A method and system for dynamically acquiring historical viewport coordinates and predicting future viewports

Abstract

The present invention relates to the field of intelligent trajectory prediction technology, and discloses a method and system for dynamically acquiring viewport historical coordinates to predict future viewports. The method comprises: collecting viewport trajectory data of watching immersive panoramic videos in real time through VR equipment, preprocessing the viewport trajectory data to construct a data set, and splitting the data set into a training set, a verification set, and a test set; constructing a viewport prediction model based on an implicit ensemble learning algorithm, and using the training set to train and optimize the viewport prediction model; inputting a verification set to verify the trained and optimized viewport prediction model, and inputting the test set into the viewport prediction model for evaluation, and deploying and replacing the viewport prediction model according to the evaluation result; and using the optimized and evaluated viewport prediction model to predict the future viewport of the viewport of the immersive panoramic video stream. The trajectory data of watching immersive videos is collected in real time, the problem of diversity of viewing modes is solved, and the generalization ability of the viewport prediction model is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent trajectory prediction, and in particular to a method and system for dynamically acquiring historical viewport coordinates to predict future viewports. Background Art

[0002] In the current immersive panoramic video streaming experience, viewport prediction that meets the diversity of users' viewing patterns and quality of experience (QoE) preferences is a complex problem. Existing neural adaptive methods have certain limitations in handling the diversity of users' viewing patterns and quality of experience (QoE) preferences. When users' actual viewing patterns and QoE preferences are very different from the data patterns observed during the training phase, the performance of these methods may drop significantly, resulting in poor generalization.

[0003] Specifically, some users may tend to focus on a specific area in the video. For example, in a panoramic natural scenery video, they always focus on the small area in the picture that shows unique landforms or rare plants and animals. Some users are accustomed to frequently switching perspectives, quickly browsing every corner of the panoramic picture, not missing any details, as if they want to take in the whole immersive environment. Some users may have a staged viewing mode, first browsing roughly, then selecting a key area for in-depth viewing, etc. Different viewing modes mean that there are completely different requirements for the viewport position, range, and switching frequency. In order to accurately meet these diverse needs, viewport prediction needs to have strong adaptability and accuracy.

[0004] Some users pay more attention to the clarity of the video. They hope that within the viewport they are concerned about, the image can always maintain high resolution and accurate color reproduction, even if this may require more network bandwidth or equipment resources to ensure; some users pay more attention to smoothness. Even if the picture quality is slightly compromised, there should be no lag or frame drop, otherwise it will seriously affect their immersive experience; some users pursue a balance, wanting good picture quality while also requiring the playback process to be smooth enough.

[0005] At the same time, existing neural adaptive methods have obvious limitations when dealing with these complex situations. Once the user's actual viewing mode and experience quality preference differ greatly from the data pattern observed in the training phase, the problem becomes prominent. For example, most users in the training data are mainly in a relatively stable and concentrated viewing mode, while in actual applications, there are a large number of users who frequently switch viewing angles. At this time, it is difficult for existing neural adaptive methods to accurately predict the viewport position and range of these users at the next moment. Or during training, the balance of picture clarity is mainly considered for experience quality preference, while actual users care more about smoothness. In this case, when performing viewport-related processing such as resolution allocation and buffer strategy formulation, there is no way to meet the user's real needs.

[0006] In this case, the performance of these methods will be significantly reduced, which is manifested in that the predicted viewport deviates greatly from the area that the user actually pays attention to, resulting in the user seeing the picture that is either not the content they want to pay attention to, or the quality does not meet their expected standards. The overall generalization effect is very poor, that is, it is difficult to apply the capabilities learned based on the training data to a variety of actual user viewing scenarios, which greatly limits the high-quality experience that immersive panoramic video streams bring to users. Summary of the invention

[0007] The purpose of the present invention is to solve the shortcomings of the above-mentioned prior art and to provide a method and system for dynamically obtaining historical viewport coordinates to predict future viewports, which predicts the viewport of an immersive panoramic video stream in a flexible and real-time manner, ensuring user diversity to improve generalization.

[0008] On the one hand, a method for dynamically obtaining historical viewport coordinates to predict future viewports is provided, comprising the following steps: S1: collecting viewport trajectory data of watching immersive panoramic videos in real time through VR devices, preprocessing the viewport trajectory data to construct a data set, and splitting the data set into a training set, a validation set, and a test set; S2: constructing a viewport prediction model based on an implicit ensemble learning algorithm, and using the training set to train and optimize the viewport prediction model; S3: inputting a validation set to verify the trained and optimized viewport prediction model, and inputting the test set to the viewport prediction model for evaluation, and deploying and replacing the viewport prediction model according to the evaluation result; S4: Use the optimized and evaluated viewport prediction model to predict the future viewport of the immersive panoramic video stream.

[0009] Further, in step S1, collecting viewport trajectory data of watching immersive panoramic video in real time through VR equipment includes: Create a file path and file for recording trajectory data, and record the collected current trajectory coordinates and the predicted trajectory coordinates output by the viewport prediction model separately in two files; Use VR head-mounted display device to collect user's viewport trajectory data at a set frequency and store it in the file of current trajectory coordinates, including the 3D rectangular coordinates of the user's visual focus point .

[0010] Furthermore, in step S1, preprocessing the viewport trajectory data to construct a data set further includes: S11: Convert the 3D rectangular coordinates into longitude and latitude coordinates in a spherical coordinate system , dimension and longitude The calculation formula is as follows: , , in, For a special inverse tangent function, consider and to determine the correct quadrant; S12: The longitude and latitude coordinates Normalization is performed to convert the angle into a value in the range of [-1,1], and the dimension and precision are normalized respectively, as shown below: , .

[0011] Further, in step S2, constructing a viewport prediction model based on an implicit ensemble learning algorithm includes: Designing the viewport prediction model, including an input layer, an implicit integration layer, a fusion layer, and an output layer; Receiving viewport trajectory data of multiple users through the input layer, including historical viewport coordinates and current viewport coordinates; The implicit integration layer is used to predict the viewport trajectory from different angles and using different features. The implicit integration layer includes multiple sub-models, and the sub-models are fused using a fusion policy of an attention mechanism. The final viewport prediction result is output through the output layer.

[0012] Further, in step S2, training the viewport prediction model using the training set includes: Define the loss function of the viewport prediction model , used to calculate the intersection-over-union ratio of the high-definition area of ​​the predicted viewpoint coordinates and the high-definition area of ​​the actual viewpoint coordinates, loss function The formula is as follows: in, represents the high-definition region of the predicted viewpoint coordinates, The high-definition area representing the actual viewpoint coordinates, Indicates the intersection-over-combination ratio of the high-definition area of ​​the predicted viewpoint coordinates and the high-definition area of ​​the actual viewpoint coordinates During the training process, by increasing the intersection-and-union ratio To reduce the size of the loss function, if the predicted area and the real area completely overlap, the intersection ratio is 1, then the loss function A value of 0 indicates that the prediction is accurate.

[0013] Further, in step S3, inputting the test set into the trained viewport prediction model for evaluation includes: S31: inputting the test set into the trained viewport prediction model for prediction evaluation, and outputting coordinate points of trajectory prediction; S32: converting the coordinate points of the trajectory prediction into viewport state information of the immersive panoramic video through a real-time data analysis algorithm; S32: Mapping the viewpoints into high-definition areas of the immersive video according to the designed viewpoint slice distribution; S33: Design performance evaluation indicators to comprehensively evaluate the performance of the model and obtain model evaluation results; S34: Compare the evaluation results of the new model with the original model. If there is an improvement in key performance indicators, replace the original model.

[0014] Preferably, the performance of the comprehensive evaluation model of the design performance evaluation index further includes: The accuracy and precision standards are used to measure the proportion of samples predicted correctly by the viewport prediction model to the total samples and the proportion of samples that are actually positive. By mean absolute error The mean square error is used to measure the absolute difference between the predicted value and the actual value. and RMS error The average of the squares of the differences between the predicted values ​​and the actual values ​​is measured to quantify the magnitude of the prediction error. The formula is as follows: in, is the actual viewpoint coordinate, To predict the viewpoint coordinates, is the sample size.

[0015] On the other hand, a system for dynamically obtaining historical viewport coordinates and predicting future viewports is provided, including: The acquisition and processing module is used to acquire the viewport trajectory data of watching immersive panoramic videos in real time through VR devices, pre-process the viewport trajectory data to construct a data set, and split the data set into a training set, a validation set, and a test set; A model building and training module, used to build a viewport prediction model based on an implicit ensemble learning algorithm, and to train and optimize the viewport prediction model using the training set; An optimization and evaluation module, used for inputting a verification set to verify the trained and optimized viewport prediction model, inputting the test set to the viewport prediction model for evaluation, and deploying and replacing the viewport prediction model according to the evaluation result; The application and prediction module is used to predict the future viewport of the immersive panoramic video stream using the optimized and evaluated viewport prediction model.

[0016] In addition, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the method for dynamically obtaining historical viewport coordinates and predicting future viewports as described in any one of the above is implemented.

[0017] At the same time, an electronic device is provided, including: one or more processors; a storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement any of the above-mentioned methods of dynamically obtaining viewport historical coordinates and predicting future viewports.

[0018] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a method for predicting future viewports by dynamically acquiring historical viewport coordinates of an immersive video stream, including introducing an online real-time data acquisition module to collect head motion trajectory data of a user watching an immersive video in real time; The immersive video streaming system based on slicing of the present invention aims to solve the problem of diversity in user viewing patterns, thereby significantly improving the generalization ability of the viewport prediction model; The present invention effectively reduces the prediction bias by designing an efficient viewport prediction model based on implicit ensemble learning to meet the needs of users with different browsing modes, and the calculation cost is low. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 A flow chart of a method for dynamically acquiring historical viewport coordinates and predicting future viewports according to the present invention; Figure 2 A schematic diagram of data flow in a method for dynamically acquiring historical viewport coordinates to predict future viewports according to the present invention; DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0021] The present invention can continuously collect the trajectory data of users watching immersive panoramic video streams through the intelligent viewport prediction algorithm service background. The specific implementation of the present invention is described below in conjunction with the accompanying drawings and embodiments.

[0022] Example 1 like Figure 1 As shown, this embodiment provides a method for dynamically obtaining historical viewport coordinates to predict future viewports, and the technical solution includes the following steps: S1: collecting viewport trajectory data of watching immersive panoramic videos in real time through VR devices, preprocessing the viewport trajectory data to construct a data set, and splitting the data set into a training set, a validation set, and a test set; S2: constructing a viewport prediction model based on an implicit ensemble learning algorithm, and using the training set to train and optimize the viewport prediction model; S3: inputting a validation set to verify the trained and optimized viewport prediction model, and inputting the test set to the viewport prediction model for evaluation, and deploying and replacing the viewport prediction model according to the evaluation result; S4: Use the optimized and evaluated viewport prediction model to predict the future viewport of the immersive panoramic video stream.

[0023] Among them, in step S1, collecting the viewport trajectory data of watching the immersive panoramic video in real time through the VR device includes: Create a file path and file for recording trajectory data, and record the collected current trajectory coordinates and the predicted trajectory coordinates output by the viewport prediction model separately in two files; Use a VR headset to collect the user's viewport trajectory data at a set frequency, such as 10FPS, and store it in a file of current trajectory coordinates, including the 3D rectangular coordinates of the user's visual focus point. .

[0024] Specifically, in this embodiment, the data acquisition module is designed to include two parts, one is the get_logger function for recording trajectory data, and the other is the timeTask function for cleaning abnormal data in a scheduled task. The get_logger function for recording trajectory data creates the path and file for recording trajectory data through the device-uuid and start-time fields, and records the current trajectory coordinates and the predicted trajectory coordinates separately in two files. The current trajectory coordinates are the data collected, and the predicted trajectory coordinates are the trajectory data predicted by the model, which can be used to analyze the prediction effect of the current online model in more detail.

[0025] Then, the viewport trajectory data is preprocessed to construct a data set including: S11: Convert the 3D rectangular coordinates into longitude and latitude coordinates in a spherical coordinate system , dimension and longitude The calculation formula is as follows: , , in, For a special inverse tangent function, consider and to determine the correct quadrant; S12: The longitude and latitude coordinates Normalization is performed to convert the angle into a value in the range of [-1,1], and the dimension and precision are normalized respectively, as shown below: , .

[0026] Next, the step S2 of constructing a viewport prediction model based on an implicit ensemble learning algorithm specifically includes: Designing the viewport prediction model, including an input layer, an implicit integration layer, a fusion layer, and an output layer; Receiving viewport trajectory data of multiple users through the input layer, including historical viewport coordinates and current viewport coordinates; The implicit integration layer is used to predict the viewport trajectory from different angles and using different features. The implicit integration layer includes multiple sub-models, and the sub-models are fused using a fusion policy of an attention mechanism. The final viewport prediction result is output through the output layer.

[0027] Specifically, in this embodiment, the viewport prediction model is based on the efficient MTIO-Transformer, and the specific form is: future = model.sample(history, current) Through the historical coordinates history and the current coordinates current, algorithm reasoning is performed to obtain the predicted future trajectory coordinates future.

[0028] When training the model, you can set the length window of the historical coordinates history and the length window of the predicted future trajectory coordinates future, as well as optimize the training model code and incrementally add training data.

[0029] The hyperparameters required for specific instantiation are as follows: class ViewportTransformerMTIO(nn.Module): def __init__(self, in_channel, fut_window, d_model, dim_feedforward,num_head=3, num_encoder_layers=2, num_decoder_layers=2, batch_first=True,dropout=0.2, device='cuda', repeat_prob=0.5, seed=1): super().__init__() self.in_channel = in_channel self.num_head = num_head self.fut_window = fut_window self.embedding = ViewportEmbedding(in_channels=in_channel * num_head,embedding_dim=d_model).to(device) self.transformer = Transformer(d_model=d_model, num_encoder_layers=num_encoder_layers, num_decoder_layers=num_decoder_layers, dim_feedforward=dim_feedforward, batch_first=batch_first) self.positional_embedding = PositionalEncoding(d_model=d_model,dropout=dropout).to(device) self.predictor = nn.Sequential(nn.Linear(d_model, in_channel * num_head), nn.Sigmoid()) self.device = device self.repeat_prob = repeat_prob self.seed = seed Then, the model is trained, and the viewport prediction model is trained using the training set, including: Define the loss function of the viewport prediction model , used to calculate the intersection-over-union ratio of the high-definition area of ​​the predicted viewpoint coordinates and the high-definition area of ​​the actual viewpoint coordinates, loss function The formula is as follows: in, represents the high-definition region of the predicted viewpoint coordinates, The high-definition area representing the actual viewpoint coordinates, Indicates the intersection-over-combination ratio of the high-definition area of ​​the predicted viewpoint coordinates and the high-definition area of ​​the actual viewpoint coordinates During the training process, by increasing the intersection-and-union ratio To reduce the size of the loss function, if the predicted area and the real area completely overlap, the intersection ratio is 1, then the loss function A value of 0 indicates that the prediction is accurate.

[0030] In step S3, the test set is input into the trained viewport prediction model for evaluation, including: S31: inputting the test set into the trained viewport prediction model for prediction evaluation, and outputting coordinate points of trajectory prediction; S32: converting the coordinate points of the trajectory prediction into viewport state information of the immersive panoramic video through a real-time data analysis algorithm; S32: Mapping the viewpoints into high-definition areas of the immersive video according to the designed viewpoint slice distribution; S33: Design performance evaluation indicators to comprehensively evaluate the performance of the model and obtain model evaluation results; S34: Compare the evaluation results of the new model with the original model. If there is an improvement in key performance indicators, replace the original model.

[0031] The performance of the comprehensive evaluation model of the design performance evaluation index further includes: The accuracy and precision standards are used to measure the proportion of samples predicted correctly by the viewport prediction model to the total samples and the proportion of samples that are actually positive. By mean absolute error The mean square error is used to measure the absolute difference between the predicted value and the actual value. and RMS error The average of the squares of the differences between the predicted values ​​and the actual values ​​is measured to quantify the magnitude of the prediction error. The formula is as follows: in, is the actual viewpoint coordinate, To predict the viewpoint coordinates, is the sample size.

[0032] In summary, the specific data flow is as follows Figure 2 As shown, the viewport prediction module of this embodiment uses an efficient MTIO-Transformer viewport prediction model to achieve fast and accurate prediction of head motion trajectory data through efficient real-time data analysis. Through the sharding-based immersive video streaming system, it aims to solve the diversity problem of user viewing patterns, thereby significantly improving the generalization ability of the viewport prediction model.

[0033] On the other hand, this embodiment also discloses a system for dynamically acquiring historical viewport coordinates and predicting future viewports, including: The acquisition and processing module is used to acquire the viewport trajectory data of watching immersive panoramic videos in real time through VR devices, pre-process the viewport trajectory data to construct a data set, and split the data set into a training set, a validation set, and a test set; A model building and training module, used to build a viewport prediction model based on an implicit ensemble learning algorithm, and to train and optimize the viewport prediction model using the training set; An optimization and evaluation module, used for inputting a verification set to verify the trained and optimized viewport prediction model, inputting the test set to the viewport prediction model for evaluation, and deploying and replacing the viewport prediction model according to the evaluation result; The application and prediction module is used to predict the future viewport of the immersive panoramic video stream using the optimized and evaluated viewport prediction model.

[0034] Among them, the functional implementation of each module and unit in the system for dynamically obtaining the historical coordinates of the viewport and predicting the future viewport corresponds to the steps in the above-mentioned method embodiment for dynamically obtaining the historical coordinates of the viewport and predicting the future viewport, and its functions and implementation process will not be repeated here one by one.

[0035] Finally, it should be noted that the above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

[0036] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for dynamically obtaining historical viewport coordinates to predict future viewports, characterized in that: The steps include: S1: collecting viewport trajectory data of watching immersive panoramic videos in real time through VR devices, preprocessing the viewport trajectory data to construct a data set, and splitting the data set into a training set, a validation set, and a test set; S2: constructing a viewport prediction model based on an implicit ensemble learning algorithm, and using the training set to train and optimize the viewport prediction model; S3: inputting a validation set to verify the trained and optimized viewport prediction model, and inputting the test set to the viewport prediction model for evaluation, and deploying and replacing the viewport prediction model according to the evaluation result; S4: Use the optimized and evaluated viewport prediction model to predict the future viewport of the immersive panoramic video stream.

2. The method for dynamically obtaining historical viewport coordinates and predicting future viewports according to claim 1, characterized in that: In step S1, collecting viewport trajectory data of watching immersive panoramic video in real time through VR equipment further includes: Create a file path and file for recording trajectory data, and record the collected current trajectory coordinates and the predicted trajectory coordinates output by the viewport prediction model separately in two files; Use VR head-mounted display device to collect user's viewport trajectory data at a set frequency and store it in the file of current trajectory coordinates, including the 3D rectangular coordinates of the user's visual focus point .

3. The method for dynamically obtaining historical viewport coordinates and predicting future viewports according to claim 1, characterized in that: In step S1, preprocessing the viewport trajectory data to construct a data set further includes: S11: Convert the 3D rectangular coordinates into longitude and latitude coordinates in a spherical coordinate system , dimension and longitude The calculation formula is as follows: ， ， in, For a special inverse tangent function, consider and to determine the correct quadrant; S12: The longitude and latitude coordinates Normalization is performed to convert the angle into a value in the range of [-1,1], and the dimension and precision are normalized respectively, as shown below: ， 。 4. The method for dynamically acquiring historical viewport coordinates and predicting future viewports according to claim 1, characterized in that: In step S2, constructing a viewport prediction model based on an implicit ensemble learning algorithm further includes: Designing the viewport prediction model, including an input layer, an implicit integration layer, a fusion layer, and an output layer; Receiving viewport trajectory data of multiple users through the input layer, including historical viewport coordinates and current viewport coordinates; The implicit integration layer is used to predict the viewport trajectory from different angles and using different features. The implicit integration layer includes multiple sub-models, and the sub-models are fused using a fusion policy of an attention mechanism. The final viewport prediction result is output through the output layer.

5. The method for dynamically obtaining historical viewport coordinates and predicting future viewports according to claim 1, characterized in that: In step S2, training the viewport prediction model using the training set further includes: Define the loss function of the viewport prediction model , used to calculate the intersection-over-union ratio of the high-definition area of ​​the predicted viewpoint coordinates and the high-definition area of ​​the actual viewpoint coordinates, loss function The formula is as follows: in, represents the high-definition region of the predicted viewpoint coordinates, The high-definition area representing the actual viewpoint coordinates, Indicates the intersection-over-combination ratio of the high-definition area of ​​the predicted viewpoint coordinates and the high-definition area of ​​the actual viewpoint coordinates During the training process, by increasing the intersection-and-union ratio To reduce the size of the loss function, if the predicted area and the real area completely overlap, the intersection ratio is 1, then the loss function A value of 0 indicates that the prediction is accurate.

6. The method for dynamically acquiring historical viewport coordinates and predicting future viewports according to claim 1, characterized in that: In step S3, inputting the test set into the trained viewport prediction model for evaluation further includes: S31: inputting the test set into the trained viewport prediction model for prediction evaluation, and outputting coordinate points of trajectory prediction; S32: converting the coordinate points of the trajectory prediction into viewport state information of the immersive panoramic video through a real-time data analysis algorithm; S32: Mapping the viewpoints into high-definition areas of the immersive video according to the designed viewpoint slice distribution; S33: Design performance evaluation indicators to comprehensively evaluate the performance of the model and obtain model evaluation results; S34: Compare the evaluation results of the new model with the original model. If there is an improvement in key performance indicators, replace the original model.

7. The method for dynamically acquiring historical viewport coordinates and predicting future viewports according to claim 6, characterized in that: The performance of the design performance evaluation index comprehensive evaluation model further includes: The accuracy and precision standards are used to measure the proportion of samples predicted correctly by the viewport prediction model to the total samples and the proportion of samples that are actually positive. By mean absolute error The mean square error is used to measure the absolute difference between the predicted value and the actual value. and RMS error The average of the squares of the differences between the predicted values ​​and the actual values ​​is measured to quantify the magnitude of the prediction error. The formula is as follows: in, is the actual viewpoint coordinate, To predict the viewpoint coordinates, is the sample size.

8. A system for dynamically acquiring historical viewport coordinates and predicting future viewports, characterized in that: include: The acquisition and processing module is used to acquire the viewport trajectory data of watching immersive panoramic videos in real time through VR devices, pre-process the viewport trajectory data to construct a data set, and split the data set into a training set, a validation set, and a test set; A model building and training module, used to build a viewport prediction model based on an implicit ensemble learning algorithm, and to train and optimize the viewport prediction model using the training set; An optimization and evaluation module, used for inputting a verification set to verify the trained and optimized viewport prediction model, inputting the test set to the viewport prediction model for evaluation, and deploying and replacing the viewport prediction model according to the evaluation result; The application and prediction module is used to predict the future viewport of the immersive panoramic video stream using the optimized and evaluated viewport prediction model.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for dynamically obtaining historical viewport coordinates and predicting future viewports according to any one of claims 1 to 7 is implemented.

10. An electronic device, characterized in that: include: one or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, enables the one or more processors to implement the method of dynamically obtaining historical viewport coordinates and predicting future viewports as described in any one of claims 1-7.

Images