Scene adaptive acquisition method and system of image acquisition card

Abstract

The embodiment of the invention relates to the technical field of image acquisition, in particular to a scene adaptive acquisition method and system for an image acquisition card, and the method comprises the steps: carrying out the collection and conversion of an optical signal of an optical field of view based on an image acquisition mode, and obtaining original image data; and arranging the original image data at each moment to generate an image information stream, analyzing a visual target and an environment scene for the optical field of view to obtain image acquisition scene characteristics, selecting key frames for the image information stream according to the image acquisition scene characteristics to obtain a reference information set, and sending the reference information set to a cloud server, and carrying out adaptive analysis of image acquisition on the reference information set through an optical acquisition algorithm pre-deployed in a cloud server so as to issue an optimization instruction to optimize an image acquisition mode.

Description

TECHNICAL FIELD

[0001] Embodiments of the present disclosure relate to the technical field of image acquisition, and more particularly, to a scene adaptive acquisition method and system of an image acquisition card. BACKGROUND

[0002] In the security monitoring scene, the monitoring range is wide, the environment is complex and changeable, including different light conditions and various scenes. The traditional image acquisition mode is difficult to adapt to these complex scenes, resulting in blurred images, difficult target recognition, and inability to clearly capture the characteristics of personnel or vehicles in some scenes, or loss of key information, making it difficult to acquire clear and accurate images in various environments. SUMMARY

[0003] In view of this, the present disclosure provides a scene adaptive acquisition method and system of an image acquisition card to acquire clear and accurate images in various environments.

[0004] According to a first aspect of the present disclosure, a scene adaptive acquisition method of an image acquisition card is provided, comprising: Based on the image acquisition mode, the optical signal acquisition and conversion of the optical field of view are performed to obtain raw image data; The raw image data at each time is arranged to generate an image information stream to analyze the visual target and the environment scene of the optical field of view to obtain image acquisition scene characteristics; According to the image acquisition scene characteristics, the key frames of the image information stream are selected to obtain a reference information set, and the reference information set is sent to a cloud server; The reference information set is analyzed for image acquisition adaptability by an optical acquisition algorithm pre-deployed on the cloud server, and the image acquisition mode is optimized according to the optimization instruction.

[0005] According to a second aspect of the present disclosure, a scene adaptive acquisition system of an image acquisition card is provided for implementing the scene adaptive acquisition method of the image acquisition card according to any one of the first aspect, comprising: An image acquisition module is configured to acquire and convert optical signals of an optical field of view based on an image acquisition mode to obtain raw image data; An image analysis module is configured to arrange raw image data at each time to generate an image information stream to analyze the visual target and the environment scene of the optical field of view to obtain image acquisition scene characteristics; An information reference module is configured to select key frames of the image information stream according to the image acquisition scene characteristics to obtain a reference information set, and send the reference information set to a cloud server; An acquisition optimization module is configured to perform adaptive analysis on the reference information set by an optical acquisition algorithm pre-deployed on a cloud server, and then generate optimization instructions to optimize the image acquisition mode.

[0006] The technical solution of the present disclosure has the following beneficial effects: The present disclosure optimizes the acquisition mode according to different scene characteristics, improves the image acquisition quality, analyzes the key frames according to the scene characteristics, and sends the key frames to the cloud for analysis. The optical acquisition parameters can be accurately adjusted to adapt to complex and variable environments, such as different lightings and scenes. In the fields of security and industrial detection, the system can ensure clear and accurate image acquisition, improve target recognition accuracy and detection efficiency, reduce manual intervention, enhance the degree of system automation and intelligence, reduce costs, and improve overall performance. BRIEF DESCRIPTION OF DRAWINGS

[0007] Figure 1 A step schematic diagram of a scene adaptive acquisition method of an image acquisition card in the present exemplary embodiment is shown; Figure 2 A structure schematic diagram of a scene adaptive acquisition system of an image acquisition card in the present exemplary embodiment is shown. DETAILED DESCRIPTION

[0008] To make the objectives, technical solutions, and advantages of the present disclosure clearer, the technical solutions in the embodiments of the present disclosure will be described clearly and completely below with reference to the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present disclosure. The embodiments in the present disclosure and the features in the embodiments can be combined with each other arbitrarily without conflict. Moreover, although a logical order is shown in the flowchart, in some cases, the steps shown or described can be performed in an order different from that herein.

[0009] In the specification and claims of the present disclosure, the term "comprising" and any variations thereof are intended to cover the protection without exclusivity. For example, a process, method, system, product, or device including a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed, or can optionally include other steps or units inherent to the process, method, product, or device.

[0010] One or more of the embodiments of the present disclosure, "multiple" refers to two or more than two. "And / or", describing the relationship between the associated objects, means that there can be three kinds of relationships, for example, A and / or B, which can represent: the existence of A alone, the existence of A and B at the same time, and the existence of B alone, where A and B can be singular or plural. The character " / " generally represents the relationship between the front and rear associated objects as "or". "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or multiple items. For example, at least one of a, b or c, can represent: a, b, c, a-b, a-c, b-c, or a-b-c, where a, b, and c can be single or multiple.

[0011] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-described drawings are used to distinguish similar objects, and do not necessarily describe a specific order, sequence, size, and priority, for example, the first dialog information and the second dialog information in the embodiments of the present disclosure, which are only used to distinguish different dialog information. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Rather, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0012] The exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, which are schematic illustrations of the present disclosure and are not necessarily drawn to scale. Some of the block diagrams shown in the drawings can be functional entities, which do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or in hardware modules or integrated circuits, or in networks, processors or microcontrollers. The embodiments can be implemented in various forms and should not be understood as being limited to the examples set forth herein. The features, structures or characteristics described in the present disclosure can be combined in any suitable manner in one or more embodiments. In the following description, many specific details are provided to give a sufficient understanding of the embodiments of the present disclosure. However, those skilled in the art should realize that one or more specific details can be omitted in implementing the technical solutions of the present disclosure, or other methods, components, devices, steps, etc. can be used instead of one or more specific details. It should be noted that in the embodiments of the present disclosure, the propagation, use, etc. of data complies with relevant national laws and regulations.

[0013] Please refer to Figure 1 As shown in the drawings, the embodiments of the present disclosure provide a scene adaptive acquisition method of an image acquisition card, comprising: S1: collecting and converting optical signals of the optical field of view based on an image collection mode to obtain raw image data; S2: arranging the raw image data at each time to generate an image information stream to analyze visual targets and environmental scenes of the optical field of view to obtain image collection scene characteristics; S3: selecting key frames of the image information stream according to the image collection scene characteristics to obtain a reference information set, and sending the reference information set to a cloud server; S4: performing adaptability analysis of the reference information set by an optical collection algorithm pre-deployed on the cloud server, and sending optimization instructions to optimize the image collection mode.

[0014] In step S1 of the embodiments provided in the disclosure, according to the selected image collection mode, various optical collection project parameters such as exposure time, gain, resolution, sampling frequency, etc. are accurately set. After completing the parameter deployment, the image collection card captures and collects the light in the optical field of view using its optical sensor to obtain optical signals. Then, the analog optical signals are converted into discrete digital signals by an analog-to-digital converter (ADC). Different image collection scenes and requirements require different collection parameters. For example, in a dark environment, the exposure time and gain need to be increased to improve the brightness of the image. For scenes requiring high-resolution images, a higher resolution parameter needs to be set. Reasonable parameter deployment can ensure that the collected optical signals accurately reflect the actual situation of the optical field of view, providing a good foundation for subsequent image processing. Optical signals are continuous analog signals, while computer systems can only process digital signals. Therefore, the collected optical signals need to be digitally converted for subsequent storage, transmission and processing. Discrete digital signals can be more easily quantified and analyzed, improving the accuracy and efficiency of signal processing.

[0015] The discrete digital signals are combined and arranged according to the rules specified in the pre-set data format, such as the common image file formats BMP, JPEG, PNG, etc. These digital signals are organized into two-dimensional or three-dimensional arrays, with each number representing the gray value or color value of a pixel in the image, thus forming the original image data. The pre-set data format provides a unified standard and specification for the combination of digital signals, and different devices and systems can interpret and process the original image data according to the same data format, ensuring data compatibility and universality. For example, most image processing software supports common image file formats, making it easy for users to process and analyze the collected images. Arranging digital signals into array form can visually represent the image. A two-dimensional array can represent a planar image, with each element corresponding to a pixel in the image. A three-dimensional array can be used to represent a color image or a multi-channel image. This array representation method conforms to human cognitive and processing methods of images, facilitating image display, analysis and processing.

[0016] In step S2 of the embodiments provided in the present disclosure, image semantic analysis technology is used to process the original image data at each time, analyze the features of objects, scenes, colors, textures, etc. in the image, and convert these features into information with semantic meaning, i.e. image feedback information. At the same time, a corresponding timestamp is added to each piece of image feedback information to record the specific time of image data collection. The original image data is only a collection of pixel values, which can be converted into meaningful information through semantic analysis, facilitating subsequent understanding and analysis of image content, such as recognizing objects such as people, vehicles, buildings, etc. in the image, as well as their states and mutual relationships. The timestamp provides time dimension information for the image feedback information, allowing subsequent analysis of images according to time sequence to understand the changes of visual targets and environmental scenes over time.

[0017] According to the chronological order of the timestamps, the image feedback information at each time is arranged in sequence to form a continuous data stream, i.e. image information stream. The image information stream can display the changes of the optical field of view over time in chronological order, allowing subsequent analysis to be based on continuous image sequences, more comprehensively and accurately grasping the dynamic changes of visual targets and environmental scenes. The ordered image information stream facilitates the use of various time series analysis methods, such as motion detection, trajectory tracking, etc., improving the accuracy and efficiency of visual target and environmental scene analysis.

[0018] The image feedback information in the image information stream is analyzed to identify potential visual targets and environmental scene features in the optical field of view, and potential target identification information and environmental scene features are obtained for each image feedback information. The potential target identification information is weighted and fused for analysis, and the relevance of each potential visual target is analyzed according to the environmental scene features to generate a confidence level for each potential visual target. The visual target in the optical field of view is determined according to the confidence level, and the image information stream is analyzed for target performance, such as target motion trajectory and attitude change, to obtain a performance mode of the visual target. The environmental scene performance information in each image feedback information is obtained by decomposing the image information stream based on the performance mode of the visual target, and the performance mode of the environmental scene is obtained by analyzing these information.

[0019] Analyzing the performance modes of the visual target and the environmental scene respectively can provide a deeper understanding of the various elements in the optical field of view and their mutual relationships, and provide a basis for subsequent determination of the image capture scene characteristics. Through analysis of the image information stream, the dynamic changes of the visual target and the environmental scene over time can be captured, so that their performance modes can be accurately grasped, rather than just static features.

[0020] Analyzing the mutual relationship between the performance mode of the visual target and the performance mode of the environmental scene, such as the behavior change of the visual target in different environmental scenes and the influence of the environmental scene on the visual target, can summarize the characteristics of the image capture scene. The image capture scene characteristics depend not only on the single performance of the visual target or the environmental scene, but also on the interaction between the two. Through relevance analysis, the overall characteristics of the image capture scene can be comprehensively and accurately grasped. Understanding the image capture scene characteristics can help subsequent targeted optimization of the image capture mode according to different scenes, improving the quality and effect of image capture.

[0021] In step S3 of the embodiments provided in the present disclosure, different identification forms that the visual target may exhibit in the scene are analyzed in combination with the image capture scene characteristics, such as environmental lighting, background complexity, and motion state of the visual target. For example, in a traffic monitoring scene, typical identification forms of vehicles include different perspectives such as front, side, and back, as well as different states such as driving, parking, and turning. The frequency or importance of each typical identification form in the entire image information stream is determined through statistical analysis of a large amount of image data or based on prior knowledge, so as to assign a proportion weight to each typical identification form.

[0022] Different image acquisition scenarios will affect the recognition form of the visual target. By analyzing the typical recognition form under the characteristics of the scene, the key frame representing the characteristics of the visual target can be more accurately selected, and the interference of irrelevant information can be avoided. The proportion weight reflects the importance and frequency of occurrence of each typical recognition form, which helps to select key frames of different recognition forms in a reasonable proportion in the subsequent key frame selection process, making the reference information set more representative.

[0023] For each typical recognition form, a corresponding feature template or matching rule is developed. For example, for the front recognition form of a vehicle, the features of the front face of the vehicle, such as headlights and air intake grilles, can be extracted. The image information stream is compared with the feature template of each typical recognition form, and the similarity or matching degree between them is calculated. The matching degree can be calculated by distance measurement of feature vectors, feature matching algorithm, etc. The matching degree provides a quantitative index for judging whether the original image data at a certain time can represent a certain typical recognition form. By calculating the matching degree, the image frame most consistent with the typical recognition form can be objectively selected. In the subsequent key frame selection process, the matching degree is an important reference. Prior selection of image frames with high matching degree as key frames can ensure that the image frames in the reference information set accurately reflect the typical recognition form of the visual target.

[0024] According to the proportion weight of each typical recognition form, the number of reference information allocated to each typical recognition form is determined. For example, if the proportion weight of a certain typical recognition form is 30%, and the total number of reference information set is 100 frames, then 30 frames of reference information are allocated to this typical recognition form. For each typical recognition form, the original image data is sorted according to the matching degree from high to low, and the top several frames with the highest matching degree are selected as the reference information of the typical recognition form. All the reference information of the typical recognition forms is combined to form the reference information set.

[0025] According to the proportion weight, the number of reference information is allocated, which can ensure the reasonable distribution of the reference information set in different typical recognition forms, covering various important recognition forms and avoiding excessive or insufficient information of certain recognition forms. Combined with the matching degree to select key frames, it can ensure that the image frames in the reference information set are the most representative of each typical recognition form, improving the quality and effectiveness of the reference information set and providing more accurate basis for cloud server analysis.

[0026] The generated reference information set is transmitted to the cloud server through the network communication interface. During the transmission process, encryption, compression, and other technologies can be used to ensure the security and transmission efficiency of the data. The cloud server usually has stronger computing power and storage resources, and can perform more complex and efficient analysis and processing on the reference information set. By sending the reference information set to the cloud server, the advantages of the cloud can be fully utilized to improve the accuracy and speed of image acquisition mode optimization. The cloud server can realize centralized management and sharing of data, facilitating collaborative work between different users or devices. Multiple image acquisition cards can send the reference information set to the same cloud server for unified analysis and processing, improving the overall work efficiency and quality.

[0027] In step S4 of the embodiments provided in the present disclosure, the cloud server adjusts the optical acquisition project parameters involved in the reference information set according to the optical acquisition algorithm. These parameters include but are not limited to exposure time, gain, resolution, aperture size, etc. The adjustment process can use a preset adjustment strategy, such as gradually changing the parameter value by a certain step. For the reference information set after each parameter adjustment, the image quality is evaluated using the optical acquisition algorithm. The evaluation indicators can cover multiple aspects such as image clarity, contrast, brightness uniformity, color accuracy, etc. By comparing the evaluation results under different parameter settings, the influence of parameter adjustment on image acquisition effect can be understood.

[0028] According to the specific scene and requirements of image acquisition, it is determined whether the adjusted parameters can meet the adaptability requirements. For example, in the security monitoring scene, it is necessary to ensure that the image can clearly identify the target under different lighting conditions; in the industrial detection scene, the image is required to have high resolution and accurate color restoration. According to these requirements, the effect of parameter adjustment is comprehensively evaluated.

[0029] Different image acquisition scenes and targets require different optical acquisition parameters. By adjusting and testing the parameters, the most suitable parameter combination for the current scene can be found, thereby improving the quality and effect of image acquisition. Adaptability evaluation can ensure that the adjusted parameters not only perform well in image quality, but also meet the requirements of actual application scenarios. Only after strict adaptability evaluation, can the optimized image acquisition mode have practical application value.

[0030] According to the results of effect test and adaptability evaluation, feedback is given to the parameter adjustment. If the adjustment of a certain parameter leads to a decrease in image quality or does not meet the adaptability requirements, the parameter is adjusted in the opposite direction. If the adjustment effect is good, the parameter is further fine-tuned in that direction to further optimize the image acquisition effect. The process of parameter adjustment, effect test and adaptability evaluation is repeated, and the parameters are iteratively optimized until a parameter combination that meets the adaptability requirements, i.e. the standard optical acquisition mode, is obtained.

[0031] A single parameter adjustment may not yield the optimal acquisition mode. Through a feedback adjustment mechanism, parameters can be dynamically adjusted based on actual evaluation results, gradually approaching a standard optical acquisition mode that meets adaptive requirements. This improves the accuracy and effectiveness of optimization. The image acquisition scenario may change, or the requirements for image quality may increase. The feedback adjustment mechanism allows the system to continuously adapt to these changes, continuously improving the image acquisition mode and ensuring that the acquisition effect is always at its best.

[0032] Based on the established standard optical acquisition mode, optimization instructions containing parameter settings for various optical acquisition items are generated. These instructions can use a specific data format to ensure accurate parsing by the image acquisition card. The optimization instructions are then sent to the image acquisition card via network communication. Upon receiving the instructions, the image acquisition card updates its own image acquisition mode according to the parameter settings. The optimization instructions act as a bridge between the standard optical acquisition mode and practical operation. By issuing optimization instructions, the image acquisition card can adjust its acquisition parameters in a timely manner, thereby optimizing the image acquisition mode and improving the quality and adaptability of the acquired images. The cloud server can remotely issue optimization instructions to the image acquisition card, facilitating unified management and optimization of image acquisition cards distributed in different locations. This remote control method improves the system's flexibility and maintainability.

[0033] In one possible implementation, before receiving the start command, the current time information and location information are collected in real time, and the environmental scene is predicted based on the time information and location information at predetermined intervals. Based on the prediction results, scene adaptability parameters are generated for several preset image acquisition modes. After receiving the start command, the image acquisition mode with the optimal scene adaptability parameters is selected to perform the optical signal acquisition work.

[0034] The system clock is used to obtain the current time information, including year, month, day, hour, minute, and second; the Global Positioning System (GPS) or other positioning technologies are used to obtain the device's current geographical location information, such as latitude and longitude coordinates. Time and location information are important bases for predicting environmental scenes. Different times (such as daytime, nighttime, and different seasons) and geographical locations (such as cities, rural areas, and seaside areas) will lead to significant differences in environmental scenes. These differences will have a significant impact on image acquisition. For example, in the dark at night, different acquisition modes are required; the light intensity and environmental color also vary in different regions.

[0035] The collected time and location information is associated with a pre-established environmental scene database, which stores common environmental scene features corresponding to different times and geographical locations, such as light intensity, weather conditions, and background type. Based on the results of the association analysis, machine learning algorithms or rule-based reasoning methods are used to predict the environmental scene that the current optical field of view may face. For example, if the time is night and the location is in the city center, the predicted environmental scene may be a scene with dim lighting and many artificial light sources.

[0036] By predicting environmental scenarios in advance, the image acquisition card can gain some understanding of possible scenarios before starting acquisition, preparing for the selection of appropriate acquisition modes and avoiding frequent parameter adjustments due to scene changes during acquisition, thereby improving acquisition efficiency and quality. The environmental scenario database and prediction algorithm utilize a large amount of historical data and prior knowledge, enabling them to predict environmental scenarios relatively accurately and providing a reliable foundation for generating scene adaptability parameters.

[0037] For each preset image acquisition mode, scene adaptation parameters are determined based on the predicted environmental scene. These parameters may include exposure time, gain, contrast, color correction, etc. For example, when the predicted scene is a strong light scene, the exposure time and gain are reduced; when the scene is a weak light scene, the exposure time and gain are increased. Some optimization algorithms can be combined to further optimize the generated scene adaptation parameters to ensure that the image acquisition mode can achieve the best effect in the environmental scene.

[0038] Different environmental scenarios require different image acquisition parameters to obtain the best acquisition results. By generating scene-adaptive parameters for preset acquisition modes, each acquisition mode can better adapt to the predicted environmental scenario, improving image quality and clarity. Presetting multiple image acquisition modes and generating adaptive parameters for them increases the system's flexibility, enabling it to cope with different types of environmental scenarios and meet diverse acquisition needs.

[0039] The scene adaptability parameters generated for each preset image acquisition mode are evaluated. Evaluation indicators may include image quality indicators (such as sharpness, contrast, etc.) and acquisition efficiency. Based on the evaluation results, the image acquisition mode with the optimal scene adaptability parameters is selected. The image acquisition card is then started, and optical signal acquisition is performed according to the parameter settings of this mode. Selecting the image acquisition mode with the optimal scene adaptability parameters ensures that the acquired image quality is the best under the current predicted environmental scene, meeting the user's needs. Upon receiving the start command, the appropriate acquisition mode can be quickly selected, reducing start-up time, improving system response speed, and enabling timely and efficient image acquisition.

[0040] In one possible implementation, the steps of acquiring and converting optical signals based on the image acquisition mode to obtain raw image data include: S11: Based on the image acquisition mode, the parameters of each optical acquisition item are deployed to acquire optical signals of the corresponding form in the optical field of view, and the acquired optical signals are digitally converted to obtain discrete digital signals. S12: Combine digital signals according to a predetermined data format to generate original image data formed by the arrangement of several digital signal arrays.

[0041] Multiple image acquisition modes can be preset, such as standard mode, high-resolution mode, and low-light mode. For various optical acquisition parameters, such as exposure time, gain, resolution, and sampling frequency, users can manually set them according to actual needs, or the image acquisition card can automatically adjust them according to the built-in algorithm. For example, in low-light mode, the system will automatically increase the exposure time and gain. The parameter combinations corresponding to different image acquisition modes are saved as configuration files. When a specific image acquisition mode is selected, the corresponding configuration file is directly loaded to complete the parameter deployment.

[0042] Image acquisition cards use their optical sensors (such as CCD or CMOS sensors) to capture light within an optical field of view. The pixel units in the sensor generate corresponding charges based on the intensity of the received light. These charges are converted into analog electrical signals, representing light information at different locations within the optical field of view. The analog optical signals are then converted into discrete digital signals by an analog-to-digital converter (ADC). The ADC samples the analog signals at a certain sampling frequency and quantizes the sampled values ​​into binary numbers, thus converting the continuous analog signals into discrete digital signals that can be processed by a computer.

[0043] Different image acquisition scenarios have different requirements for images. For example, when shooting fast-moving objects, a short exposure time is needed to avoid image blurring; while in low-light environments, an increased exposure time and gain are needed to improve image brightness. Reasonable parameter deployment can make the acquired images more in line with actual needs. By adjusting parameters such as resolution and sampling frequency, acquisition efficiency can be improved or specific storage and transmission requirements can be met while ensuring image quality.

[0044] Optical sensors can convert optical information into electrical signals, which is the basis for acquiring optical field information. Only by converting light into electrical signals can subsequent processing and analysis be carried out. Computer systems can only process digital signals. Converting analog optical signals into discrete digital signals makes it easier to store, transmit and further process images. Digital signals have advantages such as strong anti-interference ability and ease of processing and analysis.

[0045] Common predefined image formats include BMP, JPEG, and PNG. Different data formats have different encoding methods and storage structures. For example, BMP is an uncompressed image format that directly stores pixel data in a certain order; while JPEG uses a lossy compression algorithm, which can greatly reduce file size while maintaining a certain image quality. Depending on the selected data format, discrete digital signals are arranged according to specific rules. For two-dimensional images, the digital signals are usually arranged into a two-dimensional array, with each element representing the grayscale or color value of a pixel in the image. For example, in RGB color mode, each pixel is represented by digital signals from the red, green, and blue channels. In addition to pixel data, file header information needs to be added when generating raw image data. The file header contains basic image information, such as the image width, height, color mode, and resolution, so that other programs can correctly interpret and process the image data.

[0046] The predefined data format provides a unified standard and specification for the combination of digital signals. Different devices and software can interpret and process raw image data according to the same data format, improving data compatibility and universality. After combining digital signals into a specific data format, they can be easily stored and transmitted. A suitable data format can balance image quality and file size according to actual needs. For example, in scenarios requiring fast transmission, the JPEG format with a high compression ratio can be selected. Arranging digital signals into an array can intuitively represent the image, conforming to human image cognition and processing methods. Through the file header information, the program can accurately read and display the image, realizing image visualization.

[0047] In one possible implementation, the steps of arranging the raw image data at each moment to generate an image information stream, and analyzing the visual target and environmental scene within the optical field of view to obtain the characteristics of the image acquisition scene include: S21: Perform semantic parsing of the image feedback content on the raw image data at each time point to generate image feedback information and corresponding timestamps for each raw image data. S22: Arrange the image feedback information in sequence according to the timestamp to generate an image information stream; S23: Based on the image information stream, the visual target and environmental scene are judged in the optical field of view to obtain the representation mode of the visual target and environmental scene. S24: Analyze the correlation between the visual target and the environmental scene to obtain the characteristics of the image acquisition scene.

[0048] Using computer vision technology, various visual features are extracted from raw image data, such as color features (e.g., color histograms), texture features (e.g., gray-level co-occurrence matrices), and shape features (e.g., contours, edges). With the help of object detection and recognition algorithms (e.g., deep learning-based Faster R-CNN, YOLO, etc.), specific targets in the image, such as people, vehicles, and buildings, are identified, and their locations and categories are determined. Based on the extracted features and identified targets, semantic information describing the image content is generated, such as "In a park scene, a child is feeding pigeons." Simultaneously, a corresponding timestamp is added to each image feedback message, recording the specific time the image data was acquired. The raw image data is merely a collection of pixel values; semantic parsing transforms it into meaningful information, facilitating subsequent understanding and analysis of the image content. The timestamp provides a temporal dimension to the image feedback information, allowing for subsequent analysis of the image in chronological order to understand how visual targets and environmental scenes change over time.

[0049] By employing sorting algorithms (such as quicksort and mergesort), the image feedback information at each moment is sorted according to the chronological order of timestamps. The sorted image feedback information is then integrated into a continuous data stream, forming an image information stream. This stream can be stored as a specific data structure, such as a list or array, facilitating subsequent processing and analysis. The image information stream can sequentially display the changes in optical field of view over time, enabling subsequent analysis to be based on a continuous image sequence. This allows for a more comprehensive and accurate understanding of the dynamic changes in visual targets and environmental scenes. The ordered image information stream also facilitates the use of various time series analysis methods, such as motion detection and trajectory tracking, improving the accuracy and efficiency of visual target and environmental scene analysis.

[0050] Based on image feedback information in the image information stream, potential visual targets and environmental scene features in the optical field of view are identified. This yields potential target recognition information and environmental scene features from each image feedback message. Weighted fusion analysis is performed on the recognition information of each potential target. Simultaneously, correlation analysis is conducted on each potential visual target based on environmental scene features, generating confidence scores for each potential visual target. Based on these confidence scores, visual targets in the optical field of view are determined. The image information stream is then analyzed to interpret the target's performance, such as its motion trajectory, posture changes, and behavioral actions, resulting in a visual target performance pattern. Based on this performance pattern, environmental scene information is decomposed from each image feedback message in the image information stream, yielding environmental scene performance information in each image feedback message. This information is then combined to analyze the environmental scene performance pattern, such as changes in illumination and weather conditions. Analyzing the performance patterns of both visual targets and environmental scenes separately allows for a deeper understanding of various elements in the optical field of view and their interrelationships, providing a foundation for subsequently determining the characteristics of the image acquisition scene. Through the analysis of the image information stream, the dynamic changes of visual targets and environmental scenes over time can be captured, thus accurately grasping their performance patterns, rather than just static features.

[0051] Establish a correlation model between the visual target's performance pattern and the environmental scene's performance pattern. For example, analyze whether the movement of the visual target is affected by factors such as obstacles and lighting in the environmental scene. Extract key features that reflect the characteristics of the image acquisition scene from the correlation model, such as common behavior patterns of visual targets in specific environmental scenes and the impact of the environmental scene on visual target recognition. Based on the extracted key features, summarize the characteristics of the image acquisition scene, such as scene complexity, target recognizability, and the degree of dynamic change in the environment. The characteristics of the image acquisition scene do not depend solely on the single performance of the visual target or the environmental scene, but are the result of their interaction. Through correlation analysis, we can comprehensively and accurately grasp the overall characteristics of the image acquisition scene. Understanding the characteristics of the image acquisition scene helps to optimize the image acquisition mode in a targeted manner according to different scenes, thereby improving the quality and effect of image acquisition.

[0052] In one possible implementation, the step of judging the visual target and the environmental scene based on the image information stream to obtain the representation pattern of the visual target and the environmental scene includes: S231: Based on the image feedback information contained in the image information stream, potential visual targets and environmental scenes are identified in the optical field of view respectively, so as to obtain the potential target identification information and environmental scene features of each image feedback information. S232: Perform weighted fusion analysis on the identification information of each potential target, and at the same time perform correlation analysis on each potential visual target according to the environmental scene characteristics. Based on the results of the two analyses, generate confidence scores for each potential visual target to determine the visual targets in the optical field of view. S233: Analyze the image information stream according to the determined visual target to obtain the visual target's representation pattern; S234: Based on the performance mode, the environmental scene information of each image feedback information in the image information stream is decomposed to obtain the environmental scene performance information in each image feedback information, so as to analyze the performance mode of the environmental scene by combining the environmental scene performance information.

[0053] By using object detection algorithms, such as the YOLO (You Only Look Once) series and Faster R-CNN based on deep learning, the image feedback information of each frame in the image information stream is scanned to identify potential visual targets and record their position, size, category, and other information to form potential target recognition information. Image processing techniques are then used to extract environment-related features from the image. For example, color analysis is used to determine the lighting conditions of the environment (e.g., bright or dark), and texture analysis is used to determine whether the scene is indoors or outdoors, natural or artificial, etc., thus obtaining environmental scene features. This step is the foundation for subsequent analysis. By identifying potential visual targets and extracting environmental scene features, raw data can be provided for determining the real visual targets and analyzing the relationship between targets and the environment. Different environmental scenes may affect the recognition and performance of targets, and understanding environmental features in advance helps to analyze visual targets more accurately.

[0054] Different weights are assigned to potential target identification information at different times. The weights can be determined based on factors such as image clarity and the proportion of the target in the image. The identification information of each potential target is weighted and fused. For example, the position information of the same target in different frames is weighted and averaged to obtain a more accurate target position estimate. Combined with environmental scene features, the correlation between potential visual targets and between targets and the environment is analyzed. For example, in a street scene, if a potential target is identified as a pedestrian, and the surrounding environmental features show roads and vehicles, then the rationality of the potential target can be judged based on the positional relationship and movement trend of the pedestrian, road, and vehicles. Based on the results of weighted fusion analysis and correlation analysis, a confidence value is generated for each potential visual target. The higher the confidence value, the more likely the potential target is a real visual target. A confidence threshold is set, and potential targets with confidence values ​​higher than the threshold are identified as visual targets in the optical field of view.

[0055] In the process of image recognition, there may be cases of misidentification. By using weighted fusion analysis and correlation analysis, the influence of these interference factors can be reduced, and the accuracy of target recognition can be improved. The introduction of confidence level provides a quantitative standard for determining the true visual target, avoiding the limitations of judging based on a single frame image or a single feature, and making the determined visual target more consistent with the actual situation.

[0056] Track a identified visual target in the image information stream, record its position changes at different times, and draw the target's motion trajectory. Analyze the shape, speed, acceleration and other parameters of the motion trajectory to determine whether the target is stationary, moving at a constant speed or changing speed. Recognize the target's posture, such as the standing, walking and running postures of a human body, and the straight-moving and turning states of a vehicle. By analyzing the target's posture changes in different frames, we can understand the target's behavior patterns.

[0057] The visual target's performance patterns can reflect its behavioral characteristics and movement patterns, which is very important for many application scenarios. For example, in security monitoring, understanding the target's movement trajectory and behavior patterns can help detect abnormal behavior; in intelligent transportation, analyzing the vehicle's driving status helps optimize traffic management. The visual target's performance patterns interact with the environmental scene, and understanding the target's performance patterns can provide a reference for subsequent analysis of the environmental scene's impact on the target and the environmental scene's performance patterns.

[0058] Based on the visual target's representation pattern, a more detailed analysis of the environmental scene in the image feedback information is conducted. For example, if the target is a pedestrian walking in the rain, the intensity of the rain, wind direction, and other environmental information can be further analyzed from the image to obtain environmental scene representation information. The environmental scene representation information in each image feedback information is integrated and analyzed to observe the changes in the environmental scene over time, such as changes in lighting, weather, and dynamic changes of objects in the scene, thereby summarizing the representation pattern of the environmental scene.

[0059] The representation of an environmental scene depends not only on its own characteristics but also on the representation of the visual target. By combining the representation of the visual target with environmental scene analysis, we can gain a more comprehensive and in-depth understanding of the scene in the entire optical field of view. Understanding the representation of an environmental scene helps to optimize the image acquisition mode according to different scene characteristics. For example, in scenes with large changes in lighting, we can adjust the exposure parameters to ensure image quality.

[0060] In one possible implementation, the steps of selecting keyframes from the image information stream based on the characteristics of the image acquisition scene to obtain a reference information set, and sending the reference information set to a cloud server, include: S31: Analyze the target recognition form of the visual target in the current environment scene based on the characteristics of the image acquisition scene, so as to obtain several typical recognition forms of the visual target in the current environment scene, and the proportion and weight of each typical recognition form. S32: Based on each of the typical recognition methods, perform information matching on each of the image feedback information contained in each of the image information streams to obtain the matching degree of the original image data at each time point relative to each of the typical recognition methods; S33: Allocate the amount of reference information to each of the typical recognition forms according to their respective weights, and select a number of original image data sets for each typical recognition form by combining the matching degree of the original image data at each time point with that of each typical recognition form.

[0061] Based on the characteristics of the image acquisition scene, such as lighting conditions, background complexity, and target motion state, features of visual targets are extracted from the image information stream. For example, in a traffic monitoring scene, for a vehicle as a visual target, features such as color, shape, and license plate can be extracted. Then, the visual target is classified according to these features to determine its typical recognition form in the current environment, such as different perspectives of the vehicle, such as the front, side, and back, as well as different states such as driving, parking, and turning. The various typical recognition forms appearing in the image information stream are statistically analyzed, and the frequency of each typical recognition form is calculated. Based on the frequency, a weight is assigned to each typical recognition form, with a larger weight for recognition forms with higher frequencies.

[0062] Different image acquisition scenarios can affect the recognition of visual targets. By analyzing the typical recognition methods under the characteristics of the scenario, key frames that can represent the features of visual targets can be selected more accurately, avoiding interference from irrelevant information. The weighting reflects the importance and frequency of each typical recognition method, which helps to select key frames of different recognition methods in a reasonable proportion in the subsequent key frame selection process, making the reference information set more representative.

[0063] For each typical recognition form, key features are extracted and feature templates are constructed. For example, for the vehicle front recognition form, features of the vehicle's front face, such as headlights and grille, are extracted and corresponding feature templates are constructed. Each image feedback information in the image information stream is compared with the feature templates of each typical recognition form. Matching algorithms (such as Euclidean distance of feature vectors, cosine similarity, etc.) are used to calculate the similarity or matching degree between them. The higher the matching degree, the more representative the original image data at that moment is of the corresponding typical recognition form.

[0064] Matching degree provides a quantitative indicator for judging whether the original image data at a certain moment can represent a certain typical recognition form. By calculating the matching degree, the image frames that best match the typical recognition form can be objectively selected. In the subsequent key frame selection process, the matching degree is an important reference. Prioritizing the selection of image frames with high matching degree as key frames can ensure that the image frames in the reference information set accurately reflect the typical recognition form of the visual target.

[0065] Based on the weight of each typical recognition form, the amount of reference information allocated to each typical recognition form is determined. For example, if the weight of a certain typical recognition form is 30% and the total number of reference information frames is 100, then 30 frames of reference information are allocated to that typical recognition form. For each typical recognition form, the original image data is sorted from high to low according to the matching degree, and several frames with the highest matching degree are selected as the reference information for that typical recognition form. The reference information of all typical recognition forms is combined together to form a reference information set.

[0066] Allocating the amount of reference information according to its weight ensures a reasonable distribution of the reference information set across different typical recognition forms. This covers all important recognition forms while avoiding excessive or insufficient information for certain forms. Combining the matching degree with the selection of key frames ensures that the image frames in the reference information set are most representative of each typical recognition form, thus improving the quality and effectiveness of the reference information set and providing a more accurate basis for analysis on the cloud server.

[0067] After selecting the reference information set, it is sent to the cloud server through a network communication interface. A secure and reliable communication protocol (such as HTTPS) can be used, and the data is encrypted to ensure the security and integrity of data transmission. The reason for sending the reference information set to the cloud server is to utilize the powerful computing power and storage resources of the cloud server for more complex and efficient analysis and processing, thereby optimizing the image acquisition mode.

[0068] In one possible implementation, the step of performing adaptive analysis of image acquisition on the reference information set using an optical acquisition algorithm pre-deployed on a cloud server, and then optimizing the image acquisition mode by issuing optimization instructions, includes: S41: The optical acquisition algorithm pre-deployed on the cloud server is used to adjust the parameters of various optical acquisition items, test the effect, and evaluate the adaptability of the reference information set. Based on the evaluation results, the adjustment of the parameters of various optical acquisition items is adjusted in feedback until a standard optical acquisition mode that meets the adaptability requirements is obtained. S42: Generate corresponding optimization instructions based on the standard optical acquisition mode, and send them to the image acquisition card to optimize the image acquisition mode.

[0069] Based on the general rules and experience of optical acquisition, reasonable adjustment ranges are set for various optical acquisition parameters (such as exposure time, gain, resolution, aperture size, etc.). For example, the exposure time can be adjusted between 1 / 1000 of a second and 1 second, using different adjustment strategies, such as gradient adjustment and random search. Gradient adjustment involves gradually changing the parameter value according to a certain step size, such as increasing or decreasing the exposure time by 1 / 100 of a second each time; random search involves randomly selecting parameter values ​​within the parameter range for testing.

[0070] Determine the metrics used to evaluate the image acquisition effect, such as sharpness (which can be measured by the edge gradient and contrast of the image), brightness uniformity (calculating the brightness difference in different areas of the image), and color accuracy (comparing the deviation of colors in the image from standard colors). For each reference information set after parameter adjustment, use optical acquisition algorithms to calculate the values ​​of each evaluation metric to quantify the image acquisition effect.

[0071] Based on the specific scenarios and needs of image acquisition, adaptive evaluation standards are formulated. For example, in security monitoring scenarios, the image clarity is required to reach a certain threshold, and the target can be accurately identified under different lighting conditions. In industrial inspection scenarios, there are high requirements for the color accuracy and resolution of the image. The various index values ​​obtained from the effect test are compared with the adaptive standards to determine whether the adjusted parameters meet the adaptive requirements.

[0072] If adjusting a certain parameter leads to a decrease in image quality or fails to meet the adaptability requirements, then the parameter is adjusted in the opposite direction. If the adjustment is effective, the parameter is fine-tuned in that direction. For example, if increasing the exposure time makes the image too bright and exceeds the adaptability standard, then the exposure time is reduced. The process of parameter adjustment, effect testing, and adaptability evaluation is repeated, and the parameters are iterated and optimized until a set of parameter combinations that can meet the adaptability requirements is obtained, i.e., the standard optical acquisition mode.

[0073] Different image acquisition scenarios and targets require different optical acquisition parameters. By adjusting and testing these parameters, the most suitable parameter combination for the current scenario can be found, thereby improving the quality and effect of image acquisition. Adaptability evaluation ensures that the adjusted parameters not only perform well in terms of image quality but also meet the needs of practical application scenarios. Only through rigorous adaptability evaluation can the optimized image acquisition mode be guaranteed to have practical application value. The feedback adjustment mechanism can dynamically adjust the parameters based on the actual evaluation results, gradually approaching the standard optical acquisition mode that meets the adaptability requirements, thus improving the accuracy and effectiveness of optimization.

[0074] Define the format of the optimization command, specifying the parameter information and command code contained in the command. For example, the command can be in JSON format, containing the set values ​​of parameters such as exposure time, gain, and resolution, as well as the corresponding command code. Based on the parameter values ​​in the standard optical acquisition mode, the optimization command is generated according to the defined command format. The optimization command is sent to the image acquisition card via network communication. During the transmission process, encryption, compression, and other technologies can be used to ensure data security and transmission efficiency. After receiving the command, the image acquisition card updates its own image acquisition mode according to the parameter settings in the command.

[0075] Optimization commands serve as a bridge between standard optical acquisition modes and practical operation. By issuing optimization commands, image acquisition cards can adjust their acquisition parameters in a timely manner, thereby optimizing the image acquisition mode, improving the quality and adaptability of the acquired images. Cloud servers can remotely issue optimization commands to image acquisition cards, facilitating unified management and optimization of image acquisition cards distributed in different locations. This remote control method enhances the system's flexibility and maintainability.

[0076] Please see Figure 2 As shown, this disclosure provides a scene-adaptive acquisition system for an image acquisition card, used to implement the scene-adaptive acquisition method for an image acquisition card as described in any one of the first aspects, including: The image acquisition module is used to acquire and convert optical signals based on the image acquisition mode to obtain raw image data. The image analysis module is used to arrange the raw image data at each moment to generate an image information stream, so as to analyze the visual target and environmental scene of the optical field of view and obtain the characteristics of the image acquisition scene. The information reference module is used to select key frames of the image information stream according to the characteristics of the image acquisition scene, obtain a reference information set, and send the reference information set to the cloud server. The acquisition optimization module is used to perform adaptive analysis of image acquisition on the reference information set using an optical acquisition algorithm pre-deployed on a cloud server, and to optimize the image acquisition mode by issuing optimization commands.

[0077] In this embodiment, the specific implementation of each module in the above system embodiment is described in the above method embodiment, and will not be repeated here.

[0078] As can be seen from the above, the technical solutions disclosed herein can be implemented as methods, apparatus, systems, computer program products, storage media, electronic devices, etc. Those skilled in the art will understand that various aspects of this disclosure can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, which may be referred to as "circuit," "module," or "system," respectively.

[0079] It should be understood that this disclosure is not limited to the specific methods, steps, or structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. Those skilled in the art will readily conceive of other embodiments based on the specific implementations provided in this disclosure. Therefore, the specific implementations provided in this disclosure are merely exemplary, and the scope and spirit of this disclosure are indicated by the claims, and should cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary technical means in the art not disclosed in this disclosure.

Claims

1. A scene-adaptive acquisition method for an image acquisition card, characterized in that, include: Based on the image acquisition mode, optical signals are acquired and converted in the optical field of view to obtain raw image data; The raw image data at each moment are arranged to generate an image information stream, which is used to analyze the visual target and environmental scene in the optical field of view and obtain the characteristics of the image acquisition scene. Based on the characteristics of the image acquisition scene, key frames are selected from the image information stream to obtain a reference information set, and the reference information set is sent to the cloud server. An adaptive analysis of image acquisition is performed on the reference information set using an optical acquisition algorithm pre-deployed on a cloud server, and optimization instructions are issued to optimize the image acquisition mode.

2. The scene-adaptive acquisition method of the image acquisition card as described in claim 1, characterized in that, Before receiving the start command, the system collects the current time and location information in real time, and predicts the environmental scene based on the time and location information at predetermined intervals. Based on the prediction results, scene adaptability parameters are generated for several preset image acquisition modes. After receiving the start command, the system selects the image acquisition mode with the optimal scene adaptability parameters to perform the optical signal acquisition.

3. The scene-adaptive acquisition method of the image acquisition card as described in claim 1, characterized in that, The steps for acquiring and converting optical signals based on the image acquisition mode to obtain raw image data include: Based on the image acquisition mode, parameters for various optical acquisition items are deployed to acquire corresponding forms of optical signals in the optical field of view, and the acquired optical signals are digitally converted to obtain discrete digital signals. Digital signals are combined according to a predetermined data format to generate original image data formed by the arrangement of several digital signal arrays.

4. The scene-adaptive acquisition method of the image acquisition card as described in claim 1, characterized in that, The steps of arranging the raw image data at each moment to generate an image information stream, and then analyzing the visual target and environmental scene within the optical field of view to obtain the characteristics of the image acquisition scene include: Semantic parsing of image feedback content is performed on the raw image data at each time point to generate image feedback information and corresponding timestamps for each raw image data. The image feedback information is arranged sequentially according to the timestamp to generate an image information stream; Based on the image information stream, the visual target and environmental scene are judged in the optical field of view to obtain the representation mode of the visual target and environmental scene. The correlation between the visual target and the environmental scene is analyzed to obtain the characteristics of the image acquisition scene.

5. The scene-adaptive acquisition method of the image acquisition card as described in claim 4, characterized in that, The steps for determining the visual target and environmental scene based on the image information stream to obtain the representation mode of the visual target and environmental scene include: Based on the image feedback information contained in the image information stream, potential visual targets and environmental scenes are identified in the optical field of view to obtain potential target identification information and environmental scene features of each image feedback information. The potential target identification information is weighted and fused for analysis, and the correlation analysis of each potential visual target is performed based on the environmental scene characteristics. Based on the results of the two analyses, a confidence score is generated for each potential visual target to determine the visual target in the optical field of view. Based on the determined visual target, the image information stream is analyzed to obtain the representation pattern of the visual target. Based on the performance mode, the environmental scene information of each image feedback information in the image information stream is decomposed to obtain the environmental scene performance information in each image feedback information, so as to combine the environmental scene performance information to analyze and obtain the environmental scene performance mode.

6. The scene-adaptive acquisition method of the image acquisition card as described in claim 1, characterized in that, The steps of selecting keyframes from the image information stream based on the characteristics of the image acquisition scene to obtain a reference information set, and sending the reference information set to the cloud server include: Based on the characteristics of the image acquisition scene, the target recognition form of the visual target in the environmental scene is analyzed to obtain several typical recognition forms of the visual target in the current environmental scene, as well as the proportion and weight of each typical recognition form. Based on each of the typical recognition methods, information matching is performed on each image feedback information contained in each of the image information streams to obtain the matching degree of the original image data at each time point relative to each of the typical recognition methods; Based on the weighting of each typical recognition form, a number of reference information is allocated to each typical recognition form. Then, based on the matching degree of the original image data at each time point relative to each typical recognition form, a number of original image data are selected to form a reference information set for each typical recognition form.

7. The scene-adaptive acquisition method of the image acquisition card as described in claim 1, characterized in that, The steps of performing adaptive analysis of image acquisition on the reference information set using an optical acquisition algorithm pre-deployed on a cloud server, and then optimizing the image acquisition mode by issuing optimization instructions, include: The reference information set is subjected to parameter adjustment, effect testing and adaptability evaluation of various optical acquisition items by an optical acquisition algorithm pre-deployed on a cloud server, and feedback adjustment of the parameters of various optical acquisition items is performed based on the evaluation results until a standard optical acquisition mode that meets the adaptability requirements is obtained. Based on the standard optical acquisition mode, corresponding optimization instructions are generated and sent to the image acquisition card to optimize the image acquisition mode.

8. A scene-adaptive acquisition system for an image acquisition card, characterized in that, A scene-adaptive acquisition method for an image acquisition card according to any one of claims 1-7 includes: The image acquisition module is used to acquire and convert optical signals based on the image acquisition mode to obtain raw image data. The image analysis module is used to arrange the raw image data at each moment to generate an image information stream, so as to analyze the visual target and environmental scene of the optical field of view and obtain the characteristics of the image acquisition scene. The information reference module is used to select key frames of the image information stream according to the characteristics of the image acquisition scene, obtain a reference information set, and send the reference information set to the cloud server. The acquisition optimization module is used to perform adaptive analysis of image acquisition on the reference information set using an optical acquisition algorithm pre-deployed on a cloud server, and to optimize the image acquisition mode by issuing optimization commands.

Images