A logical evaluation method and system based on news shot switching

Abstract

The present invention discloses a logic evaluation method and system based on news shot switching, which relates to the field of video intelligent processing technology, including extracting video data from the original news video stream; detecting shot switching points in the video and analyzing to obtain shot switching rules; using a convolutional neural network to extract image features of shot switching, obtaining temporal dependencies through graph neural network modeling, and fusing to obtain switching features; using a fuzzy C-means clustering algorithm to evaluate the shot switching logic and calculate the shot switching logic score; and using a deep Q network to dynamically optimize the shot switching strategy based on the shot switching logic score and the switching features. The method of the present invention can dynamically adjust the shot switching strategy according to the user's real-time feedback by introducing a method combining deep learning and reinforcement learning, thereby improving the smoothness of shot switching and the content coherence. Flexibility, intelligence and real-time optimization are achieved in shot switching.

Description

Technical Field

[0001] The present invention relates to the technical field of video intelligent processing, and in particular to a logic evaluation method and system based on news shot switching. Background Art

[0002] In recent years, with the advancement of news video production technology, shot switching, as a crucial component of video editing, has gradually garnered widespread attention from both academia and industry. Traditional shot switching methods rely on simple rules or manual labeling, manually setting parameters such as shot duration, switching frequency, and intervals to achieve smooth video transitions. With the continuous advancement of machine learning and deep learning technologies, an increasing number of automated methods are being applied to shot switching in news videos. These technologies can help systems automatically learn the dependencies between shots, enabling more intelligent shot switching and ensuring automated and efficient video editing.

[0003] However, existing shot switching methods still face challenges, especially when dealing with shot switching rules in complex news videos. Traditional rule-based shot switching methods often rely on fixed duration and frequency settings, which are not well adapted to different types of content in news reports (such as breaking news, real-time reports, and in-depth analytical news). For example, real-time news reports may require more frequent shot switching and shorter durations, while news analysis programs require smoother shot switching and longer durations for each shot. In addition, most existing technologies cannot accurately understand the temporal dependencies between shots, resulting in adjustments to shot switching frequency and duration that cannot meet the actual needs of different news content. More importantly, these methods lack real-time dynamic optimization mechanisms and are unable to automatically adjust shot switching strategies based on audience feedback or changes in news content. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is that the existing shot switching method has the problems of inflexible switching rules, difficulty in adapting to changes in different news types, and lack of real-time feedback optimization.

[0006] To solve the above technical problems, the present invention provides the following technical solution: a logic evaluation method based on news shot switching, comprising:

[0007] Extracting video data from raw news video streams;

[0008] Detect the shot switching points in the video and analyze them to obtain the shot switching rules;

[0009] Use convolutional neural networks to extract image features of shot switching, model temporal dependencies through graph neural networks, and fuse them to obtain switching features;

[0010] Combining the switching features with the shot switching rules, the fuzzy C-means clustering algorithm is used to evaluate the shot switching logic and calculate the shot switching logic score.

[0011] The shot switching logic score is fed back in real time, and the shot switching strategy is dynamically optimized using a deep Q network according to the shot switching logic score and the switching characteristics.

[0012] As a preferred solution of the logic evaluation method based on news shot switching of the present invention, wherein: the shot switching point is the switching moment of the continuous shots;

[0013] The shot duration is the time difference between the start time of the current shot and the start time of the next shot. The shot switching frequency is the number of shot switches per unit time, which is calculated by combining each shot switching point in the video with the total duration.

[0014] By calculating the current frame With the previous frame Similarity , determine whether a lens switch occurs;

[0015] If the similarity measure If the value is lower than the set threshold, it is determined that a shot switch has occurred and the current time point is marked as a shot switch point.

[0016] As a preferred embodiment of the logic evaluation method based on news shot switching of the present invention, the shot switching rule includes obtaining the shot switching duration, frequency and sequence in the video according to the shot switching point analysis as the feature value of the news shot switching;

[0017] According to the news type, the three characteristic values ​​are adjusted respectively, and the adjusted three characteristic values ​​are used to represent the shot switching rules: for real-time reporting, the shot switching frequency is increased and the shot duration is shortened; for news analysis, the shot switching frequency is reduced and the shot duration is extended; when switching from news commentary shots to live on-site shots, the shot switching duration is extended.

[0018] As a preferred embodiment of the logic evaluation method based on news shot switching described in the present invention, the method of extracting image features of shots includes processing each frame of image using a convolutional neural network (CNN) to extract image features, texture, edge, and color distribution to describe the visual content of the shot; inputting each frame of image, extracting local features through a convolutional layer, and learning edge, color, and texture information in the image;

[0019] The downsampling operation is performed through the pooling layer, and the nonlinear features are enhanced through the ReLU activation function; finally, the feature map is output. , containing the image features of each frame.

[0020] As a preferred solution of the logic evaluation method based on news shot switching described in the present invention, wherein: the temporal dependency relationship is obtained by using a graph neural network GNN to model the dependency relationship between shots; the shot image features extracted by CNN are converted into , input into GNN as graph node features;

[0021] Define the nodes of the graph to represent shots, and the temporal dependencies between each shot switching point are used as edges in the graph. The order, duration, and frequency of shot switching are weighted and summed to obtain the result as the edge weight; construct the adjacency matrix of the graph Represents the dependency between shots; the elements of the adjacency matrix Represents a lens node and lens nodes The strength of the dependencies between

[0022] Graph neural networks propagate information through the edges of the graph, capture the temporal relationship of shot switching, and generate new temporal feature representations for each shot. , for each lens node , new temporal feature representation By using the time series features from neighboring nodes and the adjacency matrix Update by the weighted sum of It is The lens in Feature representation of the layer; It is The lens in Feature representation of the layer;

[0023] Combining the outputs of CNN and GNN, the image features and temporal features of each shot are fused together to obtain the shot switching feature representation as .

[0024] As a preferred solution of the logic evaluation method based on news shot switching of the present invention, the evaluation of shot switching logic includes constructing an objective function of fuzzy C-means clustering based on the shot switching duration, frequency and sequence constraints. and clustering parameters; the clustering parameters include the number of clusters , fuzzification factor and Category Center ; Evaluate the smoothness and coherence of shot switching through cluster membership;

[0025] Use genetic algorithm to optimize clustering parameters and set the objective function As a fitness function, multiple individuals are generated, each of which represents a set of clustering parameters; the combination of clustering parameters is optimized through the selection, crossover and mutation operations of the genetic algorithm;

[0026] Clustering is performed using parameters optimized by genetic algorithms to obtain the membership and category center of each shot switching; the characteristics of each shot switching are analyzed through cluster centers and membership. ,According to the clustering results, the shot switching is assigned a smoothness and coherence score, and the score is input into the reward function database.

[0027] As a preferred solution of the logic evaluation method based on news shot switching described in the present invention, the dynamic optimization shot switching strategy includes optimizing the shot switching strategy through the deep Q network DQN, dynamically adjusting the duration, frequency, and content coherence characteristics of the shot switching; defining a state space representation, state It consists of the image features and temporal features of each shot. , indicating the The state of the moment;

[0028] in, Represents the image features extracted by CNN;

[0029] Define action space, action is the decision of lens switching, indicating the The actions taken at all times include adjusting the shot duration and switching frequency; updating the Q value and using the deep Q network according to the state and actions Calculate the Q value and update the strategy through the reinforcement learning algorithm, represents the discount factor;

[0030] Define the reward function Feedback on shot switching strategies, Indicates the Rewards of the moment; It is obtained by the weighted summation of fluency rewards and content coherence rewards; through training iterations, the deep Q network will learn the optimal shot switching strategy and automatically adjust the smoothness and content coherence of shot switching.

[0031] As a preferred solution of the logic evaluation system based on news shot switching described in the present invention, it includes: a video data extraction module, an analysis module, a graph visual fusion module, and a strategy adjustment module;

[0032] The video data extraction module is used to extract images from the original news video stream, perform preprocessing, and input the processed images into the analysis module;

[0033] The analysis module is used to detect shot switching points and perform rule analysis based on extracted features;

[0034] The image vision fusion module is used to fuse the image features extracted by CNN with the temporal features extracted by GNN to obtain a comprehensive feature representation of each shot;

[0035] The strategy adjustment module is used to dynamically optimize the shot switching strategy based on the visual features and temporal features of the shot using a deep Q network (DQN).

[0036] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of a logic evaluation method based on news shot switching.

[0037] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a logic evaluation method based on news shot switching.

[0038] Beneficial Effects of the Invention: The logical evaluation method for news shot switching provided by this invention addresses these technical challenges by integrating deep learning and reinforcement learning. Specifically, the combination of convolutional neural networks and graph neural networks enables the system to simultaneously consider the visual content and temporal dependencies of shots, optimizing shot switching strategies. Furthermore, through the introduction of a deep Q-network, the system can dynamically adjust shot switching strategies based on real-time user feedback, improving shot switching fluidity and content coherence. This enables flexibility, intelligence, and real-time optimization during shot switching. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0040] Figure 1 This is an overall flow chart of a logic evaluation method based on news shot switching provided by the first embodiment of the present invention. DETAILED DESCRIPTION

[0041] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0042] Example 1, with reference to Figure 1 , as an embodiment of the present invention, provides a logic evaluation method based on news shot switching, comprising:

[0043] S1: Extract video data from the original news video stream; detect the shot switching points in the video and analyze them to obtain the shot switching rules.

[0044] Furthermore, shot cut detection relies on a similarity metric between adjacent frames. The similarity between the current frame and the previous frame is calculated to determine whether a shot cut has occurred. The similarity calculation formula uses the Structural Similarity Index (SSIM), which measures the similarity between two frames. The specific formula is:

[0045]

[0046] in: and Represent two frames of images respectively; and is the average brightness of the two frames; and is the brightness variance of the two frames; is the covariance of the two frames; and is a small constant added for stability. Usually chosen:

[0047]

[0048]

[0049] in and is a constant, is the dynamic range of the image pixels (e.g., for an 8-bit image, ). When the similarity When it is lower than the set threshold, it is determined that a shot switch has occurred and the current time point is marked as a shot switch point.

[0050] Based on the similarity between adjacent frames, we can accurately identify the occurrence of shot cuts, and further analyze the shot duration, cut frequency, and other related features. The duration of each shot is determined by the time difference between the cut time of the current shot and the cut time of the next shot.

[0051] The shot switching rules include analyzing the shot switching points to obtain the shot switching duration, frequency, and sequence in the video as the characteristic values ​​of news shot switching.

[0052] Indicates the The start time of the shot. Indicates the The start time of the shot. Indicates the The length of a shot. The formula for calculating shot length is:

[0053]

[0054] in, Indicates lens duration of and They are the start time of the current shot and the next shot respectively.

[0055] Shot Switching Frequency Calculation The shot switching frequency reflects the number of shot switching events per unit time. The frequency is determined by calculating the ratio of each shot switching point to the total duration of the video. Indicates the total number of camera cuts. Indicates the total length of the video. Indicates the lens switching frequency. The lens switching frequency calculation formula is:

[0056]

[0057] in, is the total number of lens switching times, is the total duration of the video, The frequency of camera switching.

[0058] The shot switching frequency reflects the number of shots switched per unit time. A high switching frequency may result in a fast pacing, while a low frequency may make the content appear bland and uninteresting. Adjust parameters such as shot switching smoothness and duration, select shot type, and adjust the switching rhythm.

[0059] Furthermore, according to the news type, the three characteristic values ​​are adjusted respectively, and the adjusted three characteristic values ​​are used to represent the shot switching rules. For real-time reporting, the shot switching frequency is increased and the shot duration is shortened; for news analysis, the shot switching frequency is reduced and the shot duration is extended; when switching from news commentary shots to live on-site shots, the shot switching duration is extended.

[0060] Based on the feature values, shot switching rules are defined. These rules are based on the duration, frequency, and content type of shot switching. The specific rules are as follows:

[0061] Rule 1: The minimum interval between lens switching When switching from a "long shot" to a "close shot", the time interval is Should not be less than Seconds. The time interval when switching from "close-up" to "close-up" Should not be less than Second.

[0062] Rule 2: Lens switching frequency. If the lens switching frequency per unit time is Greater than a threshold , then the shot duration It should be extended to prevent the camera from switching too frequently.

[0063] Rule 3: The association between shot switching and content logic.

[0064] In news reporting, shot switching should take into account the continuity of the content. For example, when switching from a reporter's explanation to a live shot of the scene, too fast a switch should be avoided.

[0065] Graph neural networks are used to model temporal dependencies and combine them with image features to determine the rationality of shot cuts. For example, the cut from a "news figure" to a "background" shot should be based on a specific temporal structure. Similar criteria are used to analyze shot coherence and the naturalness of cuts.

[0066] The rules are adjusted based on the characteristics of news reports for quantitative representation, and the rules are further dynamically adjusted according to the type of news.

[0067] Rule adjustment (camera switching frequency): For sports news, the camera switching frequency Increase, threshold It can be set to a higher value. For in-depth news analysis, the frequency of camera switching should be reduced and the threshold It can be set to a lower value (for example, 1 time / second). Quantization expression:

[0068]

[0069] Rule Adjustment (Shot Length): In sports coverage, shot length Can be shortened. In in-depth analysis news, the length of the shot Should be extended (e.g. set to 5 seconds). Quantization adjustment:

[0070]

[0071] Among them, Scale is the scaling factor corresponding to the news type (e.g., 0.8 for sports news and 1.2 for in-depth analysis news).

[0072] in, Indicates the news shot switching interval threshold, represents the fine switching interval threshold, Indicates the maximum frequency of shot switching allowed per unit time; fluency is evaluated by shot switching duration and frequency, and coherence is evaluated by shot switching sequence and similarity index.

[0073] It should be noted that the camera switching rules are dynamically adjusted based on the type of news report (e.g., sports coverage, breaking news, or in-depth news analysis). For sports coverage, the camera switching frequency is increased and the shot duration is appropriately shortened. For in-depth news analysis, the camera switching frequency is reduced and the shot duration is extended to make the content more immersive and coherent. This process is accomplished by adjusting feature values ​​in a deep learning model. Combined with a feedback mechanism, the system can automatically optimize according to different news contexts.

[0074] S2: Use a convolutional neural network to extract the image features of shot switching, obtain temporal dependencies through graph neural network modeling, and fuse them to obtain switching features.

[0075] Furthermore, to improve the understanding and optimization of shot switching, we introduce a graph-based visual fusion method, combining a convolutional neural network (CNN) to extract image features and a graph neural network (GNN) to model the temporal dependencies between shots. CNN extracts image features and inputs each frame. It outputs the deep features of each frame. CNN extracts low-level and high-level features of each frame through multi-layer convolution and pooling operations, including texture, edge and color information. Input the image features of each shot And switching timing relationship. Output timing dependency between shots represents .

[0076] Extracting the image features of the lens includes using a convolutional neural network (CNN) to process each frame of the image, extracting image features, texture, edges, and color distribution to describe the visual content of the lens; inputting each frame of the image, extracting local features through the convolution layer, and learning the edge, color, and texture information in the image.

[0077] The downsampling operation is performed through the pooling layer, and the nonlinear features are enhanced through the ReLU activation function; finally, the feature map is output. , containing the image features of each frame.

[0078] Combining graph neural networks (GNNs) with convolutional neural networks (CNNs), we introduced a graph-vision fusion approach. This approach combines the CNN to extract visual features of shots with the GNN to model the temporal dependencies and graph structure between shots. The goal is to comprehensively improve the ability to understand shot cuts, not just relying on single image features but also considering the temporal relationships and overall structure of shot cuts.

[0079] A convolutional neural network (CNN) is used to process each image frame and extract low-level and high-level features of the image. These features include texture, edges, color distribution, etc., which are mainly used to describe the visual content of the shot.

[0080] Each frame of an image is input. The convolutional layer extracts local features and learns information such as edges, colors, and textures. The pooling layer downsamples the feature map, reducing its size while retaining the most important features. The ReLU activation function enhances nonlinear features.

[0081] CNN output: the final output feature map Contains the visual characteristics of each frame.

[0082] Obtaining temporal dependencies involves using graph neural networks (GNNs) to model the dependencies between shots; extracting shot image features from CNNs , which is input into GNN as graph node features.

[0083] Define the nodes of the graph to represent shots, and the temporal dependencies between each shot switching point are used as edges in the graph. The order, duration, and frequency of shot switching are weighted and summed to obtain the result as the edge weight; construct the adjacency matrix of the graph Represents the dependency between shots; the elements of the adjacency matrix Represents a lens node and lens nodes The strength of the dependency relationship between them.

[0084] Graph neural networks propagate information through the edges of the graph, capture the temporal relationship of shot switching, and generate new temporal feature representations for each shot. , for each lens node , new temporal feature representation By using the time series characteristics of neighboring nodes and the adjacency matrix The weighted sum of is updated. Indicates the The lens in Feature representation of the layer; is the temporal feature extracted by GNN, expressed in The lens in Feature representation of the layer.

[0085] GNN input: image features of each shot (Extracted by CNN) as node features are input into GNN. The temporal dependencies between each shot switching point are used as edges in the graph. The order, duration and frequency of shot switching will be used as edge weights. Adjacency Matrix: Construct the adjacency matrix of the graph , represents the dependency between shots. The information between shots is propagated through the edges of the graph to obtain the temporal relationship between shots. The calculation formula of the graph neural network is:

[0086]

[0087] in: It is The lens in Feature representation of the layer. Indicates lens ’s adjacent nodes (i.e., the shots related to it). It is The weight matrix of the layer, is the bias term. It is an activation function (such as ReLU) used to enhance the nonlinear modeling ability of the network.

[0088] Combining the outputs of CNN and GNN, the image features and temporal features of each shot are fused together to obtain the shot switching feature representation as .

[0089] Through information propagation, graph neural networks can capture the temporal relationship of shot switching and generate new temporal feature representations for each shot. , which contains the context information of the shot switching.

[0090] It should be noted that CNNs are used to extract visual features of shots, such as texture, color, and edges, providing rich visual information for each shot. Graph Neural Networks (GNNs) model temporal relationships: By modeling the temporal dependencies between shots through GNNs, the contextual relationships and temporal structure of shot cuts can be captured, ensuring the rationality of shot cuts. The image features extracted by CNNs are combined with the temporal features learned by GNNs to generate a comprehensive feature representation for each shot. This not only understands the visual features of the shots, but also the temporal dependencies of shot cuts, making cut decisions more intelligent. Graph-visual fusion technology improves the system's perception capabilities, accurately grasping the temporal and content relationships of shot cuts, and avoiding inappropriate cuts.

[0091] S3: Combining the switching features with the shot switching rules, the fuzzy C-means clustering algorithm is used to evaluate the shot switching logic and calculate the shot switching logic score.

[0092] Furthermore, the fuzzy C-means clustering model is defined. The fuzzy C-means clustering FCM algorithm performs clustering by minimizing the objective function. In this step, the core objective function of FCM is defined and the parameters that need to be optimized are determined. The objective function of the FCM algorithm is expressed as:

[0093]

[0094] in: is a data point In category The degree of membership indicates the degree to which the data point belongs to the category. is a category center. It is the fuzzification factor, which controls the fuzziness of the membership. It is the number of cluster categories, indicating the number of categories that need to be set in FCM. is the total number of data points, is the data point, the fuzzification factor , controls the fuzziness of membership, usually takes the value .

[0095] Constructing the objective function of fuzzy C-means clustering based on the constraints of shot switching duration, frequency and sequence and clustering parameters; the clustering parameters include the number of clusters , fuzzification factor and Category Center ; Smoothness is evaluated by the duration and frequency of shot switching, and coherence is evaluated by the shot switching order and similarity index; the smoothness and coherence of shot switching are evaluated by the cluster membership.

[0096] Genetic Algorithm (GA) Optimization of Fuzzy C-means Clustering Genetic algorithm can be used to optimize the parameters in FCM, such as the number of clusters C, the fuzzification factor m and the initialization of the category center .

[0097] Define the individuals and populations of the genetic algorithm. Each individual represents a set of FCM parameters. Specifically, the individual includes: the number of clusters C, the fuzzification factor m, and the initial value of the cluster center (the center point of each category). .

[0098] Initialize the population and randomly generate a population of a certain number of individuals, each of which contains a set of possible parameter settings. The goal of the genetic algorithm to design the fitness function is to optimize the objective function of FCM through selection, crossover and mutation operations to make the clustering results more accurate. Therefore, the fitness function will be designed based on the clustering results and objective function of FCM. As a fitness function: the lower the fitness value, the better the clustering effect of the current individual. Fitness function:

[0099]

[0100] in, is the objective function value of FCM. The goal is to minimize , so take its reciprocal as the fitness value.

[0101] The fuzzy C-means clustering algorithm performs fuzzy C-means clustering using the parameters optimized by the genetic algorithm. The steps of the clustering process are as follows:

[0102] Initialize the membership matrix U and category centers Initialization stage: The membership matrix U is an N×C matrix, where each element Represents data points The degree of membership in class j. The initial membership value is usually random.

[0103] Update the membership matrix Update the membership matrix according to the current category center. Calculated by the following formula:

[0104]

[0105] This formula updates the membership based on the distance between each data point and each category center. The smaller the distance, the greater the membership. Update category center Based on the updated membership matrix U, update the center of each category :

[0106]

[0107] The class center is the weighted average of all data points in each class. Repeatedly update the membership matrix and class center until the objective function Convergence, or reaching the maximum number of iterations.

[0108] Clustering is performed using parameters optimized by genetic algorithms to obtain the membership and category center of each shot switching; the characteristics of each shot switching are analyzed through cluster centers and membership. ,According to the clustering results, the shot switching is assigned a smoothness and coherence score, and the score is input into the reward function database.

[0109] It should be noted that fuzzy clustering provides a "soft" rating that better reflects the audience's subjective experience than simple right / wrong judgments. DQN online learning can dynamically adjust based on actual results, automatically adapting to different news types and emergency scenarios. Overall, this approach improves viewing fluency and information reception efficiency, enabling intelligent, adaptive optimization of news shot switching.

[0110] S4: Feedback the shot switching logic score in real time, and dynamically optimize the shot switching strategy using a deep Q network based on the shot switching logic score and the switching features.

[0111] Furthermore, by combining the outputs of CNN and GNN, the image features and temporal features of each shot are fused together to obtain the final shot feature representation. ,This representation not only contains the visual content of the image, but also captures the temporal dependencies and overall structural relationships between shots.

[0112] Dynamically adjust shot switching strategies based on real-time user feedback to ensure that shot switching meets audience expectations and enhances the viewing experience of the program.

[0113] During video playback, users, whether news editors or viewers, can provide real-time feedback through a graphical interface. Users can adjust the speed or duration of shot transitions, select shot types, adjust the tempo of shot transitions, and provide ratings on the smoothness and frequency of shot transitions.

[0114] Reinforcement learning optimizes the camera switching strategy through the deep Q network (DQN) to optimize the camera switching strategy and dynamically adjust the duration, frequency, content coherence and other characteristics of the camera switching. State representation: State It consists of the image features and temporal features of each shot, specifically:

[0115]

[0116] in, is the image feature extracted by CNN, It is the temporal feature extracted by GNN. Action representation: action Represents the decision of shot switching, including adjusting shot duration, switching frequency, etc. Q value update: Use deep Q network to update the state and actions Calculate the Q value and update the strategy through the reinforcement learning algorithm:

[0117]

[0118] in, It is a reward calculated based on user feedback. is the discount factor.

[0119] Reward Function Design Reward Function Used to provide feedback on the quality of shot switching and help the model learn the optimal strategy. Fluency reward: High rewards are given when the shot switching is natural, and low rewards are given when it is not. Content coherence reward: High rewards are given when the shot switching is consistent with the logic of the news content, and low rewards are given when the switching is inappropriate. Reward function design:

[0120]

[0121] in, Indicates the weight coefficient of fluency reward, Represents the weight coefficient of the content coherence reward, which is used to balance the impact of content coherence on the total reward. Represents the smoothness reward function, reflecting the smoothness of the current camera switch. When switching cameras, smoother and more natural switches will receive higher rewards, while abrupt or overly fast switches will receive lower rewards. Represents the content coherence reward function, which evaluates whether the shot switching conforms to the content logic and news structure.

[0122] If the shot switching conforms to the logical order of the content (such as a natural transition from a news reporter to a background shot), a higher reward will be given. If the switching is illogical (such as jumping directly from a background shot to an unrelated shot), a lower reward will be given. Indicates at time The action taken, i.e. the current shot switching decision. This action may include adjusting shot duration, switching frequency, or selecting a new shot type.

[0123] Through continuous training, the Deep Q Network will learn the optimal shot switching strategy and automatically adjust the duration, frequency and content consistency of shot switching, making the shot switching smoother, more natural and in line with audience needs.

[0124] It should be noted that the system has effectively enhanced its logical evaluation and optimization capabilities for news shot switching. These improvements enable the system to not only automatically extract and analyze video data but also intelligently adjust shot switching rules based on different news types. Deep learning techniques such as CNN and GNN are combined to optimize shot feature extraction and temporal modeling. Furthermore, real-time feedback mechanisms and reinforcement learning (DQN) optimize shot switching strategies. Ultimately, this system delivers smoother, more coherent, and rhythmic shot switching, enhancing the viewing experience.

[0125] Example 2 is an embodiment of the present invention, which provides a logic evaluation method based on news shot switching. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0126] First, the goal of this example is to verify the effectiveness of a logical evaluation method based on news shot switching, specifically how to extract visual features of shots using a convolutional neural network (CNN), model the temporal dependencies between shots using a graph neural network (GNN), and optimize shot switching strategies using a deep Q-network (DQN). The experiment will cover the following aspects: extracting data from raw news video streams, detecting shot switching points, analyzing and deriving shot switching rules, making rule-based judgments, and adjusting these rules based on the characteristics of news reports.

[0127] At the beginning of the experiment, each frame of image data was extracted from the news video and preprocessed. Each frame was fed into a convolutional neural network (CNN) to extract visual features such as edges, texture, and color distribution. Through pooling layers and activation functions, the CNN extracted high-level features of each frame, which served as input for subsequent time series modeling.

[0128] Next, we use a graph neural network (GNN) to model the temporal dependencies between shots. The image features of each shot serve as node features in the GNN, and the temporal dependencies between shot transitions are represented by constructing a graph adjacency matrix. The GNN learns the contextual dependencies between shots through information propagation.

[0129] Shot switching rules are generated based on the detected shot switching points. The duration of each shot is calculated by the time difference between the current shot switching point and the next shot switching point. The shot switching frequency is evaluated by counting the number of shot switching times per unit time.

[0130] Adjust the frequency and duration of shot cuts for different news types (such as sports news, in-depth news analysis, and real-time news). For example, real-time news coverage may require increased shot cut frequency and shorter shot duration, while in-depth news analysis requires reduced shot cut frequency and longer shot duration. Ultimately, the shot cut strategy is optimized using a Deep Q-Network (DQN). Based on real-time user feedback, parameters such as shot duration and cut frequency are dynamically adjusted to improve shot cut quality.

[0131] A series of news video clips will be used for verification in the experiment. The goal is to test whether the proposed method can effectively optimize shot switching and improve the fluency and logic of news programs under different news types.

[0132] Using the proposed logical evaluation method based on news shot switching, experimental videos significantly optimized their shot switching strategies across different news coverage types. In sports news, due to its dynamic nature, shot switching is relatively frequent and short (average switching time is 2.5 seconds). For example, the "long shot to close shot" switch in Experimental Video 1 lasted 2.5 seconds, with a switching frequency of 3.5 times per second. The user fluency rating was 9, and the content coherence rating was 8, indicating that the video's shot switching performed well in terms of both rhythm and smoothness.

[0133] In contrast, in-depth news analysis programs require fewer shot cuts, and each shot is longer (for example, the "close-up" to "close-up" cut in Experimental Video 2 takes 3.5 seconds). This allows viewers to better focus on the content without being interrupted by frequent shot cuts. Experimental Video 2 has a lower switching frequency (1.2 times / second) and a high user fluency score (9), demonstrating that the system can dynamically adjust the shot cut rules based on the characteristics of the news report to adapt to different rhythm requirements.

[0134] Real-time news reports, such as those in Experiments 3 and 6, typically require a high frequency of camera cuts to highlight the dynamics of the news event. In these experiments, the camera cut frequency was generally higher than 3 times per second, and the cut duration was short (for example, the transition from the "foreground to the background" in Experiment 3 took 2 seconds). These settings ensured the news' compactness and timeliness, and received positive ratings in user feedback.

[0135] Example 3 is an embodiment of the present invention, which provides a logic evaluation system based on news shot switching, including a video data extraction module, an analysis module, a graph visual fusion module, and a strategy adjustment module.

[0136] The video data extraction module is used to extract images from the original news video stream, perform preprocessing, and input the processed images into the analysis module.

[0137] The analysis module is used to detect shot switching points and perform rule analysis based on extracted features;

[0138] The graph vision fusion module is used to fuse the image features extracted by CNN with the temporal features extracted by GNN to obtain a comprehensive feature representation of each shot.

[0139] The strategy adjustment module is used to dynamically optimize the shot switching strategy based on the visual features and temporal features of the shot using a deep Q network (DQN).

[0140] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the method of the present invention. The aforementioned storage medium includes various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0141] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0142] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting, or processing it in another suitable manner as necessary, and then storing it in a computer memory.

[0143] It should be understood that various aspects of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, it may be implemented using a combination of any of the following technologies known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc. It should be noted that the above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to be limiting. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art will understand that modifications or equivalent substitutions may be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and such modifications are intended to be encompassed by the claims of the present invention.

Claims

1. A logic evaluation method based on news shot switching, characterized in that: include: Extracting video data from raw news video streams; Detect the shot switching points in the video and analyze them to obtain the shot switching rules; Use convolutional neural networks to extract image features of shot switching, model temporal dependencies through graph neural networks, and fuse them to obtain switching features; Combining the switching features with the shot switching rules, the fuzzy C-means clustering algorithm is used to evaluate the shot switching logic and calculate the shot switching logic score. The shot switching logic score is fed back in real time, and the shot switching strategy is dynamically optimized using a deep Q network according to the shot switching logic score and the switching characteristics.

2. The logic evaluation method based on news shot switching according to claim 1, wherein: The shot switching point is the switching moment of the continuous shots; The shot duration is the time difference between the start time of the current shot and the start time of the next shot; The shot switching frequency is the number of shots switched per unit time. It is calculated by combining each shot switching point and the total duration of the video. By calculating the current frame With the previous frame Similarity , determine whether a lens switch occurs; If the similarity measure If the value is lower than the set threshold, it is determined that a shot switch has occurred and the current time point is marked as a shot switch point.

3. The logic evaluation method based on news shot switching according to claim 2, characterized in that: The shot switching rules include analyzing the shot switching points to obtain the shot switching duration, frequency and sequence in the video as feature values ​​of the news shot switching; According to the news type, the three characteristic values ​​are adjusted respectively, and the adjusted three characteristic values ​​are used to represent the shot switching rules: for real-time reporting, the shot switching frequency is increased and the shot duration is shortened; for news analysis, the shot switching frequency is reduced and the shot duration is extended; when switching from news commentary shots to live on-site shots, the shot switching duration is extended.

4. The logic evaluation method based on news shot switching according to claim 3, wherein: Extracting image features of a shot includes processing each frame of image using a convolutional neural network (CNN) to extract image features, including texture, edges, and color distribution, for describing the visual content of the shot; inputting each frame of image, extracting local features through a convolutional layer, and learning edge, color, and texture information in the image; The downsampling operation is performed through the pooling layer, and the nonlinear features are enhanced through the ReLU activation function; finally, the feature map is output. , containing the image features of each frame.

5. The logic evaluation method based on news shot switching according to claim 4, characterized in that: The temporal dependency relationship is obtained by using a graph neural network (GNN) to model the dependency relationship between shots; and the shot image features extracted by CNN are transformed into the shot image features. , input into GNN as graph node features; Define the nodes of the graph to represent shots, and the temporal dependencies between each shot switching point are used as edges in the graph. The order, duration, and frequency of shot switching are weighted and summed to obtain the result as the edge weight; construct the adjacency matrix of the graph Represents the dependency between shots; the elements of the adjacency matrix Represents a lens node and lens nodes The strength of the dependencies between Graph neural networks propagate information through the edges of the graph, capture the temporal relationship of shot switching, and generate new temporal feature representations for each shot. , for each lens node , new temporal feature representation By using the time series features from neighboring nodes and the adjacency matrix Update by the weighted sum of It is The lens in Feature representation of the layer; It is The lens in Feature representation of the layer; Combining the outputs of CNN and GNN, the image features and temporal features of each shot are fused together to obtain the shot switching feature representation as .

6. The logic evaluation method based on news shot switching according to claim 5, characterized in that: The evaluation of the shot switching logic includes constructing an objective function of fuzzy C-means clustering based on the shot switching duration, frequency and sequence. and clustering parameters; the clustering parameters include the number of clusters , fuzzification factor and Category Center The fluency of shot switching is evaluated by the duration and frequency of shot switching, and the coherence is evaluated by the shot switching sequence and similarity index; the fluency and coherence of shot switching are evaluated by the cluster membership; Use genetic algorithm to optimize clustering parameters and set the objective function As a fitness function, multiple individuals are generated, each of which represents a set of clustering parameters; the combination of clustering parameters is optimized through the selection, crossover and mutation operations of the genetic algorithm; Clustering is performed using parameters optimized by genetic algorithms to obtain the membership and category center of each shot switching; the characteristics of each shot switching are analyzed through cluster centers and membership. ,According to the clustering results, the shot switching is assigned a smoothness and coherence score, and the score is input into the reward function database.

7. The logic evaluation method based on news shot switching according to claim 6, characterized in that: The dynamic optimization of the shot switching strategy includes optimizing the shot switching strategy through a deep Q-network (DQN) to dynamically adjust the duration, frequency, and content coherence characteristics of the shot switching; Define the state space representation, state It consists of the image features and temporal features of each shot. , indicating the The state of the moment; in, Represents the image features extracted by CNN; Define action space, action is the decision of lens switching, indicating the The actions taken at all times include adjusting the shot duration and switching frequency; updating the Q value and using the deep Q network according to the state and actions Calculate the Q value and update the strategy through reinforcement learning algorithm; Define the reward function Feedback on shot switching strategies, Indicates the Rewards of the moment; It is obtained by the weighted summation of fluency rewards and content coherence rewards; through training iterations, the deep Q network will learn the optimal shot switching strategy and automatically adjust the smoothness and content coherence of shot switching.

8. A system using the logic evaluation method based on news shot switching according to any one of claims 1 to 7, characterized in that: It includes video data extraction module, analysis module, image and visual fusion module, and strategy adjustment module; The video data extraction module is used to extract images from the original news video stream, perform preprocessing, and input the processed images into the analysis module; The analysis module is used to detect shot switching points and perform rule analysis based on extracted features; The image vision fusion module is used to fuse the image features extracted by CNN with the temporal features extracted by GNN to obtain a comprehensive feature representation of each shot; The strategy adjustment module is used to dynamically optimize the shot switching strategy based on the visual features and temporal features of the shot using a deep Q network (DQN).

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the logic evaluation method based on news shot switching according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the logic evaluation method based on news shot switching according to any one of claims 1 to 7 are implemented.

Images