Open-pit mine blasting effect optimization method and system based on video target recognition

Abstract

This invention discloses a method and system for optimizing blasting effects in open-pit mines based on video target recognition, belonging to the fields of mining and image or video recognition. The method includes: acquiring digital borehole design data and blasting design parameters for the target blasting area; collecting video data from before to the end of blasting using a drone; extracting video frames before blasting and identifying unblasted boreholes using a first target detection model; determining the borehole number of each unblasted borehole; extracting the video sequence after blasting, identifying punching targets using a second target detection model, and determining the trajectory of the punching targets using a multi-target tracking algorithm; matching the positions of the punching targets with the unblasted boreholes whose numbers have been matched to determine the borehole number of each punching target; acquiring the actual coordinates, actual hole depth, blasting design parameters, and punching event feature parameters for each punching target; and obtaining the borehole diagnosis result based on a preset diagnostic rule base. This invention can efficiently and accurately diagnose and optimize blasting effects.

Description

Technical Field

[0001] This invention belongs to the fields of mining, digital blasting, and image or video recognition, specifically relating to a method and system for optimizing blasting effects in open-pit mines based on video target recognition. Background Technology

[0002] In open-pit mining, blasting effectiveness directly impacts subsequent mining, loading, transportation efficiency, and overall costs. Currently, the evaluation and optimization of blasting effectiveness primarily rely on manual on-site investigation and measurement after blasting, such as assessing the blast pile morphology, block size distribution, and backflush distance. This method has significant drawbacks: first, it is severely delayed, failing to provide immediate feedback for processes like charging and networking after the blast; second, it is highly subjective and has low coverage, making it difficult to conduct detailed inspections of high-risk areas; and third, it cannot accurately attribute macroscopic blasting quality issues (such as large blocks, foundation defects, and excessive flyrock) to a specific blast hole, resulting in a lack of targeted optimization of subsequent blasting parameters. Often, large-scale adjustments are made based on experience, leading to low efficiency and unstable results.

[0003] With the development of drones and image processing technology, methods have emerged for using drones to scan the terrain after blasting to assess volume changes and block size distribution. However, these methods are still post-explosive assessments and fail to capture crucial dynamic information during the blasting process. "Breakhole punching" is the most direct dynamic phenomenon reflecting the effectiveness of borehole plugging and the rationality of the charge. Real-time, automatic identification and location of borehole punching would provide a revolutionary technical means for accurate and immediate diagnosis and optimization of blasting effects. Summary of the Invention

[0004] To address the aforementioned problems in the existing technology, this invention provides a method and system for optimizing blasting effects in open-pit mines based on video target recognition. The technical problem to be solved by this invention is achieved through the following technical solution:

[0005] In a first aspect, embodiments of the present invention provide a method for optimizing blasting effects in open-pit mines based on video target recognition, the method comprising:

[0006] Obtain digital borehole design data and blasting design parameters for the target blasting area in an open-pit mine; wherein, the digital borehole design data includes the borehole number, design coordinates, and design depth for each borehole; the blasting design parameters include the charge amount and plugging length for each borehole;

[0007] The drone above the target blasting area was activated to collect video data from before the blast to the end of the blast;

[0008] The video frames before the blasting are extracted from the video data, and the unblasted blast holes are identified using a pre-trained first target detection model. The blast hole numbers corresponding to each identified unblasted blast hole are determined by position matching.

[0009] The video sequence after the blast is extracted from the video data. The pre-trained second target detection model is used to identify the punching targets in the sequence. The trajectory of each punching target is determined by combining the preset multi-target tracking algorithm. Based on the trajectory of each punching target, the punching targets are matched with the unblasted blast holes with matched blast hole numbers to determine the blast hole number corresponding to each punching target.

[0010] For each blasting target, based on the determined borehole number, the corresponding actual coordinates, actual borehole depth, and blasting design parameters are retrieved from the database. Additionally, the blasting event characteristic parameters determined by the multi-target tracking algorithm when determining the trajectory of the blasting target are also retrieved, forming the basic data for single-hole diagnosis. The basic data for single-hole diagnosis is then used to infer the diagnosis results for the corresponding borehole based on a preset diagnosis rule base. These diagnostic results include the borehole number, diagnostic conclusion, and optimization suggestions. The diagnostic results are stored in the database as historical data to guide blasting design for future boreholes under similar conditions. Similar conditions include similar geological conditions or spatial proximity. Similar geological conditions mean that, among multiple preset geological parameters, at least two geological parameters between the new blasting area and the historical blasting area meet the corresponding requirements. Spatial proximity means that the new blasting area and the historical blasting area meet any preset spatial location requirement.

[0011] Secondly, embodiments of the present invention provide a system for optimizing blasting effects in open-pit mines based on video target recognition, used to implement the method for optimizing blasting effects in open-pit mines based on video target recognition described in the first aspect. The system includes a drone module, a data management module, a visual analysis engine module, an intelligent diagnosis and optimization module, and a comprehensive management and control platform; wherein...

[0012] The drone module is used to collect video data of the target blasting area in the open-pit mine using a drone and transmit the data back.

[0013] The data management module is used to store and manage digital drilling design data, blasting design parameters, video data, and historical data.

[0014] The visual analysis engine module integrates a first target detection model, a second target detection model, and a preset multi-target tracking algorithm. It uses the first target detection model to identify unexploded blast holes in the video frame before blasting in the video data, and determines the blast hole number corresponding to each identified unexploded blast hole through position matching. It also uses the second target detection model to identify punching targets in the video sequence after blasting in the video data, and combines this with the multi-target tracking algorithm to determine the trajectory of each punching target. Based on the trajectory of each punching target, it performs position matching between the punching targets and the identified unexploded blast holes to determine the blast hole number corresponding to each punching target.

[0015] The intelligent diagnosis and optimization module has a built-in preset diagnosis rule base. For each punching target, based on the determined borehole number, the module retrieves the corresponding actual coordinates, actual hole depth, and blasting design parameters from the database, and obtains the punching event characteristic parameters determined by the multi-target tracking algorithm when determining the trajectory of the punching target, thus forming the basic data for single-hole diagnosis. Based on the diagnosis rule base, the module infers from the basic data to obtain the diagnosis result for the corresponding borehole of the punching target. The diagnosis result includes the borehole number, diagnosis conclusion, and optimization suggestions. The diagnosis result is stored in the database as historical data.

[0016] The integrated management and control platform is used to provide a human-computer interaction interface, perform task scheduling, process monitoring, display of diagnostic results, and output of diagnostic result reports.

[0017] The beneficial effects of this invention are:

[0018] The method and system for optimizing blasting effects in open-pit mines based on video target recognition provided in this invention have the following beneficial effects:

[0019] 1. Real-time and automation: It changes the traditional lagging and manual evaluation mode, and can output a diagnostic report that locates the specific blast hole within minutes after the blasting is completed, which greatly improves the feedback efficiency.

[0020] 2. Precise location and attribution: By using the "two-step correlation method" (first locate the blast hole, then correlate the punch hole), the macroscopic blasting phenomenon is precisely attributed to the microscopic single hole design parameters, making the optimization measures more targeted.

[0021] 3. Data-driven and closed-loop optimization: The process data, analysis results, and optimization suggestions of each blasting are stored in a structured manner to form valuable digital assets for mine blasting. This provides data support for subsequent intelligent and refined blasting design and promotes the transformation of blasting technology from experience-driven to data-driven.

[0022] 4. Enhance safety: By automatically identifying and recording the punching phenomenon, objective basis is provided for blasting safety assessment and optimization of protective measures. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating a method for optimizing blasting effects in open-pit mines based on video target recognition, as provided in an embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram of the result of step S3 (identification and matching of unexploded blast holes) in the open-pit mine blasting effect optimization method based on video target recognition provided in the embodiment of the present invention.

[0025] Figure 3 This is a schematic diagram of the results of step S4 (blasting hole identification and association) in the open-pit mine blasting effect optimization method based on video target recognition provided in the embodiment of the present invention.

[0026] Figure 4 This is a schematic diagram of the structure of an open-pit mine blasting effect optimization system based on video target recognition provided in an embodiment of the present invention;

[0027] Figure 5 This is a schematic diagram of an example structure of an open-pit mine blasting effect optimization system based on video target recognition provided in an embodiment of the present invention. Detailed Implementation

[0028] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0029] Firstly, embodiments of the present invention provide a method for optimizing blasting effects in open-pit mines based on video target recognition. This method can identify blasting holes in real time and automatically, accurately locate them to specific designed blasting holes, and combine this with background data to achieve single-hole-level blasting effect diagnosis and parameter optimization, forming an intelligent closed loop. For example... Figure 1 As shown, the method may include the following steps:

[0030] S1, acquire digital borehole design data and blasting design parameters for the target blasting area in the open-pit mine;

[0031] In this context, the target blasting area in an open-pit mine refers to an area on the mine to be blasted. When designing the blasting for this area, multiple boreholes are pre-designed to form a borehole array. The digital borehole design data includes the borehole number, design coordinates, and design depth for each borehole; and each borehole number is unique. The blasting design parameters include the charge amount and plugging length for each borehole. The digital borehole design data and blasting design parameters in this embodiment can be stored in a database and retrieved when needed.

[0032] S2, activate the drone above the target blasting area to collect video data from before the blasting to the end of the blasting;

[0033] In this embodiment of the invention, the drone can be equipped with a high-definition camera. Before the target blasting area is blasted, it is controlled to fly to a safe airspace above the blast pile in the target blasting area, ensuring that the high-definition camera lens covers the entire target blasting area. Then, video recording is started before the target blasting area is blasted and continues until the blasting is completed. The collected video data is then transmitted back in real time via a network (such as a 5G / 4G wireless network) for corresponding video data processing.

[0034] S3, extract the video frames before blasting from the video data, identify the unblasted blasting holes in them using the pre-trained first target detection model, and determine the blasting hole number corresponding to each identified unblasted blasting hole through position matching;

[0035] S3-S5 are the video data processing stages. Extracting the pre-blast video frame from the video data can involve extracting the last clear video image before the blast.

[0036] The first object detection model can be any existing object detection model, obtained through training.

[0037] In one optional implementation, the first target detection model is a YOLOv8-based blast hole recognition model, which is trained using several pre-blasting sample images with labeled unblasted blast hole openings and their pixel positions. The first target detection model is used to identify unblasted blast hole openings and their pixel positions in the input image.

[0038] For details on the YOLOv8 network structure and processing, please refer to the relevant technical explanations. The training set used for training the first object detection model includes several pre-blasting sample images, all extracted from corresponding video data. These images represent the corresponding blasting area before blasting, such as thousands of open-pit mine blast hole images under different lighting and angles. The images are labeled with the unblasted blast hole openings and their pixel positions, which can be pixel coordinate bounding boxes. The specific training process of the first object detection model can be found in existing YOLOv8 training process explanations and will not be detailed here. The trained first object detection model can identify the unblasted blast hole openings and determine their pixel positions from the input pre-blasting images (pre-blasting video frames).

[0039] Therefore, in S3, the first target detection model can be used to identify all or part of the unexploded blast holes in the video frame before the blast.

[0040] In S3, determining the borehole number corresponding to each identified unexploded borehole through position matching includes:

[0041] 1) For each identified unexploded blast hole, using the positioning and attitude data of the UAV and the camera intrinsic parameter data, the pixel position of the unexploded blast hole is converted from pixel coordinates to geodetic coordinates through photogrammetry.

[0042] The core of this step is to use the "collinearity equation + known elevation constraint" to project the pixels corresponding to the unexploded blast holes on the image (i.e., the video frame before blasting) onto a specific plane in three-dimensional space, thereby obtaining their geodetic coordinates. This completes the conversion of image coordinates to geographic coordinates, resulting in the geodetic coordinates of each identified unexploded blast hole. This process is a mature and standard photogrammetric procedure; please refer to relevant technical explanations for details, which will not be elaborated upon here.

[0043] 2) Using the nearest neighbor algorithm, the obtained geodetic coordinates are matched with the three-dimensional coordinates in the digital borehole design data. The borehole number corresponding to the matched three-dimensional coordinate is determined as the borehole number corresponding to the unblasted borehole.

[0044] Since the digital borehole design data contains the borehole numbers and 3D coordinates of all boreholes in the target blasting area, this step utilizes a nearest neighbor algorithm for spatial matching. This matches the nearest neighbor 3D coordinates for the geodetic coordinates of each identified unblasted borehole opening. The borehole number corresponding to the matched 3D coordinates is then determined as the borehole number for that identified unblasted borehole opening. The purpose is to uniquely match each borehole opening in the image (i.e., the video frame before blasting), completing the identification and matching of unblasted boreholes, establishing an accurate correspondence between the video footage and the real-world design drawings, and obtaining the "bombhole location". Design borehole number mapping table.

[0045] The nearest neighbor algorithm matches by calculating the Euclidean distance. Specifically, it calculates the Euclidean distance between the obtained geodetic coordinates and each three-dimensional coordinate in the digital borehole design data. The three-dimensional coordinate with the smallest Euclidean distance in the digital borehole design data is the matched three-dimensional coordinate, which is most similar to the obtained geodetic coordinates. The borehole number corresponding to this matched three-dimensional coordinate is determined as the borehole number corresponding to the unblasted borehole.

[0046] S4, extract the post-blast video sequence from the video data, identify the punching targets in it using the pre-trained second target detection model, and determine the trajectory of each punching target by combining it with the preset multi-target tracking algorithm; according to the trajectory of each punching target, match the position of the punching target with the unblasted blast hole with the matched blast hole number to determine the blast hole number corresponding to each punching target.

[0047] Extracting the post-explosion video sequence from the video data involves extracting a continuous video sequence from the start to the end of the explosion, which includes multiple frames.

[0048] The second object detection model can be any existing object detection model, obtained through training.

[0049] In one optional implementation, the second target detection model is a YOLOv8-based punching recognition model, which is trained using several post-blasting sample images with punching targets and their pixel positions already labeled. The second target detection model is used to identify punching targets and their pixel positions in the input image.

[0050] For details on the YOLOv8 network structure and processing, please refer to the relevant technical explanations. The second target detection model is trained using explosion video frames labeled "punching". The training set includes several post-explosion sample images extracted from corresponding continuous post-explosion video sequences. These images contain multiple frames of the corresponding explosion area from the start to the end of the explosion, with the punching target and its pixel location labeled on the images. The punching target is rock or dust thrown vertically upwards during the explosion, and the pixel location can be a pixel coordinate bounding box. For a detailed explanation of the training process of the second target detection model, please refer to the existing YOLOv8 training process explanations; details will not be provided here. The trained second target detection model can identify punching targets in each frame of the input post-explosion video sequence and determine their pixel locations, i.e., it detects punching targets frame-by-frame in the "explosive punching" scene.

[0051] Furthermore, a pre-defined multi-target tracking algorithm is then introduced to assign or associate a unique trajectory ID to each punching target detected in each frame, thereby achieving stable tracking across frames and distinguishing multiple independent punching events that occur simultaneously or sequentially, and determining the trajectory of each punching target.

[0052] In one optional implementation, the preset multi-target tracking algorithm includes the ByteTrack algorithm. Of course, other multi-target tracking algorithms can also be used, and there is no limitation here.

[0053] In S4, based on the trajectory of each punching target, the punching target is matched with the unexploded blast hole openings with matched blast hole numbers to determine the blast hole number corresponding to each punching target, including:

[0054] For each punching target trajectory, the nearest neighbor algorithm is used to match the initial occurrence position of the punching target trajectory with the pixel position of the unexploded blast hole with the matched blast hole number to determine the unexploded blast hole that generated the punching target and the corresponding blast hole number.

[0055] Specifically, this section analyzes the initial appearance position of each punching target trajectory (which can be the center point or bottom center point of the detection box in the first frame of the trajectory), and compares the coordinates of the initial appearance position of the punching target trajectory with the "bomb hole position" generated in S3. The target is compared and associated with the "designed borehole number" mapping table. If the initial location coordinates of the target's trajectory fall into or are extremely close to the image location of a borehole with a matching number, the target is determined to originate from that borehole, thus completing the precise association between the "bombing event" and the "designed borehole number".

[0056] Therefore, in S4, the blasting hole identification and association can be used to determine the blast hole number corresponding to each blasting target and to determine the blast hole from which each blasting target originates.

[0057] Furthermore, it should be noted that during the processing using the multi-target tracking ByteTrack algorithm, the punching event characteristic parameters of each punching event will also be identified; the punching event characteristic parameters include: punching duration and maximum punching height.

[0058] Understandably, the punching duration is the time from when the punching target first appears on the target trajectory to when it disappears. The maximum punching height refers to the maximum height the punching target can punch upwards on the target trajectory; both the punching duration and the maximum punching height together reflect the intensity of the punching.

[0059] S5. For each punching target, based on the determined borehole number, retrieve the corresponding actual coordinates, actual hole depth, and blasting design parameters from the database, and retrieve the punching event characteristic parameters determined by the multi-target tracking algorithm when determining the trajectory of the punching target, thus forming the basic data for single-hole diagnosis; perform reasoning on the basic data for single-hole diagnosis according to the preset diagnosis rule base to obtain the diagnosis result of the borehole corresponding to the punching target, wherein the diagnosis result includes the borehole number, diagnosis conclusion, and optimization suggestions; wherein, the diagnosis result is used to store in the database as historical data to guide the blasting design of boreholes under similar conditions in the future.

[0060] The digital borehole design data represents the design parameters of the boreholes, including the borehole number, design coordinates, and design depth for each borehole. The actual coordinates and actual depth are data generated during the borehole inspection stage in the blasting process, representing the actual parameters of the boreholes. There is a certain difference between the actual coordinates and actual depth and the design coordinates and design depth (due to errors caused by drilling rigs). For greater accuracy, S5 uses the actual parameters of the boreholes here. Furthermore, the charge quantity and plugging length of the boreholes in the blasting design parameters are based on the designed charge quantity and plugging length, since the corresponding real data cannot be verified in most mines at present.

[0061] S5 performs single-hole effect diagnosis and optimization. It analyzes the blasting effect based on a preset diagnostic rule base and generates design optimization suggestions for the single hole (mainly optimization suggestions for blasting design parameters).

[0062] Specifically, in this embodiment of the invention, the preset diagnostic rule base contains several rules. Each rule contains two parts: conditions and conclusions. The conditions are obtained based on the combination of relevant blasting data, and the conclusions include the diagnostic results of the blasting problem under the corresponding conditions.

[0063] Specifically, the pre-set diagnostic rule base in this embodiment of the invention is determined after evaluating the blasting effect based on a large amount of historical data. The diagnostic rule base includes relevant data for each blast, such as the geological conditions at the time of each blast, the coordinates of the blasting area, the borehole number, the actual hole depth, the actual coordinates, the blasting design parameters (charge amount and plugging length for each borehole), the number of perforations, the perforation duration for each perforation, and the maximum perforation height, etc.

[0064] Furthermore, the diagnostic rule base is designed with conditions based on blasting-related data. These conditions are combinations of blasting-related data, such as combinations of perforation event characteristic parameters, digital drilling design data, and blasting design parameters, as well as the magnitude relationship between these data and corresponding thresholds. The diagnostic rule base also designs diagnostic conclusions for blasting problems corresponding to these conditions, forming several "condition-conclusion" rules. For example, a rule could be: "If the design value of the hole plugging length is less than the threshold L0, and its associated perforation duration is greater than the threshold T0 and the maximum perforation height is greater than the threshold H0, then the diagnosis is 'insufficient plugging length,' and it is recommended to increase the plugging length to L1."

[0065] When performing step S5, for a single punching target, based on its determined borehole number, the corresponding actual coordinates, actual hole depth and blasting design parameters are obtained from the database, and the corresponding punching event characteristic parameters are obtained to form the corresponding single hole diagnostic basic data.

[0066] Then, based on a preset diagnostic rule base, the single-hole diagnostic baseline data is automatically inferred. The system searches for target rules matching the single-hole diagnostic baseline data from the rule base, and based on the target rules, obtains the diagnostic result for the corresponding borehole. The diagnostic result includes the borehole number, diagnostic conclusion, and optimization suggestions, and can be output as a diagnostic report. The diagnostic results are also stored in a database as historical data to guide borehole blasting design under similar conditions in the future. All diagnostic records, along with their corresponding blasting-related data, video data, and process data, are stored in the database as historical data to guide blasting design under similar conditions in the future, forming a closed loop of continuous optimization.

[0067] In this embodiment of the invention, similar conditions include similar geological conditions or spatial proximity.

[0068] When designing new blasting operations, diagnostic results from relevant historical data can be retrieved from the database as a design reference based on similar geological conditions or spatial proximity (e.g., blasting under the same geological conditions or in adjacent blasting areas), thereby achieving automatic optimization of blasting parameters.

[0069] Among these, similar geological conditions mean that, among multiple preset geological parameters, at least two geological parameters between the new blasting area and the historical blasting area meet the corresponding requirements. The multiple preset geological parameters and their corresponding requirements include:

[0070] 1. Rock category: Classified into categories I to V according to the "Engineering Rock Mass Classification Standard" (GB / T 50218), rocks belonging to the same category are considered similar;

[0071] 2. Uniaxial saturated compressive strength of rock: relative difference between measured values ​​≤ 15%;

[0072] 3. Rock mass integrity coefficient: The absolute difference between measured values ​​≤ 0.15;

[0073] 4. Angle between the main fracture direction and the borehole alignment: absolute difference ≤ 15°;

[0074] 5. Terrain slope: Absolute difference ≤ 10°.

[0075] Among them, spatial proximity means that the new blasting area and the historical blasting area meet any of the preset spatial location requirements, including the following:

[0076] 1. Within the same step plane, the horizontal distance between the center points of two areas is ≤ 50 meters;

[0077] 2. Adjacent steps have a vertical height difference of ≤ 15 meters and overlapping horizontal projections or an edge distance of ≤ 20 meters;

[0078] 3. Similar resistance lines refer to two blast holes whose resistance line length is less than 0.5 meters. If the two blast holes are considered to have similar resistance lines, then the two blast holes are considered to have similar resistance lines.

[0079] 4. Belongs to the same blasting zone number defined in the mine's annual blasting plan.

[0080] To facilitate understanding of the processing procedure of the method in the embodiments of the present invention, a specific example is given below. The method of the present invention is applied in bench blasting of an iron mine, and the specific processing procedure includes the following steps:

[0081] Step 1, Data Preparation:

[0082] This step is explained in the previous section, S1.

[0083] Retrieve digital borehole design data and blasting design parameters for the target blasting area from the database;

[0084] The target blasting area is designed with 150 boreholes; the digital borehole design data is the borehole design table for these 150 boreholes, including the borehole number, design coordinates (three-dimensional coordinates, including X / Y / Z coordinates), and design depth for each borehole; the blasting design parameters are the blasting design table for these 150 boreholes, including the charge amount and plugging length for each borehole;

[0085] Step 2, Video Capture:

[0086] This step is explained in step S2 above.

[0087] Three minutes before detonation, a drone (which can be a DJI Matrice 350 RTK) is controlled to fly to a safe observation point approximately 150 meters above the blast zone in the target blasting area. The drone adjusts the angle of its gimbal camera to ensure complete coverage of the blast pile. After the detonation command is issued, the drone automatically begins recording 4K high-definition video from before the detonation to the end of the blast, and transmits it back to the integrated management platform in real time via the 5G private network within the mining area.

[0088] Step 3, Identification and matching of unexploded ordnance holes:

[0089] This step is explained in step S3 above.

[0090] Reference Figure 2 After receiving the video data, the integrated control platform automatically captures the last panoramic image frame before the detonation command is issued. This image is then input into the trained first target detection model (a YOLOv8s-based borehole recognition model). The model successfully identifies 148 obvious boreholes in the image and outputs 148 bounding boxes as the pixel positions of the boreholes. The high-precision RTK position and attitude data of the UAV corresponding to this frame are read, and the geodetic coordinates of these 148 boreholes are obtained through forward intersection calculation. Using the nearest neighbor algorithm, 145 of these boreholes are successfully matched to the borehole numbers designed in the digital drilling design data, achieving a matching success rate of 97.3%.

[0091] Step 4, Identification and Association of Explosive Punching Holes:

[0092] This step is explained in step S4 above.

[0093] Reference Figure 3 It demonstrates the process by which the ByteTrack algorithm tracks the trajectories of multiple punching targets and associates them with the location of the blast holes.

[0094] After the blast began, the video analysis engine was activated. The trained second target detection model (a punch detection model based on YOLOv8) began frame-by-frame analysis. The first punch target was detected in frame 10 (approximately 0.33 seconds after the blast). The multi-target tracking algorithm ByteTrack immediately assigned it trajectory ID=1. Subsequently, multiple punch targets were detected and assigned different IDs. The multi-target tracking ByteTrack algorithm successfully overcame smoke and dust obstruction, maintaining the continuity of each punch target's trajectory. Analyzing the initial location of punch target trajectory ID=1, its coordinates deviated from the center position of the matched numbered borehole "A-32" by only 3 pixels, indicating that the punch target originated from borehole "A-32". Similarly, punch targets with trajectory IDs of IDs 2 and 3 were associated with boreholes "B-15" and "C-07," respectively. A total of 9 significant punch events were identified and associated in this blast.

[0095] Step 5, Single-hole effect diagnosis and optimization:

[0096] This step is explained in step S5 above.

[0097] For example, for borehole "A-32", the automatic blasting design parameters show a designed plugging length of 2.0 meters. The associated perforation duration is as long as 2.5 seconds, with an estimated maximum perforation height exceeding 30 meters. A rule in the diagnostic rule base is triggered: "Pluging length < 2.5 meters, perforation duration > 2.0 seconds, and maximum perforation height > 20 meters, diagnosed as: severely insufficient plugging, with primary energy released from the borehole opening." The recommended optimization is: "It is suggested to increase the plugging length of boreholes in similar rock areas and under similar resistance conditions to 3.5-4.0 meters."

[0098] Furthermore, the method in this embodiment of the invention can achieve closed-loop optimization, specifically:

[0099] All analytical data, video clips, diagnostic reports, etc., were archived and stored in the database along with the geological and lithological description of the blast area.

[0100] Two weeks later, when designing blasting in a new blasting area within the same step and a horizontal distance of 30 meters, the system automatically matched historical data and pushed a prompt: "Historical data from the adjacent area shows that the rock type in this area is Class III, the rock mass integrity coefficient is 0.65, and the uniaxial saturated compressive strength of the rock is 65MPa. By calculating the differences, the similarity of the current geological conditions is determined to be 0.92 (which can be achieved by designing an algorithm using multiple preset geological parameters and corresponding requirements). The historical optimization record 20231025-report shows that under these conditions, it is easy for violent blasting to occur due to insufficient plugging. It is recommended to preset the plugging length to be no less than 3.5 meters." This helps designers make better decisions.

[0101] The open-pit mine blasting effect optimization method based on video target recognition provided in this invention has the following beneficial effects:

[0102] 1. Real-time and automation: It changes the traditional lagging and manual evaluation mode, and can output a diagnostic report that locates the specific blast hole within minutes after the blasting is completed, which greatly improves the feedback efficiency.

[0103] 2. Precise location and attribution: By using the "two-step correlation method" (first locate the blast hole, then correlate the punch hole), the macroscopic blasting phenomenon is precisely attributed to the microscopic single hole design parameters, making the optimization measures more targeted.

[0104] 3. Data-driven and closed-loop optimization: The process data, analysis results, and optimization suggestions of each blasting are stored in a structured manner to form valuable digital assets for mine blasting. This provides data support for subsequent intelligent and refined blasting design and promotes the transformation of blasting technology from experience-driven to data-driven.

[0105] 4. Enhance safety: By automatically identifying and recording the punching phenomenon, objective basis is provided for blasting safety assessment and optimization of protective measures.

[0106] Secondly, corresponding to the above-described method embodiments, this invention also provides an open-pit mine blasting effect optimization system based on video target recognition, used to implement the open-pit mine blasting effect optimization method based on video target recognition described in the first aspect, such as... Figure 4 As shown, the system includes a drone module, a data management module, a visual analysis engine module, an intelligent diagnosis and optimization module, and a comprehensive control platform; wherein,

[0107] The drone module is used to collect video data of the target blasting area in the open-pit mine using a drone and transmit the data back.

[0108] The data management module is used to store and manage digital drilling design data, blasting design parameters, video data, and historical data.

[0109] The visual analysis engine module integrates a first target detection model, a second target detection model, and a preset multi-target tracking algorithm. It uses the first target detection model to identify unexploded blast holes in the video frame before blasting in the video data, and determines the blast hole number corresponding to each identified unexploded blast hole through position matching. It also uses the second target detection model to identify punching targets in the video sequence after blasting in the video data, and combines this with the multi-target tracking algorithm to determine the trajectory of each punching target. Based on the trajectory of each punching target, it performs position matching between the punching targets and the identified unexploded blast holes to determine the blast hole number corresponding to each punching target.

[0110] The intelligent diagnosis and optimization module has a built-in preset diagnosis rule base. For each punching target, based on the determined borehole number, the module retrieves the corresponding actual coordinates, actual hole depth, and blasting design parameters from the database, and obtains the punching event characteristic parameters determined by the multi-target tracking algorithm when determining the trajectory of the punching target, thus forming the basic data for single-hole diagnosis. Based on the diagnosis rule base, the module infers from the basic data to obtain the diagnosis result for the corresponding borehole of the punching target. The diagnosis result includes the borehole number, diagnosis conclusion, and optimization suggestions. The diagnosis result is stored in the database as historical data.

[0111] The integrated management and control platform is used to provide a human-computer interaction interface, perform task scheduling, process monitoring, display of diagnostic results, and output of diagnostic result reports.

[0112] Furthermore, the integrated management and control platform is also used to provide a blasting design assistance interface. When a user is designing a blasting operation in a new area, the platform can match historical data in the database and output historical optimization suggestions that are similar to the conditions of the current design area.

[0113] Among them, similarity with the current design area conditions includes similar geological conditions or spatial proximity; similar geological conditions means that among multiple preset geological parameters, the differences between the new blasting area and the historical blasting area in at least two geological parameters meet the corresponding requirements; spatial proximity means that the new blasting area and the historical blasting area meet any preset spatial location requirement; see the above for details.

[0114] For details, please refer to [link / reference]. Figure 5 An exemplary system architecture is shown, with alternative implementations as follows:

[0115] The drone module may include a drone, a high-definition stabilized gimbal camera, an RTK positioning submodule, and a wireless image / data transmission submodule. The drone module collects video data above the target blasting area from before the blast to the end of the blast, and transmits the video data back to the integrated management and control platform in real time via wireless network. The integrated management and control platform sends the video data to the visual analysis engine module to perform tasks such as borehole identification, coordinate matching, hole detection, and tracking association. The intelligent diagnosis and optimization module performs single-hole data association, effect analysis, and optimization suggestion generation. The video data, diagnostic results, and related process data are stored in the database as historical data.

[0116] The data management module is deployed on a file server and includes a database and a file server. The database adopts a relational database and object storage method to manage structured design data (including digital drilling design data and blasting design parameters), unstructured video data, and diagnostic result data, respectively.

[0117] The visual analysis engine module is deployed on a server or workstation with GPU computing power. It encapsulates the first target detection model, the second target detection model, and the multi-target tracking ByteTrack algorithm of YOLOv8, and provides standard API interfaces for the integrated management and control platform to call.

[0118] The intelligent diagnosis and optimization module includes a configurable diagnostic rule base (rule engine), a data correlator, and a report generator (used to generate reports of diagnostic results).

[0119] The integrated management and control platform is developed based on Web technology and provides a visual task management interface (for task management), a real-time video monitoring window (for video monitoring), an analysis result display panel (for result display), and a blasting design assistance interface (for design assistance). For example, the integrated management and control platform can automatically pop up relevant prompts in the sidebar of the design interface to assist in the design.

[0120] For details on the specific processing procedures of each module of this system, please refer to the relevant content in the first section, which will not be elaborated here.

[0121] The open-pit mine blasting effect optimization system based on video target recognition provided in this invention, targeting open-pit mine blasting scenarios, can complete the entire process of automated processing from data acquisition and intelligent analysis to decision output through the collaborative work of various modules. This achieves real-time, automatic, and accurate location and attribution of blasting problems, forming a "monitoring" system. analyze The "optimization" closed loop significantly improved blasting effectiveness and operational safety.

[0122] It should be noted that, in the description of this invention, the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0123] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0124] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. In addition, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.

[0125] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0126] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of protection of the present invention.

Claims

1. A method for optimizing blasting effects in open-pit mines based on video target recognition, characterized in that, include: Obtain digital borehole design data and blasting design parameters for the target blasting area in an open-pit mine; wherein, the digital borehole design data includes the borehole number, design coordinates, and design depth for each borehole; the blasting design parameters include the charge amount and plugging length for each borehole; The drone above the target blasting area was activated to collect video data from before the blast to the end of the blast; The video frames before the blasting are extracted from the video data, and the unexploded blasting holes are identified using a pre-trained first target detection model. The hole numbers corresponding to each identified unexploded blasting hole are determined by position matching. The first target detection model is used to identify the unexploded blasting holes and their pixel positions in the input image. The video sequence after the blast is extracted from the video data. A pre-trained second target detection model is used to identify the punching targets in the sequence, and a preset multi-target tracking algorithm is used to determine the trajectory of each punching target. For each punching target trajectory, the nearest neighbor algorithm is used to match the initial appearance position of the punching target trajectory with the pixel position of the unblasted blast hole with the matched blast hole number to determine the unblasted blast hole that generated the punching target and the corresponding blast hole number. The second target detection model is used to identify the punching targets and their pixel positions in the input image. For each blasting target, based on the determined borehole number, the corresponding actual coordinates, actual borehole depth, and blasting design parameters are retrieved from the database. Additionally, the blasting event characteristic parameters determined by the multi-target tracking algorithm when determining the trajectory of the blasting target are also retrieved, forming the basic data for single-hole diagnosis. The basic data for single-hole diagnosis is then used to infer the diagnosis results for the corresponding borehole based on a preset diagnosis rule base. These diagnostic results include the borehole number, diagnostic conclusion, and optimization suggestions. The diagnostic results are stored in the database as historical data to guide blasting design for future boreholes under similar conditions. Similar conditions include similar geological conditions or spatial proximity. Similar geological conditions mean that, among multiple preset geological parameters, at least two geological parameters between the new blasting area and the historical blasting area meet the corresponding requirements. Spatial proximity means that the new blasting area and the historical blasting area meet any preset spatial location requirement.

2. The method according to claim 1, characterized in that, The first target detection model is a YOLOv8-based borehole recognition model, which is trained using several pre-blasting sample images with labeled unblasted boreholes and their pixel positions.

3. The method according to claim 2, characterized in that, The process of determining the borehole number corresponding to each identified unexploded borehole through location matching includes: For each identified unexploded blast hole, using the positioning and attitude data of the UAV and the camera intrinsic parameter data, the pixel position of the unexploded blast hole is converted from pixel coordinates to geodetic coordinates through photogrammetry principles. Using the nearest neighbor algorithm, the obtained geodetic coordinates are matched with the three-dimensional coordinates in the digital borehole design data. The borehole number corresponding to the matched three-dimensional coordinate is determined as the borehole number corresponding to the unblasted borehole.

4. The method according to claim 2, characterized in that, The second target detection model is a YOLOv8-based perforation recognition model, which is trained using several post-blasting sample images with labeled perforation targets and their pixel positions; wherein, the perforation target is rock or dust thrown vertically upward during the blasting process.

5. The method according to claim 1, characterized in that, The preset multi-target tracking algorithm includes the ByteTrack algorithm.

6. The method according to claim 1, characterized in that, The punching event characteristic parameters include: punching duration and maximum punching height.

7. The method according to claim 6, characterized in that, The preset diagnostic rule base contains several rules. Each rule contains two parts: conditions and conclusions. The conditions are obtained based on the combination of relevant blasting data, and the conclusions include the diagnostic results of the blasting problem under the corresponding conditions.

8. A system for optimizing blasting effects in open-pit mines based on video target recognition, characterized in that, The system, used to implement the open-pit mine blasting effect optimization method based on video target recognition as described in any one of claims 1-7, comprises a UAV module, a data management module, a visual analysis engine module, an intelligent diagnosis and optimization module, and a comprehensive control platform; wherein... The drone module is used to collect video data of the target blasting area in the open-pit mine using a drone and transmit the data back. The data management module is used to store and manage digital drilling design data, blasting design parameters, video data, and historical data. The visual analysis engine module integrates a first target detection model, a second target detection model, and a preset multi-target tracking algorithm. It uses the first target detection model to identify unexploded blasting holes in the video frame before blasting in the video data, and determines the corresponding blasting hole number through position matching. It also uses the second target detection model to identify punching targets in the video sequence after blasting in the video data, and determines the trajectory of each punching target using the multi-target tracking algorithm. For each punching target trajectory, it uses a nearest neighbor algorithm to match the initial appearance position of the punching target trajectory with the pixel position of the unexploded blasting hole with the matched blasting hole number, determining the unexploded blasting hole that generated the punching target and its corresponding blasting hole number. The first target detection model is used to identify unexploded blasting holes and their pixel positions in the input image; the second target detection model is used to identify punching targets and their pixel positions in the input image. The intelligent diagnosis and optimization module has a built-in preset diagnosis rule base. For each punching target, based on the determined borehole number, the module retrieves the corresponding actual coordinates, actual hole depth, and blasting design parameters from the database, and obtains the punching event characteristic parameters determined by the multi-target tracking algorithm when determining the trajectory of the punching target, thus forming the basic data for single-hole diagnosis. Based on the diagnosis rule base, the module infers from the basic data to obtain the diagnosis result for the corresponding borehole of the punching target. The diagnosis result includes the borehole number, diagnosis conclusion, and optimization suggestions. The diagnosis result is stored in the database as historical data. The integrated management and control platform is used to provide a human-computer interaction interface, perform task scheduling, process monitoring, display of diagnostic results, and output of diagnostic result reports.

9. The system according to claim 8, characterized in that, The integrated management and control platform is also used to provide a blasting design assistance interface. When a user designs a blasting operation in a new area, it matches historical data in the database and outputs historical optimization suggestions that are similar to the conditions of the current design area. The similarity of conditions includes similar geological conditions or spatial proximity. Similar geological conditions mean that among multiple preset geological parameters, the differences between the new blasting area and the historical blasting area meet the corresponding requirements. Spatial proximity means that the new blasting area and the historical blasting area meet any preset spatial location requirement.

Images