Dynamic identification and alarm system for illegal construction activities of 500kV transmission lines

Abstract

This invention relates to the field of image recognition and power construction safety monitoring technology, and discloses a dynamic identification and alarm system for construction violations in 500kV transmission line construction. The system includes a video acquisition module, a target identification and relationship generation module, a rule effectiveness status generation module, a cross-stage judgment module, an evidence window control module, and an alarm output module. The video acquisition module is used to acquire continuous video frames from the construction site of the 500kV transmission line. This system identifies construction physical objects, personnel targets, equipment targets, work area targets, and safety protection equipment targets from continuous video frames, generates candidate results and suspected violation relationships for each construction stage, and generates the rule effectiveness status for each construction stage based on the proportion of candidate results, the continuous appearance of key construction physical objects, spatial combination relationships, and preset stage transition relationships.

Description

Technical Field

[0001] This invention relates to the fields of image recognition and power construction safety monitoring technology, and in particular to a dynamic identification and alarm system for violations during the construction of 500kV transmission lines. Background Technology

[0002] Existing power transmission line construction violation alarm systems typically collect images or videos of the construction site using camera equipment. They then utilize target detection models to identify targets such as personnel, safety helmets, safety belts, construction equipment, and work areas. Based on preset violation rules, they determine whether there are violations such as personnel not wearing protective equipment, entering hazardous areas, lingering in hoisting areas, or approaching tensioning work areas. Some systems also incorporate edge computing devices or backend servers to continuously verify the identification results and push suspected violations to the management terminal for on-site management personnel to handle.

[0003] However, 500kV transmission line construction sites are characterized by continuous changes in construction stages, alternating physical objects on site, temporary changes in work areas, and overlapping operations by personnel and equipment. During the transition from foundation construction to tower erection and from tower erection to line stringing, different construction physical objects from different stages may coexist in the same frame, such as foundation pits, tower materials, scaffolding, hoisting slings, and tensioning equipment. Existing systems typically trigger alarms directly based on single-frame recognition results, fixed construction stage models, or single violation rules. This makes it difficult to determine whether the violation rule corresponding to the current construction stage has met the conditions for effectiveness, and it is also difficult to handle situations where the current candidate construction stage rule and the rules of adjacent construction stages produce different judgment results for the same suspected violation relationship item. This easily leads to false alarms caused by premature activation of stage rules, short-term false detection of targets, fluctuations in spatial relationships, or cross-stage judgment conflicts during stage transitions. Summary of the Invention

[0004] The technical problem to be solved by this invention is that the existing technology has the drawback of false alarms in the identification of violations during the construction phase transition period. To address this, we propose a dynamic identification and alarm system for construction violations in 500kV transmission lines.

[0005] To achieve the above objectives, this application adopts the following technical solution: a dynamic identification and alarm system for illegal construction activities in 500kV transmission line construction, including:

[0006] It includes a video acquisition module, a target recognition and relationship generation module, a rule effective status generation module, a cross-stage judgment module, an evidence window control module, and an alarm output module;

[0007] The video acquisition module is used to acquire continuous video frames of the 500kV transmission line construction site; the target recognition and relationship generation module is used to identify construction physical objects, personnel targets, equipment targets, work area targets, and safety protection equipment targets from the continuous video frames, extract the category, continuous occurrence status, and spatial combination relationship of the construction physical objects, determine key construction physical objects from the construction physical objects, and generate candidate results for the construction stage and suspected violation relationship items based on the recognition and extraction results; the rule effective status generation module is used to generate the rule effective status for the construction stage based on the occurrence ratio of the candidate results for the construction stage within the time sliding window, the continuous occurrence status of the key construction physical objects, the spatial combination relationship, and the preset stage transition relationship, the rule effective status for the construction stage includes rule effective status, rule pending verification status, and rule invalid status; The cross-stage determination module is used to, when the construction stage rule effective state is the rule pending verification state, substitute the suspected violation relationship item into the violation rule of the current construction stage corresponding to the candidate construction stage result and the violation rule of the adjacent construction stage determined based on the preset stage transfer relationship for determination, and generate a cross-stage determination state, the cross-stage determination state including a cross-stage consistency state and a cross-stage conflict state; the evidence window control module is used to, according to the construction stage rule effective state and the cross-stage determination state, control the start, continuation, pause or reset of the continuous frame evidence window for the suspected violation relationship item, and determine the alarm permission state of the suspected violation relationship item according to the state of the continuous frame evidence window; the alarm output module is used to, when the alarm permission state is the alarm allowed state, generate a valid violation event and output a structured alarm evidence package.

[0008] Preferably, the target identification and relationship generation module is used to generate structured target data for each identified target. The structured target data includes target identifier, target category, target two-dimensional bounding box coordinates, target center point coordinates, and target timestamp. The spatial combination relationship is determined based on the target two-dimensional bounding box coordinates, the target center point coordinates, or the calibrated and converted construction plane distance.

[0009] Preferably, the target identification and relationship generation module is used to determine the key construction physical objects from the construction physical objects according to the preset mapping relationship between the construction stage and the construction physical objects; the key construction physical objects include the foundation pit objects corresponding to the foundation construction stage, the tower material hoisting objects corresponding to the tower erection stage, and the tension conductor objects corresponding to the line stringing stage.

[0010] Preferably, the rule validity status generation module is used to determine the rule validity status of the corresponding construction stage as the rule valid status when the occurrence ratio of the candidate results of the construction stage, the continuous occurrence status of the key construction physical object, the spatial combination relationship, and the preset stage transfer relationship all meet the validity conditions of the corresponding construction stage; when at least one of the validity conditions is not met, the rule validity status of the corresponding construction stage is determined as the rule pending verification status; when the candidate results of the construction stage conflict with the key construction physical object, the spatial combination relationship, or the preset stage transfer relationship, the rule validity status of the corresponding construction stage is determined as the rule invalid status.

[0011] Preferably, the suspected violation relationship item includes personnel target information, associated construction physical object information, associated work area information, safety protection equipment identification results, target trajectory sequence, suspected violation type, timestamp sequence, and spatial relationship parameters, which are used to characterize the suspected violation relationship formed by the personnel target relative to the construction physical object, the work area, or the safety protection equipment within the corresponding time slice.

[0012] Preferably, the cross-stage determination module is used to obtain the first determination result of the violation rule output of the current construction stage and the second determination result of the violation rule output of the adjacent construction stage; when the first determination result and the second determination result point to the same violation type, the cross-stage consistency state is generated; when the first determination result and the second determination result are inconsistent, the cross-stage conflict state is generated.

[0013] Preferably, the evidence window control module is used to start or extend the continuous frame evidence window when the construction phase rule is in the effective state of the rule and the suspected violation relationship item satisfies the violation rule corresponding to the current construction phase; to pause the continuous frame evidence window when the construction phase rule is in the effective state of the rule pending verification and the cross-phase judgment state is the cross-phase conflict state; and to reset the continuous frame evidence window when the construction phase rule is in the effective state of the rule invalid.

[0014] Preferably, the evidence window control module is used to stop accumulating the number of valid alarm frames corresponding to the suspected violation relationship item when pausing the continuous frame evidence window, and to retain the target trajectory sequence, spatial relationship parameters, on-site image fragments and the cross-stage conflict state corresponding to the suspected violation relationship item; and to clear the number of valid alarm frames corresponding to the suspected violation relationship item when resetting the continuous frame evidence window.

[0015] Preferably, the evidence window control module is used to determine the alarm permission state as the allowed alarm state when the continuous frame evidence window continuously records the suspected violation relationship item and no preset exclusion condition is triggered; to determine the alarm permission state as the pending review state when the continuous frame evidence window is paused or there is a cross-stage conflict state; and to determine the alarm permission state as the prohibited alarm state when the continuous frame evidence window is reset or the construction stage rule is ineffective.

[0016] Preferably, the structured alarm evidence package output by the alarm output module includes a stable construction phase, the effective status of construction phase rules, suspected violation relationship items, cross-phase judgment status, continuous frame evidence window time information, triggering rules, preset exclusion condition judgment results, on-site image fragments, and alarm level.

[0017] The technical effects and advantages of this invention are as follows:

[0018] In this invention, by identifying construction physical objects, personnel targets, equipment targets, work area targets, and safety protection equipment targets from continuous video frames, candidate results and suspected violation relationship items for each construction stage are generated. Based on the occurrence ratio of candidate results, the continuous occurrence status of key construction physical objects, spatial combination relationships, and preset stage transition relationships, the effective status of construction stage rules is generated. When a rule is pending verification, suspected violation relationship items are substituted into the violation rules corresponding to the current construction stage and adjacent construction stages for cross-stage judgment. Then, based on the effective status of construction stage rules and the cross-stage judgment status, the continuous frame evidence window is controlled to start, continue, pause, or reset, ultimately determining the alarm permission status and outputting a structured alarm evidence package. This reduces false alarms caused by premature rule activation, short-term false detection of targets, or cross-stage judgment conflicts during the switching of construction stages in 500kV transmission lines, ensuring that the output valid violation events have more stable construction stage basis, spatial relationship basis, and continuous frame evidence basis. Attached Figure Description

[0019] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts:

[0020] Figure 1 This is a block diagram of the system modules of the present invention;

[0021] Figure 2 This is a block diagram of the target identification and relationship generation module of the present invention;

[0022] Figure 3 This is a flowchart of the rule effective status generation module of the present invention;

[0023] Figure 4 This is a flowchart of the cross-stage determination module of the present invention;

[0024] Figure 5 This is a flowchart of the evidence window control module of the present invention;

[0025] Figure 6 This is a schematic diagram of the alarm output module and the structured alarm evidence package of the present invention. Detailed Implementation

[0026] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0027] Reference Figure 1-6 As shown, this invention provides a technical solution: a dynamic identification and alarm system for construction violations in 500kV transmission lines. This system is applicable to construction sites where foundation construction, tower erection, and line stringing of 500kV transmission lines are continuously changing. The system includes a panoramic video acquisition terminal deployed at the construction site, an edge computing device communicatively connected to the panoramic video acquisition terminal, a cloud server communicatively connected to the edge computing device, and a field management terminal for receiving alarm information.

[0028] The panoramic video acquisition terminal is used to acquire continuous video frames from the construction site. It can be deployed at construction area boundaries, tower erection observation points, tensioning fields, cross-area locations, or temporary command points. Edge computing devices are used to perform video frame preprocessing, construction physical object identification, construction status data generation, suspected violation relationship item generation, construction phase rule effectiveness determination, cross-phase determination, continuous frame evidence window control, and alarm permission status determination. The cloud server receives structured alarm evidence packages, manual review results, and new samples, and updates construction phase identification rules, violation determination rules, threshold parameters, and identification model parameters based on the received results. The site management terminal receives valid violation alarms, events awaiting review, and review suggestions.

[0029] In one implementation, the panoramic video acquisition terminal includes a rotatable camera or a multi-view panoramic camera, and the video acquisition frame rate can be 30fps. The edge computing device can be an embedded computing device with image inference capabilities, or it can be a temporary server set up at the construction site. The panoramic video acquisition terminal and the edge computing device can be integrated into a portable panoramic monitoring terminal, or they can be set up separately.

[0030] The system performs target detection and construction status extraction on consecutive video frames. (The sentence is incomplete and requires more context.) Frame video image For example, the system can accurately identify multiple elements such as construction physical objects, personnel targets, equipment targets, work area targets, and safety protection equipment targets.

[0031] The physical objects of construction include the foundation pit boundary, foundation formwork, rebar cage, tower components, derricks, hoisting slings, tensioning equipment, conductors, pulleys, crossing areas, and conductor traction paths. Personnel targets are further categorized into general construction workers, workers working at heights, personnel within the hoisting area, and personnel surrounding the tensioning area. Equipment targets encompass hoisting equipment, tensioning equipment, transport vehicles, various construction machinery, and temporary work facilities. Furthermore, safety protection equipment targets primarily involve safety helmets, safety belts, fall arrestors, insulating protective equipment, and warning signs.

[0032] For each valid target captured in the image, the system encapsulates it into structured target data. This data not only includes the target identifier, category, confidence level, frame number, and timestamp, but also precisely records the target's geometric position in the two-dimensional plane. The target's two-dimensional bounding box coordinates are represented as coordinate tuples:

[0033] ;

[0034] In the formula, and These represent the x-coordinate and y-coordinate of the top-left corner of the target bounding box in the image coordinate system, respectively. and Let x and y represent the x and y coordinates of the lower right corner of the target bounding box in the image coordinate system, respectively. Based on the target's two-dimensional bounding box coordinates, the... The center point coordinates of the target It can be represented as:

[0035] ;

[0036] To reduce the impact of instantaneous fluctuations in single-frame image recognition on construction phase judgments, the system uses a time-sliding window. As the unit of analysis. Time sliding window. Including continuous Frame video image. The settings can be configured based on the capture frame rate, ambient lighting conditions, construction area scale, and edge computing capabilities. For example, when the video capture frame rate is 30fps, A value of 150 can be used, corresponding to approximately a 5-second video clip. This parameter is for illustrative purposes only and can be adjusted according to actual site conditions in practical applications.

[0037] Furthermore, in order to quantify the candidate results for a certain construction stage Set a time sliding window Total number of frames: The number of frames in which this candidate result appears is The percentage of candidate results appearing in this construction stage is... The calculation formula is:

[0038] ;

[0039] in, Sliding window for time The total number of video frames contained within. Candidate results for the construction phase In the time sliding window The number of frames identified within. In one implementation, when the measured... When the percentage is greater than or equal to the first percentage threshold (e.g., preset to 80%, or fine-tuned according to the model performance of the on-site video quality), the system considers the candidate results for that construction stage to be within the current time sliding window. It exhibits a tendency to appear continuously.

[0040] Along with the percentage statistics, the system further tracks the continuous occurrence of key construction physical objects. For the foundation construction phase, key construction physical objects can include the foundation pit boundary, foundation formwork, reinforcement cage, and concrete pouring area. For the tower erection phase, key construction physical objects can include tower components, derricks, hoisting slings, and partially formed areas of the tower body. For the overhead line construction phase, key construction physical objects can include tensioning equipment, conductors, pulleys, crossing areas, and conductor traction paths. If the key construction physical objects corresponding to a certain construction phase appear within a time sliding window... If the number of consecutive occurrences of a key construction physical object reaches a preset number of frames, or if the proportion of occurrences reaches a second proportion threshold, the system will mark the key construction physical object as continuously occurring.

[0041] The system also calculates the spatial combination relationships between physical objects under construction. These spatial combination relationships are used to base the determination of construction stages on image recognition results and geometric relationship calculations. In one implementation, the system calculates the intersection-union ratio, centroid distance, shortest distance of the bounding box, normalized image distance, or calibrated and converted construction plane distance based on the target two-dimensional bounding box.

[0042] For two objectives and Their intersection and comparison The calculation formula is:

[0043] ;

[0044] in, Indicate target With the goal The overlapping area of ​​the bounding boxes. Indicate target With the goal The combined area of ​​the bounding boxes.

[0045] For two objectives and The center point, if the target The coordinates of the center point are ,Target The coordinates of the center point are Then the centroid distance between the two in the image coordinate system It can be represented as:

[0046] ;

[0047] If the system has already established a mapping relationship between image coordinates and construction plane coordinates through camera calibration parameters, ground reference points, or work area calibration frames, then This can be converted to the actual distance in the construction plane coordinate system; if no actual dimensional calibration has been performed, then... Distance can be based on the normalized width and height of the image.

[0048] For example, when determining the tower assembly construction status, the system checks whether a preset spatial combination relationship has been formed between the tower components and the gantry. If the normalized distance between the boundary frame of the tower component and the boundary frame of the gantry is less than a first distance threshold, and the tower component shifts along the hoisting direction in consecutive video frames, then the spatial relationship between the tower component and the gantry can be considered valid. Similarly, when determining the transition from foundation construction to tower assembly construction, if the pit boundary continues to appear, and tower components and hoisting slings begin to appear, but the spatial combination relationship between the tower component and the gantry is not yet stable, the system determines that the tower assembly construction rules do not yet meet the conditions for direct effectiveness.

[0049] The system generates the effective status of construction stage rules based on the candidate results of the construction stage, the continuous occurrence of key construction physical objects, the spatial combination relationships of construction physical objects, and the preset stage transition relationships. The effective status of construction stage rules includes rule valid status, rule pending verification status, and rule invalid status.

[0050] The preset stage transition relationship is used to characterize the permissible transition sequence between foundation construction, tower erection, and line stringing during the construction of a 500kV transmission line. In one implementation, foundation construction can transition to tower erection, and tower erection can transition to line stringing. When the system identifies a candidate construction stage, it does not directly set the violation rule corresponding to that candidate construction stage to a valid state. Instead, it combines the previous stable construction stage to determine whether the candidate construction stage conforms to the permissible transition sequence, and makes a judgment based on the state of the physical construction objects in consecutive video frames.

[0051] Specifically, if the candidate results for a certain construction stage are within the time sliding window If a construction phase continuously occurs within a given construction phase, and the key physical objects corresponding to that construction phase are in a state of continuous occurrence, and the spatial combination relationship between the key physical objects satisfies the site morphology of that construction phase, and the construction phase conforms to the preset phase transition relationship with the previous stable construction phase, then the system sets the violation rule corresponding to that construction phase to a valid rule state.

[0052] If a candidate result for a certain construction stage has appeared, but the key physical objects corresponding to that construction stage are insufficient, or the spatial combination relationship between the physical objects is not yet stable, or the transfer relationship between that construction stage and the previous stable construction stage is not yet fully satisfied, then the system will set the violation rule corresponding to that construction stage to a rule pending verification state.

[0053] If there are conflicts between the candidate results of a construction phase, the continuous occurrence of key construction physical objects, the spatial combination relationship of construction physical objects, and the preset phase transition relationship, the system will set the violation rule corresponding to that construction phase to an invalid state. For example, if the system identifies candidate results for overhead line construction in consecutive video frames but does not identify key construction physical objects such as tensioning equipment, conductor traction paths, or pulleys, and the previous stable construction phase is still the basic construction phase, the system can set the violation rule corresponding to the overhead line construction to an invalid state.

[0054] In this implementation, the valid, unverified, and invalid states of a rule are calculated jointly by the target category, the continuous occurrence of the target, spatial combination relationships, and stage transition relationships in consecutive video frames. Therefore, whether a construction stage rule is effective is associated with identifiable physical objects and image geometric relationships on site, thereby reducing misjudgments caused by relying solely on construction plan information or manual stage markers.

[0055] While identifying the construction status, the system correlates personnel targets, equipment targets, work area targets, and safety protection equipment targets in continuous video frames to form suspected violation relationship items. These suspected violation relationship items circulate in the system as structured data elements, used to represent the suspected violation relationships between a certain personnel target and construction equipment, work area, or safety protection equipment within a certain time slice.

[0056] In one implementation, suspected violation relationship item This can be represented as a structured data packet:

[0057] ;

[0058] in, Indicates the target identifier. Indicates the target category. Represents the coordinates of the target's two-dimensional bounding box. Indicates the coordinates of the target's center point. Indicates the identifier of the associated construction physical object. Indicates the associated work area identifier. This indicates the identification results of safety protective equipment. Represents the target trajectory sequence. Indicates the type of suspected traffic violation. Indicates the associated construction stage. Represents a timestamp sequence, Indicates spatial relationship parameters.

[0059] For example, a suspected violation relationship item can be represented as a structured data packet, whose specific fields include: target identifier as... The target category is construction workers, and the target's 2D bounding box coordinates are coordinate tuples. The associated physical object of construction is the hoisting equipment, the associated work area is the hoisting work area, the suspected violation type is the personnel staying in the hoisting area, the timestamp sequence is the time corresponding to several consecutive frames, and the spatial relationship parameter is the distance between the center point of the personnel target and the boundary of the hoisting area.

[0060] For suspected violation relationships that arise when the rules are valid, the system substitutes the suspected violation relationship into the violation rules corresponding to the current stable construction stage for judgment. If the suspected violation relationship satisfies the work area boundary rules, target space relationship rules, safety distance rules, and action continuity rules, the system activates the continuous frame evidence window corresponding to the suspected violation relationship.

[0061] For suspected violation relationships generated under the rule-pending verification state, the system does not directly initiate a valid alarm. Instead, it substitutes the same suspected violation relationship into the violation rules corresponding to the current candidate construction stage and the violation rules corresponding to the adjacent construction stage, generating a cross-stage judgment state. Cross-stage judgment states include cross-stage consistent states and cross-stage conflict states.

[0062] In one implementation, the violation rules corresponding to the current candidate construction stage output a first judgment result for suspected violation relationships, and the violation rules corresponding to adjacent construction stages output a second judgment result for the same suspected violation relationship. Both the first and second judgment results can include violation type, legal operation, inability to confirm, and confidence level. If both the first and second judgment results are of the same violation type and both confidence levels reach a preset confidence threshold, the system generates a cross-stage consistent state. If the first judgment result is a violation type, and the second judgment result is legal operation, a different violation type, or inability to confirm, the system generates a cross-stage conflict state. If neither the first nor the second judgment result can be confirmed, the system maintains a rule pending verification state and prohibits suspected violation relationships from directly generating valid alarms.

[0063] For example, if the current candidate construction stage rule output is "lifting violation" with a confidence level of 0.85, and the adjacent construction stage rule output is also "lifting violation" with a confidence level of 0.90, then the system sets the cross-stage judgment status corresponding to this suspected violation relationship item to a cross-stage consistent state. As another example, if the current candidate construction stage rule output is "lifting violation" with a confidence level of 0.85, and the adjacent construction stage rule output is "legal operation" with a confidence level of 0.75, then the system sets the cross-stage judgment status corresponding to this suspected violation relationship item to a cross-stage conflict state.

[0064] In other embodiments, the cross-stage determination state may not rely on a single confidence threshold, but may be jointly determined by the consistency of violation type, the consistency of target spatial relationship, the consistency of work area, and the occurrence of consecutive frames. As long as the processing object is the same suspected violation relationship item, and a cross-stage consistent state or a cross-stage conflict state is generated based on the output differences between the current candidate construction stage rules and the rules of adjacent construction stages, it can be used to achieve the technical objective of this invention.

[0065] The system controls the state transition of the continuous frame evidence window based on the effective status of the construction phase rules and the cross-phase judgment status. The continuous frame evidence window includes the start state, continuation state, paused state, and reset state.

[0066] When the construction phase rules are in a valid state, and the suspected violation relationship item meets the violation rules corresponding to the current stable construction phase, the system sets the continuous frame evidence window to the active state. When the continuous frame evidence window has been activated, and the same suspected violation relationship item is continuously identified in subsequent video frames, the target trajectory is not interrupted, the spatial relationship continues to meet the corresponding violation rules, and no exclusion conditions are triggered, the system sets the continuous frame evidence window to the continuation state.

[0067] When a construction phase rule is in a pending verification state and the cross-phase judgment state is consistent across phases, the system allows the continuous frame evidence window to continue in a pending confirmation mode. In the pending confirmation mode, the system records the target trajectory, on-site image fragments, and rule judgment results, but does not generate a valid violation alarm until the corresponding construction phase rule is converted to a valid rule state. Then, the system determines whether the alarm permission conditions are met based on the accumulated evidence.

[0068] When a construction phase rule is in a pending verification state and the cross-phase judgment state is a cross-phase conflict state, the system will put the continuous frame evidence window into a paused state. In the paused state, the system will pause the accumulation of valid alarm frames corresponding to the suspected violation relationship item, while maintaining the evidence window number, starting frame number, current frame number, and recorded timestamp sequence of the suspected violation relationship item unchanged. Simultaneously, the system will store the recorded target trajectory, spatial relationship parameters, on-site image fragments, and cross-phase conflict results in the temporary cache of the edge computing device as evidence for subsequent verification, and record the paused frame number, pause time, conflict output result, and the construction phase corresponding to the conflict. Suspected violation relationship items in the paused state are prohibited from triggering valid violation alarm push commands.

[0069] When a construction phase rule is invalid, or when a suspected violation does not match the rules governing the current construction phase's work area, target space relationship, or safety distance, the system resets the continuous frame evidence window. In the reset state, the system clears the number of valid alarm frames corresponding to the suspected violation and excludes the suspected violation from the valid alarm judgment process.

[0070] The system further determines the alarm permission status of suspected violation items based on the status of the continuous frame evidence window. The alarm permission status includes alarm allowed, pending review, and alarm prohibited.

[0071] When the continuous frame evidence window is in a continuous state, and the same suspected violation relationship item is continuously recorded within a preset number of continuous video frames, and the suspected violation relationship item meets the work area rules, target space relationship rules, safety distance rules and action continuity rules corresponding to the current stable construction stage, and no exclusion conditions are triggered, the system sets the alarm permission status of the suspected violation relationship item to the alarm permission status.

[0072] In one implementation, the preset number of consecutive video frames can be 30 frames, 60 frames, or other values. Taking a 30fps acquisition frame rate as an example, 30 consecutive frames correspond to approximately 1 second. If a construction worker remains within the hoisting operation area for 30 consecutive frames, and the spatial distance between them and the hoisting equipment is less than the corresponding safe distance threshold, and no legal operation status is detected, the system can set the alarm permission status of this suspected violation to an allowed alarm status. This value is only an example; in actual applications, it can be adjusted according to the on-site shooting distance, operation type, and alarm sensitivity.

[0073] When the continuous frame evidence window is paused, or when the construction phase rule is in a rule-pending-verification state and there is a cross-phase conflict, the system sets the alarm permission status of the suspected violation relationship item to a pending-review state. Suspected violation relationship items in the pending-review state do not generate valid violation alarms, but can generate pending-review events with conflict output results, on-site image clips, and review suggestions.

[0074] When the continuous frame evidence window is in a reset state, or the construction phase rule is in an invalid state, or the suspected violation relationship item does not match the work area, target space relationship, or safety distance rule corresponding to the current construction phase, the system sets the alarm permission status of the suspected violation relationship item to the alarm prohibition status. Suspected violation relationship items in the alarm prohibition status will not push valid violation alarms to the site management terminal.

[0075] In this embodiment, the exclusion criteria include image occlusion, image blurring, short-term target loss, target trajectory interruption, legal operation status, and cross-stage conflict status. Each exclusion criterion can be determined through image processing and target tracking results.

[0076] Specifically, the system uses image gradient values ​​or Laplacian variance to keenly detect "image blur." Once the image quality falls below a certain threshold, the current frame immediately becomes invalid. For "image occlusion," the system quantifies the impact based on the visible area ratio of the target bounding box and the keypoint missing rate. If the image quality falls below the visibility threshold or exceeds the missing threshold, occlusion is decisively determined. In the face of "short-term target loss," the system no longer mechanically resets but instead uses tracking algorithms to predict the trajectory. If the target is only hidden in a few frames but its predicted location remains within a reasonable neighborhood, the system leniently presses the cumulative "pause button." Once the target reappears and successfully matches within the recovery period, the evidence record is fully restored. Conversely, if the target's center point displacement undergoes an illogical abrupt change or the tracking number is completely broken, the "trajectory interruption" clause will be triggered. Furthermore, any compliant protective gear, legal work station authorization, and actions that meet the current stage's safety regulations constitute strong "legal work exclusion conditions," serving as a solid shield against false alarms. Finally, the aforementioned "cross-stage conflict state" is itself a crucial exclusion condition; once triggered, the relevant item will be diverted to the pending review channel.

[0077] The following describes the working process of this invention using a scenario of transitioning from foundation construction to tower assembly.

[0078] At a 500kV transmission line construction site, the previous stable construction phase involved foundation construction. Continuous video frames captured by the panoramic video acquisition terminal still showed unfilled foundation pit boundaries, while the first batch of tower components and hoisting rigging arrived at the construction site. Within a time sliding window W, the system detected the continued appearance of the foundation pit boundary and the initial emergence of tower components. However, the spatial arrangement between the tower components and the gantry crane was not yet stable, and the proportion of candidate tower assembly construction results did not meet the conditions for a valid rule state. Therefore, the system maintained the foundation construction rule as valid and set the tower assembly construction rule to a pending verification state.

[0079] Subsequently, the system detected a construction worker active near the tower material unloading area and generated a suspected violation relationship item. This suspected violation relationship item includes the worker's target identifier, the worker's target bounding box, the tower material component bounding box, the target trajectory sequence, the associated work area, the suspected violation type, and the timestamp sequence. Since the tower assembly construction rules are in a pending verification state, the system does not directly generate a valid violation alarm. Instead, it substitutes the same suspected violation relationship item into both the tower assembly construction rules and the foundation construction rules.

[0080] In this scenario, the tower erection construction rules might detect hoisting slings and personnel approaching the tower materials and output a suspected violation result indicating personnel lingering in the hoisting area; the foundation construction rules, on the other hand, would output a legal operation result based on the personnel being in the material unloading auxiliary area, the pit boundary not being intruded upon, and the personnel maintaining a safe distance from the dangerous boundary of the pit. Because the two sets of rules output different judgment results for the same personnel target, the same work area, and the same time slice, the system generates a cross-stage conflict state.

[0081] After generating a cross-stage conflict state, the system pauses the continuous frame evidence window corresponding to the suspected violation relationship item, halts the accumulation of valid alarm frames corresponding to the suspected violation relationship item, and stores the recorded target trajectory, spatial relationship parameters, on-site image fragments, and cross-stage conflict results in the temporary buffer of the edge computing device. Simultaneously, the system records the pause frame number, pause time, tower construction rule output results, foundation construction rule output results, and corresponding on-site image fragments. Subsequently, the system sets the alarm permission status of the suspected violation relationship item to a pending review status and outputs the pending review event to the on-site management terminal, without pushing valid violation alarms.

[0082] When tower components, scaffolding, and hoisting slings continuously appear in subsequent video frames, and the spatial arrangement between tower components and scaffolding remains stable, the candidate results for tower assembly construction reach a preset percentage within the time sliding window, and the preset stage transition relationship from foundation construction to tower assembly construction is met, the system sets the tower assembly construction rule to a valid state. At this point, for any newly generated suspected violation items, the system restarts the corresponding continuous frame evidence window according to the tower assembly construction rule, and allows the generation of a valid violation alarm after satisfying the spatial relationship, continuous frame count, and exclusion conditions.

[0083] Through the above process, the system reduces false alarms during the transition from foundation construction to tower erection due to the short-term appearance of tower materials, premature activation of stage rules, or conflicts in the output of rules from different stages. Compared with methods that output alarms based solely on single-frame recognition results or fixed construction stage models, this implementation establishes effective violation alarms on the continuous control of the construction stage rule activation status, cross-stage judgment status, continuous frame evidence window status, and alarm permission status, ensuring that the output alarm results have stable stage basis, spatial relationship basis, and continuous frame evidence basis. Therefore, the system can reduce the continuous false alarms caused by target feature mismatch, premature activation of stage rules, or short-term recognition fluctuations during periods of rapid change in the construction scene when existing alarms are based on single-stage models or single-frame target detection results.

[0084] When a suspected violation is in an alert-allowed state, the system generates a valid violation event and outputs a structured alarm evidence package. The structured alarm evidence package includes the stable construction phase, the effective status of construction phase rules, suspected violation items, cross-phase judgment status, start and end times of continuous frame evidence windows, personnel target trajectories, target spatial relationships, trigger rules, exclusion condition judgment results, on-site image clips, alarm levels, and review suggestions. After receiving the structured alarm evidence package, the on-site management terminal can display the location of the violation, the violation type, the trigger rules, video clips, and review suggestions. After receiving the structured alarm evidence package and the manual review results, the cloud server can update the construction phase identification rules, violation rules, threshold parameters, and identification model parameters.

[0085] In other implementations, the system can also receive hoisting equipment attitude sensing data, tensioning equipment operating status data, or positioning data to assist in verifying the status of construction physical objects obtained from image recognition. However, this auxiliary data is not the sole basis for generating the effective status of construction stage rules; the system still uses the category, spatial distribution, continuous occurrence status, and change sequence of construction physical objects in continuous video frames as the main basis for judgment.

[0086] The time sliding window length, percentage threshold, number of consecutive frames, distance threshold, sharpness threshold, and confidence threshold described in this embodiment are all exemplary parameters. Those skilled in the art can adjust these parameters based on the camera's frame rate, shooting distance, construction area scale, on-site lighting conditions, target recognition model accuracy, and alarm sensitivity. As long as such adjustments are still based on the state transition relationship between the construction phase rule's effective state, cross-phase judgment state, consecutive frame evidence window state, and alarm permission state, and effectively control alarms for suspected violation items, they should fall within the scope of this invention.

[0087] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

A dynamic identification and alarm system for illegal construction activities in 1.500kV transmission line construction is characterized by, include: It includes a video acquisition module, a target recognition and relationship generation module, a rule effective status generation module, a cross-stage judgment module, an evidence window control module, and an alarm output module; The video acquisition module is used to acquire continuous video frames of the 500kV transmission line construction site; the target recognition and relationship generation module is used to identify construction physical objects, personnel targets, equipment targets, work area targets, and safety protection equipment targets from the continuous video frames, extract the category, continuous occurrence status, and spatial combination relationship of the construction physical objects, determine key construction physical objects from the construction physical objects, and generate candidate results for the construction stage and suspected violation relationship items based on the recognition and extraction results; the rule effective status generation module is used to determine the occurrence ratio of the candidate results for the construction stage and the key construction physical objects based on the time sliding window. The continuous occurrence state of the image, the spatial combination relationship, and the preset stage transition relationship generate the effective state of the construction stage rules. The effective state of the construction stage rules includes the rule valid state, the rule pending verification state, and the rule invalid state. The cross-stage judgment module is used to substitute the suspected violation relationship item into the violation rules of the current construction stage corresponding to the candidate result of the construction stage and the violation rules of the adjacent construction stage determined based on the preset stage transition relationship when the effective state of the construction stage rule is the rule pending verification state, and to make a judgment to generate a cross-stage judgment state. The cross-stage judgment state includes a cross-stage consistent state and a cross-stage conflict state. The evidence window control module is used to control the start, continuation, pause, or reset of the continuous frame evidence window for the suspected violation relationship item according to the effective status of the construction stage rules and the cross-stage judgment status, and to determine the alarm permission status of the suspected violation relationship item according to the status of the continuous frame evidence window; the alarm output module is used to generate a valid violation event and output a structured alarm evidence package when the alarm permission status is an allowed alarm status.

2. The dynamic identification and alarm system for illegal construction activities of 500kV transmission lines according to claim 1, characterized in that: The target identification and relationship generation module is used to generate structured target data for each identified target. The structured target data includes target identifier, target category, target two-dimensional bounding box coordinates, target center point coordinates, and target timestamp. The spatial combination relationship is determined based on the target two-dimensional bounding box coordinates, the target center point coordinates, or the calibrated and converted construction plane distance.

3. The dynamic identification and alarm system for construction violations of 500kV transmission lines according to claim 1, characterized in that: The target identification and relationship generation module is used to determine the key construction physical objects from the construction physical objects according to the preset mapping relationship between the construction stage and the construction physical objects; the key construction physical objects include the foundation pit objects corresponding to the foundation construction stage, the tower material hoisting objects corresponding to the tower erection stage, and the tension conductor objects corresponding to the line stringing stage.

4. The dynamic identification and alarm system for illegal construction activities of 500kV transmission lines according to claim 1, characterized in that: The rule validity status generation module is used to determine the rule validity status of the corresponding construction stage as the rule valid status when the occurrence ratio of candidate results in the construction stage, the continuous occurrence status of the key construction physical object, the spatial combination relationship, and the preset stage transfer relationship all meet the validity conditions of the corresponding construction stage; when at least one of the validity conditions is not met, the rule validity status of the corresponding construction stage is determined as the rule pending verification status; when there is a conflict between the candidate results in the construction stage and the key construction physical object, the spatial combination relationship, or the preset stage transfer relationship, the rule validity status of the corresponding construction stage is determined as the rule invalid status.

5. The dynamic identification and alarm system for construction violations of 500kV transmission lines according to claim 1, characterized in that: The suspected violation relationship items include personnel target information, associated construction physical object information, associated work area information, safety protection equipment identification results, target trajectory sequence, suspected violation type, timestamp sequence, and spatial relationship parameters, which are used to characterize the suspected violation relationship formed by the personnel target relative to the construction physical object, the work area, or the safety protection equipment within the corresponding time slice.

6. The dynamic identification and alarm system for construction violations of 500kV transmission lines according to claim 1, characterized in that: The cross-stage determination module is used to obtain the first determination result of the violation rule output of the current construction stage and the second determination result of the violation rule output of the adjacent construction stage. When the first determination result and the second determination result point to the same violation type, the cross-stage consistent state is generated; when the first determination result and the second determination result are inconsistent, the cross-stage conflict state is generated.

7. The dynamic identification and alarm system for illegal construction activities in 500kV transmission line construction as described in claim 1, characterized in that: The evidence window control module is used to start or extend the continuous frame evidence window when the construction phase rule is in the effective state and the suspected violation relationship item satisfies the violation rule corresponding to the current construction phase; to pause the continuous frame evidence window when the construction phase rule is in the unverified state and the cross-phase judgment state is the cross-phase conflict state; and to reset the continuous frame evidence window when the construction phase rule is in the invalid state.

8. The dynamic identification and alarm system for construction violations of 500kV transmission lines according to claim 7, characterized in that: The evidence window control module is used to stop accumulating the number of valid alarm frames corresponding to the suspected violation relationship item when pausing the continuous frame evidence window, and to retain the target trajectory sequence, spatial relationship parameters, on-site image fragments and cross-stage conflict status corresponding to the suspected violation relationship item; when resetting the continuous frame evidence window, it clears the number of valid alarm frames corresponding to the suspected violation relationship item.

9. The dynamic identification and alarm system for construction violations of 500kV transmission lines according to claim 7, characterized in that: The evidence window control module is used to determine the alarm permission state as the alarm permission state when the continuous frame evidence window continuously records the suspected violation relationship item and no preset exclusion condition is triggered. When the continuous frame evidence window is paused or the cross-stage conflict exists, the alarm permission status is determined to be pending review; when the continuous frame evidence window is reset or the construction stage rule is ineffective, the alarm permission status is determined to be prohibited alarm status.

10. The dynamic identification and alarm system for construction violations of 500kV transmission lines according to claim 9, characterized in that: The structured alarm evidence package output by the alarm output module includes the stable construction stage, the effective status of construction stage rules, suspected violation relationship items, cross-stage judgment status, continuous frame evidence window time information, triggering rules, preset exclusion condition judgment results, on-site image fragments, and alarm level.

Images