Open operation area dust tracing and disposal method and system based on multi-source visual perception

Abstract

The application discloses a kind of open operation area dust tracing and disposal method and system based on multi-source visual perception, including the image information of collection operation area;Image information is perceived and analyzed, if dust generating equipment and its associated area exist dust characteristics when identifying, then trigger abnormal event;Abnormal event is dynamically analyzed, generates and issues dispatching instruction;Scheduled equipment collects the multi-view spatio-temporal data of operation area;Image information and spatio-temporal data are fused and analyzed, generate the spatio-temporal correlation evidence chain around the abnormal event, and generate trace report and carry out collaborative disposal.The application can realize the all-weather, blind area-free, automated high-precision identification and tracing of open operation area dust.

Description

Technical Field

[0001] This invention relates to the field of dust monitoring, specifically to a method and system for tracing and handling dust pollution in open work areas based on multi-source visual perception. Background Technology

[0002] At present, construction dust, open-pit mining dust, and dust from easily dispersed material storage yards have become major sources of air pollution. Construction areas, open-pit mining sites, and easily dispersed material storage yards are typical open operation areas, characterized by complex terrain, large land area, multiple operation processes, a large number of vehicles and equipment, and high dust generation intensity.

[0003] Current regulatory practices primarily rely on manual inspections, fixed-point dust sensors, or single-source video surveillance. Manual inspections are inefficient and have limited coverage; fixed-point sensors can only provide single-point concentration data and cannot pinpoint diffusion paths or sources; traditional single-source video surveillance algorithms, such as those using YOLO or optical flow methods, are ineffective at two things: first, they struggle to effectively identify the shapeless, semi-transparent dust itself; second, they cannot distinguish dust movement from other interfering movements and cannot correlate it with pollution sources. Furthermore, existing dust source tracing technologies largely depend on fixed-location sensor arrays for monitoring and estimation, lacking the ability to actively track and accurately locate mobile dust sources; while other solutions, although constructing monitoring networks, focus primarily on data correlation and prediction, failing to rapidly deploy heterogeneous equipment for multi-view on-site evidence collection and preservation after an incident.

[0004] Therefore, in order to solve the technical problems of high missed detection rate and inability to dynamically trace the source of dust in open operation areas based on multi-source visual perception, a method and system for tracing and handling dust in open operation areas is needed, which can realize all-weather, blind-spot-free, automated high-precision identification and tracing of dust in open operation areas. Summary of the Invention

[0005] In view of this, the purpose of this invention is to overcome the defects in the prior art and provide a method and system for tracing and handling dust in open work areas based on multi-source visual perception, which can realize all-weather, blind-spot-free, automated, high-precision identification and tracing of dust in open work areas.

[0006] The present invention provides a method for tracing and controlling dust pollution in open work areas based on multi-source visual perception, comprising:

[0007] Acquire image information of the work area;

[0008] If the image information is perceptually analyzed and dust characteristics are identified in the dust-generating equipment and its associated area, an abnormal event is triggered.

[0009] Dynamically analyze abnormal events, generate and issue scheduling instructions;

[0010] The scheduled equipment collects multi-view spatiotemporal data of the work area;

[0011] The image information and the spatiotemporal data are fused and analyzed to generate a spatiotemporal evidence chain surrounding the abnormal event, and a source tracing report is generated and collaborative handling is carried out.

[0012] Furthermore, an abnormal event is identified by combining a target detection algorithm with a dynamic analysis algorithm. The target detection algorithm is used to locate potential dust-generating equipment and related areas. The dynamic analysis algorithm is used to analyze pixel motion features in image sequences within the area associated with dust-generating equipment to determine whether they conform to the dynamic characteristics of dust, thereby triggering the abnormal event.

[0013] Furthermore, abnormal events are dynamically analyzed to generate and issue scheduling instructions, specifically including:

[0014] If the abnormal event is assessed as large-scale or high-altitude dust, then a drone will be dispatched to the work area.

[0015] If an abnormal event is assessed as localized fine dust or requires manual confirmation, the nearest manager to the work area will be dispatched to the work area wearing a wearable mobile evidence collection device.

[0016] Furthermore, the scheduled equipment collects multi-view spatiotemporal data of the work area, specifically including:

[0017] Managers using drones or wearable mobile evidence collection devices are directed to the work area to collect evidence and take photos. The visual analysis module set up on the drone or evidence collection device tracks the dust range and diffusion path, locates the source machinery or source area, and transmits the evidence collection data stream, including time and space stamps, back to the target device in real time via wireless network.

[0018] Furthermore, the image information and the spatiotemporal data are fused and analyzed, specifically including: aligning the image information and the spatiotemporal data in time and space based on timestamps and geographic location information.

[0019] Furthermore, the source tracing report includes the event number, the time and geographical location of the event, the source of the event, the event level, and key time segments of multi-angle image evidence.

[0020] Furthermore, coordinated action will be taken, specifically including:

[0021] A tiered warning is triggered based on the event level. Managers take on-site action based on the warning, update the event status after the action is completed, and archive and save all relevant data to form a closed management loop, waiting for the next abnormal event to be triggered.

[0022] A dust source tracing and disposal system for open work areas based on multi-source visual perception includes a perception layer, a network layer, a decision layer, and an application layer.

[0023] The perception layer is used to collect image information and multi-view spatiotemporal data of the work area, perform perception analysis on the image information, identify abnormal events and upload them.

[0024] The decision-making layer is used to dynamically analyze abnormal events and generate and issue scheduling instructions to the perception layer.

[0025] The application layer is used to perform fusion analysis on the image information and the spatiotemporal data, generate a spatiotemporal correlation evidence chain surrounding the abnormal event, and generate a source tracing report and a collaborative handling plan.

[0026] The network layer is used to support information transmission between the perception layer, decision layer, and application layer.

[0027] Furthermore, the perception layer includes fixed monitoring cameras installed in the work area, airborne cameras on drones that can move in the air, and wearable mobile cameras worn by management personnel who conduct mobile patrols on the ground.

[0028] The beneficial effects of this invention are as follows: The method and system for tracing and handling dust pollution in open work areas based on multi-source visual perception disclosed in this invention solves the problem of the inability to accurately locate pollution sources from a single perspective by using fixed and mobile multi-source visual monitoring for routine monitoring; it utilizes a hybrid visual model to intelligently identify dust-generating equipment and dust dynamics in related areas to trigger abnormal events, significantly reducing false alarm and false alarm rates; it makes intelligent decisions based on the type of abnormal event, dispatching drones or wearable mobile evidence collection devices to the site to collect data, realizing on-demand and efficient allocation of limited resources and avoiding idle or blind resource use; it transmits the collected multi-view data back, and through spatiotemporal alignment and fusion analysis of the transmitted data, it achieves accurate source tracing and collaborative handling. Attached Figure Description

[0029] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0030] Figure 1 This is a schematic diagram illustrating the implementation process of the traceability and disposal method of the present invention;

[0031] Figure 2 This is a schematic diagram of the overall architecture of the traceability and disposal system of the present invention. Detailed Implementation

[0032] The present invention will be further described below with reference to the accompanying drawings, as shown in the figures:

[0033] This embodiment discloses a method for tracing and handling dust pollution in open work areas based on multi-source visual perception, including the following steps:

[0034] Acquire image information of the work area;

[0035] If the image information is perceptually analyzed and dust characteristics are identified in the dust-generating equipment and its associated area, an abnormal event is triggered.

[0036] Dynamically analyze abnormal events, generate and issue scheduling instructions;

[0037] The scheduled equipment collects multi-view spatiotemporal data of the work area;

[0038] The image information and the spatiotemporal data are fused and analyzed to generate a spatiotemporal evidence chain surrounding the abnormal event, and a source tracing report is generated and collaborative handling is carried out.

[0039] This invention provides a fully automated solution for the entire process, from automatic identification, intelligent scheduling, on-site evidence collection to analysis reports and disposal archiving. Compared with existing technical solutions that rely on fixed sensors or predictive models, this invention achieves dynamic monitoring of the entire process, enabling clear on-site observation, accurate source identification, and timely disposal, greatly improving the efficiency and effectiveness of dust monitoring in open operation areas.

[0040] In this embodiment, during the initial monitoring process, fixed surveillance cameras can be used to collect image information of the work area. Alternatively, drones and wearable video equipment can be used by management personnel to conduct round-the-clock, blind-spot-free, automated routine monitoring of the open work area to continuously collect video image information. The open work area includes at least one of the following: construction site, open-pit mine, material storage yard, spoil disposal site, and bulk material logistics storage yard.

[0041] In this embodiment, an abnormal event is identified by combining a target detection algorithm and a dynamic analysis algorithm. The target detection algorithm is used to locate potential dust-generating equipment and associated areas. The dynamic analysis algorithm analyzes pixel motion features in image sequences within the associated areas of dust-generating equipment to determine if they match dust dynamic characteristics, thereby triggering an abnormal event. The target detection algorithm is implemented using a deep learning-based target detection model, which performs real-time analysis of video images to quickly locate targets. The dynamic analysis algorithm is implemented using a motion field analysis algorithm based on continuous frame sequences, determining the presence and diffusion dynamics of dust by calculating the statistical characteristics of motion vectors within the associated areas.

[0042] The dust-generating equipment includes at least one of the following: excavator, bulldozer, mining dump truck, stacker-reclaimer, loader, crushing and screening equipment, and open-pit drilling equipment.

[0043] Specifically, firstly, target detection algorithms, such as neural networks based on YOLO, SSD, and Faster R-CNN architectures, are used to locate potential dust-generating devices and their associated areas within the image. Then, for each detected target, dynamic analysis algorithms, such as optical flow, frame difference, and background subtraction based on continuous frame sequences, are activated within its associated area to analyze the dust movement characteristics within the associated area. When a specific piece of machinery is detected and its associated area simultaneously exhibits high-intensity, diffuse, and irregular movement, it is identified as an abnormal event and uploaded to the system.

[0044] In this embodiment, upon receiving an abnormal event report, an assessment is performed based on preset rules. The abnormal event is dynamically analyzed, and scheduling instructions are generated and issued, specifically including:

[0045] If the abnormal event is a large-scale, high-altitude dust event, then a drone will be dispatched to the work area to collect evidence.

[0046] If the abnormal event is localized, a ground-source event, or requires personnel intervention, the nearest management personnel will be dispatched to the work area to wear wearable mobile evidence collection equipment to conduct evidence collection and investigation.

[0047] The dispatch instructions include information such as target location and evidence collection requirements.

[0048] In this embodiment, when the dispatched equipment performs the evidence collection task, the drone takes pictures from high altitude to obtain macroscopic images of the dust range and diffusion trend; while the personnel wearing wearable video equipment record from a first-person perspective to obtain microscopic images such as the dust source and operation details.

[0049] The scheduled equipment collects multi-view spatiotemporal data of the work area, specifically including:

[0050] Managers using drones or wearable mobile evidence collection devices are directed to the work area to collect evidence and take photos. The visual analysis module set up on the drone or evidence collection device tracks the dust range and diffusion path, locates the source machinery or source area, and transmits the evidence collection data stream, including time and space stamps, back to the target device (such as the dispatch center) in real time via wireless network.

[0051] In this embodiment, after receiving the returned evidence data, the dispatch center uses a data fusion and analysis platform to align and correlate the image information collected from monitoring and the multi-source video image data collected by the evidence collection equipment in time and space, based on the time and space stamp, to form an evidence chain surrounding the same abnormal event.

[0052] Secondly, by analyzing multi-view images, the source of dust pollution is accurately located, such as specific machinery and operating areas, and the event level is assessed. Based on this, a graded warning is triggered, and personnel are dispatched to coordinate on-site handling, automatically generating a source tracing report. Finally, all data and handling results are archived, forming a traceable management loop. The source tracing report includes the event number, the time and geographical location of the event, the event source, the event level, and key time segments of multi-angle image evidence.

[0053] The collaborative handling process includes: triggering tiered warnings based on the event level, having managers handle the situation on-site based on the warnings, updating the event status after the handling is completed, archiving and saving all relevant data to form a closed management loop, and waiting for the next abnormal event to be triggered.

[0054] The present invention also relates to a dust source tracing and disposal system for open work areas based on multi-source visual perception. The source tracing and disposal system can be understood as a system that implements the source tracing and disposal method of the above embodiments. The source tracing and disposal system includes a perception layer, a network layer, a decision layer and an application layer.

[0055] The perception layer is used to collect image information and multi-view spatiotemporal data of the work area, perform perception analysis on the image information, identify abnormal events, and upload them; the decision layer is used to perform dynamic analysis on abnormal events, generate and issue scheduling instructions to the perception layer; the application layer is used to perform fusion analysis on the image information and the spatiotemporal data, generate a spatiotemporal correlation evidence chain around the abnormal event, and generate a source tracing report and a collaborative handling plan; the network layer is used to support information transmission between the perception layer, the decision layer, and the application layer.

[0056] The perception layer includes fixed monitoring cameras installed in the work area, airborne cameras mounted on unmanned aerial vehicles (UAVs), and wearable mobile cameras worn by management personnel conducting ground patrols. The network layer includes at least one of 5G, 4G, Wi-Fi, and private networks. The decision-making layer includes a dispatch center or dispatch server. The application layer includes a data fusion and analysis platform and an event management platform; both the data fusion and analysis platform and the event management platform utilize existing data servers, which will not be elaborated further here.

[0057] To better understand the dust source tracing and disposal method and system of the present invention, the following examples are provided:

[0058] 1. Taking an earthwork construction area of ​​a certain project as an example, the working process of this invention is explained:

[0059] Three fixed smart cameras with edge computing capabilities (perception layer) are deployed at the highest point on the construction site boundary. At least one inspection drone and a wearable recorder for administrators (perception layer) are also provided. All devices are connected to the collaborative scheduling platform and data center (network layer, decision layer, application layer) via a dedicated 5G / 4G / WIFI network.

[0060] Camera A's built-in hybrid vision analysis model (e.g., using a target detection submodule based on the YOLO architecture combined with a dynamic analysis submodule based on Farneback optical flow) runs continuously. Its target detection submodule locates the excavator in the scene in real time. Once the excavator is detected, the dynamic analysis submodule is immediately activated within a preset extended area around the device to calculate the optical flow field characteristics. An abnormal event is only determined when the outputs of the two submodules simultaneously meet preset threshold conditions (e.g., mechanical confidence > 0.8 and regional motion diffusion entropy > 0.5) for at least 5 frames.

[0061] The dispatch center analyzed the incident and found that it was caused by an excavator, the dust source was clearly located, but the spread was relatively large. According to the rules, a drone was prioritized for dispatch. The optimal flight path for drone B was automatically planned, and instructions were issued to approach and take pictures to confirm the area.

[0062] Drone B automatically flew to the target area, captured video from an overhead angle, confirmed the dust generation area, tracked the direction of spread, and calculated the dust coverage area. Simultaneously, the dispatch center notified the nearest administrator to wear a head-mounted device C and proceed to the site to record the excavator's operations and the generation and spread of dust from a ground perspective.

[0063] The data center simultaneously receives initial footage from camera A, overhead stream from drone B, and ground video from head-mounted device C. The timelines are automatically aligned, and the three perspectives are fused onto an electronic map. An analysis report is automatically generated: 14:30:05, Earthwork Zone 3, excavator (e.g., EX-03) digging operations causing dust pollution, moderate pollution, lasting 120 seconds. The report also includes keyframe evidence from the three perspectives.

[0064] The report triggered platform alarms and SMS notifications to the supervisors. Based on clear evidence, the supervisors immediately halted operations and ordered water spraying to suppress dust. The handling process and outcome were recorded and linked to the event file, completing the closed loop.

[0065] 2. Taking an open-pit mining area of ​​a certain mine as an example, the working process of this invention is explained as follows:

[0066] Fixed intelligent monitoring cameras (perception layer) with edge computing capabilities are deployed at commanding heights along the boundaries of key areas such as mining areas and spoil heaps. These cameras employ wide-angle and zoom lenses to accommodate the large mining areas and long target distances. Simultaneously, inspection drones are deployed, and vehicle-mounted mobile cameras are provided for large equipment such as electric shovels and mining dump trucks. Furthermore, management personnel wear wearable intelligent recorders. All equipment is connected to the collaborative dispatch center (decision layer) and the data fusion platform (application layer) via 5G / private network.

[0067] The target recognition module, deployed on fixed cameras and vehicle-mounted terminals, is based on a deep learning-based target detection model (e.g., a target detection model based on the YOLO architecture). This model is specifically trained using image datasets of mining-specific equipment such as mining dump trucks, electric shovels, and drilling rigs to optimize the recognition accuracy of equipment in long-distance, dusty environments.

[0068] The dynamic analysis module, deployed on a fixed camera, is based on a motion field analysis algorithm using continuous frame sequences (e.g., a motion field analysis algorithm based on Farneback optical flow). For different types of dust commonly found in mines, such as blasting dust (instantaneous, large-scale) and transportation dust (linear, continuous), the module sets differentiated analysis parameters, such as motion amplitude thresholds and diffusion determination durations.

[0069] When a fixed monitoring system identifies an electric shovel and its continuous high-intensity, diffused motion characteristics in the loading area through its hybrid visual analysis model (i.e., the collaboration of the aforementioned target recognition and dynamic analysis modules), it determines it as a dust event during loading operations and reports the event.

[0070] After receiving the event, the dispatch center analyzes it according to preset rules:

[0071] If the event involves large-scale, high-altitude dust (such as dust after an explosion), drones should be immediately dispatched to the airspace to capture images of the macroscopic dispersion pattern.

[0072] If the incident is localized and involves ground dust (such as the loading operation in this example), then dispatch the nearest vehicle-mounted camera (such as a nearby mining truck) or notify the management personnel wearing wearable recorders to go there to obtain details and evidence of the near-field operation.

[0073] All forensic data (raw camera streams, drone aerial videos, and near-field videos from vehicle-mounted or wearable devices) are accompanied by high-precision spatiotemporal stamps. The data fusion platform uses these spatiotemporal stamps and geographic information as a basis for automatic alignment and correlation. In this example, the location of the electric shovel captured by the camera, the dust dispersion outline captured by the drone, and the specific loading actions captured by the wearable device are fused to accurately pinpoint the source as the loading operation of electric shovel No. 3 and estimate the impact range. Simultaneously, a platform warning is triggered.

[0074] Based on the platform notification, on-site management personnel immediately took measures such as activating water spraying, adjusting the loading height, and halting operations, and reported the incident to the dispatch center. The handling process and results were recorded and linked to the event file, completing the closed loop.

[0075] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for tracing and controlling dust pollution in open work areas based on multi-source visual perception, characterized in that: include: Acquire image information of the work area; If the image information is perceptually analyzed and dust characteristics are identified in the dust-generating equipment and its associated area, an abnormal event is triggered. Dynamically analyze abnormal events, generate and issue scheduling instructions; The scheduled equipment collects multi-view spatiotemporal data of the work area; The image information and the spatiotemporal data are fused and analyzed to generate a spatiotemporal evidence chain surrounding the abnormal event, and a source tracing report is generated and collaborative handling is carried out.

2. The method for tracing and handling dust pollution in open work areas based on multi-source visual perception according to claim 1, characterized in that: An abnormal event is identified by combining a target detection algorithm with a dynamic analysis algorithm. The target detection algorithm is used to locate potential dust-generating equipment and related areas. The dynamic analysis algorithm is used to analyze pixel motion features in image sequences within the area associated with dust-generating equipment to determine whether they conform to the dynamic characteristics of dust, thereby triggering the abnormal event.

3. The method for tracing and handling dust pollution in open work areas based on multi-source visual perception according to claim 1, characterized in that: Dynamic analysis of abnormal events is performed to generate and issue scheduling instructions, specifically including: If the abnormal event is assessed as large-scale or high-altitude dust, then a drone will be dispatched to the work area. If an abnormal event is assessed as localized fine dust or requires manual confirmation, the nearest manager to the work area will be dispatched to the work area wearing a wearable mobile evidence collection device.

4. The method for tracing and handling dust pollution in open work areas based on multi-source visual perception according to claim 1, characterized in that: The scheduled equipment collects multi-view spatiotemporal data of the work area, specifically including: Managers using drones or wearable mobile evidence collection devices are directed to the work area to collect evidence and take photos. The visual analysis module set up on the drone or evidence collection device tracks the dust range and diffusion path, locates the source machinery or source area, and transmits the evidence collection data stream, including time and space stamps, back to the target device in real time via wireless network.

5. The method for tracing and handling dust pollution in open work areas based on multi-source visual perception according to claim 1, characterized in that: The fusion analysis of the image information and the spatiotemporal data specifically includes: aligning the image information and the spatiotemporal data in time and space based on timestamps and geographic location information.

6. The method for tracing and handling dust pollution in open work areas based on multi-source visual perception according to claim 1, characterized in that: The source tracing report includes the event number, the time and geographical location of the event, the source of the event, the event level, and key time segments of multi-angle image evidence.

7. The method for tracing and handling dust pollution in open work areas based on multi-source visual perception according to claim 6, characterized in that: Coordinated action will be taken, specifically including: A tiered warning is triggered based on the event level. Managers take on-site action based on the warning, update the event status after the action is completed, and archive and save all relevant data to form a closed management loop, waiting for the next abnormal event to be triggered.

8. A system for tracing and managing dust pollution in open work areas based on multi-source visual perception, characterized in that: It includes the perception layer, network layer, decision layer, and application layer; The perception layer is used to collect image information and multi-view spatiotemporal data of the work area, perform perception analysis on the image information, identify abnormal events and upload them. The decision-making layer is used to dynamically analyze abnormal events and generate and issue scheduling instructions to the perception layer. The application layer is used to perform fusion analysis on the image information and the spatiotemporal data, generate a spatiotemporal correlation evidence chain surrounding the abnormal event, and generate a source tracing report and a collaborative handling plan. The network layer is used to support information transmission between the perception layer, decision layer, and application layer.

9. The open-area dust source tracing and disposal system based on multi-source visual perception according to claim 8, characterized in that: The perception layer includes fixed monitoring cameras installed in the work area, airborne cameras on unmanned aerial vehicles (UAVs) that can move in the air, and wearable mobile cameras worn by management personnel who conduct mobile patrols on the ground.

Images