Search

CN-122023654-A - Three-dimensional visual integrated management platform for ore body

CN122023654ACN 122023654 ACN122023654 ACN 122023654ACN-122023654-A

Abstract

The application relates to the technical field of image data processing, in particular to a three-dimensional visualized integrated management platform for a mineral body, which comprises a sensor data acquisition module for acquiring sensor data flow, a space-time feature extraction module for extracting mutation feature vectors comprising rock mass displacement acceleration, event duration and space diffusion coefficients, a continuous transparency weight parameter in the range of 0.2 to 0.9 is generated in real time through an ST-AM core module, an alpha channel value of a corresponding mineral body three-dimensional model layer in an OpenGL rendering pipeline is modified by a layer weight adjustment engine, and automatic highlighting is implemented on a high risk area by the three-dimensional visualized rendering engine. Therefore, the problems of structural mismatch and the like existing between a static layer weight mechanism and nonlinear mutation characteristics of dynamic risk events in the related technology are solved, millisecond response and mine safety early warning with low misjudgment rate can be realized, and the dynamic adaptability and emergency response precision of the three-dimensional visualization system are improved.

Inventors

  • Duan Xuezi
  • Jia Baoqian
  • ZHAO ZHENDONG
  • YU BIN

Assignees

  • 山东金鼎矿业有限责任公司

Dates

Publication Date
20260512
Application Date
20260130

Claims (10)

  1. 1. A three-dimensional visual integrated management platform for ore bodies is characterized by comprising a sensor data acquisition module, a space-time feature extraction module, a lightweight space-time attention model ST-AM core module, a layer weight adjustment engine and a three-dimensional visual rendering engine, wherein, The sensor data acquisition module is used for acquiring sensor data streams comprising rock mass displacement acceleration, three-dimensional coordinate offset and event time stamps in real time through a mine monitoring network; the space-time feature extraction module is connected with the sensor data acquisition module and is used for receiving the sensor data stream and extracting mutation feature vectors containing the rock mass displacement acceleration, event duration and space diffusion coefficients when the rock mass displacement acceleration is larger than a first preset threshold value; the ST-AM core module is connected with the space-time feature extraction module and comprises a space-time encoder and a weight generator, wherein the space-time encoder is used for space-time encoding the mutation feature vector and generating an attention thermodynamic diagram, and the weight generator is used for generating dynamic transparency weight according to the mutation feature vector and the attention thermodynamic diagram; The layer weight adjustment engine is connected with the ST-AM core module and is used for directly modifying an alpha channel value of a corresponding ore body three-dimensional model layer in an OpenGL rendering pipeline through a CUDA kernel function according to the dynamic transparency weight, and the modification comprises fusion calculation based on the dynamic transparency weight and a preset basic transparency; The three-dimensional visual rendering engine is connected with the layer weight adjustment engine and is used for rendering the ore body three-dimensional model in real time according to the modified alpha channel value and triggering an acousto-optic alarm when the dynamic transparency weight is greater than a second preset threshold.
  2. 2. The ore body three-dimensional visualization integrated management platform according to claim 1, wherein the sensor data acquisition module comprises a microseismic sensor array deployed on a roadway roof, a displacement monitor installed at a fault position of a rock body and an environment monitoring sensor deployed in the ore body; The sensor data acquisition module sends the sensor data stream to the space-time feature extraction module through a 5G special communication network, and the unidirectional transmission delay is not more than 50ms.
  3. 3. The ore body three-dimensional visualization integrated management platform according to claim 2, wherein the spatiotemporal feature extraction module receives the sensor data stream through a APACHE KAFKA message queue, the throughput of the APACHE KAFKA message queue is not lower than 10000 messages per second, and the spatiotemporal feature extraction module is deployed in an FPGA accelerator card; The first preset threshold value is 5mm/s 2 , and the space-time feature extraction module is used for completing extraction of the mutation feature vector within 50ms when the rock mass displacement acceleration is detected to be larger than 5mm/s 2 .
  4. 4. The ore body three-dimensional visualization integrated management platform according to claim 1, wherein the space-time encoder in the ST-AM core module is a three-layer convolution long-short-time memory network, the weight generator comprises a two-layer fully-connected neural network and a Softmax activation function, and the model parameter number of the ST-AM core module is smaller than 500KB.
  5. 5. The ore body three-dimensional visualization integrated management platform of claim 4, wherein the formula for calculating the dynamic transparency weight in the weight generator is: ; Wherein, the For the dynamic transparency weight to be used, For the displacement acceleration of the rock mass, To indicate a function, when the spatial diffusion coefficient is greater than 0.5, =1, Otherwise =0; The value range of the dynamic transparency weight is 0.2-0.9.
  6. 6. The ore body three-dimensional visualization integrated management platform according to claim 1, wherein the second preset threshold value is 0.7, the value of the basic transparency is 0.3, and the fusion calculation is implemented by: When the dynamic transparency weight is not more than 0.7, calculating the final transparency through linear interpolation based on the dynamic transparency weight and the preset basic transparency; and when the dynamic transparency weight is greater than 0.7, starting a pulsation animation mode to periodically change the final transparency within the range of 0.75-0.9.
  7. 7. The ore body three-dimensional visual integrated management platform of claim 1, wherein the three-dimensional visual rendering engine is implemented based on CesiumJS frames and runs in a client browser as WebGL instances; the layer weight adjustment engine is in communication connection with the three-dimensional visual rendering engine and is used for synchronizing the dynamic transparency weight or the alpha channel state based on modification to the three-dimensional visual rendering engine; And the three-dimensional visual rendering engine is used for setting the rendering color of the corresponding ore body three-dimensional model layer as a first color attribute when the dynamic transparency weight is greater than 0.7, controlling the transparency of the layer to periodically change within the range of 0.75-0.9 and synchronously triggering an audible and visual alarm according to the synchronous weight or state.
  8. 8. The ore body three-dimensional visualization integrated management platform of claim 7, wherein the three-dimensional visualization rendering engine is further configured to automatically retrieve and display borehole log information associated with a highlighted red pulse layer when a user clicks on the highlighted red pulse layer by performing reverse mapping in a geological database according to a layer identification of the highlighted red pulse layer.
  9. 9. The ore body three-dimensional visualization integrated management platform of claim 1, wherein the layer weight adjustment engine is further configured to process a plurality of ore body three-dimensional model layers with overlapping spatial locations; For the same spatial position, the layer weight adjustment engine acquires dynamic transparency weights of all layers covering the position, and calculates a comprehensive transparency weight according to a preset fusion rule for modifying the final performance of the position in rendering.
  10. 10. The ore body three-dimensional visualization integrated management platform of claim 5, wherein the weight generator further comprises a temperature adjustment parameter, and the modified calculation formula of the dynamic transparency weight is: ; Wherein, the For the dynamic transparency weight to be used, As a function of the Sigmoid, For the displacement acceleration of the rock mass, In order to adjust the parameters of the temperature, To indicate a function, when the spatial diffusion coefficient is greater than 0.5, =1, Otherwise =0。

Description

Three-dimensional visual integrated management platform for ore body Technical Field The application relates to the technical field of image data processing, in particular to a three-dimensional visualized integrated management platform for ore bodies. Background The ore body three-dimensional visual integrated management platform is used for deeply fusing a static geological model and dynamic safety monitoring data, so that real-time, visual and accurate presentation of the structure state of the underground rock body is realized. Currently, such platforms have evolved from traditional auxiliary design tools to critical infrastructure that ensures life safety and production continuity for the downhole operators. Especially under complex working conditions such as deep mining, high-stress surrounding rock and the like, the visual response capability of the platform to sudden geological disasters (such as rock burst, spalling and fault activation) directly determines the timeliness and accuracy of emergency decision. Currently mainstream ore body three-dimensional visualization platforms, such as Surpac, mineSight and Leapfrog Geo, generally adopt static layer weight strategies based on preset rules to solve the problem of multi-source information superposition. Specifically, the system assigns fixed transparency values (typically between 0.3 and 0.5) to different layers of geologic structures, roadway engineering, sensor points and the like in order to visually compromise background information retention and foreground event highlighting. The mechanism has certain rationality in the early stage mainly comprising static modeling, and can meet the basic expression requirement of the conventional exploration and planning stage on the spatial relationship. The technical logic is established on the premise of 'relatively stable geological environment and smooth change of monitored data', and the readability and the operation stability of a visual interface are maintained under the limited data updating frequency through weight parameters set by manual experience. However, with the continuous upgrading of mine safety supervision standards and the exponential increase of sensor network density, the static weighting mechanism exposes an inherent contradiction that is difficult to reconcile at a principle level. It is at its root in the structural mismatch between its rigid weight distribution logic and the nonlinear mutation characteristics of dynamic risk events. According to the mandatory requirement of the 5.2 th item of mine safety regulations (AQ 2033-2020), rock mass displacement exceeds 5mm to form an early warning threshold, and actual disaster evolution is often expressed as an instantaneous jump (such as 6.2mm/s 2) of displacement acceleration, and the events have strong burstiness, short duration (often lower than 100 ms) and local space aggregation. In such a scenario, the fixed transparency layer cannot visually enhance the high risk area, resulting in the critical pre-warning signal being submerged in a complex geological background. Actual measurement data in 2023 years of China national institute of safety and production science shows that in 287 displacement events of 12 metal mines, the early warning false alarm rate caused by unadjustable layer weight is up to 32.7%, namely a typical case is a mine collapse accident in 2022 years of Anhui, namely the layer transparency of a fault area is only 0.3, so that displacement abnormality cannot be recognized in time, and evacuation decision is delayed for over two minutes. Still further, the schemes in the related art generally suffer from response delay problems during the multi-source heterogeneous data fusion process. The sensor flow such as microseismic, displacement, temperature and humidity and the static geologic model have coordinate system difference, after a fixed weight strategy is overlapped, the system needs to rely on manual intervention to adjust parameters to adapt to new events, and literature on mine three-dimensional visualization technology bottleneck research indicates that 1200 millisecond delay is averagely introduced in the process, and the 200 millisecond response upper limit defined by the safety regulations is far exceeded. Even if some platforms attempt to introduce deep learning models (such as U-Net partition networks) to improve feature extraction capability, the huge parameter (super 5 megabytes) also makes it severely depend on cloud GPU (Graphics Processing Unit, graphics processor) resources, which cannot be deployed at the downhole edge computing nodes, so that the possibility of achieving millisecond-level response is lost in the physical architecture. It is noted that the improvements in the related art have not yet reached the nature of the contradiction. For example, patent CN119915350a (a mine safety monitoring information acquisition system and method) proposes that data is cooperatively acquired through vibratio