CN117632313B

CN117632313B - Software driving processing method and system based on artificial intelligence

Info

Publication number: CN117632313B
Application number: CN202410105626.2A
Authority: CN
Inventors: 陈平; 王者师; 徐巍; 韩少非
Original assignee: Shenzhen M2micro Co ltd
Current assignee: Shenzhen M2micro Co ltd
Priority date: 2024-01-25
Filing date: 2024-01-25
Publication date: 2024-03-26
Anticipated expiration: 2044-01-25
Also published as: CN117632313A

Abstract

The invention relates to the technical field of data processing, in particular to a software driving processing method and system based on artificial intelligence. The method comprises the following steps: extracting equipment characteristics of the hyperspectral remote sensing data to generate hardware characteristic data; performing feature compression on the hardware feature data to generate hardware feature sparse data; similar data extraction is carried out on the hardware information data, and heterogeneous entity relation data is generated; constructing a fusion map according to the heterogeneous entity relationship data to generate a heterogeneous information fusion map; acquiring user demand data; carrying out hardware type identification on the hardware feature sparse data by utilizing the heterogeneous information fusion map to generate hardware parameter data; and performing driver design based on the hardware parameter data and the user demand data to generate software driver data. According to the invention, the heterogeneous information fusion map is constructed so as to accurately identify the hardware equipment lacking training data.

Description

Software driving processing method and system based on artificial intelligence

Technical Field

The invention relates to the technical field of data processing, in particular to a software driving processing method and system based on artificial intelligence.

Background

The diversity of hardware devices and the ever-updated model make conventional approaches difficult to handle, and the release of new hardware devices typically requires the rapid generation of new drivers. The method based on artificial intelligence can be suitable for various hardware devices and generate drivers applicable to different models. The artificial intelligence method can quickly adapt to new hardware by learning and reasoning without waiting for a developer to manually write a new driver. The method based on artificial intelligence can reduce the problem of a driver caused by human errors and improve the stability and reliability. However, the existing software driving processing method and system based on artificial intelligence have low accuracy or poor stability of the generated driving program due to high data dependency.

Disclosure of Invention

Based on this, it is necessary to provide a software driven processing method based on artificial intelligence to solve at least one of the above technical problems.

To achieve the above object, a software-driven processing method based on artificial intelligence, the method comprising the steps of:

step S1: acquiring hyperspectral remote sensing data of hardware equipment to be identified; performing data preprocessing on the hyperspectral remote sensing data to generate standard remote sensing data; extracting equipment characteristics of the standard remote sensing data to generate hardware characteristic data;

Step S2: carrying out data block division processing on the hardware characteristic data to generate a hardware characteristic data block; performing feature compression on the hardware feature data block to generate hardware feature sparse data;

step S3: acquiring hardware information data; similar data extraction is carried out on the hardware information data, and heterogeneous entity relation data is generated; constructing a fusion map according to the heterogeneous entity relationship data to generate a heterogeneous information fusion map;

step S4: acquiring user demand data; carrying out hardware type identification on the hardware feature sparse data by utilizing the heterogeneous information fusion map to generate hardware parameter data; and performing driver design based on the hardware parameter data and the user demand data to generate software driver data.

According to the invention, by acquiring hyperspectral remote sensing data, which contains abundant spectral information, fine features including invisible spectrum can be captured from the surface of hardware equipment. Helping to identify hardware devices more accurately, especially under different lighting conditions. By adopting multi-band data, the diversity of hardware devices can be considered, so that the method is better suitable for hardware of different models and manufacturers. The data block division is helpful to reduce the data size of the processing, thereby reducing the computational complexity and improving the efficiency of data processing. The hardware feature data block partitioning may emphasize the local features of the hardware device, making it easier to capture information related to certain parts of the device. Fusing hardware information data and heterogeneous entity relationship data can provide more comprehensive information for hardware devices, including manufacturer, specification, historical performance, and the like. Context information of similar hardware devices is added to facilitate a more accurate understanding of the devices. The fusion map construction can carry out multidimensional information analysis on similar hardware, including hardware characteristics, manufacturer information, market trend and the like. Helping to analyze more deeply the characteristics and performance of the hardware device. Based on the user demand data, a personalized driver can be generated to meet the specific requirements of the user on the performance and the functions of the hardware equipment. User satisfaction is improved, and unnecessary configuration work is reduced. By carrying out the design of the driving program according to the data of the user demand, the compatibility problem between the hardware equipment and the operating system can be reduced, and the stability and the reliability of the system are improved. Therefore, the software driving processing method and system based on artificial intelligence, disclosed by the invention, obtain the hardware feature sparse data through feature compression of the hardware feature data, and identify the hardware feature sparse data lacking training data by utilizing the constructed heterogeneous atlas, so that the corresponding hardware equipment is accurately identified, and the software driving program is stably generated.

Preferably, step S1 comprises the steps of:

step S11: acquiring hyperspectral remote sensing data of hardware equipment to be identified; performing data preprocessing on the hyperspectral remote sensing data to generate standard remote sensing data;

step S12: extracting equipment characteristics of the standard remote sensing data to generate initial characteristic data; performing mutation evaluation on the initial characteristic data to generate robustness evaluation data;

step S13: feature data screening is conducted on the robustness assessment data according to a preset robustness scoring threshold value, and when the robustness assessment data is larger than or equal to the preset robustness scoring threshold value, high robustness feature data are generated; when the robustness evaluation data is smaller than a preset robustness grading threshold value, eliminating the robustness evaluation data;

step S14: performing feature association analysis on the high-robustness feature data to generate hardware feature association data;

step S15: mutual information analysis is carried out through hardware equipment to be identified, and mutual information analysis data are generated;

step S16: performing association correction on the hardware feature association data by utilizing mutual information analysis data to generate feature association correction data;

step S17: and performing necessary feature screening on the high-robustness feature data by utilizing the feature association correction data to generate hardware feature data.

The invention provides the detailed spectral characteristics of the hardware equipment through hyperspectral remote sensing data. Helping to more accurately distinguish hardware devices, especially under a variety of environmental conditions. By acquiring hyperspectral data, the system can adapt to different illumination conditions, weather conditions and environmental changes, and the robustness of hardware equipment identification is improved. Preliminary features of the hardware device are extracted from the standard remote sensing data, and the features can be used for preliminary identification of the hardware device. The extracted initial characteristic data is relatively less, and the data size of subsequent processing can be reduced, so that the efficiency is improved. The robustness assessment data is used to determine the robustness of the feature. By setting the robustness scoring threshold, features with robustness can be screened out, thereby reducing susceptibility to noise and outliers. And high-robustness characteristic data are generated, so that accuracy and stability of hardware equipment identification are improved, and possibility of false identification is reduced. Through feature correlation analysis, correlations and interactions between different features can be captured. Helping to more fully understand the characteristics and performance of hardware devices. The feature association analysis can reduce the data dimension, improve the efficiency of subsequent processing, and retain the most relevant information. Mutual information analysis can provide personalized identification for different devices by comparing the hardware device with data of known hardware devices. Helping to distinguish between similar hardware devices. Mutual information analysis data provides more information, thereby improving the recognition accuracy of hardware devices, especially in complex environments. By analyzing the data with mutual information, the feature association data can be modified to more accurately reflect the association between hardware devices. And helps to reduce misleading information in the associated data. The correlation correction improves the quality of the correlation data and facilitates more accurate hardware device identification. By correlating the correction data with features, the most relevant and informative features can be selected to generate higher quality hardware feature data. The generated hardware characteristic data can more accurately reflect the characteristics of the hardware equipment, so that the identification accuracy and stability of the hardware equipment are improved.

Preferably, step S2 comprises the steps of:

step S21: carrying out data block division processing on the hardware characteristic data to generate a hardware characteristic data block;

step S22: performing discrimination optimization on the hardware characteristic data blocks based on the characteristic association correction data to generate optimized characteristic data blocks;

step S23: performing sparsification processing on the optimized feature data blocks to generate feature dimension reduction data blocks;

step S24: performing feature compression on the feature reduced data block to generate a feature compressed data block;

step S25: and carrying out data block combination on the characteristic compressed data blocks to generate hardware characteristic sparse data.

The invention can promote parallel processing through dividing the data blocks, can process a plurality of data blocks at the same time, and improves the processing speed. The hardware characteristic data block division is favorable for better managing and organizing data, and the large-scale data is divided into smaller blocks, so that the processing efficiency and maintainability of the data are improved. Optimizing the feature data blocks reduces redundancy between features and improves information density of data. The differentiation optimization is based on the feature association correction data, so that the differentiation degree of different features in the hardware feature data block is enhanced. Helping to better capture the characteristics of each data block. The sparsification process can preserve critical information while reducing redundant information, helping to better capture critical features of the data block. The sparsification process helps to reduce the data dimension and reduce the complexity of the data. The processing and storage costs are reduced and the data processing speed is increased. The compressed data block occupies fewer memory and storage resources, and can more effectively utilize the computing resources, thereby improving the processing and transmission efficiency of the data. The combined data can be subjected to global feature extraction, so that the overall characteristics and performances of the hardware equipment can be comprehensively known.

Preferably, step S24 comprises the steps of:

step S241: performing low-dimensional mapping on the feature dimension reduction data block based on preset mapping coding times to generate a multi-low-dimensional mapping data block; mapping and decoding the multi-low dimensional mapping data to generate a multi-structure characteristic data block;

step S242: performing data clustering processing according to the multi-low-dimensional mapping data block and the multi-construction characteristic data block to generate a low-dimensional mapping data block and a construction characteristic data block;

step S243: performing reconstruction error calculation on the low-dimensional mapping data block and the reconstruction characteristic data block by using a reconstruction error calculation formula to generate reconstruction error data;

step S244: performing compression rate evaluation according to the low-dimensional mapping data block and the reconstruction characteristic data block to generate compression rate data;

step S245: performing compression dimension selection through the reconstruction error data and the compression rate data to generate compression dimension selection data; and performing feature compression on the reconstructed feature data block according to the compression dimension selection data to generate a feature compressed data block.

The invention can reduce the feature dimension to a lower dimension through multiple mapping encoding and decoding, reduces the complexity of data and improves the processing efficiency. The generation of the multi-constituent feature data blocks ensures that as much information as possible is retained, providing important information about the hardware device characteristics even in low dimensions. Data clustering facilitates the integration of multiple low-dimensional map data blocks and reconstructed feature data blocks into larger clustered data blocks, providing global feature information. Through clustering, similarity among data blocks can be analyzed, and the similarity and the difference among different hardware devices can be better understood. The reconstruction error data provides information about the accuracy of the reduction and reconstruction, helping the system to know the degree of loss of information in the low dimension. By reconstructing the error data, the quality of the degradation and reconstruction can be evaluated and adjusted as necessary to improve the usability of the data. The compression rate data reflects the degree of data size reduction of the data block during the de-mapping and reconstruction. Helping the system to determine the compression rate required in different situations. By compressing the rate data, storage and transmission resources can be better managed, and the storage and transmission cost of the data is reduced. Based on the reconstruction error data and the compression rate data, the system can intelligently select appropriate dimensions for feature compression to reduce the dimensions of the data while maintaining important information. The generation of the feature compressed data blocks reduces the dimensionality of the data, thereby reducing storage and transmission costs and improving processing efficiency.

Preferably, the reconstruction error calculation formula in step S243 is as follows:

in the method, in the process of the invention,reconstructing error value, < >>For the spatial domain of the data block +.>For position->Is used for mapping the data blocks in the low dimension,for position->Is a reconstructed feature data block of->Weight value for gradient error, +.>For position->Gradient of low-dimensional map data block, +.>For position->Gradient of reconstructed characteristic data block, +.>Weight value for advanced feature, +.>For position->High-level characteristic data block of the low-dimensional map of +.>For position->Is used for reconstructing the high-level characteristic data block.

The invention constructs a reconstruction error calculation formula which is used for carrying out reconstruction error calculation on the low-dimensional mapping data block and the reconstruction characteristic data block to generate reconstruction error data. The formula fully considers the spatial domain of the data blockPosition->Is +.>Position->Is a reconstructed feature data block->Weight value of gradient error +.>Position->Gradient of low-dimensional map data block +.>Position->Gradient of reconstructed characteristic data block +.>Weight value of advanced feature +.>Position->High-level characteristic data block of low-dimensional mapping of +.>Position->Reconstructed advanced feature data block +.>And the interaction relationship between the variables, constitute the following functional relationship:

By aligningIntegration is performed taking into account each point within the data block. The method is beneficial to globally measuring the reconstruction errors, and ensures that not only the errors of single points are small, but also the overall errors are controlled. By->Representing the error between the low-dimensional map data block and the reconstructed feature data block. The L2 norm is used to minimize the error result to ensure that the reconstructed feature data block is as close as possible to the low dimensional map data block, thereby preserving the overall shape and numerical information of the data block. And a larger penalty is applied to the points with large errors through square calculation, so that the influence of local outliers is reduced, and the robustness of the algorithm to noise and outliers is improved. By->Representing the difference between the gradients of the low-dimensional map data block and the reconstructed feature data block. By minimizing the difference result, it is ensured that the gradient of the reconstructed feature data block is as close as possible to the gradient of the low-dimensional map data block. Weight value of gradient error->Controlling the relative importance of the gradient error in the overall loss function. By adjusting->The importance between euclidean distance errors and gradient errors can be balanced. Greater->The importance of the gradient error increases. The whole spatial domain is integrated to globally scale the gradient error, ensuring that the gradient remains smooth across the whole data block. The squaring calculation helps to maintain the spatial smoothness of the data block, ensuring that excessive noise or sharp edges are not introduced to prevent over-processing. By- >Representing the difference between the high-level features of the low-dimensional map data block and the reconstructed feature data block. By minimizing this portion, it is beneficial to ensure that the advanced features of the reconstructed feature data block are as close as possible to the advanced features of the low-dimensional map data block. Weight value of advanced feature +.>For balancing the importance between euclidean distance errors and advanced feature errors. The whole spatial domain is integrated taking into account each point within the data block. The method is beneficial to globally measuring high-level characteristic errors, and ensures that not only is the error of a single point small, but also the overall error is controlled. The calculation of the square helps to ensure that important feature information in the low-dimensional map data block is preserved in the reconstructed feature data block. The functional relation can ensure that the data block can be accurately reconstructed, maintain the spatial smoothness and keep key high-level characteristic information. By adjusting the weight values, the relative importance of different parts in the whole loss function can be balanced to meet the requirements of specific applications, and the method is used for different reconstruction characteristic data blocks, so that the flexibility and applicability of the algorithm are improved.

Preferably, step S3 comprises the steps of:

step S31: acquiring hardware information data; building an entity relation model according to the hardware information data to generate a hardware entity relation model;

Step S32: similar data extraction is carried out on the hardware information data according to the hardware entity relation model, and heterogeneous entity relation data is generated;

step S33: performing component matching degree evaluation on the heterogeneous entity relationship data to generate component matching score data;

step S34: carrying out combined feasibility analysis on the heterogeneous entity relationship data to generate feasibility analysis data;

step S35: carrying out heterogeneous fitness evaluation through the component matching score data and the feasibility analysis data to generate heterogeneous fitness data;

step S36: and constructing a fusion map according to the isomerism fit data to generate an isomerism information fusion map.

The invention is helpful for visualizing and modeling the relationship between different entities in the hardware information data through the construction of the hardware entity relationship model. To facilitate a better understanding of the associations and dependencies between hardware devices. By extracting the similarity data, similarities and important connections between hardware devices can be identified. Helping to better understand the commonalities and differences in hardware devices. The heterogeneous entity relationship data integrates different data types in the hardware information data, so that the diversity of the hardware equipment is better known. Component matching evaluation helps determine the degree of matching between components of different hardware devices. Similarities and differences between hardware devices may be determined. Component relationships between system hardware devices are facilitated to support hardware identification and driver generation. The combined feasibility analysis helps determine whether a combination of different hardware devices is feasible or interoperable. Is critical to the integration and interoperability of hardware devices. Generating feasibility analysis data helps to understand the synergistic effect between hardware devices to support optimization of system performance and hardware selection. Heterogeneous fitness evaluation combines data from component matching and feasibility analysis to determine fitness between hardware devices. And the applicability of the hardware device can be better evaluated. The fusion map integrates heterogeneous fitness data, and the association, similarity and feasibility among hardware devices are visually presented. Helping to better understand the interrelationship between hardware devices.

Preferably, step S32 comprises the steps of:

step S321: performing similar hardware parameter evaluation on the hardware information data to generate hardware similar data;

step S322: performing context information matching on the hardware similar data to generate multidimensional information data; performing information source evaluation through the multidimensional information data to generate multidimensional information source data;

step S323: performing error calculation on the multidimensional information source data by using an information error evaluation formula to generate information error data; performing error correction on the multidimensional information data by using the information error number to generate multidimensional information correction data;

step S324: performing similarity evaluation based on the multidimensional information correction data to generate similar hardware information data;

step S325: performing comprehensive similarity calculation according to the similar hardware information data to generate comprehensive similar hardware data;

step S326: and extracting similar data of the comprehensive similar hardware data according to the hardware entity relation model to generate heterogeneous entity relation data.

The invention facilitates determining parameter similarity between different hardware devices through similar hardware parameter evaluation. To identify commonalities and differences between hardware devices. The context information matching is helpful to match the hardware similar data with other context information, generate multidimensional information data, and integrate hardware information from different dimensions. The system is aided in determining the trustworthiness of the hardware information to aid in evaluating the source of the hardware device information. The information error calculation helps the system evaluate the accuracy of the hardware information to determine the quality of the information. The multidimensional information correction data generation is beneficial to correcting information errors and improving the accuracy and the credibility of information. The similarity assessment corrects the data based on the multidimensional information, helping to determine similarity and correlation between hardware devices. To help the system filter and select the hardware information that best matches the user's needs. The comprehensive similarity calculation combines multidimensional similarity information among different hardware devices to provide a comprehensive hardware device similarity measurement. The system is facilitated to be matched with hardware equipment more accurately, and the accuracy of software driver generation is improved. Similar data extraction facilitates correlating the comprehensive similar hardware data with a hardware entity relationship model to generate heterogeneous entity relationship data to better understand relationships between different hardware devices.

Preferably, the information error evaluation formula in step S323 is as follows:

in the method, in the process of the invention,for information error value, ++>For the number of dimensions of the information data +.>Is the%>Individual dimension weight values +.>For the matching->Information data value->Is the anticipated->Information data value->Is->Standard deviation of individual dimension difference +.>As an exponential function +.>Diffusion rate parameter for information error, +.>For information diffusion parameter, ++>Adjusting the value for the deviation of the information diffusion parameter, +.>Attenuation rate parameter for information error, +.>For information error density value, +.>Is the rate of change of the information error density value over time.

The invention constructs an information error evaluation formula which is used for carrying out error calculation on multidimensional information source data to generate information error data. The formula fully considers the dimension number of the information dataInformation data->Individual dimension weight value +.>Matching->Personal information data value->Anticipated->Personal information data value->First->Standard deviation of individual dimension difference->Exponential function->Diffusion rate parameter of information error->Information diffusion parameter->Deviation adjustment value of information diffusion parameter +.>Attenuation rate parameter of information error +. >Information error Density value->Information error Density value change Rate with time +.>And the interaction relationship between the variables, constitute the following functional relationship:

by adjusting the dimension weight values, the contribution of different dimensions to the error term can be controlled, and if a dimension has a higher weight, its error has a greater impact on the total error. By passing throughThe deviation between the measured value and the expected value is measured to describe the magnitude of the error between the actual value and the expected value. Divided by->For normalizing the error in each dimension so that the errors between different dimensions can be compared. The larger the standard deviation, the less this dimension contributes to the total error. This section represents the error weighted sum of the different dimensions. The second part of the formula, by a larger +.>The value represents a more information error diffusion rateFast, the effect on errors is more pronounced. Less->The value indicates a slower diffusion rate with less impact on the error. />The value of (2) determines the extent of the spread of the information error, greater +.>The values mean a wider range of error diffusion and a greater impact on the overall error. By dividing by +.>The speed and the amplitude of information error correction are controlled, so that the formula is more flexibly adapted to different application scenes. / >And (3) introducing deviation for adjusting information diffusion parameters to enable the deviation to be closer to actual information diffusion conditions. Can be determined by sensitivity analysis>The value, i.e. by changing->Is to observe the change of the information error to determine +.>The degree of influence on the information error, thereby adjusting the deviation of the information diffusion parameter. />The distribution of information errors is shown. />For capturing the time evolution of the information error to take into account the effect of the change in the information error over time on the information error assessment. By taking into account->Dynamic correction of information errors and prediction of the trend of future information errors can be achieved. And the information error evaluation is more comprehensive, and the complexity of the information error in space and time can be captured. Meanwhile, the weight value and the deviation adjustment value in the formula can be adjusted according to actual conditions, and the method is applied to error calculation of data of different information sources, so that the flexibility and applicability of an algorithm are improved.

Preferably, step S4 comprises the steps of:

step S41: performing equipment matching evaluation on the hardware feature sparse data by utilizing the hardware information data to generate hardware matching degree data;

step S42: dividing the matching degree of the hardware matching degree data based on a preset matching degree threshold value, and generating hardware parameter data when the hardware matching degree data is larger than or equal to the preset matching degree threshold value; when the hardware matching degree data is smaller than the preset matching degree threshold value, executing step S43;

Step S43: carrying out hardware type matching on the hardware feature sparse data by utilizing the heterogeneous information fusion map to generate hardware parameter data;

step S44: designing input parameters based on hardware parameter data to generate program input parameters;

step S45: acquiring user demand data; user demand analysis is carried out based on the user demand data, and demand input parameters are generated;

step S46: carrying out parameter optimization design on the program input parameters by utilizing the demand input parameters to generate program input optimization parameters;

step S47: and performing driver design based on the program input optimization parameters to generate software driver data.

The method is helpful for determining the matching degree between the hardware feature sparse data and the known hardware equipment through equipment matching evaluation. Generating hardware matching degree data can be used for integrating hardware feature sparse data with known hardware information data to provide more accurate hardware equipment information. The matching degree division is helpful for screening out hardware feature sparse data with high matching degree with the known hardware equipment according to the matching data. This can improve accuracy of hardware identification. Through the heterogeneous information fusion map, the hardware equipment type corresponding to the hardware feature sparse data with low matching degree can be determined. Facilitating automatic matching and identification of hardware types. Generating hardware parameter data supports more accurate hardware device type matching, and is beneficial to improving the accuracy of driver design. Based on the hardware parameter data, the system may design input parameters of the program to ensure efficient interaction of the program with the hardware device, supporting efficient generation of the software driver. Helping to automatically configure software to suit the needs of a particular hardware device. The user demand data is used to analyze the user's desires and demands. Helping to determine the functions and configurations required by the software drivers. Generating the demand input parameters facilitates personalizing the software driver to meet the specific demands of the user. The parameter optimization design is helpful to adjust the program input parameters to meet the requirements of users and hardware equipment. The performance and adaptability of the driver are improved. Generating program input optimization parameters helps to optimize resource utilization to improve the efficiency and performance of the software driver. Based on the program input optimization parameters, a customized software driver can be automatically generated to meet the requirements of users and hardware equipment. The generation of software driver data supports an automatic software development flow, and development efficiency and accuracy are improved.

A software driven processing system based on artificial intelligence, comprising:

the hardware feature extraction module is used for acquiring hyperspectral remote sensing data of the hardware equipment to be identified; performing data preprocessing on the hyperspectral remote sensing data to generate standard remote sensing data; extracting equipment characteristics of the standard remote sensing data to generate hardware characteristic data;

the hardware characteristic compression module is used for carrying out data block division processing on the hardware characteristic data to generate a hardware characteristic data block; performing feature compression on the hardware feature data block to generate hardware feature sparse data;

the heterogeneous map construction module is used for acquiring hardware information data; similar data extraction is carried out on the hardware information data, and heterogeneous entity relation data is generated; constructing a fusion map according to the heterogeneous entity relationship data to generate a heterogeneous information fusion map;

the driver programming module is used for acquiring user demand data; carrying out hardware type identification on the hardware feature sparse data by utilizing the heterogeneous information fusion map to generate hardware parameter data; and performing driver design based on the hardware parameter data and the user demand data to generate software driver data.

The method has the beneficial effects that the noise and redundancy in the data can be eliminated through data preprocessing, the quality of the generated standard remote sensing data is ensured to be high, and the accuracy of the hardware characteristic data is improved. The generation of the hardware feature data can accurately extract key features of the hardware devices, including spectral features, which is critical for accurate distinction of different hardware devices. The data block division and the feature compression reduce the dimension of the data, reduce the complexity of subsequent calculation and accelerate the hardware type recognition process. Generating hardware feature sparse data is helpful for reducing the data density, so that the data is easier to process, and subsequent hardware feature analysis and recognition are supported. The heterogeneous information fusion map integrates hardware information data from different sources, establishes a relation between hardware entities and provides more comprehensive hardware information. The heterogeneous information fusion map provides a visual way to explore the relationships between hardware devices, helping to better understand the hardware ecosystem. The method is beneficial to ensuring the consistency of hardware information data, and avoiding contradiction and repetition among different data sources, thereby accurately identifying hardware equipment lacking training data. Hardware type identification using heterogeneous information fusion maps helps determine the type of hardware device for which to generate the appropriate drivers. The driver design based on the hardware parameter data and the user demand data can ensure that the generated software driver meets the user demand, and the user experience is improved. And a customized software driver is automatically generated, so that the requirement of manual programming is reduced, and the development efficiency is improved. Therefore, the software driving processing method and system based on artificial intelligence, disclosed by the invention, obtain the hardware feature sparse data through feature compression of the hardware feature data, and identify the hardware feature sparse data lacking training data by utilizing the constructed heterogeneous atlas, so that the corresponding hardware equipment is accurately identified, and the software driving program is stably generated.

Drawings

FIG. 1 is a schematic flow chart of steps of a software driven processing method based on artificial intelligence;

FIG. 2 is a flowchart illustrating the detailed implementation of step S2 in FIG. 1;

FIG. 3 is a flowchart illustrating the detailed implementation of step S3 in FIG. 1;

the achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.

Detailed Description

The following is a clear and complete description of the technical method of the present patent in conjunction with the accompanying drawings, and it is evident that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, are intended to fall within the scope of the present invention.

Furthermore, the drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. The functional entities may be implemented in software or in one or more hardware modules or integrated circuits or in different networks and/or processor methods and/or microcontroller methods.

It will be understood that, although the terms "first," "second," etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another element. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element, without departing from the scope of example embodiments. The term "and/or" as used herein includes any and all combinations of one or more of the associated listed items.

To achieve the above object, referring to fig. 1 to 3, an artificial intelligence-based software driven processing method includes the following steps:

In the embodiment of the present invention, as described with reference to fig. 1, a schematic flow chart of steps of an artificial intelligence-based software driving processing method of the present invention is provided, and in this example, the method includes the following steps:

in the embodiment of the invention, hyperspectral remote sensing data of the hardware equipment to be identified are acquired. This can be accomplished by various sensors and remote sensing devices to obtain the spectral information of the hardware device. And carrying out data preprocessing on the acquired hyperspectral remote sensing data so as to ensure the quality and usability of the data. The preprocessing step, such as removing noise, correcting spectral shifts, correcting atmospheric disturbances, etc., generates standard remote sensing data. And extracting the characteristics of the hardware equipment from the standard remote sensing data. Feature extraction methods, such as spectral feature extraction, texture feature extraction, shape feature extraction, etc., to describe unique features of the hardware device. Features extracted from the devices are combined into hardware feature data that will be used for subsequent hardware identification and driver generation.

in an embodiment of the invention, the hardware characteristic data is divided into smaller blocks for further processing. The block partitioning may be based on spatial or frequency domain characteristics of the data, ensuring that each block contains local information about the hardware device. After the block division, blocks of hardware feature data are generated, which contain local features of the hardware device. Each block may be considered as a partial representation of a hardware feature. And performing feature compression on the hardware feature data block to reduce the dimension and redundancy of the data. This can be achieved by methods such as Principal Component Analysis (PCA), singular Value Decomposition (SVD), wavelet transformation, etc. By feature compression, hardware feature sparse data is generated that has a low dimensionality, but retains critical hardware feature information.

In the embodiment of the invention, the hardware information data is collected, and the hardware information data comprises the specification, the characteristics, the technical parameters and the like of the hardware equipment. And performing similar data extraction on the hardware information data to determine the similarity between the hardware devices. Such as text similarity analysis, feature ratio peering methods may be used. After the similar data extraction, heterogeneous entity relationship data is generated, which describes relationships between hardware devices, such as similarities, dependencies, and the like between hardware devices. And constructing a fusion map by utilizing the heterogeneous entity relationship data, wherein the map correlates information of different hardware devices together to provide comprehensive hardware information and relationship.

In the embodiment of the invention, the requirement data of the user is obtained, and the data comprise the requirement and the expectation of the user on the hardware equipment, such as performance requirements, functional requirements and the like. And carrying out hardware type identification by utilizing the heterogeneous information fusion map and the hardware feature sparse data. Methods of machine learning models, pattern recognition, etc. may be used to determine the type of hardware device. After hardware type identification, hardware parameter data is generated, which includes detailed parameters and specifications of the hardware device. And carrying out driver programming based on the hardware parameter data and the user demand data so as to meet the demands of users. Including writing driver code, configuring hardware parameters, etc. Software driver data is generated based on the driver design, which data will be used to connect and control the hardware devices.

Preferably, step S1 comprises the steps of:

In the embodiment of the invention, the hyperspectral remote sensing data of the hardware equipment to be identified is collected, for example, the image or the spectrum data is acquired from a sensor or a camera. Preprocessing the acquired hyperspectral data, including denoising, radiation correction, geometric correction and the like, to generate standard remote sensing data. Device features are extracted from standard remote sensing data, for example, using methods such as spectrum analysis, feature point extraction, texture analysis, and the like. Mutation evaluation is performed on the extracted initial feature data to detect outliers or outliers in the data. Statistical methods or anomaly detection algorithms may be used to calculate anomaly values. From the mutation evaluation result, robustness evaluation data are generated, which describe the robustness of the initial feature data. Robustness may take into account data dispersion, noise, etc. According to a preset robustness scoring threshold, the robustness scoring threshold can be obtained by comprehensively considering factors such as data noise level, data quantity and task targets and is used for screening robustness evaluation data. And when the robustness evaluation data is larger than or equal to a preset threshold value, the robustness evaluation data is regarded as high robustness characteristic data, otherwise, the high robustness characteristic data is rejected. And carrying out association analysis on the high-robustness characteristic data to determine the correlation and the relation between the characteristics. For example, correlation coefficient analysis, covariance analysis, and the like are used. And carrying out association analysis on the high-robustness characteristic data to determine the correlation and the relation between the characteristics. This may include correlation coefficient analysis, covariance analysis, and the like. Mutual information between different features is calculated using a mutual information analysis method to measure the correlation and degree of information sharing between them. Mutual information can be calculated by a formula. And generating mutual information analysis data according to the calculated mutual information values, wherein the mutual information analysis data comprise the mutual information values among different features. And carrying out association correction on the hardware characteristic association data by using the mutual information analysis data. For example, the correlation coefficient is adjusted, the correlation is corrected, and the characteristic correlation correction data is generated, wherein the characteristic correlation correction data comprises the adjusted correlation and the correlation information. And carrying out necessary feature screening on the high-robustness feature data according to the feature association correction data. And selecting the characteristic with high correlation with the characteristic of the hardware equipment to generate hardware characteristic data.

Preferably, step S2 comprises the steps of:

As an example of the present invention, referring to fig. 2, the step S2 in this example includes:

in the embodiment of the invention, by dividing the hardware characteristic data into blocks, a sliding window or other data dividing technology is generally adopted. Each block contains a number of hardware characteristic data points. For each block, the hardware feature data points therein are combined into one hardware feature data block. This block will contain a subset of the hardware feature data.

in the embodiment of the invention, the hardware characteristic data block is optimized through the characteristic association correction data. For example by adjusting feature weights in the data blocks to improve the degree of discrimination of different features. And generating an optimized hardware characteristic data block according to the result of applying the characteristic association correction data, wherein the optimized hardware characteristic data block comprises the adjusted hardware characteristic data so as to improve the distinction degree.

In the embodiment of the invention, the feature quantity in the hardware feature data block is reduced by adopting the feature selection or feature dimension reduction technology, so that the calculation efficiency is improved and the redundant information is reduced. And generating a feature dimension reduction data block according to the result of the sparsification processing, wherein the feature dimension reduction data block contains hardware feature data subjected to dimension reduction processing so as to reduce the data dimension.

in the embodiment of the invention, by selecting an appropriate feature compression method, a linear method such as Principal Component Analysis (PCA) and a nonlinear method such as a self-encoder (Autoencoder) may be used. Compressing the features in the feature-reduced data block. For example, for PCA, feature-reduced data blocks are mapped to new feature space by linear transformation to preserve the most important features. The compressed data is combined into a feature compressed data block, which contains compressed features that are lower in dimension than the original data block.

In the embodiment of the invention, the compressed data blocks with different characteristics are combined into a large data block so as to contain the information of all the characteristics. During the merging process, data normalization or normalization may be required to ensure similar dimensions and distribution between different data blocks. And finally generating a hardware feature sparse data block which contains compression and combination information of all hardware features.

Preferably, step S24 comprises the steps of:

In the embodiment of the invention, the number of mapping codes is preset by determining how many low-dimensional maps with different dimensions are generated. And performing mapping encoding on the characteristic dimension-reduced data blocks for a plurality of times, and generating one dimension-reduced mapping data block at a time. Linear methods such as Principal Component Analysis (PCA), nonlinear methods such as self-encoder (Autoencoder), etc. may be used. And combining all the generated low-dimensional mapping data blocks to form multi-low-dimensional mapping data blocks, wherein each data block corresponds to one mapping coding frequency. And selecting a proper clustering method, such as K-means clustering or hierarchical clustering, for carrying out clustering analysis on the multi-low-dimensional mapping data blocks. The data points in the multi-low dimensional map data block are clustered, and are divided into different clusters. This can be done by calculating the similarity between the data points. For each cluster, the corresponding data points are assigned to a low-dimensional map data block and a reconstructed feature data block, respectively. Thus, each cluster corresponds to a low-dimensional map and reconstruction feature, resulting in an averaged low-dimensional map data block and reconstruction feature data block. And using a preset reconstruction error calculation formula, wherein the formula fully considers the data error, the gradient error and the advanced characteristic error to accurately and globally measure the reconstruction error, and performing reconstruction error calculation on the low-dimensional mapping data block and the reconstruction characteristic data block to measure the difference between the original data point and the reconstruction version thereof. Or by means of mean square error, structural similarity index, cross validation, etc. The reconstruction errors for each data point are summed to form reconstruction error data. To provide information about the quality of the reconstruction of the data points. And the accuracy of reconstruction is ensured. For each low-dimensional map data block and its corresponding reconstructed feature data block, a compression ratio is calculated. And associating the compression rate of each low-dimensional mapping data block with the corresponding reconstruction characteristic data block to generate compression rate data. According to the compression rate data, low-dimensional mapping data blocks with high compression rate are selected, which shows that the low-dimensional mapping data blocks have better compression effect. And selecting reconstruction characteristic data blocks with low reconstruction errors according to the reconstruction error data, wherein the reconstruction characteristic data blocks are better in the aspect of information retention. And finding a balance point by combining the compression rate and the reconstruction error data, and determining the compression dimension selection. Feature compression is performed for selected low-dimensional map data blocks to reduce dimensionality. This can be achieved by methods such as Principal Component Analysis (PCA), linear Discriminant Analysis (LDA), t-SNE, etc. The compressed data block will contain features of lower dimension.

The invention constructs a reconstruction error calculation formula for mapping the low-dimensional data block and the reconstruction characteristicsAnd carrying out reconstruction error calculation on the data block to generate reconstruction error data. The formula fully considers the spatial domain of the data blockPosition->Is +.>Position->Is a reconstructed feature data block->Weight value of gradient error +.>Position->Gradient of low-dimensional map data block +.>Position->Gradient of reconstructed characteristic data block +.>Weight value of advanced feature +.>Position->High-level characteristic data block of low-dimensional mapping of +.>Position->Reconstructed advanced feature data block +.>And the interaction relationship between the variables, constitute the following functional relationship:

Preferably, step S3 comprises the steps of:

As an example of the present invention, referring to fig. 3, the step S3 in this example includes:

in the embodiment of the invention, the information data related to the hardware equipment is collected, wherein the information data comprises hardware specifications, component information and the like. The collected data is cleaned, and missing values, noise and inconsistencies are processed to ensure the quality and consistency of the data. Useful features are extracted from the data to construct an entity relationship model. Text analysis, data mining, natural language processing, and the like may be used. And constructing a hardware entity relation model based on the extracted features. Tools such as graph databases or knowledge maps may be employed to represent relationships between entities.

in the embodiment of the invention, the similarity between different hardware information data is calculated based on the hardware entity relation model. This may use text similarity measures, graph algorithms, or deep learning models. And screening out the hardware information data with the similarity higher than the threshold value to generate heterogeneous entity relationship data. These similar data may be directed to the same type of hardware device or share similar features.

in the embodiment of the invention, the matching degree between different hardware components is calculated according to the heterogeneous entity relation data. The degree of matching may be evaluated based on attribute similarity, sharing relationships, or other relevant features. The result of the computation of the degree of matching of the components is mapped to a matching score for representing the degree of similarity between the hardware components. May be a continuous value or a discrete fraction.

in embodiments of the present invention, it is assessed whether hardware components can be connected together by based on their physical characteristics. This may require consideration of the type of interface, electrical connections, size and location, etc. It is checked whether the hardware components are compatible in technical specifications, protocols and standards. Including aspects that take into account communication protocols, data formats, and operating systems. It is evaluated whether the performance of the hardware component is sufficient to meet the user's needs. Factors such as processing power, storage capacity, transmission speed, and energy efficiency may need to be considered.

In the embodiment of the invention, the relative importance of different hardware components is determined by combining the component matching score and the feasibility analysis result. The fitness score is calculated using an appropriate algorithm (e.g., a weighted sum or other combination function). This score may represent the overall adaptability of the hardware component. According to actual requirements, a threshold value of the fitness score can be set to determine which hardware components are suitable for a specific task, and component data with low fitness is screened out.

In the embodiment of the invention, the hardware component and the fitness information are represented by selecting a proper map model, such as a knowledge map or a map database. The hardware components, fitness data, and other information related to the hardware feature data are integrated into the atlas. Such as assigning nodes and edges to graphs. Ensuring that the atlases support complex query operations for subsequent hardware selection and driver design. Graph query languages or APIs may be used to support these query operations.

Preferably, step S32 comprises the steps of:

In the embodiment of the invention, the similarity is found by comparing and matching each hardware parameter in the hardware information data. May include hardware specifications, functional characteristics, model numbers, etc. Similarity measures (such as cosine similarity or Jaccard similarity) are used to evaluate the degree of similarity between hardware parameters. The definition and scope of the context information is determined. Such as the environment of the hardware parameters, the usage scenario, related devices, etc. Matching similar hardware parameters with corresponding context information. This may be accomplished by associating context information for the hardware parameters or using a matching algorithm, such as a rule-based matching or machine learning method. And associating the matched context information with similar hardware parameters to generate multidimensional information data, wherein the multidimensional information data comprises the hardware parameters and the related context information. The preset information error evaluation formula is utilized, and the formula fully considers the variables such as errors, error diffusion rate, error attenuation rate and the like, so that the information error value is comprehensively and accurately evaluated to correct multidimensional information data, and a basis is provided for acquiring heterogeneous entity relations. Alternatively, a mean square error (Mean Squared Error) or other suitable error calculation method may be used. And comparing the multidimensional information data with the related information, and calculating an information error. I.e. the actual data is compared with the expected or reference data and then the selected error metric method is applied. The multidimensional information data is corrected using the error values. For example, the data is adjusted or optimized to generate multidimensional information correction data. The hardware parameters in the multidimensional information modification data are compared with other hardware information using an appropriate similarity measure (e.g., cosine similarity, euclidean distance, jaccard similarity, etc.). The similarity scores between each hardware parameter are combined into a composite similarity score. This may be achieved by simple averaging, weighted averaging or other combination methods. And associating the comprehensive similar hardware data with the hardware information data by using a hardware entity relation model. Information about the type, function, relationship, etc. of hardware may be included. And associating the comprehensive similar hardware data with the hardware information data to extract the hardware relationship with similarity. Techniques such as database querying, correlation operations, or using graph databases may be used. Heterogeneous entity relationship data is generated, which includes hardware information and related hardware relationships.

by adjusting the dimension weight values, the contribution of different dimensions to the error term can be controlled, and if a dimension has a higher weight, its error has a greater impact on the total error. By passing throughThe deviation between the measured value and the expected value is measured to describe the magnitude of the error between the actual value and the expected value. Divided by->For normalizing the error in each dimension so that the errors between different dimensions can be compared. The larger the standard deviation, the less this dimension contributes to the total error. This section represents the error weighted sum of the different dimensions. The second part of the formula, by a larger +.>The value indicates that the information error diffusion rate is faster and the effect on the error is more pronounced. Less->The value indicates a slower diffusion rate with less impact on the error. />The value of (2) determines the extent of the spread of the information error, greater +.>The values mean a wider range of error diffusion and a greater impact on the overall error. By dividing by +.>The speed and the amplitude of information error correction are controlled, so that the formula is more flexibly adapted to different application scenes. / >And (3) introducing deviation for adjusting information diffusion parameters to enable the deviation to be closer to actual information diffusion conditions. Can be determined by sensitivity analysis>The value, i.e. by changing->Is to observe the change of the information error to determine +.>The degree of influence on the information error, thereby adjusting the deviation of the information diffusion parameter. />Representing the distribution of information errorsThe condition is as follows. />For capturing the time evolution of the information error to take into account the effect of the change in the information error over time on the information error assessment. By taking into account->Dynamic correction of information errors and prediction of the trend of future information errors can be achieved. And the information error evaluation is more comprehensive, and the complexity of the information error in space and time can be captured. Meanwhile, the weight value and the deviation adjustment value in the formula can be adjusted according to actual conditions, and the method is applied to error calculation of data of different information sources, so that the flexibility and applicability of an algorithm are improved.

Preferably, step S4 comprises the steps of:

In the embodiment of the invention, the hardware information data and the hardware feature sparse data are subjected to matching analysis. Various matching algorithms and techniques, such as pattern matching, data mining algorithms, or deep learning methods, may be used to evaluate how well hardware parameters match hardware features. An appropriate matching metric is designed to quantify the degree of matching. The hardware matching degree data can be generated by capturing multiple aspects of matching and can be a single numerical index or can comprise multiple metrics. And setting a threshold value of the hardware matching degree according to application requirements. The choice of threshold depends on the distribution of the match data and the tolerance of hardware identification. Dividing the hardware matching degree data according to a threshold value to generate two categories: match success and match failure. And generating corresponding hardware parameter data for the successfully matched hardware features. And matching and classifying the hardware feature sparse data by using information in the heterogeneous information fusion map, including types, characteristics, relations and the like of hardware. The hardware features are compared with the profile information by selecting an appropriate matching algorithm, e.g., a method based on graph matching, semantic matching, or knowledge-graph matching. And generating corresponding hardware parameter data according to the matching result. The physical characteristics and specifications of the hardware parameters are converted into parameter forms acceptable to the program. Proper program input parameters are designed to meet the requirements of hardware equipment. Including determining the name, type, scope, and default values of the parameters. User demand data is collected, including demands in terms of desired functionality, performance, configuration, etc. of the hardware device. The user needs are analyzed, such as analyzing priorities, weights, constraints, etc. of the user needs. And outputting the analysis result as a demand input parameter. And optimizing program input parameters by using the requirement input parameters so as to meet the requirements of users and hardware parameters. The method can be realized by adjusting parameters, applying an optimization algorithm and the like. The generating program inputs optimization parameters that meet the dual requirements of hardware requirements and user requirements. The software driver is designed to communicate and control with the hardware device based on the program input optimization parameters. And coding and testing the designed driving program to ensure the normal operation and meet the hardware requirement. Complete software driver data is generated, including program code, configuration files, documents, and the like.

The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

The foregoing is only a specific embodiment of the invention to enable those skilled in the art to understand or practice the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. The software driving processing method based on artificial intelligence is characterized by comprising the following steps:

step S3: acquiring hardware information data; similar data extraction is carried out on the hardware information data, and heterogeneous entity relation data is generated; constructing a fusion map according to the heterogeneous entity relationship data to generate a heterogeneous information fusion map; wherein, step S3 includes the following steps:

step S36: constructing a fusion map according to the heterogeneous fitness data to generate a heterogeneous information fusion map;

Step S4: acquiring user demand data; carrying out hardware type identification on the hardware feature sparse data by utilizing the heterogeneous information fusion map to generate hardware parameter data; and (3) carrying out driver design based on the hardware parameter data and the user demand data to generate software driver data, wherein the step S4 comprises the following steps:

2. The artificial intelligence based software driven processing method according to claim 1, wherein the step S1 comprises the steps of:

3. The artificial intelligence based software driven processing method according to claim 1, wherein the step S2 comprises the steps of:

4. The artificial intelligence-based software driving processing method according to claim 3, wherein the step S24 comprises the steps of:

5. The method according to claim 4, wherein the reconstruction error calculation formula in step S243 is as follows:

in the method, in the process of the invention,reconstructing error value, < >>For the spatial domain of the data block +.>For position->Is a low-dimensional map data block of->For position->Is a reconstructed feature data block of->Weight value for gradient error, +.>For position->Gradient of low-dimensional map data block, +.>For position->Gradient of reconstructed characteristic data block, +. >Weight value for advanced feature, +.>For position->High-level characteristic data block of the low-dimensional map of +.>For position->Is used for reconstructing the high-level characteristic data block.

6. The artificial intelligence-based software driving processing method according to claim 1, wherein the step S32 comprises the steps of:

7. The artificial intelligence-based software driving processing method according to claim 6, wherein the information error evaluation formula in step S323 is as follows:

in the method, in the process of the invention,for information error value, ++>For the number of dimensions of the information data +.>Is the%>The weight value of each dimension is set to be equal to the weight value of each dimension,for the matching->Information data value->Is the anticipated->Information data value->Is->The standard deviation of the difference in the individual dimensions,as an exponential function +.>Diffusion rate parameter for information error, +.>For information diffusion parameter, ++>Adjusting the value for the deviation of the information diffusion parameter, +.>Attenuation rate parameter for information error, +.>For information error density value, +.>Is the rate of change of the information error density value over time.

8. An artificial intelligence based software driven processing system for executing the artificial intelligence based software driven processing method according to claim 1, the artificial intelligence based software driven processing system comprising: