Tailings pond emergency space on-site rapid verification method, system and device

Abstract

The present application belongs to the technical field of environmental emergency management, and particularly relates to a tailing pond emergency space on-site rapid verification method, system and device, the method comprising: identifying emergency space as a candidate object based on remote sensing image and digital elevation model to construct a pre-identification database; entering on-site verification data through a structured collection form running on a mobile terminal, each record being associated with a unique number UID, positioning coordinates and photos; comparing real-time positioning coordinates with the boundary of the candidate object in space to push pre-identified information to guide verification; performing real-time image reasoning and geometric parameter extraction on the candidate object through an edge computing device, and comparing the parameters with preset parameters to output hierarchical on-board verification results; matching the on-board verification results and ground verification results to the corresponding UID through coordinates to fuse them, comprehensively evaluating the capability of each emergency space and outputting an emergency disposal priority recommendation sequence. The present application realizes rapid verification and data fusion of tailing pond emergency space.

Description

Technical Field

[0001] This invention belongs to the field of environmental emergency management technology, specifically relating to a method, system and device for rapid on-site verification of emergency spaces in tailings dams. Background Technology

[0002] Tailings ponds are storage sites for tailings waste from mining operations. Leaks downstream of tailings ponds pose a serious threat to downstream water environments and the safety of drinking water for residents. my country currently has a large number of tailings ponds, a significant portion of which are located upstream of centralized drinking water sources or near inter-provincial border sections, resulting in significant environmental risk control pressures. In accordance with the requirements of the Ministry of Ecology and Environment's "Technical Guidelines for the Implementation of the Nanyang Practice in Environmental Emergency Response to Sudden Water Pollution Incidents in River Basins," it is necessary to systematically identify, verify, and assess the emergency spaces and facilities within a certain range along the downstream river channels and banks of tailings ponds, and to develop a tailings-specific emergency plan with a map for each tailings pond to support rapid response and scientific handling of sudden water pollution incidents.

[0003] Currently, on-site verification of emergency spaces and facilities mainly relies on manual methods. Staff, equipped with GPS devices, conduct on-site surveys, recording the location, type, size, and status of each emergency space. Verifying a single tailings dam typically takes several days and poses safety risks to personnel. Furthermore, different investigators use different recording methods, resulting in inconsistent data formats and completeness, making it difficult to create standardized verification archives and effective traceability evidence. While research has explored using drone aerial photography for emergency space identification in recent years, existing solutions largely rely on offline processing after transmitting raw images back to the ground, making it difficult to output verification conclusions in real-time during flight. They also lack dedicated comparison logic for verifying the geometric parameters of tailings dam emergency spaces. In addition, there is a lack of automatic fusion mechanisms between structured attribute data collected by mobile terminals and geometric parameter data acquired by drones, preventing the formation of an integrated verification process combining ground data collection and aerial verification. Therefore, there is an urgent need to develop a technical solution that can rapidly, systematically, and collaboratively complete on-site verification of tailings dam emergency spaces. Summary of the Invention

[0004] The purpose of this invention is to provide a method, system, and device for rapid on-site verification of emergency space in tailings dams, in order to solve the technical problems of low efficiency and difficulty in automatically linking and integrating ground-collected data with aerial verification data caused by the reliance on manual methods in existing tailings dam emergency space verification.

[0005] The present invention achieves the above objectives through the following technical solutions: Firstly, this invention proposes a method for rapid on-site verification of emergency spaces in tailings dams. The method utilizes ground data collection via mobile terminals and aerial verification via drones equipped with edge computing devices. The method includes: S1: Identify the emergency spaces within a set range on both banks of the downstream river channel of the tailings dam as candidate objects, extract the type label, boundary coordinates, size parameters and shape features of each candidate object, and construct a remote sensing pre-identification database; S2: Run the structured data collection form through the mobile terminal, input on-site verification data for each candidate object and generate ground verification results. Each ground verification record is associated with a unique ID, location coordinates and on-site photos. S3: The positioning coordinates obtained in real time by the mobile terminal are spatially compared with the boundary of the candidate object. When the positioning coordinates enter the set range based on the boundary of the candidate object, the corresponding pre-identification information is pushed to guide the verification, and the pre-identification label of the candidate object is corrected according to the on-site verification results. S4: The edge computing device performs real-time image reasoning and geometric parameter extraction on each candidate object, compares it with preset parameters, and outputs the airborne verification results in a hierarchical manner. S5: The airborne verification results and the ground verification results are fused by matching the location coordinates to the corresponding UIDs, a comprehensive capability assessment of each emergency space is performed, and a recommended sequence of emergency response priorities is output.

[0006] Furthermore, step S1 specifically includes: S101: According to the classification standard for environmental emergency spaces in the event of sudden water pollution in tailings dams, environmental emergency spaces are divided into interception facilities, storage spaces, diversion facilities, treatment facilities, and monitoring facilities. S102: Based on remote sensing images and digital elevation models, a method combining deep learning target detection with DEM-based D8 confluence analysis and topographic hydrological analysis using topographic humidity index (TWI) is adopted to identify the emergency space in the downstream river channel of the tailings dam and within a set range on both banks. S103: Extract the type labels, boundary coordinates, size parameters and shape features of each emergency space to construct a remote sensing pre-identification database; S104: A weighted fusion method is used to comprehensively evaluate the confidence level of each identification result, and the pre-identification fusion score of each emergency space is obtained. The calculation formula is: ; In the formula, C DL For deep learning detection confidence, N(TWI) is the normalized TWI value, and N(A) is the TWI value. c () represents the normalized cumulative inflow. + + =1; S105: Based on the pre-identification fusion score, emergency spaces with scores higher than a set threshold are selected and retained as candidate objects and stored in the remote sensing pre-identification database for use by mobile terminals when performing spatial comparisons and as reference data when edge computing devices load preset parameters.

[0007] Furthermore, step S2 specifically includes: S201: Deploy structured forms on mobile terminals. Each verification record is automatically generated with a unique UID. The encoding rule is "tailings dam number + location number + timestamp". Real-time location coordinates are automatically collected and associated with the UID. The structured forms include a tailings dam site survey form, a downstream environmentally sensitive target investigation form, and an emergency space site verification form, which are used to input basic information about the tailings dam, sensitive target information, and emergency space site information, respectively. S202: After the data entry is completed, a ground verification result is generated. The ground verification result includes the UID, real-time positioning coordinates, on-site photos, and corresponding structured form data of each candidate object.

[0008] Furthermore, step S3 specifically includes: S301: The mobile terminal obtains GPS coordinates in real time as real-time positioning coordinates and performs spatial comparison with the boundary coordinates of candidate objects in the remote sensing pre-identification database; S302: When the real-time positioning coordinates are within the boundary range of any candidate object or are less than or equal to a set distance threshold from its boundary, the pre-identification information of the candidate object is pushed. The pre-identification information includes type label, estimated size and shape features. S303: Provide three options for investigators to confirm or correct the pre-identified label: "Confirmed Correct", "Needs Correction", and "Does Not Exist". If it is determined to be "Does Not Exist", it will be marked as deleted from the candidate. S304: For emergency spaces not covered by the pre-identified database, create a new verification record and enter attribute information through a mobile terminal. The data source is marked as "new on-site verification". The new emergency space is added to the candidate list.

[0009] Furthermore, step S4 specifically includes: S401: The multispectral camera on the UAV acquires real-time image frames. The high-precision timestamp module binds the camera shutter event with the output data of the inertial measurement unit and the global positioning system to obtain the UAV pose data corresponding to the frame. S402: The edge computing device runs a point-quantized lightweight deep learning model to perform real-time inference on image frames; S403: Inversely map the emergency space targets identified by reasoning from the image coordinate system to the geographic coordinate system to obtain the actual geographic boundaries and physical dimensions of each emergency space. The inverse mapping equation is as follows: ; in,( , , () represents the target geographic coordinates of the emergency space. The rotation matrix constructed for the roll angle, pitch angle, and yaw angle of the UAV. , (X0, Y0, Z0) represents the image pixel coordinates, K represents the camera intrinsic parameter matrix, R represents the rotation matrix constructed from the flight control pose, d represents the depth from the pixel to the target plane, and (X0, Y0, Z0) represents the UAV position coordinates. S404: Based on the real-time location of the UAV, dynamically load a subset of preset parameters corresponding to the emergency space from the remote sensing pre-identification database using the R-tree spatial index; S405: Compare the actual geographic boundary and physical dimensions with the preset parameter subset, calculate the size deviation and shape overlap, and calculate the comprehensive deviation index based on the weighted average of the size deviation and shape overlap. S406: Classify the levels according to the comprehensive deviation index, and dynamically select the corresponding communication protocol to transmit the airborne verification results to the mobile terminal; The airborne verification results include the actual geographical boundaries, physical dimensions, and deviation levels of each emergency space, and are matched to the corresponding ground verification record UID via GPS coordinates.

[0010] Furthermore, the dimensional deviation in step S405 And shape overlap ratio through intersection-union ratio The overall deviation index D is obtained by the following formula: ; ; D = w1·| | + w2·(1-IoU) + w3·d H in, The actual size for airborne identification. For preset size, This represents the actual boundary region for airborne identification. For the preset boundary region, d H The normalized Hausdorff distance is represented by w1, w2, and w3, which are weighting coefficients.

[0011] Furthermore, step S5 specifically includes: Using a coordinate distance less than a set threshold as the criterion for determining the same emergency space, the actual geographical boundary and physical dimensions in the airborne verification results are attached to the UID of the ground verification record with the same location coordinates to form a merged emergency space archive. The integrated emergency spaces are comprehensively evaluated based on their spatial capacity, accessibility, applicability, and reliability to obtain a comprehensive capability score. As shown in the following formula: ; in, Let i be the weight of the i-th evaluation dimension. Let be the number of specific indicators included in the i-th evaluation dimension. Let be the weight of the j-th indicator in the i-th dimension. Let be the score value of the j-th indicator under the i-th dimension; Based on a multi-objective optimization algorithm, with the objective functions of maximizing emergency response effectiveness, minimizing response costs, and minimizing environmental impact, and with the upper limit of emergency space capacity and pollutant transport time as constraints, a recommended sequence of emergency response priorities is output.

[0012] Secondly, the present invention proposes an edge computing device for an unmanned aerial vehicle (UAV) platform, applied to implement the aerial verification step in the method described above. The device is mounted on the UAV platform and includes: The image acquisition unit is used to carry a multispectral camera to acquire real-time image frames. It binds the camera shutter events with the output of the inertial measurement unit and the global positioning system through a high-precision timestamp module, and outputs real-time image frames with spatiotemporal labels. The edge computing unit is used to receive the real-time image frames, run a lightweight deep learning model with fixed-point quantization for real-time inference, identify emergency space targets in the image, perform image-geographic coordinate inverse mapping, transform the identified emergency space targets from the image coordinate system to the geographic coordinate system, and output the actual geographic boundaries and physical dimensions of each emergency space. Storage unit, used to store a subset of emergency spatial preset parameters loaded from the remote sensing pre-identification database; The comparison unit is used to receive the actual geographic boundary and physical size, as well as the preset parameters, perform joint comparison, and output the comparison result; A communication unit is used to output airborne verification results in a hierarchical manner based on the comparison results, and to transmit the airborne verification results to the mobile terminal; The UID association module is used to match the airborne verification results with the ground verification results to the corresponding UIDs through location coordinates for fusion, to conduct a comprehensive capability assessment of each emergency space, and to output a recommended sequence of emergency response priorities.

[0013] Thirdly, this invention proposes a rapid on-site verification system for emergency spaces in tailings dams, applied to implement the method described above. The system includes: Mobile terminals and edge computing devices mounted on drone platforms; The mobile terminal includes: The data acquisition unit is used to run structured data acquisition forms, input on-site verification data for each candidate object, and associate each verification record with a unique ID, location coordinates and on-site photos to generate ground verification results. The spatial comparison unit is used to spatially compare the real-time acquired positioning coordinates with the boundary of the candidate object. When the positioning coordinates enter the set range based on the boundary of the candidate object, the corresponding pre-identification information is pushed to guide the verification. The fusion assessment unit is used to receive the airborne verification results transmitted by the edge computing device, fuse the airborne verification results with the ground verification results by matching the location coordinates to the corresponding UID, perform a comprehensive capability assessment of each emergency space, output a recommended sequence of emergency response priorities, and generate a verification report. The edge computing device is used to perform over-the-air verification steps and transmit the airborne verification results to the mobile terminal.

[0014] The beneficial effects of this invention are as follows: (1) The ground mobile terminal of the present invention reduces the on-site verification time of a single tailings dam from 2 to 3 days to several hours by navigating through structured forms and pre-identified databases; the UAV completes aerial verification of multiple emergency spaces at one time along a preset route without taking off and landing separately for each space; the airborne edge computing outputs verification conclusions in real time during flight without waiting for images to be transmitted back to the ground for offline processing, thus significantly shortening the time from data collection to the output of verification conclusions.

[0015] (2) In this invention, the ground verification record is bound to the on-site photos and positioning coordinates through UID, and the airborne verification result is matched to the corresponding ground UID through GPS coordinates, so that the attribute data and geometric parameters of the same emergency space are automatically merged into a complete record without manual matching, thus avoiding the problem of two types of data being recorded separately and difficult to match later in the traditional method.

[0016] (3) In this invention, the shape-quantity joint comparison quantifies the actual state of the emergency space from two dimensions: size deviation and shape overlap. The deviation level directly informs the investigators whether the space has changed and the degree of change, providing quantifiable data basis for the preparation and updating of emergency plans, and avoiding the subjectivity of human judgment. Attached Figure Description

[0017] Figure 1This is a flowchart illustrating a rapid on-site verification method for emergency space in tailings dams, as described in one embodiment of the present invention. Figure 2 This is another flowchart illustrating a rapid on-site verification method for emergency tailings dam spaces according to an embodiment of the present invention. Figure 3 This is a structural block diagram of an unmanned aerial vehicle (UAV) onboard edge computing device in one embodiment of the present invention; Figure 4 This is a block diagram of the overall architecture of a rapid on-site verification system for emergency tailings dam spaces, according to one embodiment of the present invention. Detailed Implementation

[0018] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0019] Example 1

[0020] Please see Figure 1 and Figure 2 This embodiment provides a method for rapid on-site verification of emergency spaces in tailings dams. The method is based on ground data collection using mobile terminals and aerial verification using UAVs equipped with edge computing devices. The method includes: S1: Identify the emergency spaces within a defined range on both banks of the downstream river channel of the tailings dam as candidate objects, extract the type label, boundary coordinates, size parameters and shape features of each candidate object, and construct a remote sensing pre-identification database.

[0021] In this embodiment, emergency space refers to various facilities and terrain spaces within a certain range on both banks of the downstream river channel of the tailings dam, which can be used to intercept, store, divert, dispose of, or monitor pollutants in the event of a sudden water pollution incident. These include specific objects such as dams, flood storage areas, diversion channels, emergency pools, and water quality monitoring sections. Candidate objects refer to targets pre-identified through remote sensing imagery and digital elevation models that are preliminarily determined to have emergency space potential and require on-site verification; that is, a set of emergency space candidates selected and retained from all identification results. The remote sensing pre-identification database is a data set formed by storing the type labels, boundary coordinates, size parameters, shape characteristics, and other information of the aforementioned candidate objects in a unified data format. This database is used, on the one hand, for mobile terminals to perform spatial comparison and navigation push during on-site verification; and on the other hand, it serves as a reference data source for loading preset parameters when the UAV edge computing device performs aerial verification.

[0022] S2: Run a structured data collection form via a mobile terminal to enter on-site verification data for each candidate object and generate ground verification results. Each ground verification record is associated with a unique ID, location coordinates, and on-site photos.

[0023] In this step, the mobile terminal has three types of structured data collection forms built-in. The first type is the tailings dam site survey form, which includes fields such as tailings dam name, number, affiliated company, administrative division, geographical coordinates, mineral type, form of entry into the dam, capacity, dam height, grade, dam type, operating status, characteristic pollutants, wastewater treatment process, location of discharge outlet, safety level, and emergency plan filing. The second type is the downstream environmentally sensitive target investigation form, which includes fields such as associated tailings dam number, sensitive target name, type, grade, location coordinates, distance from tailings dam clarification, associated water body, water flow direction, flow rate and velocity, water quality target, service population, water supply scale, and protection requirements. The third type is the emergency space site verification form, which includes fields such as number, associated tailings dam, name, type, location coordinates, river section, distance from tailings dam, capacity, area, length, usage status, ownership, emergency purpose, accessibility, data source, close-up photos, panoramic photos, drone aerial photos, and remarks.

[0024] Each verification record generates a unique UID, encoded using the tailings dam number, location number, and timestamp. Simultaneously, the current GPS coordinates are collected and associated with this UID as positioning coordinates. After data entry, ground verification results are generated, including the UID, positioning coordinates, on-site photos, and corresponding structured form data for each candidate object.

[0025] S3: Spatial comparison is performed between the real-time location coordinates obtained by the mobile terminal and the boundary of the candidate object. When the location coordinates enter the set range based on the boundary of the candidate object, the corresponding pre-identification information is pushed to guide the verification, and the pre-identification label of the candidate object is corrected according to the on-site verification results.

[0026] In this step, the mobile terminal acquires its current GPS location at a fixed frequency as real-time positioning coordinates and continuously compares them spatially with the boundary coordinates of candidate objects in the remote sensing pre-identification database. The real-time positioning coordinates are the spatial coordinates of the investigator's current location. When these coordinates are within the boundary of a candidate object or less than or equal to a set threshold, it is considered to have entered the designated range. At this time, the mobile terminal automatically pushes the pre-identification information of the candidate object, guiding the investigator to that location for on-site verification. After arriving at the site, the investigator confirms or corrects the pre-identification labels according to the actual situation, including confirming the type label is correct, correcting the type label, marking it as non-existent, etc. Simultaneously, new emergency spaces not covered by the pre-identification database can be entered.

[0027] Through the above guidance mechanism, investigators can complete the verification in sequence without having to memorize the location of each candidate space in advance, reducing the time spent on navigation and searching.

[0028] S4: Real-time image reasoning and geometric parameter extraction are performed on each candidate object through edge computing devices, and the results are compared with preset parameters to output the airborne verification results in a hierarchical manner.

[0029] In this step, as the UAV flies along a preset route, its onboard multispectral camera continuously acquires ground image frames. An edge computing device runs a quantized deep learning model to perform real-time target detection on these image frames, identifying emergency space targets within the images. Pixel coordinates are then converted to geographic coordinates through coordinate inverse mapping, yielding the actual geographic boundaries and physical dimensions of each emergency space. Subsequently, the actual values ​​are compared to the preset boundary coordinates and dimensions of the emergency space stored in the remote sensing pre-identification database to calculate the degree of deviation. Based on the magnitude of the deviation, different levels are categorized, with different communication transmission strategies corresponding to different levels, ensuring that verification results with larger deviations are prioritized for transmission to ground mobile terminals. The airborne verification results include the actual geographic boundaries, physical dimensions, and deviation levels of each emergency space, and are matched to the corresponding ground verification record UID using GPS coordinates.

[0030] S5: The airborne verification results and the ground verification results are matched with the corresponding UIDs through location coordinates and then integrated to conduct a comprehensive capability assessment of each emergency space and output a recommended sequence of emergency response priorities.

[0031] Understandably, the priority recommendation sequence refers to a list of spaces to be activated in descending order of their comprehensive scores. This sequence indicates to decision-makers which emergency spaces should be prioritized to respond to sudden water pollution incidents, thus providing them with a clear basis for space selection within a limited emergency response time.

[0032] It should be noted that, in this invention, the mobile terminal refers to a portable device carried by investigators, equipped with GPS positioning, photography, data entry, and wireless communication functions. It is used to achieve structured data entry for on-site verification, spatial comparison navigation, and photo acquisition and association. Examples include smartphones, tablets, and industrial-grade handheld data acquisition terminals. The edge computing device refers to an embedded computing device mounted on a drone platform, capable of processing aerial images in real time during flight, completing tasks such as target recognition, geometric parameter extraction, and deviation comparison without transmitting the original images back to the ground.

[0033] In one implementation, step S1 specifically includes: S101: According to the classification standard for environmental emergency spaces in the event of sudden water pollution from tailings dams, environmental emergency spaces are divided into interception facilities, storage and retention spaces, diversion facilities, treatment facilities, and monitoring facilities. Interception facilities include subcategories such as dams, control gates, and rubber dams; storage and retention spaces include subcategories such as flood storage areas, reservoirs, and ponds; diversion facilities include subcategories such as diversion channels, flood diversion channels, and water diversion canals; treatment facilities include subcategories such as emergency pools, sedimentation pools, and adsorption dams; and monitoring facilities include subcategories such as automatic water quality monitoring stations and manual monitoring sections.

[0034] S102: Based on remote sensing imagery and digital elevation models (DEMs), a method combining deep learning-based target detection with DEM-based D8 runoff analysis and topographic hydrological analysis using the topographic humidity index (TWI) is employed to identify emergency spaces within a designated area on both banks of the downstream river channel of the tailings dam. Deep learning-based target detection utilizes a Faster R-CNN network to detect and identify visible ground features such as dams, flood detention areas, and emergency pools in the remote sensing imagery. Topographic hydrological analysis employs DEM-based D8 runoff analysis to calculate the cumulative runoff volume of each grid cell, and combines this with the topographic humidity index (TWI) to identify topographic depressions and runoff areas with water retention functions. The formula for calculating the topographic humidity index (TWI) is as follows: ; In the formula, α is the upstream catchment area per unit contour line length, and β is the local slope. The higher the TWI value, the easier it is for the area to accumulate water, and the more suitable the terrain conditions are for it to serve as a water retention space.

[0035] S103: Extract the type labels, boundary coordinates, size parameters and shape features of each emergency space to construct a remote sensing pre-identification database; S104: A weighted fusion method is used to comprehensively evaluate the confidence level of each identification result, and the pre-identification fusion score of each emergency space is obtained. The calculation formula is: ; In the formula, C DL For deep learning detection confidence, N(TWI) is the normalized TWI value, and N(A) is the TWI value. c () represents the normalized cumulative inflow. + + =1.

[0036] This fusion approach allows for the supplementation of ground features that are detected by deep learning but may be missed by topographic and hydrological analysis. Potential stagnant areas identified by topographic and hydrological analysis can also be confirmed by deep learning results, thereby improving the recall and accuracy of pre-identification.

[0037] S105: Based on the pre-identification fusion score, emergency spaces with scores higher than a set threshold are selected and retained as candidate objects and stored in the remote sensing pre-identification database. These are used for spatial comparison by mobile terminals and as reference data when edge computing devices load preset parameters.

[0038] In one implementation, step S2 specifically includes: S201: Deploy structured forms on mobile terminals. Each verification record automatically generates a unique number UID. The encoding rule is "tailings dam number + location number + timestamp". Real-time location coordinates are automatically collected and associated with the UID.

[0039] The structured forms include a tailings dam site survey form, a downstream environmentally sensitive target investigation form, and an emergency space site verification form, which are used to enter basic information about the tailings dam, sensitive target information, and emergency space site information, respectively.

[0040] Each verification record is automatically generated with a unique UID upon creation. The UID is automatically generated by the system according to the rule of tailings dam number plus location number plus timestamp, used to uniquely identify each verification record and serve as the primary key for data association in subsequent fusion steps. The mobile terminal also automatically collects the current GPS coordinates as positioning coordinates and associates them with this UID. The structured data entered by investigators and the on-site photos taken are all bound to this UID to ensure the data integrity and traceability of each verification record. During subsequent fusion, all data in the same emergency space can be quickly retrieved and associated using the UID.

[0041] S202: After data entry is complete, a ground verification result is generated. This result includes the UID, location coordinates, on-site photos, and corresponding structured form data for each candidate object. The ground verification result, as a set of structured attribute data collected on the ground side, will be integrated with the airborne verification result in subsequent steps.

[0042] In one implementation, step S3 specifically includes: S301: The mobile terminal obtains GPS coordinates in real time as real-time positioning coordinates and performs spatial comparison with the boundary coordinates of candidate objects in the remote sensing pre-identification database; S302: When the real-time positioning coordinates are within the boundary of any candidate object or are less than or equal to a set distance threshold from its boundary, push the pre-identification information of the candidate object. The pre-identification information includes type label, estimated size and shape features. S303: Provides three options for investigators to confirm or correct the pre-identified label: "Confirmed Correct", "Needs Correction", and "Does Not Exist". If it is determined to be "Does Not Exist", it will be marked as deleted from the candidate. S304: For emergency spaces not covered by the pre-identified database, create a new verification record and enter attribute information through a mobile terminal. The data source is marked as "new on-site verification". The new emergency space is added to the candidate list.

[0043] In one implementation, step S4 specifically includes: S401: The multispectral camera on the UAV acquires real-time image frames. The high-precision timestamp module binds the camera shutter event with the output data of the inertial measurement unit and the global positioning system to obtain the UAV pose data corresponding to the frame. The high-precision timestamp module can be implemented based on the IEEE 1588 PTP protocol to ensure that each frame of the image has a precise corresponding UAV position coordinates and attitude angle.

[0044] S402: The edge computing device runs a point-quantized lightweight deep learning model to perform real-time inference on image frames; S403: Inversely map the emergency space targets identified by reasoning from the image coordinate system to the geographic coordinate system to obtain the actual geographic boundaries and physical dimensions of each emergency space. The inverse mapping equation is as follows: ; in,( , , () represents the target geographic coordinates of the emergency space. The rotation matrix constructed for the roll angle, pitch angle, and yaw angle of the UAV. , (X0, Y0, Z0) represents the image pixel coordinates, K represents the camera intrinsic parameter matrix, R represents the rotation matrix constructed from the flight control pose, d represents the depth from the pixel to the target plane, and (X0, Y0, Z0) represents the UAV position coordinates. This inverse mapping converts the target detection boxes identified in the pixel domain into spatial objects in the geographic coordinate system. Each object contains geometric parameters such as type label, actual geographic boundary, and physical size.

[0045] S404: Based on the real-time location of the UAV, a subset of preset parameters for the corresponding emergency space is dynamically loaded from the remote sensing pre-identification database through the R-tree spatial index; with the current location of the UAV as a reference, only the emergency space parameters within the range of the forward flight path are loaded, while the parameters of the areas already flown over are released, so as to avoid the onboard memory being filled with all the data and ensure that the edge computing device operates stably with limited storage resources.

[0046] S405: Compare the actual geographic boundaries and physical dimensions with a preset parameter subset, calculate the size deviation and shape overlap, and calculate the comprehensive deviation index based on the weighted average of the size deviation and shape overlap. S406: Classify the levels according to the comprehensive deviation index, and dynamically select the corresponding communication protocol to transmit the airborne verification results to the mobile terminal; The airborne verification results include the actual geographical boundaries, physical dimensions, and deviation levels of each emergency space, and are matched to the corresponding ground verification record UID via GPS coordinates.

[0047] In one implementation, the dimensional deviation in step S405 And shape overlap ratio through intersection-union ratio The overall deviation index D is obtained by the following formula: ; ; D = w1·| | + w2·(1-IoU) + w3·d H in, The actual size for airborne identification. For preset size, This represents the actual boundary region for airborne identification. For the preset boundary region, d H The normalized Hausdorff distance is represented by w1, w2, and w3, which are weighting coefficients.

[0048] In one implementation, step S5 specifically includes: Using a coordinate distance less than a set threshold as the criterion for determining the same emergency space, the actual geographical boundary and physical dimensions in the airborne verification results are attached to the UID of the ground verification record with the same location coordinates to form a merged emergency space archive. The integrated emergency spaces are comprehensively evaluated based on their spatial capacity, accessibility, applicability, and reliability to obtain a comprehensive capability score. As shown in the following formula: ; in, Let i be the weight of the i-th evaluation dimension. Let be the number of specific indicators included in the i-th evaluation dimension. Let be the weight of the j-th indicator in the i-th dimension. Let be the score value of the j-th indicator under the i-th dimension; Among them, space capacity reflects the total amount of sewage or pollutants that the emergency space can hold, which directly determines the upper limit of the scale of emergency response; accessibility reflects whether the emergency space has good transportation and passage conditions, affecting the speed at which emergency personnel and equipment arrive at the scene; applicability reflects the degree of matching between the type of emergency space and the type of pollutants; and reliability reflects the structural integrity and maintenance status of the emergency space, affecting whether it can be safely activated in emergency scenarios.

[0049] Based on a multi-objective optimization algorithm, this method aims to maximize the effectiveness of emergency response, minimize response costs, and minimize environmental impact, while using the upper limit of emergency space capacity and pollutant transport time as constraints. The resulting algorithm outputs a recommended sequence of emergency response priorities. This implementation employs the NSGA-II multi-objective optimization algorithm, which yields a set of Pareto optimal solutions through multi-objective optimization. These solutions allow decision-makers to select the optimal combination of emergency response spaces based on their specific scenario preferences.

[0050] According to the above embodiments, this invention collects ground verification data such as attribute information, positioning coordinates, and on-site photos of the emergency space via mobile terminals, and constructs a remote sensing pre-identification database to guide investigators in completing on-site verification. Simultaneously, an UAV equipped with an edge computing device performs real-time image inference, coordinate inverse mapping, and geometric parameter extraction of the emergency space during flight. The extracted actual dimensions and boundaries are compared with preset parameters in the database, outputting the deviation level and verification results. Finally, the ground-collected data and the airborne verification results are fused by matching coordinates to the same UID to form a complete emergency space archive. These technical means solve the technical problem of the separation between ground-collected data and aerial verification data in existing tailings dam emergency space verification, achieving integrated data fusion of ground collection and aerial verification.

[0051] Example 2

[0052] Please see Figure 3 This embodiment provides an edge computing device for a drone platform. The device is mounted on the drone platform and is used to perform the over-the-air verification step as described in Embodiment 1, including: The image acquisition unit is equipped with a multispectral camera, which includes RGB and near-infrared bands, enabling it to acquire aerial images under various lighting conditions. A high-precision timestamp module binds camera shutter events to the outputs of the inertial measurement unit and the global positioning system, achieving synchronization accuracy better than 1 millisecond based on the IEEE 1588 PTP protocol. Real-time image frames with spatiotemporal tags are then output to the edge computing unit.

[0053] The edge computing unit is used to receive real-time image frames, run a lightweight deep learning model with INT8 fixed-point quantization for real-time inference, perform image-geographic coordinate inverse mapping, and output the actual geographic boundaries and physical dimensions of each emergency space.

[0054] The storage unit is used to store a subset of preset parameters for emergency spaces loaded from the remote sensing pre-identification database, including the type label, preset boundary coordinates, preset size thresholds, and preset shape feature parameters for each emergency space.

[0055] The comparison unit is used to receive the actual geographic boundary and physical dimensions, as well as preset parameters, calculate the size deviation, crossover ratio and comprehensive deviation index, and output the comparison results; The communication unit is used to dynamically select the communication protocol based on the deviation level in the comparison results to transmit the airborne verification results to the mobile terminal. The UID association module is used to match the airborne verification results with the corresponding UID generated by the mobile terminal through coordinate matching, thereby realizing the association between ground and air data. This module uses GPS coordinate distance as the threshold for determining the same emergency space, and writes the actual geographical boundaries and physical dimensions in the airborne verification results into the ground verification record of the corresponding UID.

[0056] In practical applications, the aforementioned device is mounted on a drone platform and flies along a preset route. During flight, the image acquisition unit continuously acquires aerial images of the ground, while the edge computing unit performs target recognition and coordinate mapping in real time on the airborne end. This eliminates the need to wait for images to be transmitted back to the ground for processing, allowing the output of the actual geographical boundaries and physical dimensions of each emergency space. The comparison unit calculates the deviation based on preset parameters in the database, and the communication unit selects the corresponding communication strategy for transmission according to the deviation level. Larger deviations result in higher transmission priority, ensuring that information on significantly changing emergency spaces is delivered to the ground in a timely manner. The UID association module merges the airborne verification results into the corresponding ground verification records through coordinate matching, automatically associating aerial verification data and ground-acquired data for the same emergency space without manual matching. This device enables the drone to complete the geometric parameter verification of multiple emergency spaces in a single flight, reducing the number of times drones or manual re-surveys are required in traditional methods.

[0057] Example 3

[0058] Please see Figure 4 This embodiment provides a rapid on-site verification system for emergency tailings dam spaces. The system includes a mobile terminal and an edge computing device. In this invention, the mobile terminal may also be referred to as a ground mobile terminal to emphasize its role in performing data acquisition and processing functions on the ground side.

[0059] The system includes: a mobile terminal and an edge computing device mounted on a drone platform; Mobile terminals include: The data acquisition unit is used to run structured data acquisition forms, input on-site verification data for each candidate object, and associate each verification record with a unique ID, location coordinates and on-site photos to generate ground verification results. The spatial comparison unit is used to spatially compare the real-time acquired positioning coordinates with the boundary of the candidate object. When the positioning coordinates enter the set range based on the boundary of the candidate object, the corresponding pre-identification information is pushed to guide the verification. The fusion assessment unit is used to receive the airborne verification results transmitted by the edge computing device, merge the airborne verification results with the ground verification results by matching the location coordinates to the corresponding UID, conduct a comprehensive capability assessment of each emergency space, output a recommended sequence of emergency response priorities, and generate a verification report. Edge computing devices are used to perform over-the-air verification procedures and transmit the airborne verification results to mobile terminals.

[0060] To more clearly illustrate the present invention and its advantages, the method provided by the present invention will be further explained below in conjunction with specific embodiments.

[0061] Taking a copper mine tailings dam as an example, this tailings dam is located upstream of a river, with a centralized drinking water source within 5 kilometers downstream. The method of this invention is used for rapid on-site verification of emergency spaces. The specific process is as follows: Pre-identification stage: Based on GF-2 remote sensing imagery (0.8 meters resolution) and 1:10000 digital line maps, a method combining deep learning target detection and topographic and hydrological analysis was adopted to pre-identify 23 candidate emergency spaces within a 1-kilometer radius of the downstream river channel and both banks of the tailings dam. These include 5 interception facilities, 8 storage spaces, 4 diversion facilities, 3 disposal facilities, and 3 monitoring facilities. A remote sensing pre-identification database was then constructed.

[0062] Ground data collection phase: Investigators used a mobile terminal verification application to sequentially enter basic information about the tailings dam, information about downstream environmentally sensitive targets, and on-site verification information for emergency spaces according to three types of structured forms. Each record automatically generated a UID and collected WGS84 coordinates. As investigators moved along the river, the mobile terminal compared their GPS coordinates with the boundary coordinates of candidate objects in the pre-identification database in real time, automatically pushing nearby candidate emergency space information to guide investigators to verify them one by one. On-site verification confirmed 18 locations with correct pre-identification types, corrected type labels for 3 locations, and added 2 emergency spaces not included in the pre-identification database. Investigators took close-up and panoramic photos at each emergency space site, and the system automatically bound the photos to the corresponding UIDs, linking a total of 156 photos.

[0063] Aerial verification phase: The drone flew along a preset route for 45 minutes. The onboard edge computing device performed image inference and geometric parameter extraction in real time, identifying and verifying a total of 25 emergency spaces (including 2 newly added ground-based spaces). Among them, the actual boundaries of 2 flood storage and detention areas deviated from the preset boundaries by more than 50%, triggering a Level 1 alarm; 5 spaces had deviations of 10% to 50%, triggering a Level 2 alarm; the remaining 18 spaces were in normal condition and were recorded as ordinary records. The verification results were transmitted to the mobile terminal in real time.

[0064] In the fusion assessment phase, mobile terminals merged ground-collected data with airborne verification data via UIDs, performed AHP-weighted comprehensive capability scoring on 25 emergency spaces, and output priority recommendation sequences based on the NSGA-II model. The recommendation sequences showed that Flood Detention Area No. 1 had a comprehensive score of 0.87, and Dam No. 3 had a comprehensive score of 0.82, making them the preferred spaces for emergency response.

[0065] In this embodiment, the total time from pre-identification data preparation to final verification report generation is approximately 6 hours, including 1 hour for pre-identification data preparation, 3 hours for ground data collection, 1 hour for UAV verification, and 1 hour for assessment and report generation. Compared to the traditional manual verification cycle of 2 to 3 days, this method significantly shortens the verification cycle and provides unified and traceable data formats, effectively supporting the development of emergency plans with a single database, policy, and map.

[0066] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated.

[0067] The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0068] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0069] In addition, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0070] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for rapid on-site verification of emergency spaces in tailings dams, characterized in that, The method is based on ground data collection using mobile terminals and aerial verification using edge computing devices mounted on UAVs. The method includes: S1: Identify the emergency spaces within a set range on both banks of the downstream river channel of the tailings dam as candidate objects, extract the type label, boundary coordinates, size parameters and shape features of each candidate object, and construct a remote sensing pre-identification database; S2: Run the structured data collection form through the mobile terminal, input on-site verification data for each candidate object and generate ground verification results. Each ground verification record is associated with a unique ID, location coordinates and on-site photos. S3: Spatial comparison of the location coordinates obtained in real time by the mobile terminal with the boundary of the candidate object. When the location coordinates enter the set range based on the boundary of the candidate object, push the corresponding pre-identification information to guide the verification, and correct the pre-identification label of the candidate object according to the on-site verification results. S4: The edge computing device performs real-time image reasoning and geometric parameter extraction on each candidate object, compares it with preset parameters, and outputs the airborne verification results in a hierarchical manner. S5: The airborne verification results and the ground verification results are fused by matching the location coordinates to the corresponding UIDs, a comprehensive capability assessment of each emergency space is performed, and a recommended sequence of emergency response priorities is output.

2. The method for rapid on-site verification of emergency space in tailings dams according to claim 1, characterized in that, Step S1 specifically includes: S101: According to the classification standard for environmental emergency spaces in the event of sudden water pollution in tailings dams, environmental emergency spaces are divided into interception facilities, storage spaces, diversion facilities, treatment facilities, and monitoring facilities. S102: Based on remote sensing images and digital elevation models, a method combining deep learning target detection with DEM-based D8 confluence analysis and topographic hydrological analysis using topographic humidity index (TWI) is adopted to identify the emergency space in the downstream river channel of the tailings dam and within a set range on both banks. S103: Extract the type labels, boundary coordinates, size parameters and shape features of each emergency space to construct a remote sensing pre-identification database; S104: A weighted fusion method is used to comprehensively evaluate the confidence level of each identification result, and the pre-identification fusion score of each emergency space is obtained. The calculation formula is: ； In the formula, C DL For deep learning detection confidence, N(TWI) is the normalized TWI value, and N(A) is the TWI value. c () represents the normalized cumulative inflow. + + =1; S105: Based on the pre-identification fusion score, emergency spaces with scores higher than a set threshold are selected and retained as candidate objects and stored in the remote sensing pre-identification database for use by mobile terminals when performing spatial comparisons and as reference data when edge computing devices load preset parameters.

3. The method for rapid on-site verification of emergency space in tailings dams according to claim 1, characterized in that, Step S2 specifically includes: S201: Deploy structured forms on mobile terminals. Each verification record is automatically generated with a unique UID. The encoding rule is "tailings dam number + location number + timestamp". Real-time location coordinates are automatically collected and associated with the UID. The structured forms include a tailings dam site survey form, a downstream environmentally sensitive target investigation form, and an emergency space site verification form, which are used to input basic information about the tailings dam, sensitive target information, and emergency space site information, respectively. S202: After the data entry is completed, a ground verification result is generated. The ground verification result includes the UID, real-time positioning coordinates, on-site photos, and corresponding structured form data of each candidate object.

4. The method for rapid on-site verification of emergency space in tailings dams according to claim 1, characterized in that, Step S3 specifically includes: S301: The mobile terminal obtains GPS coordinates in real time as real-time positioning coordinates and performs spatial comparison with the boundary coordinates of candidate objects in the remote sensing pre-identification database; S302: When the real-time positioning coordinates are within the boundary range of any candidate object or are less than or equal to a set distance threshold from its boundary, the pre-identification information of the candidate object is pushed. The pre-identification information includes type label, estimated size and shape features. S303: Provides three options for investigators to confirm or correct the pre-identified label: "Confirmed Correct", "Needs Correction", and "Does Not Exist". If it is determined to be "Does Not Exist", it will be marked for deletion from the candidate. S304: For emergency spaces not covered by the pre-identified database, create a new verification record and enter attribute information via a mobile terminal. The data source is marked as "new on-site verification". The new emergency space is then added to the candidate list.

5. A method for rapid on-site verification of emergency space in tailings dams according to claim 1, characterized in that, Step S4 specifically includes: S401: The multispectral camera on the UAV acquires real-time image frames. The high-precision timestamp module binds the camera shutter event with the output data of the inertial measurement unit and the global positioning system to obtain the UAV pose data corresponding to the frame. S402: The edge computing device runs a point-quantized lightweight deep learning model to perform real-time inference on image frames; S403: Inversely map the emergency space targets identified by reasoning from the image coordinate system to the geographic coordinate system to obtain the actual geographic boundaries and physical dimensions of each emergency space. The inverse mapping equation is as follows: ； in,( , , () represents the target geographic coordinates of the emergency space. The rotation matrix constructed for the roll angle, pitch angle, and yaw angle of the UAV. , (X0, Y0, Z0) represents the image pixel coordinates, K represents the camera intrinsic parameter matrix, R represents the rotation matrix constructed from the flight control pose, d represents the depth from the pixel to the target plane, and (X0, Y0, Z0) represents the UAV position coordinates. S404: Based on the real-time location of the UAV, dynamically load a subset of preset parameters corresponding to the emergency space from the remote sensing pre-identification database using the R-tree spatial index; S405: Compare the actual geographic boundary and physical dimensions with the preset parameter subset, calculate the size deviation and shape overlap, and calculate the comprehensive deviation index based on the weighted average of the size deviation and shape overlap. S406: Classify the levels according to the comprehensive deviation index, and dynamically select the corresponding communication protocol to transmit the airborne verification results to the mobile terminal; The airborne verification results include the actual geographical boundaries, physical dimensions, and deviation levels of each emergency space, and are matched to the corresponding ground verification record UID via GPS coordinates.

6. A method for rapid on-site verification of emergency space in tailings dams according to claim 5, characterized in that, Dimensional deviation in step S405 And shape overlap ratio through intersection-union ratio The overall deviation index D is obtained by the following formula: ； ； D = w1·| | + w2·(1-IoU) + w3·d H in, The actual size for airborne identification. For preset size, This represents the actual boundary region for airborne identification. For the preset boundary region, d H The normalized Hausdorff distance is represented by w1, w2, and w3, which are weighting coefficients.

7. The method for rapid on-site verification of emergency space in tailings dams according to claim 1, characterized in that, Step S5 specifically includes: Using a coordinate distance less than a set threshold as the criterion for determining the same emergency space, the actual geographical boundary and physical dimensions in the airborne verification results are attached to the UID of the ground verification record with the same location coordinates to form a merged emergency space archive. The integrated emergency spaces are comprehensively evaluated based on their spatial capacity, accessibility, applicability, and reliability to obtain a comprehensive capability score. As shown in the following formula: ； in, Let i be the weight of the i-th evaluation dimension. Let be the number of specific indicators included in the i-th evaluation dimension. Let be the weight of the j-th indicator in the i-th dimension. Let be the score value of the j-th indicator under the i-th dimension; Based on a multi-objective optimization algorithm, with the objective functions of maximizing emergency response effectiveness, minimizing response costs, and minimizing environmental impact, and with the upper limit of emergency space capacity and pollutant transport time as constraints, a recommended sequence of emergency response priorities is output.

8. An edge computing device for an unmanned aerial vehicle (UAV) platform, characterized in that, The device, used to implement the aerial verification step in the method as described in any one of claims 1-7, is mounted on an unmanned aerial vehicle (UAV) platform and includes: The image acquisition unit is used to carry a multispectral camera to acquire real-time image frames. It binds the camera shutter events with the output of the inertial measurement unit and the global positioning system through a high-precision timestamp module, and outputs real-time image frames with spatiotemporal labels. The edge computing unit is used to receive the real-time image frames, run a lightweight deep learning model with fixed-point quantization for real-time inference, identify emergency space targets in the image, perform image-geographic coordinate inverse mapping, transform the identified emergency space targets from the image coordinate system to the geographic coordinate system, and output the actual geographic boundaries and physical dimensions of each emergency space. Storage unit, used to store a subset of emergency spatial preset parameters loaded from the remote sensing pre-identification database; The comparison unit is used to receive the actual geographic boundary and physical size, as well as the preset parameters, perform joint comparison, and output the comparison result; A communication unit is used to output airborne verification results in a hierarchical manner based on the comparison results, and to transmit the airborne verification results to the mobile terminal; The UID association module is used to match the airborne verification results with the ground verification results to the corresponding UIDs through location coordinates for fusion, to conduct a comprehensive capability assessment of each emergency space, and to output a recommended sequence of emergency response priorities.

9. A rapid on-site verification system for emergency spaces in tailings dams, characterized in that, The system, used to implement the method as described in any one of claims 1-7, comprises: Mobile terminals and edge computing devices mounted on drone platforms; The mobile terminal includes: The data acquisition unit is used to run structured data acquisition forms, input on-site verification data for each candidate object, and associate each verification record with a unique ID, location coordinates and on-site photos to generate ground verification results. The spatial comparison unit is used to spatially compare the real-time acquired positioning coordinates with the boundary of the candidate object. When the positioning coordinates enter the set range based on the boundary of the candidate object, the corresponding pre-identification information is pushed to guide the verification. The fusion assessment unit is used to receive the airborne verification results transmitted by the edge computing device, fuse the airborne verification results with the ground verification results by matching the location coordinates to the corresponding UID, perform a comprehensive capability assessment of each emergency space, output a recommended sequence of emergency response priorities, and generate a verification report. The edge computing device is used to perform over-the-air verification steps and transmit the airborne verification results to the mobile terminal.

Images