor 802 can be a CPU, ASIC, FPGA, or GPU.

[0168] 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 text crossings in automotive drawing dimension lines can be accomplished by hardware related to program instructions, and the program can be stored in a readable medium.

[0169] 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 embodiment of the automatic detection method for text crossing dimension lines on automotive drawings 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.

[0170] 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 uses computing power to perform each process of the automatic detection method embodiment for the dimension lines crossing text in automotive drawings, and can achieve the same technical effect. To avoid repetition, it will not be described in detail here.

[0171] 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.

[0172] 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.

[0173] 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.

[0174] 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 forms under the guidance of the present invention without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of the present invention.

Claims

1. A method for automatically detecting text crossings by dimension lines on automotive drawings using a smart computing cloud platform, characterized in that: The method includes: Step S1: Based on positive and negative examples, construct a sample dataset and divide the sample dataset into a training set and a validation set. The positive example example indicates that there are no intersection, overlap, or coverage anomalies between dimension lines and text in the car drawing. The negative example example indicates that there are at least one of the following anomalies between dimension lines and text in the car drawing: intersection, overlap, or coverage anomalies. 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 detection of dimension lines crossing text on the received automotive drawings; Step S2 includes: Step S21: Select a lightweight deep learning detection network as the initial model. The initial model is used to output the position information of the text in the car drawing, as well as the detection results of the dimension lines and text in the car drawing. Step S22: The intelligent computing cloud platform utilizes computing power and adopts a distributed data parallel framework to input a preset number of sample data randomly selected from the training set into the initial model for feature extraction. Then, the initial model outputs the predicted location information and the predicted detection results. Step S23: 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 position information of the text in the car drawing in the sample data output by the initial model and the actual position information of the text in the car drawing in the sample data. The second loss value represents the error between the predicted detection results of the dimension lines and text in the car drawing in the sample data output by the initial model and the actual detection results of the dimension lines and text in the sample data. The first loss value is calculated by the intersection-union ratio (IoU) loss, and the second loss value is calculated by the multi-class cross-entropy loss. Step S24: 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, and the learning rate is reduced according to the cosine annealing algorithm in each iteration. Step S25: If the total loss value of the initial model after the Mth iteration of training is less than the total 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.

2. The method as described in claim 1, characterized in that, Step S1 includes: Step S11: Generate positive example patterns according to preset rules, the preset rules including drafting specifications and enterprise testing standards; Step S12: Perform cross processing on the positive example styles to obtain the first negative example style set. The first negative example style set includes negative example styles where the dimension line intersects the text at any angle. Step S13: Overlap the positive example styles to obtain a second set of negative example styles. The second set of negative example styles includes negative example styles with overlapping dimension lines, negative example styles with overlapping text, and negative example styles with overlapping dimension lines and text. Step S14: Override the positive example style to obtain the third negative example style set, which includes negative example styles where the dimension line and text overlap. Step S15: Construct a sample dataset based on the positive example sample set and the negative example sample set. The negative example sample set includes the first negative example pattern set, the second negative example pattern set, and the third negative example pattern set. The positive example sample set includes the positive example patterns.

3. The method as described in claim 2, characterized in that, Step S15 includes: Step 151: Perform preset processing on each positive example pattern in the positive example sample set and each negative example pattern in the negative example sample set to obtain the sample data after preset processing. The preset processing includes at least one of adding scanning noise and blurring. Adding scanning noise is to randomly apply optical character recognition (OCR) noise to the positive example pattern or the negative example pattern. Step 152: Set labels for each of the pre-processed sample data to obtain the sample dataset.

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 batches of car drawings and outputting the detection results of the dimension lines of the car drawings crossing the text.

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 detection results. The review report includes the position information of the text in the car drawing, the anomaly type of dimension lines crossing the text in the car drawing, and the confidence level corresponding to the anomaly type. 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 abnormal patterns in the car drawings corresponding to the abnormal results are used as new sample data.

6. A device for automatically detecting text crossings in automotive blueprint dimension lines using computing power on an intelligent computing cloud platform, characterized in that... The device includes: The construction module is used to construct a sample dataset based on positive example patterns and negative example patterns, and divide the sample dataset into a training set and a validation set. The positive example patterns represent that the dimension lines and text in the car drawings have no intersection anomalies, no overlap anomalies, and no coverage anomalies. The negative example patterns represent that the dimension lines and text in the car drawings have at least one of the intersection anomalies, overlap anomalies, and coverage anomalies. 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 deployment module is used to deploy the final model on the intelligent computing cloud platform to perform batch detection of dimension lines crossing text on the received automotive drawings; The training module is specifically used for: A lightweight deep learning detection network was selected as the initial model. The initial model was used to output the position information of text in the car drawing, as well as the detection results of dimension lines and text in the car drawing. The intelligent computing cloud platform utilizes computing power and adopts a distributed data parallel framework to input a preset number of sample data randomly selected from the training set into the initial model for feature extraction. Then, the initial model outputs the predicted location information and the predicted detection results. 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 position information of the text in the car drawing in the sample data output by the initial model and the actual position information of the text in the car drawing in the sample data. The second loss value represents the error between the predicted detection results of the dimension lines and text in the car drawing in the sample data output by the initial model and the actual detection results of the dimension lines and text in the sample data. The first loss value is calculated using the intersection-union ratio (IoU) loss, and the second loss value is calculated using the multi-class cross-entropy loss. 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, and the learning rate is reduced according to the cosine annealing algorithm in each iteration. If the total loss value of the initial model after the Mth iteration of training is less than the total 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.

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