Dynamic transmission and dynamic loading method, device and system based on prediction model

Abstract

The invention provides a dynamic transmission and dynamic loading method, device and system based on a prediction model, and relates to the field of mining area model loading.The dynamic transmission method comprises the steps that a user request and an initial view are obtained, and ore block grade model data in the initial view are transmitted to a client; user operation and a real-time visual field are obtained in real time, and ore block grade model data in the real-time visual field are transmitted to the client side; and inputting the user operation and the real-time visual field into the trained prediction model to obtain a prediction visual field, and transmitting ore block grade model data in the prediction visual field to a client. Through local transmission and local loading of the model, the bandwidth requirement of transmission and the calculation amount during rendering are reduced, meanwhile, the prediction area possibly concerned by the user is obtained through the prediction model, model data of the prediction area is rendered in advance, and the time for the user to wait for rendering is shortened.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of mine area model loading, and in particular to a dynamic transmission and dynamic loading method, device and system based on a prediction model. BACKGROUND

[0002] Under the background of today's digital mine construction, the demand for visualization of ore block grade models is increasing. WebGL, as a technology for high-performance 3D graphics rendering on web pages, is widely used in the display of ore block grade models. However, the existing method of rendering ore block grade models in the WebGL environment has many problems.

[0003] With the continuous progress of geological exploration technology, the data volume of ore block grade models is growing explosively. A complete ore block grade model may contain a large amount of triangular facets, vertex information and grade attribute data. The traditional rendering method usually loads the entire ore block grade model data into the GPU for processing and rendering at one time. This approach will cause the GPU to face a huge computing pressure instantly when facing large-scale data. Because the GPU needs to process a large amount of graphics data, including vertex transformation, lighting calculation, texture mapping and other operations, the rendering frame rate drops sharply, the picture freezes or even cannot respond for a long time, which seriously affects the user experience.

[0004] In addition, loading the entire model at one time will also bring a lot of data transmission problems. The network bandwidth from the server to the client is often limited, and a large amount of data transmission will not only occupy valuable network resources, but also cause long transmission time and data loss or loading failure due to network fluctuations. Moreover, in actual application scenarios, when users browse the ore block grade model, they can only focus on the local area of the model at a certain moment, and full loading and rendering of the entire model is undoubtedly a great waste of computing resources and network resources.

[0005] Furthermore, the existing loading method lacks intelligent analysis of user behavior and perspective. During the browsing process, the user's perspective will change constantly, but the traditional method does not dynamically adjust the loaded model data according to the user's real-time perspective. This results in the model parts that are no longer in the field of view still occupying GPU computing resources and network bandwidth even after the user's perspective moves, further reducing the overall performance of the system.

[0006] In summary, the existing technology for rendering ore block grade models in the WebGL environment has problems such as large GPU computing load, heavy network transmission burden and unreasonable resource utilization when dealing with large data volumes. There is an urgent need for a new method to optimize the loading and rendering process of ore block grade models to meet the efficient and smooth visualization needs in digital mine construction. SUMMARY

[0007] The application provides a dynamic transmission and dynamic loading method, device and system based on a prediction model, which is used for dynamic transmission and dynamic loading of a block grade model, and can realize local dynamic loading of the block grade model and early loading of a prediction view by predicting the prediction view of user attention through the prediction model, so as to shorten the waiting time of loading the model after the user changes the view while saving transmission bandwidth and rendering resources.

[0008] The first aspect of the embodiment of the application provides a dynamic transmission method based on a prediction model, which is used for dynamic transmission of a block grade model, and the dynamic transmission method comprises the following steps: obtaining a user request and an initial view, and transmitting block grade model data in the initial view to a client; obtaining a user operation and a real-time view in real time, and transmitting block grade model data in the real-time view to the client; inputting the user operation and the real-time view into a trained prediction model to obtain a prediction view, and transmitting block grade model data in the prediction view to the client.

[0009] In some embodiments, before the step of obtaining the user request and the initial view, the dynamic transmission method further comprises the following step: inputting a training data set into a prediction model to train the prediction model until the training is completed, wherein the training data set comprises a plurality of groups of training data, and each group of training data comprises: block grade model data in a current view range, a user view angle change operation and block grade model data in an operation view range, wherein the operation view range is obtained based on the current view and the user view angle change operation.

[0010] In some embodiments, the step of inputting the training data set into the prediction model to train the prediction model until the training is completed comprises the following step: inputting the training data set into a long short-term memory model to train the long short-term memory model until the training is completed.

[0011] In a second aspect, the embodiment of the present application provides a dynamic loading method based on a prediction model, which is used for dynamically loading a block grade model. The dynamic loading method comprises the following steps: sending a user request to a server and sending an initial field of view to the server, obtaining, by the server, block grade model data in the initial field of view; rendering the block grade model data in the initial field of view and displaying the rendered graphics in a predetermined area of a browser window; listening to user operations in real time, calculating a real-time field of view based on the user operations, and sending the user operations and the real-time field of view to the server, obtaining, by the server, block grade model data in the real-time field of view and block grade model data in a prediction field of view; rendering the block grade model data in the real-time field of view and displaying the rendered graphics in the predetermined area of the browser window, and rendering the block grade model data in the prediction field of view.

[0012] In some embodiments, after the step of listening to user operations in real time, calculating a real-time field of view based on the user operations, and sending the real-time field of view to the server, the dynamic loading method further comprises the following steps: obtaining, by the server, block grade model data in the real-time field of view and block grade model data in an extended real-time field of view, wherein the extended real-time field of view is obtained based on the real-time field of view and an extended distance; the step of rendering the block grade model data in the real-time field of view and displaying the rendered graphics in the predetermined area of the browser window comprises the following steps: calculating a coincidence degree of a basic field of view and the real-time field of view, wherein the basic field of view is a field of view before a user field of view change operation; when the coincidence degree is greater than a preset coincidence degree threshold, obtaining a new area based on the basic field of view and the real-time field of view, obtaining block grade model data of the new area from the block grade model data in the extended real-time field of view, rendering the block grade model data of the new area, and displaying the rendered graphics in the predetermined area of the browser window; when the coincidence degree is less than the preset coincidence degree threshold, rendering the block grade model data in the real-time field of view as a whole, and displaying the rendered graphics in the predetermined area of the browser window.

[0013] In some embodiments, the extended distance is obtained based on a model complexity of the block grade model, a number of blocks in the block grade model, a number of different grade areas in the block grade model, and a historical average view angle movement distance, wherein the historical average view angle is obtained through historical operation data of a user.

[0014] In some embodiments, after the rendering of the ore block grade model data in the predicted view and the displaying of the rendered graphics in the predetermined area of the browser window, the dynamic loading method further comprises: obtaining, by the server, overall data of the ore block grade model, and pre-rendering ore block model data in a view area other than the current view and the predetermined area, wherein the pre-rendering has a lower accuracy than the rendering of the ore block grade model in the current view and the predicted view.

[0015] The third aspect of the embodiments of the present application provides a dynamic transmission device based on a prediction model, used for dynamically transmitting an ore block grade model. The dynamic transmission device comprises: an obtaining module, configured to obtain a user request and an initial view; a transmission module, configured to transmit ore block grade model data in the initial view to a client; the obtaining module is further configured to obtain a user operation and a real-time view in real time; the transmission module is further configured to transmit ore block grade model data in the real-time view to the client; and a prediction module, configured to input the user operation and the real-time view into a trained prediction model to obtain a predicted view; and the transmission module is further configured to transmit ore block grade model data in the predicted view to the client.

[0016] The fourth aspect of the embodiments of the present application provides a dynamic loading device based on a prediction model, used for dynamically loading an ore block grade model. The dynamic loading device comprises: a sending module, configured to send a user request to a server and send an initial view to the server; a receiving module, configured to obtain, by the server, ore block grade model data in the initial view; a rendering module, configured to render ore block grade model data in the initial view and display the rendered graphics in a predetermined area of a browser window; a determination module, configured to listen to a user operation in real time and calculate a real-time view based on the user operation; the sending module is further configured to send the user operation and the real-time view to the server; the receiving module is further configured to obtain, by the server, ore block grade model data in the real-time view and ore block grade model data in the predicted view; and the rendering module is further configured to render ore block grade model data in the real-time view and display the rendered graphics in the predetermined area of the browser window, and render ore block grade model data in the predicted view.

[0017] In a fifth aspect, the embodiment of the present application provides a dynamic loading system based on a prediction model, which is used for dynamically loading a block grade model, and the dynamic loading system comprises a client, a server, and the client.

[0018] The present application reduces the amount of data to be transmitted in a single transmission process and the amount of calculation to be processed in a single rendering process by locally loading and locally transmitting the whole block grade model, thereby reducing the demand for network bandwidth resources and the demand for computing resources, and predicting the area that the user may focus on, i.e., the prediction view, by using a prediction model, and transmitting and loading the block grade model data in the prediction view in advance, so as to shorten the loading time of the area when the user's view moves to the prediction view and make the loading of the model more smooth. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 A flowchart of a dynamic transmission method based on a prediction model is provided for the embodiment of the present application. Figure 2 A flowchart of a training method of a prediction model is provided for the embodiment of the present application. Figure 3 A flowchart of a first dynamic loading method based on a prediction model is provided for the embodiment of the present application. Figure 4 A flowchart of a second dynamic loading method based on a prediction model is provided for the embodiment of the present application. Figure 5 A schematic diagram of a view cone and a real-time view is provided. Figure 6 A schematic diagram of a view cone, a real-time view and an extension distance is provided. Figure 7 is a schematic diagram of a field of view cone, a real-time field of view and an extended real-time field of view; Figure 8 is a structural schematic diagram of a dynamic transmission device based on a prediction model provided by an embodiment of the present application; Figure 9 is a structural schematic diagram of a dynamic loading device based on a prediction model provided by an embodiment of the present application; Figure 10 is a framework and data flow diagram of a dynamic loading system based on a prediction model provided by an embodiment of the present application. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be described in detail below in combination with the drawings and specific embodiments.

[0021] In the specific embodiments, each specific technical feature in each of the various embodiments described can be combined in various combinations, for example, different specific technical features can form different embodiments through combination, in order to avoid unnecessary repetition, various possible combinations of each specific technical feature in the present application will not be described again.

[0022] It should also be noted here that, in order to avoid obscuring the present application due to unnecessary details, only structures and / or processing steps closely related to the solutions of the present application are shown in the drawings, and other details not closely related to the present application are omitted.

[0023] In addition, it should also be noted that the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment. In the following description, the terms "first, second,... " are only used to distinguish different objects, and do not mean that the objects have the same or relationship. It should be understood that the orientation described by the orientation terms "above", "below", "inner", "outer" and the like is the orientation in the normal use state.

[0024] In some embodiments, as shown in Figure 1 , Figure 1 is a flowchart of a first dynamic transmission method based on a prediction model provided by an embodiment of the present application, the dynamic transmission method comprising: Step S101, obtaining a user request and an initial field of view, and transmitting the ore block grade model data in the initial field of view to the client.

[0025] The user request is initialization data triggered by the user based on the client, and is used to request to call a related resource to start transmission and loading of the ore block grade model. Meanwhile, the client generates a preset initial field of view or prompts the user to select and generates an initial field of view according to the selection of the user. After the initial field of view is generated, the initial field of view is sent to the server to enable the server to transmit the ore block grade model data in the initial field of view. It can be understood that only part of the ore block grade model data in the initial field of view is transmitted, that is, the data of a local region in the model is transmitted, so that the amount of data to be transmitted is reduced, and the bandwidth required for transmitting the model data is reduced. The ore block grade model data includes geometric information (vertex coordinates, face connection relationships, etc.) of blocks in the region and grade attribute information. The ore block grade model needed can be rendered through the information. Subsequent ore block grade model data also includes the above information, which will not be described in detail hereinafter.

[0026] In step S102, the user operation and the real-time field of view are acquired in real time, and the ore block grade model data in the real-time field of view is transmitted to the client.

[0027] Specifically, the client can determine the real-time field of view of the user according to the operation of the user. After the client transmits the real-time field of view to the server, the server transmits the ore block grade model data in the real-time field of view to the client, so that the client can render and display the ore block grade model in the field of view needed by the user. It should be noted that the client transmits the user operation to the server to provide data for the subsequent prediction step.

[0028] In step S103, the user operation and the real-time field of view are input into the trained prediction model to obtain a predicted field of view, and the ore block grade model data in the predicted field of view is transmitted to the client.

[0029] The predicted field of view output by the prediction model predicts other regions that the user may focus on based on the real-time field of view of the user and the operation habit of the user, and transmits the ore block grade model data in the predicted field of view to the client in advance, so that the client can load the ore block grade model in the region in advance. If the user shifts the real-time view angle to the region through the view angle operation, the client can quickly display the model of the region, thereby shortening the waiting time of the user.

[0030] It should be noted that the prediction model not only outputs the predicted field of view through the real-time field of view, but also needs to predict through user operation, which can be understood as that the prediction model not only extracts the internal relationship between the ore block grade model data in the real-time field of view before and after the user operation, for example, extracts the spatial structure characteristics such as the overall shape of the model, the distribution of different grade regions, the connection relationship between the blocks, etc., which helps the prediction model to better understand the characteristics of the model, so as to more accurately predict the area that the user may pay attention to; also extracts the internal relationship between the user's operation habits and the change of the field of view, for example, by collecting a large amount of user operation data, including but not limited to the moving track of the user's perspective, the stay position, the zoom operation and other information, these data will be an important basis for the training of the prediction model, to analyze the user's behavior pattern and preference in the browsing process, that is, the prediction model comprehensively predicts the area that the user may pay attention to through the two aspects of model data relationship and user operation habits, so as to make the prediction of the prediction model more accurate.

[0031] In some embodiments, Figure 1 The prediction model in the step S101 is a neural network model with deep learning function, and the prediction model is trained through historical data before the step S101, that is, the training data set is input into the prediction model to train the prediction model until the training is completed. Figure 1 Figure 2 ​The training process is exemplarily illustrated as follows. In a first step, user time-series browsing behavior data is collected, i.e., a training data set is formed by historical data, which includes multiple groups of training data, each group of training data including: ore grade model data in a current field of view, a user's view change operation, and ore grade model data in a field of view after the operation, wherein the field of view after the operation is obtained based on the current field of view and the user's view change operation, for example, the user's view change operation is mouse translation, and the field of view after the operation is the field of view after the field of view before the operation is translated by a preset distance in the mouse movement direction. The training data sets are arranged in time sequence according to the user's operation to form a training data set. In a second step, the weight coefficients and bias vectors in the training model are initialized, for example, each weight coefficient and bias vector in the training model is randomly assigned an initial value. In a third step, the user's operation and the ore grade model data in the field of view before the operation in each group of training data are input into the prediction model in time sequence, and the prediction result output by the neural network is obtained by layer-by-layer calculation of the hidden layer of the neural network model, which is the predicted field of view. In a fourth step, the loss value is determined according to the difference between the output predicted field of view and the actual field of view after the operation in the corresponding training data, and the accuracy of the prediction is determined according to the loss value. In a fifth step, before inputting new training data each time, the weight matrix and bias vector of the neural network model are optimized by gradient descent method until the loss value converges to a minimum value, which is called one training. In a sixth step, after each training is completed, it is determined whether the loss value after the training meets the training completion condition, i.e., whether the loss value after the training continuously converges to a range smaller than a predetermined loss value threshold. If the training completion condition is not met, the next group of training data is input into the neural network model to continue iterative training. If the training completion condition is met, it is considered that the prediction model is trained.

[0032] Optionally, the neural network model is a recurrent neural network; optionally, the neural network model is a long short-term memory model. When the long short-term memory model is trained according to the training data set arranged in time sequence, the long short-term memory network can not only associate adjacent two groups of training data, but also associate interval training data, and the smaller the interval of the training data, the higher the degree of association of the long short-term memory network to the training data, so as to further mine the association between multiple groups of training data sets according to the time sequence relationship, and thus more accurate prediction can be performed. The basic unit of the long short-term memory network and its calculation formula are introduced as follows. The long short-term memory network includes: forget gate f_t: f_t=σ(W_f·[h_{t-1},x_t]+b_f) input gate i_t: i_t=σ(W_i·[h_{t-1},x_t]+b_i) Candidate memory unit widetilde{C}_t: widetilde{C}_t=tanh(W_C·[h_{t-1},x_t]+b_C) Memory unit C_t: C_t=f_t·C_{t-1}+i_t·widetilde{C}_t Output gate o_t: o_t=σ(W_o·[h_{t-1},x_t]+b_o) Hidden state h_t: h_t = o_t·tanh(C_t) Where σ() represents the Sigmoid function, tanh() represents the hyperbolic tangent function, W_f, W_i, W_C, W_o are weight matrices, b_f, b_i, b_C, b_o are bias vectors, x_t is the input at the current time step, h_{t-1} is the hidden state at the previous time step, and C_{t-1} is the memory unit at the previous time step.

[0033] This invention also provides a dynamic loading method based on a prediction model. This dynamic loading method is applied to dynamically loading a ore grade model. The execution subject of this dynamic loading method is a client, which has a built-in WebGL environment. Through the WebGL environment, the ore grade model can be displayed in the form of a visual model. The following is an exemplary description of this dynamic loading method.

[0034] In some embodiments, such as Figure 3 As shown, Figure 3 This invention provides a flowchart illustrating a dynamic loading method based on a prediction model, which includes: Step S201: Send a user request to the server and send the initial field of view to the server, so that the server can obtain the ore grade model data within the initial field of view.

[0035] That is, by sending a request to the server, the ore grade model data within the initial field of view is obtained in order to facilitate the subsequent display of the ore grade model within the initial field of view. It should be noted that the initial field of view is expressed in a parameterized form, so that it can be transmitted to the server as data. The following is an example of the formula for the parameterized representation of the user's perspective: Assuming the user's viewpoint position is E(x_E, y_E, z_E), and the mouse click position on the screen is (x_m, y_m), the screen coordinates can be converted to world coordinates through the projection matrix P and the view matrix V of WebGL. First, convert the screen coordinates (x_m, y_m) to normalized device coordinates (x_{ndc}, y_{ndc}) using the following formulas: x_{ndc}=\frac{2x_m}{width}-1 y_{ndc}=1-\frac{2y_m}{height} where width and height are the width and height of the WebGL canvas, respectively.

[0036] Then convert the normalized device coordinates to world coordinates P_w(x_w, y_w, z_w) through the inverse projection matrix P^{-1} and the inverse view matrix V^{-1}: \begin{pmatrix}x_w\\y_w\\z_w\\1\end{pmatrix}=V^{-1}\cdotP^{-1}\cdot\begin{pmatrix}x_{ndc}\\y_{ndc}\\-1\\1\end{pmatrix} The line-of-sight vector \vec{v} is then P_w-E, i.e. \vec{v}(x_w-x_E, y_w-y_E, z_w-z_E).

[0037] Field of view range (field of view angle): Assuming the vertical field of view angle set in the WebGL scene is \theta_{v}, the horizontal field of view angle \theta_{h} can be calculated through the aspect ratio aspect=\frac{width}{height}: \theta_{h}=2\arctan(\tan(\frac{theta_{v}}{2})·aspect) Current viewpoint position: The viewpoint position is directly determined by the user's observation point coordinates in the three-dimensional scene, i.e. E(x_E, y_E, z_E).

[0038] It should be noted that the above formulas express matrices or vectors in Latex format, and the subsequent parameterization methods of the field of view can refer to the above formulas, so they will not be described in detail in the following.

[0039] Step S202, render the ore grade model data within the initial field of view and display the rendered graphics in a predetermined area of the browser window.

[0040] It can be understood that only loading the local ore block grade model data within the initial field of view and only displaying the local model reduces the amount of data rendered, speeding up the loading of the model.

[0041] Step S203, real-time monitoring of user operation and calculating real-time field of view based on user operation, and sending user operation and real-time field of view to the server to obtain the ore block grade model data within the real-time field of view and the ore block grade model within the predicted field of view from the server.

[0042] It can be understood that by monitoring the user operation, if the user's operation changes the field of view, the changed field of view is calculated according to the user operation and the parameterized user perspective, so that the real-time field of view of the user can be obtained in real time according to the user's operation. After sending the real-time field of view and operation of the user to the server, the server sends the ore grade model data within the real-time field of view to the server, and the server also sends the ore grade model data within the predicted field of view output by the prediction model to the server, that is, only the local model data within the real-time field of view and the predicted field of view needs to be transmitted, reducing the bandwidth required for data transmission.

[0043] Step S204, rendering the ore block grade model data within the real-time field of view and displaying the rendered graphics in a predetermined area of the browser window, and rendering the ore block grade model data within the predicted field of view.

[0044] It can be understood that only the ore block grade model data within the real-time field of view is rendered, thereby reducing the amount of calculation required for rendering, and by rendering the ore block grade model data within the predicted field of view in advance, the area that the user may focus on next is rendered in advance, which can reduce the user's waiting time if the user's field of view moves in this area.

[0045] The ore block grade model dynamic loading method based on the prediction model has the following beneficial effects: From the aspect of GPU calculation amount, the traditional one-time full-amount loading of ore block grade model data makes GPU need to process a large amount of graphics data, with a heavy calculation burden. The present method dynamically loads local models, only allowing GPU to process model data slightly larger than the user's viewing volume range, greatly reducing the number of vertices, the number of patches, and the amount of related lighting and texture calculation. For example, for a large data amount of ore block grade model, GPU may need to process millions of rendering tasks under the traditional method, while using the present method, GPU only needs to process tens of thousands of vertices related to the loading area at a certain moment, and the calculation amount may be only a few tenths of the original. This significantly reduces the workload of GPU, enabling it to more efficiently complete the rendering task, avoiding rendering lag caused by excessive calculation amount, thereby improving the rendering frame rate and allowing users to enjoy a smooth visual experience.

[0046] In terms of network transmission, the traditional full-amount loading method can cause a large amount of data to be transmitted between the server and the client, occupying a large amount of network bandwidth. The present method only pushes the model data in the loading area, greatly reducing the data transmission amount. For example, the data amount that originally needs to be transmitted may be several hundred MB or even larger, but now only a few MB of data needs to be transmitted. The large reduction in data transmission amount not only speeds up the data transmission speed and reduces the user waiting time, but also reduces the risk of data transmission failure caused by network bandwidth limitations or network fluctuations, improving the stability and reliability of data transmission. The advantage is more obvious especially in poor network conditions.

[0047] From the perspective of user experience, the dynamic calculation of the loading area based on the prediction model combined with the current perspective information of the user can prepare the model part that the user may focus on in advance. This means that when the user quickly moves the perspective, the required model data has been loaded or is being loaded, and there is no phenomenon of blank screen or long waiting caused by delayed data loading, ensuring the coherence and smoothness of user browsing. Moreover, the progressive loading method first displays the general outline of the model, allowing the user to quickly understand the overall situation of the loading area, and then gradually refines the details, further improving the experience during the waiting process. At the same time, this intelligent dynamic loading method can effectively apply to ore block grade models of different sizes, whether small and simple models or super large and complex models, can achieve efficient rendering through dynamic adjustment of the loading area, and provide consistent and high-quality browsing experience for users.

[0048] In addition, the present method also has good scalability and adaptability. As the data amount of the ore block grade model continues to grow and the user's demand for browsing experience continues to improve, the method can be upgraded through optimization of the prediction model, adjustment of the loading area calculation strategy, and other ways to continuously meet various needs in actual applications. At the same time, it can also be combined with other graphics optimization techniques (such as level-of-detail model technology) to further improve the rendering performance of the ore block grade model in the WebGL environment.

[0049] In some embodiments, as shown in Figure 4 , the second dynamic loading method based on the prediction model provided by the embodiments of the present application has the following steps: Figure 4 The flowchart of the second dynamic loading method based on the prediction model provided by the embodiments of the present application is shown in Figure 3 , which is different from the dynamic loading method in that, after step S203 in Figure 3 , the dynamic loading method further includes: Step S301: obtaining, by the server, the ore block grade model data within the real-time field of view and the ore block grade model data within the extended real-time field of view.

[0050] The extended real-time field of view is obtained based on the real-time field of view and the extended distance, so that the client can obtain the block grade model data slightly larger than the real-time field of view, and when the real-time field of view of the user is fine-tuned, it is not necessary to re-render a new area, but only to render the model in the new area according to the block grade model data in the extended real-time field of view, thereby further improving the model rendering speed. It should be noted that, during data transmission, the block grade model data in the real-time field of view can be transmitted first, and then the block grade model data in the extended real-time field of view can be transmitted; or the real-time block grade model data can be transmitted first, and then the extended area can be obtained by comparing the real-time field of view and the extended real-time field of view, and only the block grade model data in the extended area can be transmitted; or the block grade model data in the extended real-time field of view can be transmitted after the calculation of the extended real-time field of view is completed, and data tags are used to mark which block grade model data is located in the real-time field of view, and which block grade model data is located outside the real-time field of view and in the extended real-time field of view.

[0051] It should be noted that the extended distance is used to expand the boundary of the real-time field of view, and can be a human-preset coefficient, or can be determined by model parameters and user operation habits. The specific determination method of the extended distance will be described below. The extended distance is obtained based on the model complexity of the block grade model, the number of blocks in the block grade model, the number of different grade regions in the block grade model, and the historical average view angle moving distance. The historical average view angle is obtained by statistical analysis of user historical operation data. For example, The extended distance d can be calculated according to the model complexity C, the average view angle moving distance m obtained by statistical analysis of historical data, and an empirical coefficient k by the following formula: d=k·C·m The model complexity C can be calculated by comprehensively considering factors such as the number n of blocks in the block grade model and the number r of different grade regions, for example: C=α·n+β·r Wherein, α and β are weight coefficients adjusted according to actual conditions.

[0052] The specific determination method of the extended distance will be described below. Figures 5 to 7 The process of obtaining the extended real-time field of view based on the real-time field of view and the extended distance will be described below. It is assumed that Figure 5 The boundaries of the view frustum in the x, y, and z directions are x_1, x_2, y_1, y_2, z_1, and z_2 respectively, and the extended distance is d. As shown in Figure 6 , the range of the loading area in the x direction becomes x_1-d to x_2+d, and the same applies to the y and z directions, thereby obtaining the extended real-time field of view range as shown in Figure 7 .

[0053] It should be noted that the calculation of the above extension distance and the calculation of the extended real-time field of view can be performed by the server or by the client.

[0054] As shown in Figure 4 , Figure 1 Step S204 in includes:

[0055] The basic field of view is the field of view before the change of the field of view based on the user operation. By comparing the field of view area before and after the user operation, the degree of change of the field of view can be determined, and different rendering strategies are adopted according to the degree of change of the user's field of view.

[0056] Step S303, in the state that the coincidence degree is greater than the preset coincidence degree threshold, the new area is obtained based on the basic field of view and the real-time field of view, the block grade model data of the new area is obtained from the block grade model data in the extended real-time field of view, the block grade model data of the new area is rendered and the rendered graphics are displayed in the predetermined area of the browser window.

[0057] It can be understood that in the case that the degree of change of the user's real-time field of view is small, that is, the changed field of view still falls within the extended real-time field of view based on the real-time field of view, a new area can be obtained by comparing the field of view area before and after the change of the field of view, the block grade model data in the corresponding new area is obtained from the block grade model data in the extended real-time field of view, and then only the model data in the new area needs to be rendered, thereby further shortening the user's waiting time for model rendering in the case of small range change of the user's field of view.

[0058] Step S304, in the state that the coincidence degree is less than the preset coincidence degree threshold, the block grade model data in the real-time field of view is rendered in its entirety and the rendered graphics are displayed in the predetermined area of the browser window.

[0059] That is, if the new real-time field of view determined by the coincidence degree changes greatly, the block grade model data in the new real-time field of view needs to be rendered in its entirety to obtain the model of this part.

[0060] Step S305, rendering the block grade model data in the predicted field of view.

[0061] Optionally, after step S204 in Figure 4 , the dynamic loading method based on the prediction model further includes: the server obtains the overall data of the block grade model, and pre-renders the block model data of the field of view area except the current field of view and the predetermined area.

[0062] The pre-rendered precision is lower than that of the ore grade model within the current and predicted view areas. This can be understood as allowing for overall low-precision rendering of areas not currently of user interest or areas the user might be interested in, thus further reducing rendering time when these areas need to be rendered. Optionally, after low-precision rendering of the ore grade model data outside the current and predicted view areas, the model of these areas is displayed in a predetermined area within the browser window, allowing the model to display the intermediate areas in low precision as the user's view moves from one area to the next.

[0063] A third aspect of the present invention also provides a dynamic transmission device, such as... Figure 8 As shown, the dynamic transmission device can be, for example, a server, and includes: an acquisition module 401, a transmission module 402, and a prediction module 403. The acquisition module 401 is used to acquire user requests and initial field of view, and also to acquire user operations and real-time field of view in real time; the transmission module 402 is used to transmit the ore grade model data within the initial field of view to the client, and also to transmit the ore grade model data within the real-time field of view to the client, and also to transmit the ore grade model data within the predicted field of view to the client; the prediction module 403 is also used to input user operations and real-time field of view into the trained prediction model to obtain the predicted field of view.

[0064] A fourth aspect of the present invention also provides a dynamic loading device based on a prediction model. This loading device can be, for example, a handheld client device, such as… Figure 9 As shown, the dynamic loading device includes: a sending module 501, a receiving module 502, a rendering module 503, and a determining module 504. The sending module 501 is used to send user requests to the server and send the initial field of view to the server; it is also used to send user operations and the real-time field of view to the server. The receiving module 502 is used to obtain ore grade model data within the initial field of view from the server; it is also used to obtain ore grade model data within the real-time field of view from the server. The rendering module 503 is used to render the ore grade model data within the initial field of view and display the rendered graphics in a predetermined area of ​​the browser window; it is also used to render the ore grade model data within the real-time field of view and display the rendered graphics in a predetermined area of ​​the browser window, and to render the ore grade model data within the predicted field of view. The determining module 504 is used to monitor user operations in real time and calculate the real-time field of view based on the user operations.

[0065] The fifth aspect of this invention also provides a dynamic loading system for prediction models, such as... Figure 10As shown, the dynamic loading system comprises a client 600 and a server 700, the client 600 is configured to monitor user requests and initial view, the server 700 is configured to obtain the user requests and the initial view and send the ore block grade model data in the initial view to the client 600, the client 600 is configured to render the ore block grade data in the initial view, the client 600 is configured to initialize loading on the page of the browser and display the rendered graphics in a predetermined position of the browser window, the client 600 is also configured to listen to user operations in real time and send the user operations to the server 700, the server 700 is configured to calculate the user real-time view according to the user operations, transmit the ore block grade model data in the real-time view to the client 600, at the same time, transmit the user operations and the real-time view to the prediction model to obtain the predicted view, send the ore block grade model data in the predicted view to the client 600, and optionally, the process of calculating the real-time view based on the user operations can also be performed by the client 600, and the client 600 directly sends the user operations and the parameterized real-time view to the server 700.

[0066] The client 600 is also configured to obtain the ore block grade model data in the real-time view for rendering and display the rendered graphics in a predetermined area of the browser window of the client, and is also configured to render the obtained ore block grade model data in the predicted view.

[0067] The above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application, although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A dynamic transmission method based on a prediction model, used for dynamically transmitting ore grade models, characterized in that, The dynamic transmission method includes: Obtain the user request and initial field of view, and transmit the ore grade model data within the initial field of view to the client; Real-time acquisition of user operations and real-time field of view, and transmission of ore grade model data within the real-time field of view to the client; The prediction model trained by the user operation and the real-time vision input is used to obtain the prediction vision, and the ore grade model data within the prediction vision is transmitted to the client.

2. The dynamic transmission method according to claim 1, characterized in that, Prior to obtaining the user request and initial field of view, the dynamic transmission method further includes: The training dataset is input into the prediction model to train the prediction model until training is complete. The training dataset includes multiple sets of training data, each set of training data including: ore grade model data within the current field of view, user viewpoint change operation, and ore grade model data within the field of view after the operation. The field of view after the operation is obtained based on the current field of view and the user viewpoint change operation.

3. The dynamic transmission method according to claim 2, characterized in that, The step of inputting the training dataset into the prediction model to train the prediction model until training is complete includes: The training dataset is input into the Long Short-Term Memory (LSTM) model to train the LSM model until training is complete.

4. A dynamic loading method based on a prediction model, used for dynamically loading ore block grade models, characterized in that, The dynamic loading method includes: A user request is sent to the server, and the initial field of view is sent to the server, which then obtains the ore grade model data within the initial field of view; The ore grade model data within the initial field of view is rendered and the rendered graphic is displayed in a predetermined area of ​​the browser window. The system monitors user operations in real time and calculates the real-time field of view based on the user operations. The user operations and the real-time field of view are then sent to the server, which obtains the ore grade model data within the real-time field of view and the ore grade model data within the predicted field of view. The ore grade model data within the real-time field of view is rendered and the rendered graphics are displayed in a predetermined area of ​​the browser window. The ore grade model data within the predicted field of view is also rendered.

5. The dynamic loading method according to claim 4, characterized in that, After real-time monitoring of user operations, calculation of the real-time field of view based on user operations, and sending the real-time field of view to the server, the dynamic loading method further includes: The server acquires ore grade model data within the real-time field of view and ore grade model data within the extended real-time field of view, wherein the extended real-time field of view is obtained by multiplying the real-time field of view by the extended distance. The step of rendering the ore grade model data within the real-time field of view and displaying the rendered graphic in a predetermined area of ​​the browser window includes: Calculate the overlap between the base field of view and the real-time field of view, where the base field of view is the field of view before the user's field of view changes. When the overlap is greater than a preset overlap threshold, a new area is obtained based on the basic field of view and the real-time field of view. The ore grade model data of the new area is obtained from the ore grade model data in the extended real-time field of view. The ore grade model data of the new area is rendered and the rendered graphic is displayed in a predetermined area of ​​the browser window. When the overlap is less than a preset overlap threshold, the ore grade model data within the real-time field of view is rendered as a whole, and the rendered graphics are displayed in a predetermined area of ​​the browser window.

6. The dynamic loading method according to claim 5, characterized in that, The extended distance is obtained based on the model complexity of the ore grade model, the number of blocks in the ore grade model, the number of different grade regions in the ore grade model, and the historical average perspective movement distance, wherein the historical average perspective is obtained through user historical operation data.

7. The dynamic loading method according to claim 4, characterized in that, After rendering the ore grade model data within the real-time field of view and displaying the rendered graphics in a predetermined area of ​​the browser window, and after rendering the ore grade model data within the predicted field of view, the dynamic loading method further includes: The server obtains the overall data of the ore grade model, and pre-renders the ore model data in the field of view area other than the current field of view and the predetermined area, wherein the accuracy of the pre-rendering is lower than the rendering accuracy of the ore grade model data in the current field of view and the predicted field of view.

8. A dynamic transmission device based on a prediction model, used for dynamically transmitting ore grade models, characterized in that, The dynamic transmission device includes: The acquisition module is used to acquire user requests and initial field of view; The transmission module is used to transmit the ore grade model data within the initial field of view to the client. The acquisition module is also used to acquire user operations and real-time field of view in real time; The transmission module is also used to transmit the ore grade model data within the real-time field of view to the client. The prediction module is used to obtain the predicted field of view from the prediction model trained by the user operation and the real-time field of view input. The transmission module is also used to transmit the ore grade model data within the predicted field of view to the client.

9. A dynamic loading device based on a prediction model, used for dynamically loading an ore block grade model, characterized in that, The dynamic loading device includes: The sending module is used to send user requests to the server and send the initial view to the server; The receiving module is used to obtain the ore grade model data within the initial field of view from the server; The rendering module is used to render the ore grade model data within the initial field of view and display the rendered graphics in a predetermined area of ​​the browser window. The determination module is used to monitor user operations in real time and calculate the real-time field of view based on user operations. The sending module is also used to send the user operation and real-time view to the server; The receiving module is also used to obtain ore grade model data within the real-time field of view and ore grade model data within the prediction field of view from the server. The rendering module is also used to render the ore grade model data within the real-time field of view and display the rendered graphics in a predetermined area of ​​the browser window, and to render the ore grade model data within the predicted field of view.

10. A dynamic loading system based on a prediction model, used for dynamically loading ore block grade models, characterized in that, The dynamic loading system includes: The client is used to monitor user requests and initial field of view; The server is used to obtain the user request and the initial field of view, and send the ore grade model data within the initial field of view to the client. The client is used to render the ore grade model data within the initial field of view and display the rendered graphics in a predetermined area of ​​the browser window. The client is also used to monitor user operations in real time and determine the real-time field of view based on the user operations; The server is also configured to acquire the user operation and the real-time field of view, transmit the ore grade model data within the initial field of view to the client, transmit the ore grade model data within the real-time field of view to the client, and transmit the prediction model trained by the user operation and the real-time field of view to obtain the prediction field of view, and transmit the ore grade model data within the prediction field of view to the client. The client is also used to render the ore grade model data within the real-time field of view and display the rendered graphics in a predetermined area of ​​the client's browser window, and to render the ore grade model data within the predicted field of view.

Images