Acquisition and transmission method, device and equipment of high-resolution image and storage medium

Abstract

The invention relates to the technical field of image acquisition and transmission, and discloses a high-resolution image acquisition and transmission method and device, equipment and a storage medium. The method comprises the following steps: performing multi-stage downsampling and feature extraction processing on a high-resolution original image acquired by an image acquisition card to obtain global image features and regional image features; performing random delay and data loss characteristic analysis on historical communication data of the image acquisition card to obtain a probability distribution model and a continuous data loss record; performing multi-scene prediction on the future transmission state of the image acquisition card to obtain prediction state information; and generating a self-adaptive transmission control strategy according to the prediction state information, the global image features and the regional image features, and executing a high-resolution image transmission process to obtain a high-resolution image transmission result. According to the method, the optimal distribution of transmission resources is realized on the premise of ensuring the image quality, and the problem of insufficient adaptability of traditional fixed parameter control in a complex network environment is solved.

Description

Technical Field

[0001] This invention relates to the field of image acquisition and transmission technology, and in particular to a method, apparatus, device, and storage medium for acquiring and transmitting high-resolution images. Background Technology

[0002] As image resolution continues to increase, the amount of data collected grows exponentially, posing a significant challenge to traditional image acquisition methods in processing this massive amount of data. High-resolution images are characterized by their large information content and high redundancy, making it difficult for a single global acquisition strategy to balance transmission efficiency and image quality, especially in complex scenes.

[0003] High-resolution image acquisition and transmission systems based on image acquisition cards are frequently affected by hardware interface instability in practical applications, mainly manifested as two communication defects: random sampling delay and continuous data loss. These communication defects not only lead to a decrease in image quality but may also jeopardize the stability of the entire acquisition system, with particularly severe consequences in applications requiring high precision, such as medical imaging. Traditional acquisition and transmission methods typically employ fixed bandwidth allocation and simple retransmission mechanisms, which cannot effectively cope with the uncertainties of data transmission in complex network environments. Summary of the Invention

[0004] The main objective of this invention is to provide a method, apparatus, device, and storage medium for acquiring and transmitting high-resolution images. This invention achieves optimal allocation of transmission resources while ensuring image quality, overcoming the problem of insufficient adaptability of traditional fixed parameter control in complex network environments.

[0005] To achieve the above objectives, the present invention provides a method for acquiring and transmitting high-resolution images, comprising the following steps: Multi-level downsampling and feature extraction processing are performed on the high-resolution raw images acquired by the image acquisition card to obtain global image features and regional image features; Random delay and data loss characteristics were analyzed on the historical communication data of the image acquisition card to obtain a probability distribution model and continuous data loss records; Using the probability distribution model and the continuous data loss records, multi-scenario prediction of the future transmission status of the image acquisition card is performed to obtain the predicted status information. An adaptive transmission control strategy is generated based on the predicted state information, the global image features, and the regional image features, and a high-resolution image transmission process is executed to obtain the high-resolution image transmission result.

[0006] The present invention also provides a high-resolution image acquisition and transmission device, comprising: The feature extraction module is used to perform multi-level downsampling and feature extraction processing on the high-resolution raw images acquired by the image acquisition card to obtain global image features and regional image features; The feature analysis module is used to perform random delay and data loss feature analysis on the historical communication data of the image acquisition card to obtain a probability distribution model and continuous data loss records. The multi-scenario prediction module is used to perform multi-scenario prediction of the future transmission status of the image acquisition card using the probability distribution model and the continuous data loss record, and obtain the predicted status information. The image transmission module is used to generate an adaptive transmission control strategy based on the predicted state information, the global image features, and the regional image features, and to execute the high-resolution image transmission process to obtain the high-resolution image transmission result.

[0007] The present invention 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 of any of the methods described above.

[0008] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any of the methods described above.

[0009] In summary, the technical solution provided by this invention, by constructing a pyramid image processing module and a multi-level cache transmission decision unit, achieves feature extraction and storage at different resolution levels of high-resolution original images. This solves the problem of inflexible cache management under traditional single-cache area designs and effectively improves the system's adaptability to different network environments. By combining global feature extraction with target region recognition, and using attention mechanisms and region scoring functions to perform semantic understanding of image content, it achieves accurate localization of key regions in high-resolution images, overcoming the shortcomings of traditional methods that only focus on global information while ignoring local details. Statistical analysis and modeling of historical communication data from the acquisition card accurately captures the probabilistic characteristics of random sampling delays and continuous data loss, providing a reliable theoretical basis for subsequent transmission control and solving the blindness problem of traditional fixed retransmission mechanisms. By combining interpolation polynomial algorithms, it achieves accurate prediction of future transmission states under different delay scenarios, and dynamically optimizes prediction accuracy through an adaptive node update mechanism, breaking through the limitations of traditional passive reactive control methods and achieving forward-looking control of transmission states. By optimizing the cost function and using rolling time-domain control, the optimal allocation of transmission resources is achieved while ensuring image quality, overcoming the problem of insufficient adaptability of traditional fixed parameter control in complex network environments. Attached Figure Description

[0010] Figure 1This is a schematic diagram of the steps of a high-resolution image acquisition and transmission method in one embodiment of the present invention; Figure 2 This is a structural block diagram of a high-resolution image acquisition and transmission device according to an embodiment of the present invention; Figure 3 This is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0011] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0013] Reference Figure 1 This embodiment provides a method for acquiring and transmitting high-resolution images, including the following steps: S1, performs multi-level downsampling and feature extraction processing on the high-resolution raw image acquired by the image acquisition card to obtain global image features and regional image features; The system employs a pyramid image processing module to perform structured processing on the raw high-resolution images received by the image acquisition card. Driven by a preset scaling factor, the input images undergo layer-by-layer downsampling, generating a sequence of images with decreasing resolution. This sequence starts from the original resolution image and sequentially generates multiple scale levels, effectively reducing the pressure on cache bandwidth at different levels while preserving the core image content. For each resolution image in the sequence, it is fed into a convolutional neural network for feature extraction, yielding a set of feature map data corresponding to each resolution level, reflecting the spatial texture and semantic distribution of the image at each level. A bandwidth monitoring module continuously monitors the current communication bandwidth between the image acquisition card and the host, dynamically comparing this bandwidth value with a pre-set bandwidth threshold sequence. Each pair of bandwidth threshold intervals represents a tolerable transmission capacity segment for the system. By comparing the specific interval into which the current bandwidth falls, the system can deduce the most suitable cache region configuration strategy under the current transmission environment. Based on this analysis, priority classification is performed on the multi-scale feature map data. This priority division is based on the importance of the feature data, considering factors such as image semantic saliency, region boundary clarity, and structural complexity, dividing the feature map data into different data groups such as high priority, medium priority, and low priority. To match the mapping relationship between bandwidth adaptation strategy and feature data priority, cache resources of appropriate size are allocated according to the optimal cache selection result. The cache region consists of multiple physical memory blocks, each corresponding to different levels of feature data storage requirements. These cache regions form a hierarchical structure, with capacity increasing or decreasing sequentially from high priority to low priority, constructing a multi-level cache storage layout. The system writes the feature map data into the corresponding cache units according to priority order based on this layout, thereby ensuring the complete transmission of core feature data while reducing the proportion of non-critical image information occupying the total bandwidth and effectively avoiding image quality degradation caused by resource bottlenecks. After completing the cache archiving of feature maps, the content analysis stage begins. Semantic fusion and attention weight analysis operations are performed on all feature map data in the cache to comprehensively extract the overall spatial structure and semantic coverage of the image, obtaining a global image feature representation. Simultaneously, a specific region scoring function is used to spatially score the global feature map and automatically identify several key regions in the image. These regions are concentrated at target edges, high-contrast areas, or parts with prominent semantic density. Subsequently, cropping, scaling, and local feature extraction operations are performed on these regions to extract representative and semantically clear regional image features. This yields both global image features and regional image features.

[0014] S2, perform random delay and data loss characteristics analysis on the historical communication data of the image acquisition card to obtain a probability distribution model and continuous data loss records; Specifically, a data recording and behavior analysis mechanism is constructed to continuously collect the communication performance of the image acquisition card in each transmission cycle. This mechanism uses transmission delay and data loss as the core analysis objects, corresponding to key indicators of transmission efficiency and stability, respectively. Each time the image acquisition card transmits image data to the host, the system automatically records the actual transmission time and stores it as a time-series node in the delay data sequence. By performing a holistic analysis of delay data from several past cycles, typical statistical characteristics of these delay values ​​are extracted, including their overall average level and the degree of fluctuation over time. Based on all recorded results, the changing trends and distribution characteristics of delay data in different time periods are constructed. The system also identifies and marks data loss events during data transmission. A continuous loss monitoring module is designed to determine in real time whether data has been successfully delivered to the host in each cycle. If any transmission fails to be received by the host, the current cycle is considered a loss event, and a continuous loss counter is incremented to record the consecutive length of failures since the last successful transmission. When the system detects a successful data transmission, the counter is automatically reset to zero. In this way, the system quickly identifies the occurrence cycle of continuous packet loss and forms a time-sequentially arranged loss behavior trajectory. Based on fundamental information about delay and packet loss behavior, the changing patterns of transmission states are modeled and summarized. Delay behavior modeling relies on summarizing transmission efficiency over different time periods. The system divides delay values ​​into several intervals from low to high and counts the frequency of occurrence in each interval, thus depicting the typical distribution pattern of delay in practical applications. This reveals the delay range the system operates in most of the time and the scenarios in which delay suddenly increases. Simultaneously, the records of the continuous packet loss counter are used as the core parameter for quantifying loss risk. To reflect the security level of the current communication state, a corresponding risk level is assigned to each packet loss length, which gradually increases with the increase in continuous packet loss. Especially when the system encounters multiple consecutive packet losses, the risk level will rapidly approach its upper limit, prompting the system to immediately take fault-tolerant compensation measures or switch control strategies to ensure the integrity and real-time performance of image data. After completing the above analysis, the extracted delay state model and loss risk sequence are archived together to construct a communication characteristic database. The delay state model provides an early warning reference for future bandwidth fluctuations and processing bottlenecks, while the continuous packet loss records are used to dynamically adjust cache management strategies, retransmission mechanisms, or feature priority sorting methods.

[0015] S3 uses a probability distribution model and continuous data loss records to perform multi-scenario prediction of the future transmission status of the image acquisition card and obtain the predicted status information. It should be noted that, based on the probability distribution model, the entire latency range is divided into several representative latency intervals. Each interval represents a specific transmission state scenario, such as low latency, high-frequency fluctuations, moderate latency, or extreme congestion. Each scenario corresponds to the transmission performance encountered by the system under different operating conditions. Based on this, a scenario classifier is constructed to map the real-time observed bandwidth status and latency behavior to the preset scenario intervals, thereby selecting the corresponding prediction model for computation. After scenario division, a separate historical behavior-driven prediction model is built for each latency scenario. These models constitute a prediction model set, and each model undertakes the task of fitting the trend of future transmission behavior under a specific latency state. To improve the model's response flexibility and fitting accuracy, several representative time points and corresponding historical latency values ​​are selected within each model as interpolation nodes. These nodes constitute the core input of the interpolation function. The selection of nodes not only considers the balance of time distribution but also prioritizes retaining critical moments with drastic fluctuations or prominent transitions, ensuring that the prediction model can respond promptly to nonlinear trends. Based on these nodes, the system constructs a corresponding initial interpolation function within each model. This function, using time as the independent variable, makes an initial prediction of the delay trend over several future periods using a polynomial form. As the system continues to run, the results output by the prediction model and the subsequently measured delay values ​​accumulate, providing fundamental data support for error calibration and model optimization. After each prediction period, the system calculates the deviation between the model's predicted value and the actual delay value, forming a target error index. When this error exceeds the system's set stability threshold, an interpolation node update mechanism is automatically triggered. This involves removing the oldest set of nodes, introducing the latest observation points, reconstructing the interpolation structure, and generating a new set of interpolation functions for subsequent predictions. This dynamic update mechanism enhances the model's adaptability and avoids prediction bias caused by historical data lag, ensuring high prediction accuracy even in highly volatile scenarios. After all models have been updated and the latest prediction results have been generated, these prediction values ​​are integrated. Since each model has different historical performance accuracy over different time periods, a weight reflecting its reliability is assigned to each model based on its prediction stability and error fluctuation over a past period. This weight determines the proportion of the model's predicted value in the final result, thus constructing a predicted state information output based on multi-scenario weighted fusion. The resulting predicted state information includes a quantitative estimate of the delay trend, as well as the level of potential communication risks currently faced by the system.

[0016] S4 generates an adaptive transmission control strategy based on the predicted state information, global image features, and regional image features, and executes the high-resolution image transmission process to obtain the high-resolution image transmission result.

[0017] Specifically, the predicted state information from the multi-scenario prediction module is used as the core input. This information, including future transmission delay estimates, data loss risks, and bandwidth change trends, is summarized and analyzed, and combined with real-time parameters observed during the current transmission cycle to form a multi-dimensional system state vector. Global and regional image features are input to the data scheduling unit, and all image content is classified according to an important hierarchy based on a semantic priority classification rule. This rule divides image data into core data groups and non-core data groups based on indicators such as structural clarity, semantic density, and contextual saliency. The former represents key features that must be prioritized for integrity and clarity during transmission, while the latter can be appropriately downgraded in resource-constrained or high-risk environments. Based on the cache availability, bandwidth utilization, and data loss risk level reflected in the system state vector, the data transmission ratio of each level of cache within the image acquisition card is adjusted, constructing an initial control input vector. This vector clarifies the resource share that various types of image features should occupy in different transmission channels and cache levels at the current moment. Based on the initial control parameters and the current state vector, a cost function model for transmission scheduling is constructed. This model is used to evaluate the combined costs of latency pressure, buffer load, and quality loss caused by control decisions in actual execution. To avoid policy decisions focusing only on short-term effects while ignoring medium- or long-term stability, a rolling time-domain optimization mechanism is introduced. By extrapolating the state evolution at multiple future moments, the control parameters are dynamically simulated and recalculated multiple times to form the optimal transmission control sequence. Priority allocation of data transmission resources is performed based on the optimal transmission control sequence and the loss risk of core data packets. The loss probability of various image features in the core data packets is quickly retrieved, and these results are matched with the control sequence to construct a resource priority allocation scheme based on content risk level. In this scheme, the system prioritizes the complete transmission of high-value feature data in bandwidth-constrained environments and mitigates the quality degradation risk caused by transmission congestion by flexibly adjusting the buffer call order and channel bandwidth ratio to perform delayed processing or compressed transmission of non-critical data. Based on the adaptive transmission control strategy constructed above, the image acquisition system mobilizes various buffer layers and bandwidth channels to perform scheduled transmission operations for specific image features. The transmission process of high-resolution images is guided by the importance of the image content, takes into account the availability and potential risks of the communication link, dynamically configures the distribution of transmission resources, and achieves the dual goals of ensuring image quality and stable system operation through refined management, thus obtaining high-resolution image transmission results.

[0018] In one example, multi-level downsampling and feature extraction processing are performed on the high-resolution raw image acquired by the image acquisition card to obtain global image features and regional image features, including: The high-resolution raw images acquired by the image acquisition card are downsampled step by step according to a preset scaling factor to obtain an image sequence with decreasing resolution; Perform convolution operations on each image in the image sequence with decreasing resolution to obtain the corresponding feature map data; The real-time communication bandwidth value between the image acquisition card and the host is obtained, and the real-time communication bandwidth value is compared and analyzed with the preset bandwidth threshold sequence to obtain the current optimal cache selection result. Based on the current optimal cache selection result, the feature map data is classified according to a predefined priority order to obtain feature data groups with different priorities; The feature data groups of different priorities are allocated corresponding cache areas to obtain a multi-level cache storage layout. The feature map data is then stored in the corresponding cache area according to the multi-level cache storage layout to obtain multi-level image feature data. Content analysis is performed on multi-level image feature data to obtain global image features and regional image features.

[0019] In this example, the original image frames acquired by the image acquisition card are used as input, and hierarchical downsampling processing is performed based on a set of preset scaling factors to construct an information structure with multi-scale resolution. The original image is considered as the first-level scale image, i.e., the layer with the highest resolution and largest data volume. The system gradually reduces the image along the width and height directions according to the set downsampling ratio, thereby generating several image copies with progressively decreasing resolution. These images constitute a set of image sequences with clear structure and uniform scale distribution. The image feature extraction module is called to perform convolution operations on each frame of the image sequence with decreasing resolution. By designing convolution kernel structures and feature channel settings adapted to different scales, it is ensured that multi-dimensional features such as texture, edges, contours, and local intensity that match the scale characteristics are extracted from each downsampled image, generating a set of feature map data corresponding to the image resolution level. The current communication bandwidth value between the image acquisition card and the host is obtained in real time through a bandwidth awareness mechanism, and this real-time bandwidth parameter is compared with the bandwidth threshold sequence set in the system interval by interval. By comparing the current bandwidth interval position, the approximate range of currently available transmission resources is determined, forming a decision basis for cache allocation priority. The analysis results are transformed into a current optimal cache selection result, that is, determining which levels of feature maps should be prioritized for cache allocation under the current communication environment, and what priority structure should be used for storage to maximize the feasibility of critical data transmission and avoid excessive congestion of system resources. All previously generated feature map data are prioritized according to the current optimal caching strategy. Each feature map is assigned a priority label based on its image level, feature strength, semantic importance, and adaptability to the communication environment. These labels are used to divide all feature maps into multiple feature data groups, such as high-priority, medium-priority, and low-priority groups. Each priority feature data group is mapped to the physical partition structure of the image acquisition card's internal cache resources, and a corresponding cache area of ​​different size is allocated to different priority groups. For example, high-priority feature data is allocated to a cache area with fast response speed and low access latency, while medium- and low-priority feature data are arranged in a relatively slow but sufficiently large buffer based on the current system load. When performing cache allocation, factors such as the current cache area occupancy, data importance, network transmission availability, and expected retransmission costs are comprehensively considered to construct a multi-level cache storage layout. Once this layout is established, all feature map data are systematically written to their respective cache areas according to their grouping, cache priority, and target region attributes. This generates a data buffer state with a clear structure, distinct priorities, and strong physical isolation, forming a multi-level image feature data system. This multi-level image feature data is then used as input to perform deep content analysis to extract global and regional image features.The extraction of global image features is based on a cross-scale fusion strategy of feature maps, integrating features from different levels and strengthening the response of key channels through a semantic attention mechanism to construct a descriptive vector covering the main structure, edge layout, and overall semantic composition of the image. In the extraction of regional image features, salient region detection, region scoring function evaluation, and local cropping operations are performed on the global feature map to automatically select several image regions with strong semantic expression and visual attention. These regions are then subjected to secondary convolution extraction to obtain local feature representations containing rich semantic details and region boundary information. Global image features and regional image features constitute a dual-structure description of the image content.

[0020] In one example, content analysis is performed on multi-level image feature data to obtain global image features and regional image features, including: By using a global feature extraction operator, feature fusion processing is performed on multi-level image feature data to obtain global image features; The global image features are analyzed using an attention mechanism, and the weight coefficients of each resolution level feature are calculated to obtain weighted image features. The importance of image content is calculated based on weighted image features and region scoring functions to obtain a region importance map; Based on the regional importance map, the target region is located in the high-resolution original image to obtain the target region coordinate set. The original image region corresponding to the target region coordinate set is then cropped to obtain the target region image set. The local feature extraction operator is used to perform feature calculation on the target region image set to obtain the region image features.

[0021] In this example, a global feature extraction operator is used to perform feature fusion processing on multi-level image feature data. Multi-level image feature data is a set of feature maps at multiple scales generated by progressively downsampling and convolving the original high-resolution image using an image acquisition card. It exhibits significant hierarchy and content differences. Therefore, the key to global feature extraction lies in addressing the scale inconsistency between different resolutions and constructing a cross-scale fusion structure adaptable to all levels. Through convolutional layers, residual connections, pooling operations, and feature channel alignment, high-level low-resolution abstract features are weighted and fused with low-level high-resolution detail features. Through continuous merging, fusion, and projection processing, a fused feature expression with uniform size and dimension is formed. This expression constitutes the initially constructed global image features. After obtaining the global image features, an attention mechanism analysis module is introduced to perform fine-grained channel perception and spatial response allocation on the fused global features. The attention mechanism addresses the problem of uneven contribution of multi-scale features, as the information extracted from different resolution layers has different impacts on image recognition or target extraction in specific scenarios; therefore, it dynamically weights these features. The system analyzes the importance of features at each resolution level in the final fusion output using a channel attention mechanism. Simultaneously, a spatial attention model is introduced to weight the information density of image regions in spatial structure, forming a set of weight coefficients corresponding to the feature map scale. These weight coefficients represent the expression priority of each resolution level in the fusion output; a higher weight indicates that the feature map at that resolution is more critical to the overall image representation. The system applies these weight coefficients to the fusion feature map, forming a weighted image feature representation that considers both structure and content. Based on this weighted image feature, a region scoring function is introduced to analyze the importance of image content, extracting the most valuable image regions from the fusion features. The region scoring function, based on a sliding window strategy or adaptive gridding, divides the entire image into several continuous and fully covered sub-regions. Within each sub-region, a set of structural consistency metrics, edge change intensity, semantic response intensity, and attention activity indices are calculated. These indices are combined and weighted to construct a region scoring matrix, yielding a quantitative indicator of the region's expressive power within the overall image. All scoring results are merged to form a region importance map, which is structurally a two-dimensional heatmap. The pixel value at each location represents the importance level of that region in semantic image recognition; a higher value indicates a greater contribution to image understanding. Local extrema extraction and region connectivity analysis are performed on the region importance map to identify a set of spatially important locations in the image. Based on the high-scoring hotspots in the region importance map, their corresponding spatial coordinates are traced back in the original high-resolution image to construct a target region coordinate set. This coordinate set represents the bounding boxes of all regions determined to be highly important in the original image, including the starting point coordinates, region width and height information, and scale reference.Based on this coordinate information, corresponding high-precision image sub-blocks are cropped from the original image to form a target region image set. This image set retains the resolution and image quality of the original image and possesses obvious regional feature focusing characteristics, reflecting the local structure of the target subject or semantically salient region in the image. Feature computation is performed on the target region image set using a local feature extraction operator. This operator employs a lightweight deep convolutional neural structure, combined with variable receptive fields, reconfigurable convolutional kernels, or self-attention mechanisms, to extract high-density features from multiple levels, including spatial edges, texture variations, and local geometric morphology. The local feature structure has high spatial resolution and high texture recognition capability, capable of capturing fine features suppressed by the global fusion process, such as small target boundaries, curve bends, and fine-grained information like local occlusion. The local feature vector of each region image is bound to its spatial location information to construct a region image feature set.

[0022] In one example, random latency and data loss characteristics are analyzed on historical communication data of an image acquisition card to obtain a probability distribution model and continuous data loss records, including: Statistical characteristics of the delay data from past transmission cycles of the image acquisition card are calculated to obtain the mean, variance, and distribution characteristics of the delay data. Continuous data loss event detection is performed during the data transmission process of the image acquisition card to obtain the value of the continuous data loss counter; Based on the mean, variance, and distribution characteristics of the delay data, a piecewise function model is used to model the transmission delay probability, resulting in the transmission delay probability function. Based on the value of the consecutive data loss counter, a decreasing factor is calculated to determine the risk of consecutive data loss, thus obtaining the consecutive data loss risk coefficient. The transmission delay probability function is parameterized to obtain a probability distribution model, and the continuous data loss risk coefficient and historical loss events are recorded in time series to obtain a continuous data loss record.

[0023] In this example, statistical characteristics are calculated for the delay data of the image acquisition card over past transmission cycles. A high-resolution delay trajectory dataset is constructed by recording the delay data over time. Based on this dataset, statistical characteristic analysis is performed, calculating the overall mean to characterize the average transmission delay level, and calculating the variance to measure its fluctuation range and identify the stability of the communication link across different cycles. All acquired delay values ​​are subjected to frequency statistics and normalized distribution processing to form the distribution characteristics of the delay data over the value range. This distribution reveals the concentrated areas of delay events, tail fluctuations, and outlier segments. While statistically summarizing the delay behavior, a mechanism for real-time monitoring of data transmission integrity is constructed in parallel to detect whether data loss events occur during transmission by the image acquisition card. By monitoring the received feedback signal of each data packet cycle by cycle, when the host does not receive valid image data from the acquisition card, the cycle is automatically recorded as a loss event, and a continuous loss counter is incremented. When a data transmission is successfully completed, the counter is reset to zero. This step constructs a continuous data loss counting curve, reflecting the frequency and density of data link interruptions. It also indirectly reveals the system's fault tolerance performance when encountering high latency or network fluctuations during operation, helping to identify continuous failures caused by network jitter. Based on the mean, variance, and distribution characteristics of latency, the transmission latency risk is transformed into a computable piecewise function model. This model divides the entire latency range into several intervals, such as low-latency, medium-latency, and high-latency regions, mapping the latency frequency within each interval to a risk level or probability factor, forming a piecewise function relationship with clear boundaries and approximately uniform segments. The function's role is to transform latency values ​​into risk expressions, enabling the system to make predictions based on historical patterns when facing uncertain future communication states. For example, when the system's current latency value is in the upper-middle range, the model can predict the probability of a high-latency state recurring within a subsequent time window, and accordingly trigger resource reallocation or redundancy mechanisms in advance, improving the system's feedforward response capability. Simultaneously, the loss risk is quantified based on the current continuous loss counter value. To avoid misjudgments or response delays caused by linear mapping, a decreasing factor model is adopted. This means that the higher the number of consecutive data loss events, the lower the system's tolerance for further data loss in a future period, thus calculating a consecutive loss risk coefficient. This risk coefficient exhibits a non-linear decreasing trend, representing a dynamic compression of the transmission link's trust level. When the risk coefficient approaches its lower limit, the system automatically upgrades its fault tolerance mechanism or interrupts unnecessary transmissions to ensure the priority of transmitting core image content. The transmission delay probability function is parameterized and transformed into an embeddable probability distribution model.This model uses interval probability, trend, and segmented weights as core parameter inputs, supporting rapid function interpolation or probability prediction based on real-time observations during runtime, thereby improving the system's runtime decision-making efficiency. Simultaneously, it binds the continuous loss risk coefficient to the loss event number occurring in each period, constructing a loss time-series record chain. This chain includes multiple indicators such as timestamps, number of losses, loss level, and repair status.

[0024] In one example, using a probability distribution model and continuous data loss records, multi-scenario prediction of the future transmission status of the image acquisition card is performed to obtain predicted status information, including: Based on the probability distribution model, the transmission delay range is divided into intervals to obtain multiple transmission delay scenarios. For each transmission delay scenario, a prediction model based on historical transmission data is constructed to obtain a set of prediction models. For each prediction model in the prediction model set, select multiple time points and corresponding delay values ​​as interpolation nodes to obtain an interpolation node set, and construct a first polynomial function for each prediction model based on the interpolation node set; The error between the prediction result of the first polynomial function and the actual transmission delay is calculated to obtain the target error value. The interpolation nodes are then updated based on the target error value to obtain the second polynomial function. Based on the second polynomial function, the predicted value weight allocation is calculated according to the historical prediction accuracy of the prediction model set to obtain the predicted state information.

[0025] In this example, the transmission delay probability distribution model is structurally analyzed and used as the mathematical basis for dividing transmission behavior scenarios. The entire historical transmission delay numerical domain is divided into several statistically representative sub-intervals, each corresponding to a specific transmission delay scenario. This scenario represents a typical performance of a certain type of communication state, such as a low-latency stable scenario, a medium-latency fluctuating scenario, or a high-latency congestion scenario. This interval division process combines the changing trend of the delay probability density distribution, the location of the distribution extrema, and the slope change points of the distribution curve to ensure that each interval physically reflects a stable operating state or fluctuation cycle of the real network environment. After the division is completed, an independent prediction model is established for each interval. Each model uses the delay interval to which the current scenario belongs as the input sample range and uses the fitting strategy that best matches the historical data within that interval for modeling, forming a unified and logically separated set of prediction models. This set will provide multiple perspectives for judging future transmission states through multi-path collaborative prediction during runtime. After establishing the model set, initialization processing is performed on each model before interpolation construction. Multiple representative time points are selected from the historical data of the latency scenarios covered by the model, and these time points are mapped one-to-one with their corresponding transmission latency values ​​to construct an interpolation node set. The selection of nodes follows the principles of uniform distribution, fluctuation coverage, and sensitivity to change; that is, while ensuring the integrity of the temporal sequence as much as possible, time points with drastic latency fluctuations are prioritized to ensure that the subsequent interpolation function still has sufficient fitting ability in dynamic environments. For each interpolation node set, a corresponding first interpolation function, called the first polynomial function, is constructed using a polynomial structure with high fitting stability. This function uses interpolation to predict and estimate the latency behavior within the range covered by the model, providing a preliminary reference path for short-term prediction. The error between the latency values ​​predicted by the first polynomial function and the actual latency values ​​over several future periods is calculated to obtain the target error value. This error reflects the accuracy of the current interpolation function in characterizing latency changes in this scenario. If the target error value is within a reasonable threshold range, it indicates that the model has high reliability and can be used for subsequent weight allocation; conversely, if the error exceeds the system's set tolerance threshold, the system triggers the interpolation node update mechanism. The update process involves deleting the oldest set of time points from the current node set, incorporating the most recent actual delay observations, and reconstructing the second polynomial function based on the new nodes. The new function retains the basic trend characteristics of the previous fit while incorporating the latest real-world changes in transmission status, enabling the interpolation model to adaptively adjust its prediction path in the face of sudden bandwidth changes, link instability, or network jitter. Once the second polynomial function of all models has been updated, the output results of the entire prediction model set are summarized and processed.Because different prediction models correspond to different delay scenarios, their prediction accuracy varies significantly across different time periods. Therefore, a weighting mechanism driven by historical prediction accuracy is introduced. The system tracks the average error fluctuation, prediction trend stability, and response accuracy to sudden delays for each prediction model over several prediction periods, calculating a dynamic credibility index for each model. Then, based on this credibility, a weight coefficient is assigned to each model; models with higher weights will have a larger proportion in the final prediction result. The system then sums the predicted values ​​of all second polynomial functions at the same time point according to their respective weights, obtaining a set of composite prediction state information that integrates multi-scenario prediction capabilities.

[0026] In one example, an adaptive transmission control strategy is generated based on predicted state information, global image features, and regional image features, and a high-resolution image transmission process is executed to obtain the high-resolution image transmission result, including: Based on the predicted state information, the current transmission delay, data loss status, and bandwidth information are integrated to obtain the system state vector; The importance of global image features and regional image features is classified to obtain core data groups and non-core data groups. Based on the system state vector, the control input vector for the data transmission ratio of different buffer regions is constructed to obtain the initial transmission control parameters. Based on the initial transmission control parameters and the system state vector, a cost function is used to calculate the balance between transmission delay and buffer usage to obtain the transmission control cost value. Then, rolling time-domain optimization is performed on the transmission control cost value to obtain the optimal transmission control sequence. Based on the optimal transmission control sequence and the risk of core data packet loss, priority allocation of data transmission resources is performed to obtain an adaptive transmission control strategy; The high-resolution image transmission process is executed based on an adaptive transmission control strategy to obtain the high-resolution image transmission result.

[0027] In this example, based on the predicted state information, the current transmission delay, data loss status, and bandwidth information are integrated. Delay information reflects the temporal stability of the current transmission link, data loss status indicates data continuity risk, and bandwidth data determines the upper limit of transmission resources. Through standardization and vector merging of these three state indicators, a set of system state vectors with temporal and structured characteristics is formed. Global and regional image features are graded by importance. Based on the semantic density, visual saliency, and structural coverage of features in image representation, a hierarchical priority strategy is adopted to divide all feature data into two groups: core data and non-core data. The core data group includes incompressible parts of the image such as the main structure, boundary contours, and key areas for target recognition, while non-core data includes background information, gradient textures, or redundant and repetitive areas. This data can be appropriately compressed or processed with lower priority during transmission. After grouping, the resource usage of each level of buffer within the image acquisition card is analyzed based on the current system state vector. A control input vector is constructed according to the priority requirements of the image data groups to clarify the data transmission ratio of different buffer areas. For example, when the system predicts an impending decrease in bandwidth or an increase in latency, the control input vector proactively increases the proportion of bandwidth allocated to core data and compresses the residence time of non-core data in the cache, forming a set of initial transmission control parameters with transmission adaptability and content priority awareness capabilities. Based on the initial transmission control parameters and the system state vector, a cost function is calculated to balance transmission latency and cache usage, dynamically evaluating the matching degree between the control parameters and the current system state. The cost function measures the balance between latency pressure, cache resource usage, and image integrity maintenance brought about by the current configuration in actual operation. The system uses transmission latency and cache usage as two core variables, dynamically weighting them to calculate the cost intensity of the configuration in the current system state, forming a quantitatively expressed control cost value. To improve the effectiveness and response continuity of the long-term control strategy, this cost value is used as the objective function input into the rolling temporal optimization module. Within the prediction perspective, the state evolution trend for several future cycles is recursively iterated to comprehensively generate an optimal control path sequence with the minimum cost over the entire time period. Guided by this optimal transmission control sequence, and considering the transmission loss risk levels corresponding to various image features in the core data packets, the system performs fine-grained allocation of buffer resources and transmission bandwidth. The system prioritizes allocating data corresponding to image regions with the highest loss risk to the front of the transmission queue, and sets up redundancy and fault tolerance mechanisms or replicates redundant channels for them. Core data with lower risk is transmitted according to the standard scheduling ratio specified in the optimal path. As for non-core data, if the system status indicates current resource scarcity, its transmission priority will be dynamically downgraded, or it will enter a delayed or local downsampling processing flow. This meticulously responsive resource allocation structure constitutes the system's adaptive transmission control strategy.This strategy possesses multiple control attributes, including content-level driven, state-change adaptive, path-optimized driven, and resource-elastic allocation, enabling guaranteed transmission scheduling of critical image information even under resource-constrained or drastically changing environmental conditions. During strategy execution, the image acquisition card progressively pushes image data according to the transmission plan of each data group in the control strategy. The cache manager dynamically adjusts the cache allocation ratio and update cycle based on the control input vector, while the transmission module allocates data paths according to the channel priority, bandwidth utilization, and loss tolerance defined by the strategy. The entire transmission process is a full-link closed-loop sensing control, reflecting both the image content structure and dynamically sensing the characteristics of the link environment. After each frame of image transmission, the system collects the state monitoring results for the current period and inputs them into the state update module, forming a continuously evolving state-control feedback chain. This results in high-resolution image transmission.

[0028] In one example, a high-resolution image transmission process is performed based on an adaptive transmission control strategy, resulting in the following high-resolution image transmission results: According to the adaptive transmission control strategy, transmission resources are allocated to core data packets and non-core data packets to obtain hierarchical transmission data streams. Quality evaluation indicators are calculated for the transmitted image data in the hierarchical transmission data streams to obtain the transmission image quality score. Based on the comparison between the transmitted image quality score and the preset quality threshold, the control gain of the adaptive transmission control strategy is dynamically adjusted to obtain optimized control parameters. Based on optimized control parameters, stability analysis is performed on the state changes of the transmission system to obtain system stability assessment results; Real-time monitoring and anomaly recovery are performed on the transmission latency, data loss rate, and buffer occupancy rate of the image acquisition card to obtain the corrected transmission status; Based on the system stability assessment results and the corrected transmission status, the performance indicators and quality optimization of the transmitted high-resolution images are calculated to obtain the high-resolution image transmission results.

[0029] In this example, image data is hierarchically scheduled and managed according to an adaptive transmission control strategy. This strategy specifies the transmission priority, cache ratio, and bandwidth allocation ratio for core and non-core data. During actual execution, image feature data is scheduled according to the priority order set in the control strategy based on the semantic importance of the image content. This is combined with the current cache availability and channel load capacity to construct a hierarchical transmission data stream with a clear structure. In this data stream, core data is preferentially sent to a high-speed cache and allocated to the main channel, characterized by short transmission cycles, low error tolerance, and redundant verification mechanisms. Non-core data is sent to a relatively low-speed cache or delayed transmission area, with its transmission frequency and retransmission capability constrained by the strategy, primarily scheduled when system resources are abundant. Quality assessment indicators are calculated for the transmitted image data in the hierarchical transmission data stream, based on signal-to-noise ratio, structural similarity, and edge sharpness indicators, and incorporating more realistic perceived quality indicators such as content integrity detection and regional structure preservation rate. These indicators are then aggregated across multiple time windows to form a quality score representing the actual image transmission performance. The quality score of the transmitted image is compared with a preset quality threshold to assess whether the current transmission strategy effectively utilizes system resources while ensuring the integrity of image representation. When the quality score is found to be lower than the system threshold, especially when there is image distortion, blurred details, or incomplete content in key local areas, the system activates a control gain adjustment mechanism. This mechanism dynamically corrects parameters such as bandwidth allocation ratio, cache scheduling weight, and data packet priority in the control strategy to form a set of optimized control parameters. The dynamic stability of the entire transmission system is analyzed using these optimized control parameters to verify whether the optimization strategy introduces potential risks or system oscillations. During the stability analysis, the transmission delay curve, cache usage fluctuation amplitude, and loss rate trend are observed. Window statistics and rate of change calculations are used to determine whether the system remains under control. If the adjustment of the optimized parameters is too large, causing frequent cache overflows or sudden bandwidth increases leading to channel contention imbalances, the system triggers a strategy rollback mechanism, reverting to a control state within the stable range. Conversely, if the system response is stable and the indicators slowly approach the ideal state, it indicates that the current round of control parameter adjustment is effective, and the system continues to iterate and optimize in the next cycle. Meanwhile, to prevent the system from losing control due to sudden link failures or external interference during operation, a real-time status monitoring mechanism centered on an image acquisition card is constructed to continuously sample and analyze three key indicators: transmission latency, data loss rate, and buffer occupancy rate. Through high-frequency sampling and boundary threshold judgment, abnormal states such as sudden increases in latency, continuous packet loss, or near-saturation of the buffer are detected in the first instance. Based on the anomaly type, the corresponding recovery mechanism is invoked, such as latency path switching, buffer refresh, scheduling channel adjustment, and redundant data insertion, forming a closed-loop structure covering perception, judgment, response, and repair.After each anomaly recovery, the current transmission state is recalculated based on the repair effect and input as the "corrected transmission state" into the control feedback chain to participate in the next round of strategy generation and resource scheduling. Comprehensive performance analysis and structural quality optimization are performed on the transmitted high-resolution images. The actual performance of the transmitted images in terms of sharpness, structural restoration rate, regional contrast, and semantic consistency is evaluated from a whole-image perspective, and these evaluation results are compared with the original acquired images to measure the true impact of transmission loss on content quality. Combining the system stability evaluation results and the corrected state information, image quality optimization strategies are formulated, such as adaptive sharpening for blurred edge areas, feature-based texture restoration for texture loss areas, and applying block artifact suppression algorithms to compression distortion areas, thereby improving the perceptual quality of the image output and ultimately obtaining the high-resolution image transmission result.

[0030] Reference Figure 2 This embodiment provides a high-resolution image acquisition and transmission device, including: Feature extraction module 1 is used to perform multi-level downsampling and feature extraction processing on the high-resolution raw images acquired by the image acquisition card to obtain global image features and regional image features; Feature analysis module 2 is used to perform random delay and data loss feature analysis on the historical communication data of the image acquisition card, and obtain probability distribution model and continuous data loss records; Multi-scenario prediction module 3 is used to perform multi-scenario prediction of the future transmission status of the image acquisition card by using a probability distribution model and continuous data loss records, and obtain the predicted status information. Image transmission module 4 is used to generate an adaptive transmission control strategy based on the predicted state information, global image features and regional image features, and execute the high-resolution image transmission process to obtain the high-resolution image transmission result.

[0031] In this embodiment, the specific implementation of each unit in the above device embodiment is described in the above method embodiment, and will not be repeated here.

[0032] Reference Figure 3 This invention also provides a computer device, which can be a server, and its internal structure can be as follows: Figure 3As shown, the computer device includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores the data corresponding to this embodiment. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0033] Those skilled in the art will understand that Figure 3 The structures shown are merely block diagrams of some structures related to the present invention and do not constitute a limitation on the computer devices on which the present invention is applied.

[0034] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0035] Those skilled in the art will understand that all or part of the processes in 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. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the present invention and embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0036] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0037] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for acquiring and transmitting high-resolution images, characterized in that, include: Multi-level downsampling and feature extraction processing are performed on the high-resolution raw images acquired by the image acquisition card to obtain global image features and regional image features; Random delay and data loss characteristics were analyzed on the historical communication data of the image acquisition card to obtain a probability distribution model and continuous data loss records; Using the probability distribution model and the continuous data loss records, multi-scenario prediction of the future transmission status of the image acquisition card is performed to obtain the predicted status information. An adaptive transmission control strategy is generated based on the predicted state information, the global image features, and the regional image features, and a high-resolution image transmission process is executed to obtain the high-resolution image transmission result.

2. The high-resolution image acquisition and transmission method according to claim 1, characterized in that, The process involves multi-level downsampling and feature extraction of the high-resolution raw images acquired by the image acquisition card to obtain global image features and regional image features, including: The high-resolution raw images acquired by the image acquisition card are downsampled step by step according to a preset scaling factor to obtain an image sequence with decreasing resolution; Perform a convolution operation on each image in the image sequence with decreasing resolution to obtain the corresponding feature map data; The real-time communication bandwidth value between the image acquisition card and the host is obtained, and the real-time communication bandwidth value is compared and analyzed with a preset bandwidth threshold sequence to obtain the current optimal cache selection result. Based on the current optimal cache selection result, the feature map data is classified according to a predefined priority order to obtain feature data groups with different priorities; The feature data groups of different priorities are allocated corresponding cache regions of different sizes to obtain a multi-level cache storage layout, and the feature map data is stored in the corresponding cache regions according to the multi-level cache storage layout to obtain multi-level image feature data. Content analysis is performed on the multi-level image feature data to obtain global image features and regional image features.

3. The high-resolution image acquisition and transmission method according to claim 2, characterized in that, The content analysis of the multi-level image feature data to obtain global image features and regional image features includes: The multi-level image feature data is fused using a global feature extraction operator to obtain global image features. The global image features are analyzed using an attention mechanism, and the weight coefficients of each resolution level feature are calculated to obtain weighted image features. The importance of image content is calculated based on the weighted image features and the region scoring function to obtain a region importance map; Based on the region importance map, the target region is located in the high-resolution original image to obtain the target region coordinate set, and the original image region corresponding to the target region coordinate set is cropped to obtain the target region image set. The target region image set is processed by a local feature extraction operator to obtain the region image features.

4. The high-resolution image acquisition and transmission method according to claim 1, characterized in that, The analysis of random delay and data loss characteristics of historical communication data from the image acquisition card to obtain a probability distribution model and continuous data loss records includes: Statistical characteristics of the delay data from past transmission cycles of the image acquisition card are calculated to obtain the mean, variance, and distribution characteristics of the delay data. Continuous data loss event detection is performed on the data transmission process of the image acquisition card to obtain the value of the continuous data loss counter; Based on the mean, variance, and distribution characteristics of the delay data, a piecewise function model is performed on the transmission delay probability to obtain the transmission delay probability function. Based on the value of the continuous data loss counter, a decreasing factor is calculated for the risk of continuous data loss to obtain the continuous data loss risk coefficient. The transmission delay probability function is parameterized to obtain a probability distribution model, and the continuous data loss risk coefficient and historical loss events are recorded in time sequence to obtain a continuous data loss record.

5. The high-resolution image acquisition and transmission method according to claim 1, characterized in that, The step of using the probability distribution model and the continuous data loss records to perform multi-scenario prediction of the future transmission status of the image acquisition card, and obtaining predicted status information, includes: Based on the probability distribution model, the transmission delay range is divided into intervals to obtain multiple transmission delay scenarios, and a prediction model based on historical transmission data is constructed for each transmission delay scenario to obtain a set of prediction models. For each prediction model in the prediction model set, select multiple time points and corresponding delay values ​​as interpolation nodes to obtain an interpolation node set, and construct a first polynomial function for each prediction model based on the interpolation node set; The error between the prediction result of the first polynomial function and the actual transmission delay is calculated to obtain the target error value, and the interpolation nodes are updated according to the target error value to obtain the second polynomial function; Based on the second polynomial function, the predicted value weight allocation is calculated according to the historical prediction accuracy of the prediction model set to obtain the predicted state information.

6. The high-resolution image acquisition and transmission method according to claim 1, characterized in that, The step of generating an adaptive transmission control strategy based on the predicted state information, the global image features, and the regional image features, and executing the high-resolution image transmission process to obtain the high-resolution image transmission result includes: Based on the predicted state information, the current transmission delay, data loss status, and bandwidth information are integrated to obtain the system state vector; The global image features and the regional image features are classified by importance to obtain core data groups and non-core data groups. Based on the system state vector, the data transmission ratio of different buffer regions is constructed by controlling the input vector to obtain the initial transmission control parameters. Based on the initial transmission control parameters and the system state vector, a cost function is used to calculate the balance between transmission delay and buffer usage to obtain the transmission control cost value. Then, the transmission control cost value is optimized in the rolling time domain to obtain the optimal transmission control sequence. Based on the optimal transmission control sequence and the loss risk of the core data packets, priority allocation of data transmission resources is performed to obtain an adaptive transmission control strategy; The high-resolution image transmission process is executed based on the adaptive transmission control strategy to obtain the high-resolution image transmission result.

7. The high-resolution image acquisition and transmission method according to claim 6, characterized in that, The process of performing high-resolution image transmission based on the adaptive transmission control strategy to obtain high-resolution image transmission results includes: According to the adaptive transmission control strategy, transmission resources are allocated to the core data packets and the non-core data packets to obtain hierarchical transmission data streams. Quality evaluation indicators are calculated for the transmitted image data in the hierarchical transmission data streams to obtain transmission image quality scores. Based on the comparison between the transmitted image quality score and the preset quality threshold, the control gain of the adaptive transmission control strategy is dynamically adjusted to obtain optimized control parameters. Based on the optimized control parameters, a stability analysis of the state changes of the transmission system is performed to obtain the system stability assessment results. The transmission delay, data loss rate, and buffer occupancy rate of the image acquisition card are monitored in real time and anomalies are recovered to obtain the corrected transmission status; Based on the system stability assessment results and the corrected transmission status, the performance indicators and quality optimization of the transmitted high-resolution image are calculated to obtain the high-resolution image transmission result.

8. A high-resolution image acquisition and transmission device, characterized in that, The step of implementing the high-resolution image acquisition and transmission method according to any one of claims 1 to 7, wherein the high-resolution image acquisition and transmission device comprises: The feature extraction module is used to perform multi-level downsampling and feature extraction processing on the high-resolution raw images acquired by the image acquisition card to obtain global image features and regional image features; The feature analysis module is used to perform random delay and data loss feature analysis on the historical communication data of the image acquisition card to obtain a probability distribution model and continuous data loss records. The multi-scenario prediction module is used to perform multi-scenario prediction of the future transmission status of the image acquisition card using the probability distribution model and the continuous data loss record, and obtain the predicted status information. The image transmission module is used to generate an adaptive transmission control strategy based on the predicted state information, the global image features, and the regional image features, and to execute the high-resolution image transmission process to obtain the high-resolution image transmission result.

9. 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 high-resolution image acquisition and transmission method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the high-resolution image acquisition and transmission method according to any one of claims 1 to 7.

Images