Methods for transmitting new media video data based on digital technology

Abstract

This invention discloses a method for transmitting new media video data information based on digital technology, relating to the field of information and communication technology, including the following steps: S100, extracting semantic features from received video data frames, modeling image structure information pixel by pixel, and constructing a semantic feature map; S200, matching the semantic feature map of the current frame with the semantic feature maps of adjacent reference frames, calculating semantic offset magnitude indicators and structural disturbance change indicators, and marking interference candidate regions. This invention constructs an anomaly detection and repair closed loop based on semantic feature information maps, breaking through the limitations of traditional bit-level verification, achieving multi-dimensional feature fusion and structural consistency comparison, and possessing higher recognition accuracy and response efficiency. It can achieve accurate compensation of local abnormal regions without increasing transmission load, significantly improving the semantic stability and fault tolerance of video in complex network environments.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information and communication technology, in particular to a method for transmitting new media video data information based on digital technology. BACKGROUND

[0002] "Transmitting new media video data information based on digital technology" refers to a data transmission method that uses digital coding, compression, network communication and other information processing and transmission technologies to transmit video content on a new media platform from a content source (such as a server, camera, editing terminal) to a user terminal (such as a mobile phone, television, computer). The process usually includes digital acquisition of video signals, coding processing based on compression algorithms, distribution and transmission through IP networks or wireless networks, and decoding and playing on the terminal side. Compared with traditional analog transmission methods, digital technology-based transmission has the advantages of large data capacity, high transmission quality, strong anti-interference ability, support for interaction and multi-terminal collaboration, etc., and is widely used in short video platforms, live streaming services, online video on demand, virtual reality and other new media scenarios, realizing efficient, stable and low-latency video content dissemination.

[0003] The prior art has the following disadvantages:

[0004] In the existing video data transmission process, video content is usually transmitted in the form of compressed bit stream in the network, although the system usually configures traditional check mechanisms such as cyclic redundancy check (CRC) or forward error correction coding (FEC) to prevent data damage, but in complex network environment or edge computing nodes, there may be weak transmission anomalies that are not completely covered, such as individual bit flips or unstructured disturbances of intra-frame data. Such errors often cannot be identified in time by conventional check mechanisms and do not cause obvious abnormalities in video playback, with strong concealment. However, in application scenarios such as intelligent recommendation, video content identification or violation detection that rely on image semantic feature extraction, such weak errors may cause subtle but continuous interference to the texture distribution, color features or structural patterns of video frames, thereby causing distortion of video feature vector expression, causing misjudgment of subsequent algorithm models, and causing serious consequences such as content label deviation, recommendation logic failure or distortion of audit strategy.

[0005] The above information disclosed in the background section is only used to strengthen the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0006] The purpose of the present application is to provide a method for transmitting new media video data information based on digital technology to solve the problems in the background technology.

[0007] In order to achieve the above object, the present application provides the following technical scheme: a method for transmitting new media video data information based on digital technology, comprising the following steps:

[0008] S100, performing semantic feature extraction on the received video data frame, modeling image structure information pixel by pixel, and constructing a semantic feature map;

[0009] S200, matching the semantic feature map of the current frame with the semantic feature map of the adjacent reference frame, calculating a semantic offset amplitude index and a structure disturbance change index, and marking a disturbance candidate region;

[0010] S300, selecting a historical reference frame with the highest semantic stability according to the disturbance candidate region, generating a semantic structure residual map through structure mapping inversion, and constructing an initial scale of disturbance response strength;

[0011] S400, extracting disturbance intensity parameters, disturbance distribution parameters and disturbance stability parameters based on the semantic structure residual map, generating an abnormal score result, and constructing an abnormal quantitative representation;

[0012] S500, fusing the abnormal score result, color gradient distribution features, texture direction arrangement features, semantic offset amplitude index and structure disturbance change index, performing feature clustering and classification, and outputting an abnormal type identification result;

[0013] S600, according to the spatial position binding information in the abnormal type identification result, the abnormal score result, the semantic offset amplitude index and the structure disturbance change index, adjusting the transmission scheduling strategy.

[0014] Preferably, step S100 comprises:

[0015] Obtaining the current frame and converting it into a joint expression form of a gray image and a color image;

[0016] Performing joint operation of Sobel operator and Laplace operator on the pixels of the red, green and blue channels respectively, and extracting color gradient distribution features;

[0017] Extracting texture direction arrangement features based on the direction covariance matrix and the gray level co-occurrence matrix of the local pixel region;

[0018] Using a multi-threshold Canny edge detection algorithm combined with Hough transform to extract edge contour closure features;

[0019] Fusing the color gradient distribution feature map, the texture direction arrangement feature map and the edge contour closure feature map to construct a semantic feature information map.

[0020] Preferably, step S200 comprises:

[0021] Perform multidimensional feature vector normalization to establish a semantic feature space with a unified scale;

[0022] Euclidean distances are calculated for matched pixels to construct semantic offset intensity maps, and a local window aggregation mechanism is introduced.

[0023] Extract structural perturbation change indices from edge contour closure feature maps and texture direction alignment feature maps;

[0024] Based on the spatiotemporal sliding window, the co-anomaly of semantic offset value and structural perturbation value is detected, and the interference candidate region is marked.

[0025] Preferably, step S300 includes:

[0026] Select the historical reference frame with the highest semantic stability and extract the image content at its corresponding location;

[0027] Establish structural mapping relationships and complete geometric alignment by feature point matching and affine transformation or thin plate spline transformation;

[0028] The process involves embedding and fading the mapped image to create a visually and semantically continuous restored region.

[0029] Calculate the semantic structural information residual between the restored image and the current frame, and construct an initial scale for the interference response intensity.

[0030] Preferably, step S400 includes:

[0031] The semantic offset magnitude index and the structural disturbance change index are jointly normalized to calculate the disturbance intensity parameter;

[0032] Spatial clustering is performed on highly perturbation pixels to extract perturbation distribution parameters;

[0033] Based on multi-frame residual trajectory analysis, disturbance stability parameters are extracted;

[0034] The obtained parameters are fused to generate anomaly feature scoring results, and anomaly quantification maps bound to spatial locations are constructed.

[0035] Preferably, step S500 includes:

[0036] Construct a unified and standardized set of multidimensional feature vectors;

[0037] An unsupervised clustering algorithm is used to cluster the features of abnormal regions;

[0038] Supervised classification processing is performed by combining prior knowledge of anomaly types to identify transmission interference-type semantic structure distortions.

[0039] The classification results are bound to the image spatial coordinates to output the anomaly type identification results.

[0040] Preferably, step S600 includes:

[0041] Extract the location information of the target area marked as transmission interference anomaly and establish a mapping relationship;

[0042] Construct a transmission scheduling priority model that integrates semantic offset magnitude index and structural disturbance change index;

[0043] Based on scheduling priority, video data retransmission processing is performed at the target area level;

[0044] After the reissue is completed, the semantic offset magnitude index and structural disturbance change index of the repaired area are recalculated to determine the repair effect.

[0045] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0046] This invention, based on semantic feature information graphs, constructs a complete processing chain from anomaly detection to semantic recovery and scheduling correction through multi-dimensional feature fusion and structural consistency comparison. It does not rely on traditional bit-level verification mechanisms, overcoming the limitations of existing transmission quality assurance methods in semantic layer perception. Compared with existing technologies, this invention significantly improves recognition granularity, response efficiency, and robustness. It can achieve rapid compensation and content restoration of local anomaly areas without increasing the overall data load, effectively improving the semantic stability, system fault tolerance, and user perception quality of video data in complex network environments. It is widely applicable to new media scenarios with high requirements for image semantic accuracy, such as intelligent recommendation, video review, and visual recognition, and has strong engineering application prospects and commercial value. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0048] Figure 1 This is a flowchart of the method for transmitting new media video data information based on digital technology according to the present invention. Detailed Implementation

[0049] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0050] This invention provides, for example Figure 1 The method for transmitting new media video data information based on digital technology, as shown, includes the following steps:

[0051] S100 performs semantic feature information extraction processing on the received video data frames, performs image structure information modeling pixel by pixel, extracts color gradient distribution features, texture direction arrangement features and edge contour closure features, and constructs a complete semantic feature information map.

[0052] The received video data frames undergo semantic feature extraction processing, employing a pixel-by-pixel image structure information modeling approach. This aims to provide high-resolution, highly stable structural representation information for subsequent semantic anomaly identification and transmission correction. The specific steps include:

[0053] The current video data frame from the receiving end is acquired and converted into a joint representation of grayscale and color images to account for both luminance and color channel information. For each pixel in the color image channels, its color gradient value in the color space is calculated, including the pixel change rate of the red, green, and blue channels. A multi-directional, multi-scale color gradient distribution feature map is obtained through a joint operation of the Sobel and Laplacian operators. This feature map reflects the color change trend of the image in different regions, facilitating the identification of significant regions between edges, textures, and object contours in the image. Unlike traditional single-channel processing methods, this step introduces multi-channel coupled gradient calculation, improving the sensitivity of the color gradient features to detailed regions, thereby more effectively capturing minor perturbations that may be introduced during transmission.

[0054] The Sobel and Laplacian operators are commonly used edge detection and gradient analysis methods in the field of image processing, used to capture the changing trends and structural information of pixel values ​​in an image.

[0055] In the color gradient feature extraction process involved in this step, the Sobel operator is used to detect color changes in the horizontal and vertical directions of the image. That is, by calculating the gradient response of the region around each pixel in the red, green and blue color channels respectively, it highlights the color edges, contours and other features with obvious directionality in the image. The Laplacian operator is used to detect abrupt changes in pixel grayscale or color values ​​in space, which can enhance the overall contour structure of the edges in the image, and is especially suitable for capturing boundary changes in areas with strong contrast or complex textures.

[0056] In practice, the Sobel operator is first applied to each color channel (red, green, and blue) of the color image to extract color gradient maps in the horizontal and vertical directions, forming a preliminary directional edge response map. Next, the Laplacian operator is applied to the image of the same channel to enhance the change response of the edge region in the image and identify more details that do not have obvious directionality but have strong structural changes. Finally, the response results of the two operators are fused within the channel to form a multi-directional color gradient map for each channel. The fused results of the three channels are then further superimposed and normalized to construct a color gradient distribution feature map at a unified scale.

[0057] By combining the Sobel and Laplacian operators, we can not only extract linear edges with clear directions, but also capture detailed features with no obvious direction but significant changes. This enhances the image's ability to perceive structure in multiple directions and at multiple scales, thereby constructing a more stable and complete color gradient feature representation and improving the accuracy and sensitivity of subsequent semantic analysis.

[0058] Building upon the extracted color gradient distribution features, this step further models the texture orientation arrangement features. Specifically, by calculating the orientation covariance matrix within local pixel regions, the arrangement trend of the main texture orientations is determined, and a gray-level co-occurrence matrix is ​​introduced to characterize the orientation consistency and repetition patterns of the texture. In actual processing, the image is divided into several fixed-size sub-regions, and the texture orientation distribution is statistically analyzed for each sub-region, generating a corresponding texture orientation feature map. This feature map reflects the spatial organization and directional structural changes of the object's surface texture, thus establishing a stable texture orientation reference framework within the overall image structure. Compared to methods that only use orientation gradient histograms or simple edge orientation extraction, this step has stronger orientation consistency analysis capabilities, can more accurately model the directional arrangement features of complex texture regions within a frame, and improves the sensitivity to locally disturbed regions during subsequent comparisons.

[0059] This paper models the edge and structural closure relationships of images and extracts edge contour closure features. This step employs a multi-threshold Canny edge detection algorithm to extract edge point sets from the image at multiple scales, and further uses Hough transform to detect edge segments, identifying geometric boundary regions with structural closure features. To further enhance the semantic structural representation of edge information, a boundary closure metric function is introduced. By calculating the difference in structural integrity between closed and open edges, an edge structure integrity scoring matrix is ​​established. This matrix can serve as a basis for judging region stability, assisting in identifying regions that may experience breakage, drift, or false edge diffusion during video data transmission. Compared with traditional edge detection methods, this invention incorporates structural closure analysis into the edge feature modeling process, enabling the extracted edge features to not only reflect the contours themselves but also the spatial relationships and structural dependencies between contours.

[0060] In this step, to accurately extract structural edge information from video frames, a multi-threshold Canny edge detection algorithm is used to perform multi-scale edge modeling of the image. Canny edge detection is a classic image edge extraction method with strong noise resistance and edge localization accuracy. Traditional Canny algorithms use a single pair of high and low thresholds to binarize pixel gradients, which can easily lead to loss of detail information or edge breaks. The "multi-threshold Canny edge detection algorithm" described in this invention performs multiple edge extractions on the image by setting multiple thresholds with different intensity levels, extracting strong edges (main contours) and weak edges (detail structures) respectively. Based on this, a fusion strategy is used to weight and superimpose different edge layers to generate a unified edge response map. To enhance the stability of edge detection in scale space, a Gaussian pyramid is further introduced to construct a multi-scale image version, and the above multi-threshold edge detection process is performed on images at each scale to obtain edge point sets at different scale resolutions. Through intra-scale edge completion and inter-scale edge fusion, a set of edge points with structural coherence and hierarchical features is finally obtained, providing stable input for subsequent geometric structure recognition.

[0061] After extracting the edge point set, to further analyze the geometric structural relationships in the image, this invention introduces Hough transform for edge segment identification and extraction of structurally closed regions. Hough transform is a method for detecting specific geometric features (such as lines, circles, ellipses, etc.) in an image in parameter space. It can accurately identify structural boundaries with strong continuity and conforming to geometric rules within complex edge point sets. In this step, the multi-scale fused edge map is used as input, and all edge points are transformed and mapped in polar coordinates. Each edge point in the image space is projected as a curve in the parameter space, and the intersection points of different edge points in the parameter space curves correspond to collinear edge points in the image. By detecting the peak values ​​of the parameter space accumulation array, all edge segments with strong geometric correlation in the image can be extracted. Furthermore, the identified line segments are combined according to spatial topological relationships to identify closed polygonal contour structures. A structural closure score is calculated based on boundary continuity and structural density to determine whether the current region belongs to an image region with high semantic integrity. This boundary detection method is more robust than pixel-level edge contour analysis and can effectively identify edge structure drift, breakage or pseudo-edge diffusion caused by weak perturbations in video frames, providing accurate geometric support for semantic structure anomaly localization.

[0062] By integrating color gradient distribution feature maps, texture orientation feature maps, and edge contour closure feature maps, a complete semantic feature information map of the current video frame is constructed. This map, based on spatial coordinates, jointly expresses various feature information to form a set of structural representations in a high-dimensional semantic feature space, comprehensively reflecting the structural state of the current frame image. To enhance the temporal continuity and matching robustness of the map in dynamic video frame sequences, a feature normalization mechanism based on weighted multi-feature fusion is introduced. By calculating the local rate of change and global stability of each type of feature, different weighting coefficients are assigned to each feature channel, forming a semantic feature vector map at a unified scale. This map not only possesses strong semantic expressive power but also serves as input for subsequent steps such as semantic feature comparison, anomaly region identification, and semantic structure reconstruction, ensuring that the entire methodology achieves closed-loop operation of anomaly detection and response control based on high-precision semantic modeling.

[0063] Semantic feature extraction is performed on the received video data frames, and image structure information modeling is carried out pixel by pixel. Color gradient distribution features, texture direction alignment features, and edge contour closure features are extracted to construct a complete semantic feature information map. Its core function is to establish a structured and semantic representation of the video frame image content, enabling a deeper understanding of the image content and sensitive perception of subtle perturbations. In traditional video transmission, although video frames can be successfully decoded and played, subtle pixel perturbations, such as color drift, texture distortion, or edge breakage, are often difficult to detect by low-order pixel difference recognition mechanisms. This is especially true in AI image processing systems, where these changes can potentially affect the feature extraction module, leading to model misjudgments. Therefore, this step constructs a refined image structure information modeling process to extract image features from multiple dimensions: color gradient distribution reflects local and global color change trends, texture direction alignment models the repetition and order within regions, and edge contour closure features reveal the shape, boundaries, and structural integrity of objects in the image. The joint extraction of three types of features enables the system not only to acquire the underlying physical changes of images but also to uncover the potential structural patterns at the semantic layer. By constructing a unified semantic feature information map, it provides a complete, stable, and high-resolution structural foundation for subsequent video frame-to-frame feature comparison, anomaly detection, and region repair. This processing method enhances the system's ability to discriminate and tolerate intra-frame perturbations, improves the accuracy and robustness of semantic consistency analysis and anomaly response mechanisms, and is a key prerequisite for ensuring data transmission reliability in this invention.

[0064] S200: Load the semantic feature information comparison process in chronological order, match the semantic feature information map of the current frame with the semantic feature information map of at least one adjacent reference frame, calculate the semantic offset magnitude index and structural disturbance change index, identify the spatial region corresponding to the abnormal fluctuation trend, and mark it as an interference candidate region.

[0065] Loading semantic feature information in chronological order for comparison is a crucial step in achieving the perception of structural and semantic changes between video data frames and the accurate localization of abnormal regions. This step, based on temporal matching and structural change analysis of the semantic feature information maps between the current frame and the reference frame, is used to extract semantic offset magnitude indicators and structural perturbation change indicators, identify potential abnormal fluctuation regions, and ultimately form a spatial label set of interference candidate regions. The specific process includes the following steps:

[0066] One frame before and one frame after the current received frame in the video sequence are selected as reference frames to construct a temporal sliding window structure, ensuring the continuity of temporal comparison and the stability of semantic reference. The semantic feature information maps of the current frame and the reference frames are used as inputs, and the color gradient distribution features, texture direction arrangement features, and edge contour closure features in the maps are normalized to a unified dimension, ensuring the comparability and spatial matching of different feature types. The normalization process uses a combination of zero-mean normalization and variance standardization to standardize the multiple feature vectors at each pixel, thereby constructing a multi-dimensional semantic feature space at a unified scale.

[0067] In a unified scale space, pixel-by-pixel matching is performed between the current frame and the reference frame, and the semantic offset magnitude index of corresponding pixel positions in the two frame maps is calculated. This index measures the difference between the multi-dimensional feature vectors of each pixel using the Euclidean distance function, and, combined with the continuity constraint of spatial position, constructs a complete semantic offset intensity map. To further improve the spatial consistency of the matching, a local window aggregation mechanism is introduced during the calculation process, using the average feature difference within a certain pixel neighborhood as the regional offset index to suppress local noise interference and enhance the spatial saliency expression of abnormal structural changes.

[0068] The difference measurement of the multidimensional feature vector of each pixel using the Euclidean distance function is used to measure the degree of structural change in semantics between the current frame and the reference frame. Specifically, each pixel in the image not only has a single color value, but also contains multi-dimensional semantic feature information, such as color gradient distribution, texture orientation, and edge contour closure. These features together constitute a high-dimensional feature vector. Calculating the distance between the high-dimensional vectors of corresponding pixels in the current frame and the reference frame reflects the difference in semantic expression of that point at two time points. To avoid misjudgment due to noise caused by isolated pixels, a continuity constraint of spatial location is introduced. That is, within the surrounding neighborhood of each pixel, the feature difference trend between neighboring pixels is further statistically analyzed, and the feature changes in local areas are weighted and smoothed to emphasize the consistency of the change trend. In this way, not only is the accuracy of pixel-level semantic changes preserved, but also occasional fluctuations caused by local perturbations are suppressed. Finally, a semantic shift intensity map covering the entire frame image is formed, clearly indicating which regions have significant semantic shifts in the temporal dimension, providing a quantitative basis for subsequent anomaly detection.

[0069] Based on the aforementioned semantic offset intensity map, a structural perturbation change index extraction process is performed. The structural perturbation change index is primarily used to assess the degree of fluctuation in image edge closure and texture orientation consistency. Specifically, differential analysis is performed on the edge contour closure feature maps of the current frame and the reference frame to detect the degree of geometric structural changes in closed boundary regions; simultaneously, principal direction offset analysis is performed on the texture orientation alignment feature map to identify abrupt changes in texture orientation within the region. Through this structural comparison, a set of spatially corresponding structural perturbation change index maps is generated. These indices can reveal semantic anomalies in the image structure layer caused by transmission perturbations, even in regions with relatively stable color distribution.

[0070] By combining semantic offset magnitude index maps and structural perturbation change index maps, a joint anomaly trend judgment is performed to identify anomaly trends in regions experiencing continuous changes. Specifically, a local anomaly detection method based on spatiotemporal sliding windows is employed. Co-anomaly peak points of semantic offset values ​​and structural perturbation values ​​are found in the time series of each pixel and spatially aggregated into local anomaly regions. Based on satisfying the conditions of anomaly continuity and structural strength threshold, these regions are marked as interference candidate regions, forming a spatially bound candidate region identification map. This map can be used to guide subsequent semantic restoration and anomaly correction processing, serving as a key intermediate result for achieving image perturbation recognition and accurate region localization.

[0071] The local anomaly detection method based on spatiotemporal sliding windows analyzes the semantic change trend of each pixel simultaneously in both temporal and spatial dimensions to identify anomalous regions caused by transmission perturbations. The specific steps are as follows:

[0072] In the video sequence, a time-series window is constructed for each pixel. This window covers the semantic offset and structural perturbation values ​​at the corresponding position of that pixel in the current frame and several frames before and after it. By analyzing the fluctuation trajectory of the offset and perturbation values ​​in this time window, it is possible to identify whether there are synchronously occurring abnormal peak points, that is, both indicators rise sharply at the same time position and deviate significantly from their time average and normal fluctuation range.

[0073] Spatial neighborhood analysis is performed on pixels marked as anomalous peaks to search for pixel clusters with similar anomalous behavior within their neighboring pixel range. If a group of spatially adjacent pixels exhibit coordinated anomalous responses simultaneously in the same or adjacent frames, and their anomalous patterns are consistent in both semantic and structural dimensions, then this pixel set is collectively marked as a local anomalous region. Through the dual constraints of temporal continuity and spatial consistency, this method can effectively filter isolated anomalies caused by random noise, while highlighting stable anomalous regions caused by semantic structural perturbations, providing a precise localization basis for subsequent semantic repair and transmission control.

[0074] The process loads semantic feature information in chronological order and compares the semantic feature information map of the current frame with the semantic feature information maps of at least one adjacent reference frame. Its core function is to establish a time-oriented semantic consistency analysis mechanism to accurately identify image structural changes caused by subtle disturbances during data transmission. Traditional video transmission quality assurance mechanisms typically rely on channel-level verification and retransmission mechanisms to ensure data integrity. However, these methods often only detect severe data corruption or complete frame loss, failing to detect more subtle problems such as minor disturbances or single-bit changes. This invention, through semantic feature information map comparison, performs inter-frame difference analysis at the semantic level rather than the pixel level, enabling more sensitive detection of phenomena such as texture structure changes, color gradient shifts, or edge contour breaks caused by image detail disturbances. By constructing a semantic shift magnitude index, the intensity change of the image's semantic expression in the time series is quantified; by using a structural disturbance change index, local disturbance trends at the image structural level are captured, thereby achieving precise localization of potential abnormal fluctuation trends. By further combining temporal sliding windows and spatial aggregation mechanisms, pixel clusters exhibiting anomalous behavior are marked as interference candidate regions, providing clear target areas for subsequent semantic inpainting processing. This process occupies a crucial central position in the entire video data anomaly identification and correction workflow, not only establishing the basis for anomaly identification but also ensuring the accuracy of positioning and the efficiency of compensation in subsequent response mechanisms.

[0075] S300 performs semantic structure information restoration processing based on interference candidate regions. According to the semantic offset amplitude index and structural perturbation change index, it selects the historical reference frame with the highest semantic stability, restores the content of the corresponding region in the current frame through structural mapping inversion, generates semantic structure information residual map, and constructs an initial scale of interference response intensity.

[0076] Based on the identified interference candidate regions, semantic structure information restoration processing is performed. This aims to address local semantic structure shifts or damage caused by weak perturbations during transmission. Through structural mapping inversion of historical reference frames, high-fidelity reconstruction of the content of abnormal regions in the current frame is achieved. This process not only restores the integrity of the image's semantic expression but also quantifies the impact of abnormal perturbations, providing a data foundation for subsequent interference response intensity modeling and transmission control optimization. The specific steps are as follows:

[0077] Based on the interference candidate regions marked in the previous processing steps, a set of historical frames consisting of several frames preceding the current frame is extracted from the time series. Image segments corresponding to the interference candidate regions in the current frame are then retrieved from each historical frame. To ensure the semantic continuity and stability of the reference image content used for restoration, statistical analysis is performed on the semantic shift magnitude and structural perturbation change indices of the corresponding regions in each historical frame. By setting a stability evaluation threshold, the regions in each historical frame are sorted and filtered. Finally, the frame with the smallest semantic shift and the weakest structural perturbation is selected as the reference frame with the highest semantic stability, which drives the restoration process of the interference regions in the current frame.

[0078] After identifying the historical reference frame with the highest semantic stability, a structural mapping relationship is constructed to address the subtle spatial structural changes between it and the current frame. Specifically, feature points are first extracted and matched from the interference regions of both the reference and current frames. A sparse correspondence is established using scale-invariant feature transformation. Then, affine transformation models or thin-plate spline transformations are used to perform geometric alignment and structural mapping on the interference regions, completing the semantic structural projection from the reference frame to the current frame. This structural mapping method not only maintains the texture and edge consistency of the image content but also preserves the spatial layout features of the original structure, enabling high-quality inversion and restoration of the content of the disturbed region while ensuring semantic coherence.

[0079] By using affine transformation models or thin-plate spline transformations to perform geometric alignment and structural mapping on the interference region, the purpose is to solve problems such as positional offset, deformation, and scale inconsistency between the reference frame and the current frame in the local image structure. This ensures that the semantic content in the reference frame can be accurately projected onto the interference region of the current frame in spatial geometry, thereby achieving high-fidelity semantic structure restoration. The specific steps are as follows: First, key feature points, such as corner points, edge intersections, or texture markers, are extracted from the current frame and semantically stable historical reference frames within the interference region. A point-to-point correspondence between the two frames is established using a feature matching algorithm. Second, based on the matched control point pairs, an appropriate geometric mapping model is selected—if the scene deformation is small, an affine transformation model is used to adjust the image linearly in terms of scale, rotation, and translation; if the region exhibits nonlinear deformation or complex bending structures, a thin-plate spline transformation is employed, constructing a flexible mapping function based on the principle of minimum bending energy to achieve more refined structural alignment. Third, the established transformation model is used to map and deform the corresponding image content in the reference frame to the position and shape of the interference region in the current frame, ensuring that the semantic structure maintains its original spatial relationships without distortion. Finally, edge blending processing is applied to the mapping results to eliminate boundary fragmentation and ensure a natural and coherent transition. This geometric alignment process not only restores the consistency of the image structure but also ensures the spatial correctness of the semantic content, serving as a crucial supporting step for high-quality video frame semantic restoration.

[0080] After structural mapping is completed, a semantic structural information content replacement and restoration process is performed, which involves embedding the mapped reference frame image fragment into the interference candidate region of the current frame. The embedding operation employs boundary gradient fusion technology to ensure a smooth transition in color, texture, and structural features between the restored region and the surrounding undisturbed region, avoiding new semantic discontinuities. Subsequently, the restored image is compared pixel-by-pixel with the original current frame image, and the multidimensional feature differences between the two are calculated to form a semantic structural information residual map. This map is used to intuitively characterize the structural deviation in semantic expression before and after image restoration and is an important indicator for measuring the degree of anomaly in the interference region and the restoration effect.

[0081] The embedding operation employs boundary gradient blending technology, which aims to address issues such as color abrupt changes, texture breaks, or structural misalignments that may occur at the blending edge after the reference frame is mapped to the interference area of ​​the current frame. This ensures seamless connection between the restored content and the original image at both the visual and semantic levels, avoiding the generation of new semantic discontinuities or perceptual conflicts. The specific steps are as follows: First, a transition band is set around the edge of the restored region. The width of this transition band is adaptively adjusted according to the image resolution and region scale, generally covering a boundary range of 5 to 20 pixels. Second, within this transition band, pixel-level fusion processing is performed on the mapped region and the original image of the current frame. A weighted averaging strategy is used to interpolate the color values ​​of the overlapping regions. The fusion weight gradually increases from the edge inwards and gradually decreases from the edge outwards, thus achieving a gradual transition effect. Third, texture direction consistency detection is performed on the fused region to correct any fusion inconsistencies caused by texture direction deviations. If necessary, local texture synthesis techniques are used to supplement texture continuity. Finally, after fusion, edge smoothing and structural consistency constraint analysis are performed on the entire fused region to further correct any potential boundary sharpening or structural abruptness issues, ensuring that the restored region is naturally embedded in the overall image of the current frame in terms of spatial structure. This boundary gradient fusion technique can significantly improve the overall image quality after semantic structure restoration, effectively avoiding secondary anomalies caused by abrupt region stitching, and laying a stable image foundation for subsequent anomaly intensity quantification and semantic consistency analysis.

[0082] Finally, based on the semantic structure information residual map mentioned above, the residual amplitude of each pixel in the feature space is statistically analyzed. Combined with the spatial range, offset trend, and structural change intensity of the interference region, an initial interference response intensity scale is constructed. This scale, using pixels as units, records the interference response weight of each interference region in the current frame, serving as a crucial input for subsequent anomaly quantification assessment and classification. Through this preliminary quantification of interference intensity, not only can the influence of anomaly regions be graded, but it can also assist in adjusting data retransmission strategies and priority allocation mechanisms in further processing, thus forming a complete semantically driven interference response closed loop.

[0083] The main function of semantic structure information restoration processing based on interference candidate regions is to provide a high-precision and high-stability repair mechanism for image semantic structure anomalies caused by weak perturbations during video transmission. This ensures that the image content of the current frame can be restored to a near-original expression at both the structural and semantic levels. This step introduces the semantic structure of historical reference frames as the basis for restoration. Combining semantic offset magnitude indicators and structural perturbation change indicators, it accurately selects the historical image segments with the most stable semantic expression within the corresponding regions. Then, using a structure mapping inversion method, it geometrically aligns the high-fidelity image content from the reference frames and embeds it into the abnormal region of the current frame, achieving spatial restoration and structural reconstruction of the image content. This restoration not only restores the image's color, texture, and edge features but also ensures the coherence of semantic expression, avoiding errors in video semantic understanding or misjudgments in subsequent algorithms due to local perturbations. Furthermore, after image restoration, semantic difference analysis between the original frame and the restored frame constructs a semantic structure information residual map, thereby quantifying the degree of offset of the abnormal region in the high-dimensional feature space. The residual map can intuitively reflect the range of changes in color, texture direction, and structural contour before and after restoration, providing data for further interference identification. Finally, based on the feature difference distribution of the residual map and the structural complexity of the interference region, an initial interference response intensity scale is constructed to numerically characterize the interference intensity of each anomalous region, providing key quantitative support for subsequent anomaly level assessment, classification identification, and transmission priority scheduling. Overall, this step not only undertakes the task of image structure reconstruction but also completes the detection and quantitative analysis of anomaly impacts at the semantic level, making it a core link in realizing semantically driven anomaly correction and transmission feedback closed-loop control.

[0084] S400, based on semantic structural information residual map, performs anomaly feature information evaluation processing. It uses semantic offset amplitude index and structural perturbation change index as joint input to extract perturbation intensity parameter, perturbation distribution parameter and perturbation stability parameter, generate anomaly feature scoring results bound to spatial location, and construct a comprehensive anomaly quantitative representation.

[0085] This study utilizes semantic structural information residual maps to perform anomaly feature information evaluation processing. The aim is to conduct structured, quantified, and spatially bound feature analysis of the anomaly degree of repaired regions in video data frames to support subsequent anomaly type identification and transmission response decisions. This process integrates semantic offset magnitude and structural perturbation change indices as core input parameters, comprehensively evaluates the performance differences and trends of disturbed regions in the feature space, extracts key perturbation attributes, and generates complete anomaly feature scoring results. The specific steps are as follows:

[0086] Using the semantic structural information residual map as input, preliminary statistics on anomalous features are performed on each pixel within the interference candidate region. By jointly normalizing the semantic shift magnitude index and the structural perturbation change index, the weight bias caused by scale differences between the two indices is eliminated, ensuring the equivalence of the two types of features in subsequent calculations. For each pixel, a perturbation intensity parameter is calculated based on its feature value in the residual map, representing the degree of drastic change at the semantic and structural levels. This parameter is a crucial basis for evaluating the significance of the interference, characterizing the intensity level of semantic destruction in a local region.

[0087] After calculating the intensity of a single-point disturbance, an aggregation analysis of the local disturbance region is performed to extract the disturbance distribution parameters. These parameters are used to evaluate the spatial diffusion characteristics and morphological features of the anomaly, mainly including the connectivity, density, boundary complexity, and coverage area of ​​the anomaly region. Specifically, all pixels with disturbance intensities exceeding a set threshold are clustered into several spatially continuous disturbance regions. For each region, spatial characteristic indicators are calculated, such as centroid location, shape moments, and bounding rectangle ratio, thereby constructing a distribution parameter that comprehensively reflects the disturbance morphology. This processing step helps identify whether the anomaly is an isolated point disturbance or a structural disturbance with directional and regional expansion characteristics, providing a basis for subsequent processing strategies.

[0088] To address the issues of temporal consistency and semantic continuity in anomalies, a perturbation stability parameter is extracted to measure whether the interference is sporadic noise or a structural anomaly with temporal consistency. Specifically, based on the residual evolution trajectory of the current frame and several frames before and after it at the same spatial location, a time-series perturbation curve is constructed, and its fluctuation trend, duration, and peak amplitude are statistically analyzed. Regions exhibiting high-amplitude perturbations across multiple frames are assigned higher stability scores to distinguish between instantaneous perturbations caused by noise and persistent anomalies caused by transmission errors or structural misalignments. This process further enhances the robustness and discriminative power of anomaly detection.

[0089] The three types of parameters—disturbance intensity, disturbance distribution, and disturbance stability—are weighted and fused together to generate anomaly feature scores according to preset rules. These scores are then bound to their spatial locations to form a complete anomaly quantification map. This map is a two-dimensional structure, where each pixel or pixel region is assigned a comprehensive anomaly score, intuitively reflecting its anomaly degree and spatial location. Based on this, a comprehensive anomaly quantification representation model is further established to globally evaluate the anomaly risk level of the entire video frame, providing structured and data-driven input support for subsequent anomaly identification, classification, processing, and transmission correction scheduling.

[0090] The core function of anomaly feature evaluation based on semantic structure information residual maps is to perform quantitative, precise, and spatially localized anomaly characterization analysis on regions in the image that may be affected by weak perturbations after anomaly detection and semantic repair in video transmission. This step goes beyond simple identification of the presence of anomalies; instead, it delves into the multi-dimensional properties of anomalies through two key inputs: semantic offset magnitude indicators and structural perturbation change indicators, constructing more discriminative anomaly descriptions. First, by extracting perturbation intensity parameters, the severity of the impact on a region at the semantic structure level can be determined, which helps distinguish between minor errors with low interference and high-intensity structural damage. Second, perturbation distribution parameters reveal the spatial morphology and clustering trend of anomalies, enabling analysis of whether anomalies are local random noise or widespread interference with diffusion and connectivity, providing a basis for spatial aggregation and boundary identification. Third, perturbation stability parameters are used to measure the consistency and persistence of the region in the time series, thereby distinguishing between instantaneous perturbations within a single frame and structural anomalies that continue across frames. The anomaly feature scoring results generated by fusing these three parameters are not only comprehensive and detailed but also achieve visualization and precise localization of anomaly information in the image coordinate system through spatial location binding. This scoring result, as a core component of the comprehensive anomaly quantitative characterization, not only provides a foundation for subsequent anomaly type classification and interference source tracing, but also offers precise reference for priority determination and data retransmission strategy formulation in the subsequent scheduling mechanism. Therefore, this step is a crucial bridge connecting anomaly identification and system response, and a core element in achieving an intelligent, controllable, and highly robust image transmission and restoration closed loop. It demonstrates the high precision, strong interpretability, and application-oriented capabilities of this invention at the anomaly analysis level.

[0091] S500, based on the anomaly feature scoring results, integrates color gradient distribution features, texture direction arrangement features, semantic offset magnitude index and structural perturbation change index, performs feature clustering and classification processing, identifies whether the semantic structure distortion is caused by weak interference during transmission, and outputs the anomaly type identification result.

[0092] Based on the anomaly feature scoring results, color gradient distribution features, texture direction arrangement features, semantic offset magnitude indicators, and structural perturbation change indicators are integrated to perform feature clustering and classification processing. This aims to conduct multi-dimensional semantic structure analysis on identified anomaly regions and further determine whether the anomaly is caused by weak interference during video transmission, thereby outputting accurate anomaly type identification results. This processing flow not only improves the accuracy of anomaly identification but also provides category-level decision-making basis for subsequent anomaly response strategies. The specific implementation steps are as follows:

[0093] The generated anomaly feature scores are standardized and unified with the original image's color gradient distribution features, texture orientation features, semantic offset magnitude indicators, and structural perturbation change indicators to construct a complete set of high-dimensional feature vectors. Each pixel or region fragment in a candidate anomaly region is represented as a joint vector containing multiple semantic and structural attributes. To ensure the fusion quality between features from different sources, each feature is normalized according to its numerical distribution to eliminate dimensional differences and ensure that all types of information contribute equally in subsequent clustering calculations. This process lays the numerical foundation for clustering and classification analysis in the feature space.

[0094] Based on the constructed multidimensional feature vector set, an unsupervised clustering algorithm is used to perform preliminary feature clustering on all anomalous regions. The clustering process not only groups regions based on pixel-level feature similarity but also introduces spatial connectivity constraints to prevent regions with similar features but spatially fragmented structures from being misclassified into one category. By analyzing the statistical distribution of each cluster center in terms of color gradient variation, texture direction consistency, semantic offset magnitude, and structural perturbation complexity, the inherent structural differences of anomalous features are preliminarily identified, and anomalous clusters that may represent typical interference patterns are identified.

[0095] Based on the constructed multidimensional feature vector set, an unsupervised clustering algorithm is used to perform preliminary feature clustering on all abnormal regions. Its main function is to automatically classify abnormal regions with similar semantic perturbation features, providing a clear internal structure division for subsequent abnormal type identification, thereby achieving more accurate and targeted classification judgment. The specific steps are as follows: First, each pixel or pixel region identified as an anomaly is transformed into a feature vector containing multi-dimensional attributes. This vector integrates information such as color gradient distribution features, texture direction arrangement features, semantic offset magnitude index, structural perturbation change index, and residual response score, forming a sample set in a high-dimensional feature space. Second, the feature set is standardized to unify the numerical scale of each dimension, avoiding bias caused by an excessively large feature value during clustering. Third, a suitable unsupervised clustering algorithm is selected, such as K-means, DBSCAN, or spectral clustering, to cluster the anomaly samples based on their similarity in the feature space. K-means is suitable for anomaly distributions with obvious contour structures, while DBSCAN is more suitable for irregular anomaly regions with large density variations. Fourth, based on the clustering results, the center vector and boundary range of each cluster are extracted, and the feature distribution within the clusters is analyzed to identify the performance patterns and significant feature differences of different anomaly categories in the feature dimensions. Through the above clustering steps, abnormal regions with similar shapes and consistent features can be automatically classified, which not only improves the efficiency of subsequent classification, but also enhances the contextual relevance and local consistency of anomaly type identification. It is an important intermediate step in building a highly robust anomaly identification process.

[0096] Based on the completed clustering, and combined with existing prior knowledge of anomaly types, supervised classification is performed, assigning semantic labels to each cluster. This classification process employs a multi-feature fusion discrimination strategy, judging based on the following criteria: if an anomaly region exhibits unnatural abrupt changes in color gradient features, disordered arrangement in the texture direction, and is accompanied by a large semantic shift amplitude and high-frequency structural perturbation values, it is determined to be a semantic structural distortion caused by weak interference during transmission with a high probability; conversely, if the region's feature shift is low and the structural changes are stable, it may be due to changes in the image content itself or other natural scene interference. The classifier combines the importance weights of multi-dimensional features to perform classification decisions, ensuring the model possesses strong discriminative ability and semantic interpretability.

[0097] The classification results are bound to the image coordinate space to output anomaly type identification results. Each anomaly region marked as "transmission interference-type semantic structure distortion" is not only assigned a clear semantic category label, but also includes attribute information in multiple dimensions such as location, range, and intensity, which is used to guide subsequent transmission repair scheduling and data priority strategy configuration. This anomaly type identification result can be used as an output item of the system's anomaly perception module to achieve accurate identification and highly reliable classification of the impact of hidden disturbances in video images.

[0098] Based on the anomaly feature scoring results, this method integrates color gradient distribution features, texture direction arrangement features, semantic offset magnitude indicators, and structural perturbation change indicators to perform feature clustering and classification. This enables deep semantic analysis of anomaly regions, further clarifying their causes and identifying whether semantic structural distortions are caused by weak interference during video data transmission. This provides a precise basis for subsequent anomaly correction strategies and transmission optimization mechanisms. In video images, anomaly regions can be caused by various factors, such as natural scene changes, target occlusion, lighting changes, image compression artifacts, or transmission perturbations. Relying solely on residual scoring cannot distinguish the true source of the anomaly. Therefore, this step expands the original pixel-level anomaly score into a high-dimensional feature expression with semantic richness and spatial geometric features by integrating multi-dimensional semantic and structural features. Then, using unsupervised clustering and supervised classification methods, it automatically extracts regions with color abrupt changes, texture distortion, severe semantic offset, and frequent structural perturbations, marking them as suspected transmission interference anomalies. These regions often exhibit locality, concealment, and cross-frame continuity, making them difficult to identify using traditional video quality detection mechanisms. The method of this invention can not only accurately distinguish such anomalies, but also assign them clear anomaly type labels, enabling the system to take targeted remedial measures based on the specific type of anomaly, such as refined data retransmission, local channel optimization, and content scheduling priority adjustment. In summary, this step plays a key role as a "semantic diagnostician" in the entire anomaly detection and semantic correction process, and is an indispensable part of achieving a high-resolution, intelligent, closed-loop image transmission quality assurance mechanism.

[0099] S600 performs transmission channel correction processing based on the spatial location binding information, anomaly feature scoring results, semantic offset magnitude index, and structural disturbance change index contained in the anomaly type identification results. It also jointly regulates the transmission scheduling strategy and content scheduling priority, and resends video data to the corresponding target area, thus completing the closed-loop anomaly response control process from semantic feature extraction to transmission content correction of video data.

[0100] Based on the spatial location binding information, anomaly feature scoring results, semantic offset magnitude index, and structural disturbance change index contained in the anomaly type identification results, transmission channel correction processing is performed. This aims to intelligently identify and optimize the targeted data retransmission and scheduling of regions with semantic structural anomalies caused by weak interference during video data transmission. This constructs a complete closed-loop anomaly response control process, achieving full-process closed-loop control from image semantic extraction to content correction. Specifically, it includes the following steps:

[0101] Spatial location binding information output from the anomaly type identification step is extracted to identify the location of target regions affected by semantic structural perturbations in the current frame. Each region marked as a "transmission interference anomaly" contains precise spatial coordinates, area range, anomaly level score, and contextual relationship within the image structure. Combining this information, the anomaly region is extracted from the full-frame image and a mapping relationship with the original data transmission is established, providing a spatial basis for subsequent target region-level data scheduling and repair strategies. Furthermore, anomaly feature scoring results are extracted simultaneously to construct response level labels for anomaly regions, clarifying the interference intensity of each region and its semantic impact in the current frame.

[0102] Based on the aforementioned anomaly spatial information and scoring levels, a semantically guided transmission scheduling priority model is constructed. This model integrates semantic offset magnitude and structural perturbation change indices to assign priority scores to each anomaly region and rank them according to their importance in the semantic integrity of the image. For example, high-scoring anomaly regions with complex textures or sensitive edge structures are given high priority, while low-frequency texture regions or regions with low visual attention are assigned lower scheduling weights. Furthermore, data blocks in the original video frames are reorganized, and targeted retransmission plans are formulated to maximize the benefits of semantic restoration under limited bandwidth conditions.

[0103] The process involves retransmitting video data for the target area. This process does not retransmit entire frames; instead, it selects only the area-level data requiring correction based on the location and number of the abnormal region for targeted scheduling. During scheduling, a dynamic transmission path selection strategy is employed, prioritizing redundant, low-latency paths for data retransmission. If the target area suffers severe data loss, an original encoding redundancy mechanism can be activated to extract uncompressed or lightly compressed versions of the data from the encoding end for retransmission, ensuring image restoration accuracy. Simultaneously with data retransmission, the receiving end undergoes synchronous processing, reserving reconstructable capabilities for the abnormal region's image channels to ensure seamless integration of the retransmitted content into the existing frame structure, achieving a balance between structural consistency and temporal continuity.

[0104] After the resent data arrives and the embedding and fusion processing is completed, the semantic structure consistency verification process is re-executed. The semantic offset index and structural disturbance index of the corrected region are recalculated to determine whether the preset recovery threshold is met. If the anomaly score drops below the safe range, the region is marked as "repaired." If the anomaly score is still in a critical state, the transmission priority is further increased based on the current score feedback, and the resent content is rescheduled until the semantic structure is stably restored. Through this method, a closed-loop transmission control mechanism is formed, from anomaly identification, score modeling, scheduling control to semantic backtracking, enabling self-awareness, self-repair, and semantically stable transmission of video content in complex network environments.

[0105] Based on the spatial location binding information, anomaly feature scoring results, semantic offset magnitude indicators, and structural disturbance change indicators contained in the anomaly type identification results, the transmission channel correction process is performed. This enables proactive compensation and closed-loop correction control after accurately identifying local semantic structural anomalies caused by weak interference during video transmission. This breaks the traditional "passive packet loss-retransmission-waiting" lag mode of video transmission, constructing an intelligent scheduling mechanism with semantic self-awareness capabilities. This step uses the spatial location and feature scoring of the anomaly region to accurately locate the most severely affected and structurally sensitive areas. Combined with semantic offset and structural disturbance degree, it determines the weight of the impact on semantic integrity and visual quality in the entire frame image, and dynamically adjusts the transmission scheduling strategy and content scheduling priority accordingly. Compared to traditional full-frame redundant transmission, this strategy emphasizes "on-demand retransmission and local reconstruction," significantly reducing bandwidth pressure while achieving more efficient error repair.

[0106] More importantly, this process goes beyond data retransmission itself, forming a complete closed-loop process from semantic feature extraction, anomaly detection, classification and recognition, to retransmission scheduling and repair verification. After the receiving end retransmits data, embeds it into the current frame, and performs structural fusion, the system re-evaluates the semantic consistency index of the region. If the repair standard is still not met, the priority is automatically adjusted and the next round of compensation scheduling is triggered until the anomaly score drops to an acceptable threshold. This semantic feedback-driven control method enables the system to continuously adapt to covert interference in complex network environments and has local self-healing capabilities, effectively ensuring the continuity and stability of image content at the semantic level. In summary, this step is the most critical execution link in the entire intelligent video data transmission solution, carrying the implementation of the entire "discovery-diagnosis-response-repair" process. It not only improves video transmission quality but also provides strong technical support for highly robust video communication in future application scenarios such as multi-terminal collaboration and edge intelligent computing.

[0107] The method for transmitting new media video data based on digital technology proposed in this invention can effectively solve the problem of semantic structure distortion caused by weak disturbances in existing video transmission processes, achieving accurate identification, intelligent repair, and closed-loop response control of abnormal areas. Based on semantic feature information maps, this method constructs a complete processing chain from anomaly detection to semantic recovery and scheduling correction through multi-dimensional feature fusion and structural consistency comparison. It does not rely on traditional bit-level verification mechanisms, overcoming the limitations of existing transmission quality assurance methods in semantic layer perception. Compared with existing technologies, this invention significantly improves recognition granularity, response efficiency, and robustness. It can achieve rapid compensation and content restoration of local abnormal areas without increasing the overall data load, effectively improving the semantic stability, system fault tolerance, and user perception quality of video data in complex network environments. It is widely applicable to new media scenarios with high requirements for image semantic accuracy, such as intelligent recommendation, video review, and visual recognition, and has strong engineering application prospects and commercial value.

[0108] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0109] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0110] It should be noted that, in this document, the use of relational terms such as "first" and "second" is merely for distinguishing one entity or operation from another, and does not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0111] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0112] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0113] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0114] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0115] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0116] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0117] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for transmitting new media video data information based on digital technology, characterized in that, The method comprises the following steps: S100, semantic feature extraction is performed on the received video data frame, pixel-by-pixel modeling of image structure information is performed, and a semantic feature atlas is constructed; Step S100 comprises: Obtaining a current frame and converting it into a joint expression form of a gray image and a color image; Performing joint operation of Sobel operators and Laplace operators on pixels of three channels of red, green and blue respectively to extract color gradient distribution features; Texture direction arrangement features are extracted based on a direction covariance matrix of a local pixel region and a gray level co-occurrence matrix; A multi-threshold Canny edge detection algorithm is used in combination with a Hough transform to extract edge contour closure features; Color gradient distribution feature maps, texture direction arrangement feature maps and edge contour closure feature maps are fused to construct a semantic feature information atlas; S200, the semantic feature atlas of the current frame is matched with the semantic feature atlas of a neighboring reference frame, semantic offset amplitude indicators and structure disturbance change indicators are calculated, and a disturbance candidate region is marked; Step S200 comprises: Multi-dimensional feature vector normalization processing is performed to establish a semantic feature space under a unified scale; A semantic offset intensity atlas is constructed by calculating the Euclidean distance of matched pixels, and a local window aggregation mechanism is introduced; Structure disturbance change indicators of edge contour closure feature maps and texture direction arrangement feature maps are extracted; Collaborative anomalies of semantic offset values and structure disturbance values are detected based on a spatiotemporal sliding window to mark a disturbance candidate region; S300, a historical reference frame with the highest semantic stability is selected according to the disturbance candidate region, a semantic structure residual atlas is generated by a structure mapping inversion method, and an initial disturbance response intensity scale is constructed; Step S300 comprises: A historical reference frame with the highest semantic stability is selected, and image content at a corresponding position thereof is extracted; A structure mapping relationship is established and geometric alignment is completed by feature point matching and affine transformation or thin plate spline transformation; Embedding and boundary gradual fusion of the mapping image are performed to form a restored region that is continuous in vision and semantics; Semantic structure information residuals between the restored image and the current frame are calculated to construct an initial disturbance response intensity scale; S400, disturbance intensity parameters, disturbance distribution parameters and disturbance stability parameters are extracted based on the semantic structure residual atlas, an anomaly score result is generated, and an anomaly quantification representation is constructed; Step S400 comprises: The disturbance intensity parameters are calculated by jointly normalizing the semantic offset amplitude indicators and the structure disturbance change indicators; High-disturbance pixel points are spatially clustered to extract the disturbance distribution parameters; The disturbance stability parameters are extracted based on multi-frame residual trajectory analysis; Anomaly feature score results are generated by fusing the obtained parameters, and an anomaly quantification atlas that is bound to a spatial position is constructed; S500, the anomaly score result, the color gradient distribution features, the texture direction arrangement features, the semantic offset amplitude indicators and the structure disturbance change indicators are fused, feature clustering and classification are performed, and an anomaly type identification result is output; S600, according to the spatial position binding information in the anomaly type identification result, the anomaly score result, the semantic offset amplitude indicators and the structure disturbance change indicators, a transmission scheduling strategy is regulated.

2. The method for transmitting new media video data information based on digital technology according to claim 1, wherein, Step S500 comprises: A unified standardized multi-dimensional feature vector set is constructed; An unsupervised clustering algorithm is used to perform feature clustering on the abnormal region; Supervised classification processing is performed in combination with prior knowledge of the abnormal type to identify transmission interference type semantic structure distortion; The classification result is bound with the image space coordinates to output the abnormal type identification result.

3. The method for transmitting new media video data information based on digital technology according to claim 1, wherein, Step S600 includes: Target region position information marked as a transmission interference type abnormality is extracted and a mapping relationship is established; A transmission scheduling priority model is constructed by fusing the semantic offset amplitude index and the structure disturbance change index; Target region level video data retransmission processing is performed according to the scheduling priority; After retransmission is completed, the semantic offset amplitude index and the structure disturbance change index of the repaired region are recalculated to determine the repair effect.

Images