Image scene library construction method and device, chip, computer equipment, storage medium and program product

Abstract

The invention relates to an image scene library construction method and device, a chip, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring project demand information and project image original data, and determining a plurality of types of scene libraries according to the project demand information; the project image original data is acquired through camera equipment; establishing a plurality of classification labels for each scene library; performing data extraction on the project image original data according to the classification label to obtain classification label data; and for each scene library, constructing a corresponding image scene library according to the classification label data. By adopting the method, the camera test evaluation efficiency can be improved.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to a method, apparatus, chip, computer device, computer-readable storage medium, and computer program product for constructing an image scene library. Background Technology

[0002] Cameras are crucial components in products such as smartphones, wearable devices, and automotive systems, and camera testing and verification are essential steps before mass production and market launch. During camera testing and evaluation, image scene libraries play a fundamental role, directly impacting the complexity of the testing process and the accuracy of the results.

[0003] However, the image scene library acquisition in related technologies adopts an on-demand approach, with staff acquiring images based on the temporary needs of the current test scenario. This method results in the inability to effectively inherit and reuse test sets collected from different projects, and the datasets are scattered across various disk files, reducing testing efficiency. Summary of the Invention

[0004] Based on this, it is necessary to provide a method, apparatus, chip, computer equipment, computer-readable storage medium, and computer program product for constructing an image scene library that can improve the efficiency of testing and evaluating camera equipment, in response to the above-mentioned technical problems.

[0005] Firstly, this application provides a method for constructing an image scene library, including:

[0006] The project requirements information and raw project image data are obtained, and multiple types of scene libraries are determined based on the project requirements information; the raw project image data is obtained through camera equipment.

[0007] For each of the aforementioned scenario libraries, multiple category tags are established;

[0008] Based on the classification labels, data is extracted from the original data of the project images to obtain classification label data;

[0009] For each of the aforementioned scene libraries, a corresponding image scene library is constructed based on the classification label data.

[0010] In one embodiment, the scene library includes a standard scene library; determining multiple types of scene libraries based on the project requirements information includes:

[0011] The project requirement information is parsed to obtain project requirement data and project phase identifiers;

[0012] Based on the project requirements data, match the corresponding project reference data;

[0013] A standard scenario dataset is determined based on the project reference data and the project phase identifier, and a standard scenario library is constructed based on the standard scenario dataset.

[0014] In one embodiment, the scenario library further includes a project scenario library and / or a problem scenario library; the step of determining multiple types of scenario libraries based on the project requirement information further includes:

[0015] Based on the project stage identifiers, project process data for multiple nodes in the project process is extracted from the original project image data, and a project scene library is constructed based on the project process data.

[0016] If there is abnormal data in the project process data, a problem scenario library is constructed based on the abnormal data.

[0017] In one embodiment, the establishment of multiple classification labels for each of the scene libraries includes:

[0018] Based on the environmental information of the original image data of the project, an environmental label is established, which includes one or more of the following: brightness label, color temperature label, and dynamic range label.

[0019] Based on the algorithm processing information corresponding to the original image data of the project, establish algorithm labels;

[0020] Based on the shooting metadata information of the original image data of the project, shooting condition labels are established, and the shooting condition labels include at least one of shooting time labels and ISO labels;

[0021] Based on the image content information of the original image data of the project, content tags and location tags are established.

[0022] In one embodiment, constructing a corresponding image scene library based on the classification label data for each of the scene libraries includes:

[0023] The classification label data is subjected to anomaly identification, and the anomaly data is removed to obtain cleaned classification label data;

[0024] Data distribution analysis is performed on the cleaned classification label data to generate an environment distribution map, an algorithm distribution map, and a scene content distribution map;

[0025] Based on the environmental distribution map, the algorithm distribution map, and the scene content distribution map, a data panel is constructed, and an image scene library is constructed based on the cleaned classification label data and the data panel.

[0026] Secondly, this application also provides an image scene library construction apparatus, comprising:

[0027] The data acquisition module is used to acquire project requirement information and raw project image data, and to determine multiple types of scene libraries based on the project requirement information; the raw project image data is acquired through camera equipment.

[0028] The tag generation module is used to create multiple category tags for each of the aforementioned scene libraries;

[0029] The data extraction module is used to extract data from the original data of the project image based on the classification labels to obtain classification label data;

[0030] The data processing module is used to construct a corresponding image scene library for each of the scene libraries based on the classification label data.

[0031] Thirdly, this application also provides a chip including a processor and a data interface, wherein the processor reads instructions stored in a memory through the data interface and is able to execute the steps described in the first aspect.

[0032] Fourthly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps described in the first aspect.

[0033] Fifthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps described in the first aspect.

[0034] In a sixth aspect, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps described in the first aspect.

[0035] The aforementioned image scene library construction method, apparatus, chip, computer equipment, computer-readable storage medium, and computer program product acquire project requirement information and raw project image data. Based on the project requirement information, multiple types of scene libraries are determined. The raw project image data is acquired through camera equipment, thereby establishing a standardized scene classification system. For each scene library, multiple classification labels are established, constructing a unified data organization framework for different types of scene libraries. By establishing a standardized classification label system, the originally scattered and disordered image data is classified and identified according to a unified standard. Data is extracted from the raw project image data based on the classification labels to obtain classification label data. For each scene library, a corresponding image scene library is constructed based on the classification label data. Through the scene library construction process based on classification label data, an image scene library system with clear classification and standardized management is formed, enabling different projects to share and reuse test data, and improving the testing and evaluation efficiency of camera equipment. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is an application environment diagram of an image scene library construction method in one embodiment;

[0038] Figure 2 This is a flowchart illustrating an image scene library construction method in one embodiment;

[0039] Figure 3 This is a schematic diagram of the overall framework for the digital construction of the scene library in one embodiment;

[0040] Figure 4 This is a schematic diagram illustrating the process of extracting digital tags from a scene library in one embodiment;

[0041] Figure 5 This is a flowchart illustrating step S208 of the image scene library construction method in one embodiment;

[0042] Figure 6 This is a structural block diagram of an image scene library construction device in one embodiment;

[0043] Figure 7 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0045] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0046] The image scene library construction method provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or located on the cloud or other network servers. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, drones, low-altitude aircraft, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, projection devices, etc. Portable wearable devices can include smartwatches, smart bracelets, head-mounted devices, etc. Head-mounted devices can be virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc. Server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0047] In one exemplary embodiment, such as Figure 2 As shown, an image scene library construction method is provided, which can be applied to... Figure 1 Taking server 104 as an example, the process includes steps S202 to S208. The overall framework for the digital construction of the scene library is as follows: Figure 3 As shown. Wherein:

[0048] Step S202: Obtain project requirement information and raw project image data, and determine multiple types of scene libraries based on the project requirement information.

[0049] The project requirements information refers to the specific requirements and specifications of the camera testing project, which may include project type identifier, project stage identifier, business function requirement list, quality requirement specifications, and test environment parameters. The raw image data refers to the unprocessed raw image files collected during the camera testing process; that is, the raw data obtained through the camera equipment. It can be stored in JPEG format and includes a corresponding XML (eXtensible Markup Language) parameter configuration file, recording information such as environmental parameters, algorithm parameters, and device parameters during image acquisition.

[0050] Server 104 can receive project requirement information and raw project image data transmitted from clients or other system modules. Server 104 can parse project configuration documents and extract basic project attribute information, including project number, project name, project type, and expected test objectives. Simultaneously, Server 104 can identify the current development stage of the project, including chip verification, algorithm integration, simulation debugging, or mass production testing, and convert this stage information into standardized project stage identifiers. Furthermore, Server 104 can extract a list of business function requirements from the project requirement information, identify the specific functional modules that need to be tested, such as automatic photo capture, multi-frame noise reduction, high dynamic range, night scene enhancement, and portrait optimization, and determine the type of scene library to be built according to preset scene classification rules. For projects requiring the establishment of standard test benchmarks, Server 104 can determine to build a standard scene library; for projects requiring process data management, Server 104 can determine to build a project scene library; and for projects with quality issues requiring special management, Server 104 can determine to build a problem scene library.

[0051] The scene library includes a standard scene library. The standard scene library is a foundational dataset used to provide input data for each stage of the project, from early-stage chip verification to later-stage simulation optimization. It is built based on the platform's supported functions, algorithm types, and business testing scenario dimensions. Specifically, it can include an Auto-photography full-scene image library, an MFNR (Multi-Frame Noise Reduction) scene library, an HDR (High Dynamic Range) scene library, a night scene image library, and a portrait scene library, among others.

[0052] For example, server 104 can parse project requirement information to obtain project requirement data and project phase identifiers; match corresponding project reference data based on project requirement data; determine scenario standard dataset based on project reference data and project phase identifiers; and construct a standard scenario library based on scenario standard dataset.

[0053] When parsing project requirements, server 104 can extract project requirement data from the project configuration file, including the test function list, performance indicator requirements, and test environment configuration, identify project phase identifiers, and determine which phase of the development lifecycle the project is currently in. Based on the extracted project requirement data, server 104 performs pattern matching in a pre-built project reference database to find historical project data most similar to the current project requirements as project reference data. Server 104 can combine the project reference data and project phase identifiers to generate a standard scenario dataset suitable for the current project. The dataset defines the specific scenario types, data size, and quality requirements that the standard scenario library should include. Finally, server 104 initializes the directory structure and index system of the standard scenario library according to the specification requirements of the standard scenario dataset.

[0054] The scenario library can further include a project scenario library and / or a problem scenario library. The project scenario library is a dataset derived from specific project process data, used for centralized data management and traceability. The problem scenario library is a dataset used to manage problems that occur during the project process. It is established based on the platform's supported functions, algorithm types, and problem types in business testing, and includes problem libraries for various quality evaluation dimensions such as brightness, color, and sharpness.

[0055] For example, server 104 extracts project process data for multiple nodes in the project process from the original project image data based on the project stage identifier, and constructs a project scenario library based on the project process data; if there is abnormal data in the project process data, it constructs a problem scenario library based on the abnormal data.

[0056] When constructing the project scenario library and problem scenario library, server 104 can determine the nodes that generate project data based on project stage identifiers, including chip verification nodes, algorithm integration nodes, simulation debugging nodes, and mass production testing nodes. Server 104 can identify and extract project process data corresponding to these nodes from the raw project image data, including test images, parameter configurations, and test results at each stage. Server 104 can establish a time-series index and node association relationship for project process data, forming a complete project data traceability chain, and construct the project scenario library based on this project process data. During project process data analysis, server 104 can use anomaly detection algorithms to identify abnormal patterns in the data, including image quality anomalies, parameter configuration anomalies, or test result anomalies. When abnormal data is detected, server 104 can classify these abnormal data separately and manage them hierarchically according to the anomaly type and severity.

[0057] Step S204: Create multiple category tags for each scene library.

[0058] For example, such as Figure 4As shown, server 104 can establish environmental tags based on the acquisition environment information of the original project image data. These environmental tags describe the image acquisition environment conditions and include one or more of the following: brightness tags, color temperature tags, and dynamic range tags. Brightness tags identify the image's illumination intensity level, color temperature tags identify the image's light source color temperature characteristics, and dynamic range tags identify the image's contrast range. Server 104 can also establish algorithm tags based on the algorithm processing information corresponding to the original project image data. Algorithm tags identify the type of algorithm used in the image processing and may include HDR algorithm tags, MFNR algorithm tags, night scene enhancement algorithm tags, portrait optimization algorithm tags, etc. According to the project... Based on the original image data, shooting condition tags are established. These tags record the device status and time information when the image was captured. The shooting time tag identifies the time period during which the image was captured, and the ISO tag identifies the camera sensor's light sensitivity setting. Shooting condition tags include at least one of these two tags. Based on the image content information of the original project image data, content tags and location tags are established. Content tags describe the specific objects and scene elements contained in the image, such as vegetation tags, building tags, sky tags, and road tags. Location tags are identifiers of the shooting location type inferred from the image content, such as indoor tags, outdoor tags, city street tags, and natural landscape tags.

[0059] Furthermore, server 104 can perform tag system construction operations for each established scene library, ensuring data consistency and interoperability between different scene libraries by establishing a unified classification tag framework. Server 104 can analyze the characteristics and uses of each scene library to determine the tag dimensions and tag types that need to be established for each scene library. For standard scene libraries, server 104 establishes a complete tag system including environment tags, algorithm tags, shooting condition tags, content tags, and location tags. For project scene libraries, server 104 can establish project stage tags, data source tags, and traceability relationship tags to achieve effective management of project process data. For problem scene libraries, server 104 can establish problem type tags, severity tags, and resolution status tags to support the classification management and processing tracking of problem data.

[0060] During the process of establishing environmental labels, server 104 can read the acquisition environment information from the parameter configuration file corresponding to the original image data of the project, including BV (Brightness Value) data, CT (Color Temperature) data, and DR (Dynamic Range) data. Server 104 can map BV values ​​to brightness labels according to preset grading standards, specifically including five levels: extremely low light, low light, normal light, strong light, and extremely strong light. Simultaneously, server 104 can convert CT values ​​to color temperature labels, including three categories: warm light, natural light, and cool light. Server 104 can also classify DR values ​​into dynamic range labels, including three levels: low dynamic range, medium dynamic range, and high dynamic range.

[0061] When creating algorithm tags, server 104 can parse the algorithm configuration file corresponding to the original image data of the project, identify the specific algorithm modules enabled during image processing, and extract the enabling status information of each algorithm module, including the enabling status of the HDR algorithm, the parameter settings of the MFNR algorithm, and the processing level of the night scene enhancement algorithm. Based on this algorithm enabling status information, server 104 can generate corresponding algorithm tags and supports the generation of composite algorithm tags. For example, when HDR and MFNR algorithms are enabled simultaneously, a composite algorithm tag of HDR and MFNR is generated.

[0062] When constructing shooting condition labels, server 104 can extract the shooting timestamp and ISO (International Organization for Standardization) sensitivity value by parsing the Exif (Exchangeable Image File Format) information in the header of the original JPEG image data. Server 104 can calculate the shooting time period based on the timestamp, dividing the 24 hours of a day into four time intervals: early morning, morning, afternoon, and night, generating corresponding shooting time labels. Server 104 can also generate sensitivity labels based on the ISO value range, including four levels: low sensitivity, medium sensitivity, high sensitivity, and ultra-high sensitivity.

[0063] Furthermore, when establishing content and location tags, server 104 can employ deep learning-based semantic segmentation technology to perform pixel-level content recognition and analysis on the raw project image data. Server 104 can load a pre-trained semantic segmentation model, which employs an encoder-decoder architecture. The encoder uses a ResNet101 (Residual Network 101, 101-layer residual network) structure for image feature extraction, and the decoder uses an UpperNet (Unified Perceptual Parsing Network) structure for pixel-level segmentation. Server 104 can preprocess the input image, including standardizing its size to 512 pixels by 512 pixels and normalizing pixel values, before inputting it into the semantic segmentation model. The model outputs pixel-level segmentation results, where each pixel is assigned a specific color value representing the content category to which that pixel belongs.

[0064] Furthermore, server 104 can convert the color information in the segmentation results into specific content category identifiers based on a predefined color coding mapping table. For example, green pixels correspond to vegetation categories, blue pixels to sky categories, gray pixels to building categories, and brown pixels to road categories. Server 104 can use a pixel proportion weighting algorithm to calculate the area proportion of each content category in the image. When the pixel proportion of a certain content category exceeds a preset effective proportion threshold, server 104 can add that category to the content label list, thereby avoiding invalid labels generated by noisy pixels. In the process of generating location labels, server 104 can establish a content-to-location mapping rule base and infer the type of location where the image was taken by analyzing the combination patterns of content labels. When a combination of sky, buildings, and roads is detected, server 104 can infer that it is an urban street location; when a combination of sky, vegetation, and water is detected, it can infer that it is a natural landscape location. Server 104 can also calculate a confidence score for each location label. When the confidence score is lower than a preset threshold, the data is marked as requiring manual review.

[0065] Through the above steps, server 104 has established a complete multi-dimensional classification and labeling system for each scene library, realizing the transformation of image data from its original state to structured descriptive information, and improving the manageability and retrieval of the data.

[0066] Step S206: Extract data from the original data of the project image based on the classification labels to obtain classification label data.

[0067] The classification label data is a structured dataset generated during the data extraction process, which may include complete label information, label confidence scores, and data quality assessment results for each image. For example, server 104 can establish a data extraction task queue and process the raw data of each project image one by one according to the storage path and file index order of the image files. During processing, server 104 can access the image files themselves and their corresponding parameter configuration files. Server 104 can employ a multi-threaded parallel processing mechanism to improve the processing efficiency of large-scale datasets while maintaining the consistency and reliability of the data extraction results.

[0068] During environmental label data extraction, server 104 can parse the XML parameter configuration file corresponding to the original data of each project image, locate and read the recorded environmental parameter information. Server 104 can extract BV luminance value data, converting the numerical luminance measurement results into standardized luminance labels according to a preset luminance level classification standard. Server 104 can extract CT color temperature data, mapping it to the corresponding color temperature label category based on the range of Kelvin temperature values. Server 104 can also extract DR dynamic range data, generating corresponding dynamic range labels by analyzing the brightness and darkness contrast range of the image. During the data conversion process, server 104 can apply adaptive threshold adjustment, dynamically adjusting the boundary values ​​of label classification based on the statistical distribution characteristics of the current dataset.

[0069] Regarding the extraction of shooting condition label data, server 104 can read the metadata information in the header of the original JPEG image data file through a dedicated Exif information parsing module. Server 104 can extract timestamp information, including the shooting date and specific time, and then calculate the corresponding time period label based on the time value. Server 104 divides the 24 hours into time periods such as early morning, morning, afternoon, and night, and assigns a corresponding shooting time label to each image. Server 104 can extract ISO sensitivity values ​​and generate ISO level labels based on the ISO value range. Server 104 also extracts additional metadata such as camera equipment information and lens parameters.

[0070] During the extraction of content and location tag data, server 104 can invoke the semantic segmentation processing module constructed in the above steps to perform deep learning analysis on the raw data of each project image. Server 104 can input the image into a pre-trained semantic segmentation model to obtain pixel-level segmentation result images. Based on color coding mapping rules, server 104 can convert the segmentation results into a list of content category identifiers and calculate the pixel proportion of each category. Server 104 applies a pixel proportion weighting algorithm to filter out content categories with proportions exceeding the effective threshold, generating the final content tag list. During the location tag generation process, server 104 can run a location inference engine, matching predefined location recognition rules according to the combination pattern of content tags to generate candidate location tags and calculate the corresponding confidence scores.

[0071] During the data fusion phase, server 104 can integrate the multi-dimensional label information extracted from different data sources to form a complete classification label data record for each image. It can also establish label consistency verification to check for logical conflicts between different labels of the same image. For example, server 104 can verify whether the brightness level of the environmental label matches the time of day of the shooting condition label, ensuring that there are no contradictory labels for strong light environments at night. When a label conflict is detected, server 104 can resolve the conflict based on the reliability weight of the data source.

[0072] Server 104 can also perform data quality control, calculating a comprehensive quality score for each category label data. The quality score includes multiple dimensions such as label integrity score, label confidence score, and data consistency score. Server 104 can classify and process the data according to the quality score results, directly incorporating high-quality data into subsequent processing, marking medium-quality data as requiring sampling verification, and classifying low-quality data into an anomaly queue for special processing. It can also log anomaly data, recording in detail various anomalies encountered during data extraction, including file corruption, missing parameters, and label conflicts.

[0073] Step S208: For each scene library, construct the corresponding image scene library based on the classification label data.

[0074] For example, server 104 establishes a hierarchical database storage architecture, with the top layer categorized into three scene library types, allocating independent storage space and management methods for each scene library. For the standard scene library, server 104 can divide it into multiple sub-libraries according to functional modules, establishing dedicated data storage areas for Auto photography full-scene sub-libraries, MFNR scene sub-libraries, HDR scene sub-libraries, night scene sub-libraries, and portrait scene sub-libraries. For the project scene library, server 104 can be categorized according to project stages, establishing data storage partitions for different stages such as chip verification, algorithm integration, simulation debugging, and mass production testing. For the problem scene library, server 104 can be categorized according to problem type and severity, establishing dedicated storage areas for quality dimensions such as brightness problem library, color problem library, and sharpness problem library. Server 104 can also create corresponding indexes for each scene library. Server 104 can create dedicated index tables for environment tags, algorithm tags, shooting condition tags, content tags, and location tags, and can simultaneously create composite indexes combining tags to support multi-condition queries, using a combination of hash indexes and B-tree indexes.

[0075] During the data import process, server 104 can automatically classify and store the categorized label data according to the scene library type and label characteristics. Each data record includes the storage path of the image file, complete label information, data quality score, import timestamp, and data source information. Server 104 can record the reference relationship and data inheritance between project data and the standard scene library. Between the project scene library and the problem scene library, server 104 can establish a traceability chain for problem data, recording the generation node and processing status of problem data. Between the problem scene library and the standard scene library, server 104 can establish feedback and supplementation tasks for key and common problems to the standard library.

[0076] In the aforementioned method for constructing an image scene library, by acquiring project requirement information and raw image data, and determining multiple types of scene libraries based on the project requirement information, a standardized scene classification system can be established. For each scene library, multiple classification labels are created, constructing a unified data organization framework for different types of scene libraries. By establishing a standardized classification label system, the originally scattered and disordered image data is classified and labeled according to a unified standard. Data is extracted from the raw image data of the project based on the classification labels to obtain classification label data. For each scene library, a corresponding image scene library is constructed based on the classification label data. Through the scene library construction process based on classification label data, an image scene library system with clear classification and standardized management is formed, enabling different projects to share and reuse test data, thereby improving the efficiency of camera testing and evaluation.

[0077] In one exemplary embodiment, such as Figure 5 As shown, step S208 includes steps S302 to S306. Wherein:

[0078] Step S302: Identify and remove abnormal data from the classification label data to obtain cleaned classification label data.

[0079] Anomaly identification can be the process of detecting data records in classification label data that deviate from the normal distribution pattern through statistical methods and machine learning algorithms.

[0080] For example, server 104 can establish an evaluation framework for anomaly detection, employing corresponding detection algorithms for different types of tag data. For numerical data such as BV brightness value, CT color temperature value, and DR dynamic range value in environmental tags, server 104 can use the three-standard-deviation criterion for outlier detection, calculating the mean and standard deviation of each numerical tag, and marking data exceeding the mean by plus or minus three standard deviations as outliers. For categorized data such as algorithm tags, content tags, and location tags, server 104 can use the Isolation Forest algorithm for anomaly pattern recognition, identifying anomalous records by analyzing the rarity of tag combinations and the degree of deviation from normal patterns.

[0081] Furthermore, server 104 can perform label consistency verification, cross-validating different dimensional labels of the same image, checking the temporal logical consistency between environmental labels and shooting condition labels, verifying the matching relationship between brightness level and shooting time, and identifying unreasonable label combinations such as strong light environments at night or extremely low light environments during the day. Server 104 can also verify the applicability matching between algorithm labels and environmental labels, checking the rationality of enabling specific algorithms under corresponding environmental conditions; for example, night scene enhancement algorithms should primarily be enabled in low light environments. When label conflicts are detected, server 104 can resolve conflicts based on the reliability weight and confidence score of the data source, retaining the label information with higher credibility or marking the entire record as requiring manual review.

[0082] During data uniformity analysis, server 104 can calculate the probability distribution characteristics of labels in each dimension, generate distribution histograms and statistical reports, perform uniformity tests on environmental labels, calculate the variance and skewness of the data distribution in the three dimensions of BV, CT, and DR, and identify scarce and redundant regions in the data distribution. When the amount of data in a certain environmental parameter interval is less than 5% of the total data, server 104 can mark that interval as a scarce data region, reminding users that test data under the corresponding conditions needs to be supplemented.

[0083] In some embodiments, server 104 can employ a multi-dimensional uniformity evaluation algorithm, analyzing not only the data distribution in a single dimension but also the cross-distribution uniformity between different dimensions. Server 104 can construct two-dimensional and three-dimensional cross-distribution matrices to analyze the distribution characteristics of combined scenarios such as low-light and indoor environments, and strong-light and outdoor environments, calculating the data density of various scenario combinations and identifying blind spots and overlapping areas in the test coverage. When data for certain important scenario combinations is found to be severely insufficient, server 104 can generate data supplementation suggestions to guide the focus of subsequent data collection efforts.

[0084] When performing data removal operations, server 104 can perform tiered processing. For slightly anomalous data records, server 104 can attempt to repair them, such as correcting slightly off-center numerical labels through interpolation or correcting logically conflicting classification labels through rule-based reasoning. For severely anomalous data that cannot be repaired, server 104 can remove it from the main dataset but retain it in the anomalous data archive for subsequent analysis.

[0085] Step S304: Perform data distribution analysis on the cleaned classification label data to generate an environment distribution map, an algorithm distribution map, and a scene content distribution map.

[0086] For example, server 104 can calculate the distribution characteristics and statistical indicators for different tag dimensions such as environment tag, algorithm tag, shooting condition tag, content tag, and location tag. During the environment distribution map generation process, server 104 can construct a BV-CT distribution scatter plot with BV brightness value on the x-axis and CT color temperature value on the y-axis, using the environmental parameters of each image as a data point in the scatter plot. The distribution density and clustering pattern of the points reflect the coverage of the dataset under different lighting and color temperature conditions. Server 104 can construct a BV-DR distribution scatter plot, with BV brightness value on the x-axis and DR dynamic range value on the y-axis, to analyze the correlation distribution characteristics between brightness and dynamic range. Server 104 can use color coding and point size variations to represent the density distribution of data points, and use heatmap effects to highlight concentrated and sparse areas of the dataset.

[0087] Furthermore, server 104 can use density clustering algorithms to identify clustered and blank areas of environmental parameters and automatically generate an environmental coverage heatmap. Server 104 can divide the environmental parameter space into grid areas, calculate the number of data points in each grid, and generate a density heatmap. High-density areas in the heatmap are represented by warm colors, indicating environmental conditions with sufficient test data; low-density areas are represented by cool colors, indicating environmental conditions with insufficient test coverage. Based on the distribution pattern of the heatmap, server 104 can automatically identify test blind spots and key supplementary areas and generate an environmental coverage analysis report.

[0088] Regarding algorithm distribution map generation, Server 104 can statistically analyze the frequency and usage percentage of various algorithm tags in the cleaned classification label data, calculate the usage ratio of individual algorithms such as HDR, MFNR, night scene enhancement, and portrait optimization, and generate pie charts of algorithm usage frequency. Server 104 can analyze the usage patterns of algorithm combinations, statistically analyze the frequency of different algorithm combinations, generate bar charts of algorithm combination distribution, and establish algorithm effect correlation analysis. It can also analyze the applicability and activation frequency of different algorithms under specific environmental conditions based on the combination of algorithm tags and environmental tags. For example, Server 104 can analyze the activation rate of the HDR algorithm in high dynamic range scenes and the usage pattern of the MFNR algorithm in low-light environments.

[0089] During the construction of the scene content distribution map, Server 104 can statistically analyze the frequency and proportion of various content tags in the dataset, calculate the data volume of various scene content categories such as vegetation, buildings, sky, roads, and water bodies, and generate a content distribution bar chart to display scene diversity. Server 104 can analyze the distribution ratio of location tags, calculate the proportion of location types such as indoor, outdoor, city streets, and natural landscapes, and generate a location distribution pie chart. Server 104 can also introduce content complexity evaluation metrics, calculating scene complexity scores based on the number of content categories contained in a single image and the complexity of its spatial distribution. Server 104 can also calculate corresponding statistical characteristic indicators, including mean, median, standard deviation, skewness, and kurtosis, generating a detailed data distribution analysis report. The report includes a description of the distribution characteristics of each dimension, outlier analysis, coverage assessment, and improvement suggestions. Through the above steps, Server 104 can generate comprehensive and intuitive data distribution analysis results, revealing the coverage and distribution characteristics of the dataset across various test dimensions through multi-dimensional visualization analysis.

[0090] Step S306: Construct a data panel based on the environment distribution map, algorithm distribution map, and scene content distribution map, and construct an image scene library based on the cleaned classification label data and the data panel.

[0091] The data dashboard is a comprehensive visual management interface built upon environmental distribution maps, algorithm distribution maps, and scene content distribution maps, providing interactive data analysis, query, and management functions.

[0092] For example, server 104 can integrate environment distribution maps, algorithm distribution maps, and scene content distribution maps into a unified visualization panel, providing intuitive data display and analysis functions through the client. The data panel can support multi-dimensional data filtering and combined queries, and users can select specific environmental conditions, algorithm types, or scene content to retrieve data through a graphical interface. Server 104 can also integrate real-time data statistics functions into the data panel, dynamically displaying the current dataset's size, quality indicators, and coverage evaluation results.

[0093] Furthermore, server 104 can generate project feature profiles based on the project's data collection preferences, including environment preference analysis, scenario preference statistics, and algorithm preference evaluation. Server 104 can display the feature distribution of a project across various testing dimensions using radar charts, and can support parallel comparative analysis of data from multiple projects, generating comparative radar charts and difference analysis reports to intuitively display the differences between different projects in terms of scenario coverage, algorithm usage patterns, and testing depth.

[0094] During the construction of the image scene library, server 104 can establish a hierarchical database storage structure based on the scene library type. At the first layer, server 104 can perform basic classification according to standard scene libraries, project scene libraries, and problem scene libraries, allocating independent storage space and management permissions for each scene library type. At the second layer, server 104 can further subdivide according to specific scene types, establishing dedicated data storage areas under the standard scene library, such as Auto Photography Full Scene Sub-library, MFNR Scene Sub-library, HDR Scene Sub-library, Night Scene Scene Sub-library, and Portrait Scene Sub-library. At the third layer, server 104 can establish a detailed index structure according to tag dimensions, creating efficient index tables for environment tags, algorithm tags, shooting condition tags, content tags, and location tags, supporting fast retrieval operations based on arbitrary tag combinations.

[0095] In some embodiments, server 104 can classify and store the cleaned classification label data according to a hierarchical storage structure. Each data record contains information such as the complete image file path, full-dimensional label information, data quality score, and processing timestamp, and can establish a cross-level association index system to record data reference relationships and dependencies between different scenario libraries. For example, when key issue data in the problem scenario library needs to be supplemented to the standard scenario library, server 104 can establish corresponding data flow records and version association information.

[0096] Through the above steps, server 104 can build a fully functional and clearly structured image scene library system, realizing the transformation from classification label data to a complete scene library product. By building a visual data panel and an efficient retrieval system, the usability and user experience of the image scene library can be significantly improved, and standardized management, efficient retrieval, and cross-project reuse of test data can be achieved.

[0097] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0098] Based on the same inventive concept, this application also provides an image scene library construction apparatus for implementing the image scene library construction method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more image scene library construction apparatus embodiments provided below can be found in the limitations of the image scene library construction method described above, and will not be repeated here.

[0099] In one exemplary embodiment, such as Figure 6 As shown, an image scene library construction device is provided, including: a data acquisition module 402, a label generation module 404, a data extraction module 406, and a data processing module 408, wherein:

[0100] The data acquisition module 402 is used to acquire project requirement information and raw project image data, and to determine multiple types of scene libraries based on the project requirement information; the raw project image data is acquired through camera equipment.

[0101] The tag generation module 404 is used to create multiple category tags for each scene library;

[0102] Data extraction module 406 is used to extract data from the original data of the project image based on the classification labels to obtain classification label data;

[0103] The data processing module 408 is used to construct corresponding image scene libraries based on the classification label data for each scene library.

[0104] In one embodiment, the scenario library includes a standard scenario library; the data acquisition module 402 is specifically used to parse project requirement information to obtain project requirement data and project stage identifiers; match corresponding project reference data according to the project requirement data; determine the scenario standard dataset according to the project reference data and project stage identifiers, and construct a standard scenario library according to the scenario standard dataset.

[0105] In one embodiment, the scenario library further includes a project scenario library and / or a problem scenario library; the data acquisition module 402 is also used to extract project process data of multiple nodes in the project process from the original project image data according to the project stage identifier, and to construct a project scenario library based on the project process data; and to construct a problem scenario library based on the abnormal data if there is abnormal data in the project process data.

[0106] In one embodiment, the tag generation module 404 is specifically used to: establish an environment tag based on the acquisition environment information of the original project image data, wherein the environment tag includes one or more of a brightness tag, a color temperature tag, and a dynamic range tag; establish an algorithm tag based on the algorithm processing information corresponding to the original project image data; establish a shooting condition tag based on the shooting metadata information of the original project image data, wherein the shooting condition tag includes at least one of a shooting time tag and a sensitivity tag; and establish a content tag and a location tag based on the image content information of the original project image data.

[0107] In one embodiment, the data processing module 408 is specifically used to identify abnormal data in the classification label data and remove abnormal data to obtain cleaned classification label data; to perform data distribution analysis on the cleaned classification label data to generate an environment distribution map, an algorithm distribution map, and a scene content distribution map; to construct a data panel based on the environment distribution map, the algorithm distribution map, and the scene content distribution map; and to construct an image scene library based on the cleaned classification label data and the data panel.

[0108] Each module in the aforementioned image scene library construction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0109] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 7As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media to run. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for constructing an image scene library.

[0110] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0111] In one exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to perform the following steps: acquiring project requirement information and raw project image data, and determining multiple types of scene libraries based on the project requirement information; the raw project image data is acquired through a camera device; multiple classification labels are established for each scene library; data extraction is performed on the raw project image data based on the classification labels to obtain classification label data; and a corresponding image scene library is constructed for each scene library based on the classification label data.

[0112] In one embodiment, when the processor executes the computer program, it further performs the following steps: parsing project requirement information to obtain project requirement data and project phase identifiers; matching corresponding project reference data based on the project requirement data; determining a scenario standard dataset based on the project reference data and project phase identifiers; and constructing a standard scenario library based on the scenario standard dataset.

[0113] In one embodiment, when the processor executes the computer program, it also performs the following steps: extracting project process data of multiple nodes in the project process from the original project image data according to the project stage identifier, and constructing a project scenario library based on the project process data; and constructing a problem scenario library based on the abnormal data if there is abnormal data in the project process data.

[0114] In one embodiment, when the processor executes the computer program, it further implements the following steps: establishing an environment label based on the acquisition environment information of the raw project image data, wherein the environment label includes one or more of a brightness label, a color temperature label, and a dynamic range label; establishing an algorithm label based on the algorithm processing information corresponding to the raw project image data; establishing a shooting condition label based on the shooting metadata information of the raw project image data, wherein the shooting condition label includes at least one of a shooting time label and a sensitivity label; and establishing a content label and a location label based on the image content information of the raw project image data.

[0115] In one embodiment, when the processor executes the computer program, it further performs the following steps: identifying abnormal data in the classification label data and removing abnormal data to obtain cleaned classification label data; performing data distribution analysis on the cleaned classification label data to generate an environment distribution map, an algorithm distribution map, and a scene content distribution map; constructing a data panel based on the environment distribution map, the algorithm distribution map, and the scene content distribution map, and constructing an image scene library based on the cleaned classification label data and the data panel.

[0116] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0117] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0118] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0119] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0120] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0121] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for constructing an image scene library, characterized in that, The method includes: The project requirements information and raw project image data are obtained, and multiple types of scene libraries are determined based on the project requirements information; the raw project image data is obtained through camera equipment. For each of the aforementioned scenario libraries, multiple category tags are established; Based on the classification labels, data is extracted from the original data of the project images to obtain classification label data; For each of the aforementioned scene libraries, a corresponding image scene library is constructed based on the classification label data.

2. The method according to claim 1, characterized in that, The scenario library includes a standard scenario library; based on the project requirements information, multiple types of scenario libraries are determined, including: The project requirement information is parsed to obtain project requirement data and project phase identifiers; Based on the project requirements data, match the corresponding project reference data; A standard scenario dataset is determined based on the project reference data and the project phase identifier, and a standard scenario library is constructed based on the standard scenario dataset.

3. The method according to claim 2, characterized in that, The scenario library also includes a project scenario library and / or a problem scenario library; determining multiple types of scenario libraries based on the project requirements information further includes: Based on the project stage identifiers, project process data for multiple nodes in the project process is extracted from the original project image data, and a project scene library is constructed based on the project process data. If there is abnormal data in the project process data, a problem scenario library is constructed based on the abnormal data.

4. The method according to claim 1, characterized in that, For each of the aforementioned scene libraries, multiple classification tags are established, including: Based on the environmental information of the original image data of the project, an environmental label is established, which includes one or more of the following: brightness label, color temperature label, and dynamic range label. Based on the algorithm processing information corresponding to the original image data of the project, establish algorithm labels; Based on the shooting metadata information of the original image data of the project, shooting condition labels are established, and the shooting condition labels include at least one of shooting time labels and ISO labels; Based on the image content information of the original image data of the project, content tags and location tags are established.

5. The method according to claim 4, characterized in that, The step of constructing a corresponding image scene library for each of the aforementioned scene libraries based on the classification label data includes: The classification label data is subjected to anomaly identification, and the anomaly data is removed to obtain cleaned classification label data; Data distribution analysis is performed on the cleaned classification label data to generate an environment distribution map, an algorithm distribution map, and a scene content distribution map; Based on the environmental distribution map, the algorithm distribution map, and the scene content distribution map, a data panel is constructed, and an image scene library is constructed based on the cleaned classification label data and the data panel.

6. An image scene library construction apparatus, characterized in that, The device includes: The data acquisition module is used to acquire project requirement information and raw project image data, and to determine multiple types of scene libraries based on the project requirement information; the raw project image data is acquired through camera equipment. The tag generation module is used to create multiple category tags for each of the aforementioned scene libraries; The data extraction module is used to extract data from the original data of the project image based on the classification labels to obtain classification label data; The data processing module is used to construct a corresponding image scene library for each of the scene libraries based on the classification label data.

7. A chip, characterized in that, The chip includes a processor and a data interface. The processor reads instructions stored in the memory through the data interface and can execute the steps of the method according to any one of claims 1 to 5.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images