Method and system for collaborative monitoring of illegal behaviors of ai-driven dump trucks

Abstract

This invention proposes an AI-driven collaborative monitoring method and system for illegal dump truck behavior. The method includes: collecting three-dimensional acceleration and three-dimensional angular velocity data of the dump truck using an onboard inertial sensor to construct an inertial behavior sequence; establishing an individualized behavior model based on historical normal behavior data to identify abnormal behavior periods that deviate from the norm and generate anomaly scores; when abnormal behavior is detected, triggering an onboard camera to collect environmental images and determining the legality of the scene where the dump truck is located using a visual semantic recognition model; fusing the anomaly scores and scene legality markers to generate a final violation judgment result through a joint judgment model; and triggering a multi-level linkage response mechanism based on the violation judgment result, including real-time onboard feedback, linkage with regional monitoring equipment, and uploading and storing violation data.

Description

Technical Field

[0001] This invention belongs to the field of AI-driven technology, and in particular relates to a collaborative monitoring method and system for illegal behavior of dump trucks based on AI. Background Technology

[0002] Dump trucks, as large transport vehicles commonly used in urban infrastructure construction, are responsible for transporting building materials such as slag and gravel. However, due to their large loading capacity, frequent operation, and wide operating area, often operating in areas with weak supervision, such as the outskirts of urban construction sites and nighttime construction sites, dump trucks are prone to becoming high-risk sources of traffic violations and safety hazards. In actual management, dump truck violations occur frequently, such as dumping construction waste in undesignated areas, overloading, illegally detouring on supervised routes, entering restricted areas, and illegal parking. These behaviors not only disrupt urban environmental order but also seriously threaten road traffic safety.

[0003] Current mainstream methods for regulating dump trucks typically rely on fixed video surveillance equipment deployed on urban roads or around construction sites, RFID check-in mechanisms, GPS trajectory analysis systems, and manual patrols. However, these methods generally have serious limitations. On the one hand, fixed cameras have limited coverage and cannot cover the entire transportation route, especially in remote areas, at night, or in construction dead zones where "blind spots" easily appear, making it difficult to detect key violations. On the other hand, traditional GPS route analysis lacks accuracy and cannot accurately identify whether a vehicle has actually entered the construction site or unloaded in an illegal area; even worse, some drivers bypass geofences or camera recognition by "checking in and then turning around and leaving" or "skirting the edge of the rules," rendering the monitoring system almost powerless when faced with "formal compliance but substantive violation." In addition, existing systems usually construct their analysis logic from a "single-vehicle perspective," lacking the ability to model and judge multi-vehicle coordinated violations, making it difficult to identify highly concealed and regular behaviors such as "clustered bypassing" and "group dumping."

[0004] Against this backdrop, the industry urgently needs a new technological system that can overcome the limitations of existing reliance on visual / location information, possess strong environmental adaptability, robust behavior judgment, and support vehicle-to-vehicle / vehicle-to-infrastructure cooperative perception. Furthermore, this system must not only be able to identify high-risk behaviors of individual vehicles but also perceive the semantic context in which the behavior occurs, determining whether it constitutes a genuine violation of "behavioral violation + illegal scenario," thereby improving the accuracy of judgment and the credibility of law enforcement. In summary, there is currently a lack of a technological solution that can effectively integrate multi-source perception, scene semantic understanding, and collaborative analysis of group behavior to support the intelligent monitoring and governance needs of dump truck violations. Summary of the Invention

[0005] The purpose of this invention is to propose an AI-driven collaborative monitoring method and system for illegal dump truck behavior. This system aims to overcome "perception blind spots" and "formal compliance disguises" by constructing an intelligent recognition framework with physical behavior modeling and scene semantic understanding capabilities, thereby achieving joint judgment of the actual behavior of dump trucks and their surrounding environment.

[0006] To achieve the above objectives, a first aspect of the present invention provides an AI-driven collaborative monitoring method for illegal behavior of dump trucks, the method comprising the following steps:

[0007] S1. Collect three-dimensional acceleration and three-dimensional angular velocity data of dump trucks through vehicle-mounted inertial sensors to construct an inertial behavior sequence; establish an individualized behavior model based on historical normal behavior data, identify abnormal behavior periods that deviate from the norm, and generate anomaly scores.

[0008] S2. When abnormal behavior occurs, the vehicle-mounted camera is triggered to collect environmental images, and the legality of the scene where the dump truck is located is determined by the visual semantic recognition model.

[0009] S3. Integrate the aforementioned anomaly score and scene legality marker, and generate the final violation judgment result through a joint judgment model;

[0010] S4. Based on the violation determination result, trigger a multi-level linkage response mechanism, including real-time vehicle feedback, linkage of regional monitoring equipment, and uploading and storage of violation data.

[0011] Furthermore, the individualized behavior model is trained using a Gaussian mixture model, and the anomaly score is determined by calculating the deviation of the current behavior sequence from the historical distribution; the deviation is determined based on the negative log-likelihood value.

[0012] If the deviation is higher than the average score of normal behavior, it indicates that the behavior deviates significantly from the norm and is therefore marked as a possible abnormal behavior.

[0013] Compare the deviation with the mean of historical scores to determine if it is an abnormal period:

[0014] If the deviation is greater than the sum of the historical average score and the preset threshold, the current period is marked as an abnormal behavior period.

[0015] Furthermore, the input of the visual semantic recognition model is the environmental image, and the output scene labels include construction site, no-dumping zone and road, and the environmental legality mark is dynamically generated based on the scene labels.

[0016] Furthermore, the environment legality marker is dynamically generated based on the scene tag, specifically including:

[0017] Define a scene legality marker. If the construction site is a no-dumping zone, the scene legality marker is illegal, indicating that the dump truck's behavior occurred in a no-dumping zone and violated relevant regulations. If the construction site is a construction site, the scene legality marker is legal, indicating that the area is a legal unloading site.

[0018] Furthermore, the joint judgment model generates a final judgment result by weighted behavior anomaly scoring and scene legality marking, wherein the scene illegality marking triggers additional penalty weights; the final judgment result is either violation or normal behavior.

[0019] Furthermore, the in-vehicle real-time feedback includes at least one of voice warnings and driving route restrictions, and the linkage of the area monitoring equipment includes triggering roadside cameras and drones to conduct collaborative monitoring of the violation area.

[0020] Furthermore, S4 also includes:

[0021] Calculate the regional risk index, which dynamically adjusts the allocation of monitoring resources based on the number of vehicles violating regulations and the distribution of violations in the region.

[0022] Furthermore, when the regional risk index exceeds a preset threshold, the deployment density of monitoring equipment in the region is automatically increased, and an enforcement report is generated and pushed to the management platform.

[0023] Furthermore, the inertial sensor includes a three-axis accelerometer and a gyroscope, with a data acquisition frequency of no less than once per second; the vehicle-mounted camera is a 360-degree panoramic camera, and the image acquisition is strictly synchronized with the timestamp of the abnormal behavior period.

[0024] A second aspect of the present invention provides an AI-driven collaborative monitoring system for illegal dump truck behavior, the system comprising:

[0025] The vehicle-mounted terminal module is used to collect three-dimensional acceleration and three-dimensional angular velocity data of dump trucks through vehicle-mounted inertial sensors, construct an inertial behavior sequence, establish an individualized behavior model based on historical normal behavior data, identify abnormal behavior periods that deviate from the norm, and generate anomaly scores.

[0026] The cloud-based analytics platform is used to trigger the vehicle-mounted camera to capture environmental images when abnormal behavior is detected, and to determine the legality of the scene in which the dump truck is located through a visual semantic recognition model.

[0027] The multi-level response module generates the final violation judgment result of the violation behavior by integrating the anomaly score and the scene legality mark through a joint judgment model;

[0028] The dynamic monitoring and scheduling module triggers a multi-level linkage response mechanism based on the violation determination result, including in-vehicle real-time feedback, linkage of area monitoring devices, and uploading and archiving of violation data.

[0029] The beneficial technical effects of the present invention are at least as follows:

[0030] In view of the limitations of the existing muck truck supervision system in terms of recognition range, judgment ability, and coordination mechanism, the present invention proposes a collaborative monitoring method and system for muck truck violation behaviors driven by artificial intelligence. The system aims to break through the "perception blind spot" and "formal compliance disguise" by constructing an intelligent recognition framework with the capabilities of physical behavior modeling and scene semantic understanding to achieve the joint determination of the actual behaviors of muck trucks and their occurrence environments. At the same time, the present invention introduces a multi-vehicle coordination mechanism. After a single-vehicle high-risk event occurs, it can trigger the linkage perception of surrounding vehicles and environmental nodes, thereby achieving a collaborative monitoring effect with a larger scope and stronger robustness. The core innovation of the present invention is to fuse multi-dimensional perception data (such as vehicle inertial behaviors, visual semantic features, spatio-temporal trajectories) to construct a comprehensive determination model, and by constructing a dynamic risk diffusion mechanism, achieve the chain tracking ability from "local abnormal behaviors" to "system-level high-risk events". Brief Description of the Drawings

[0031] The present invention is further described with reference to the accompanying drawings. However, the embodiments in the drawings do not constitute any limitation to the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the following drawings.

[0032] Figure 1 It is a flowchart of the collaborative monitoring method for muck truck violation behaviors driven by AI according to the present invention. Detailed Embodiments

[0033] The embodiments of the present invention are described in detail below. The examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals represent the same or similar elements or elements with the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are only used to explain the present invention and should not be construed as a limitation to the present invention.

[0034] As Figure 1 shown, the collaborative monitoring method for muck truck violation behaviors driven by AI provided by the embodiment of the present invention includes:

[0035] S1. Collect three-dimensional acceleration and three-dimensional angular velocity data of the muck truck through in-vehicle inertial sensors to construct an inertial behavior sequence; establish an individualized behavior model based on historical normal behavior data, identify abnormal behavior periods deviating from the norm, and generate an abnormal score.

[0036] Specifically, the purpose of this step is to detect abnormal behavior of dump trucks during operation, especially high-risk operations such as illegal dumping, by using real-time acceleration and angular velocity data collected by the vehicle-mounted inertial measurement unit (IMU). Since illegal behavior of dump trucks is often accompanied by significant changes in inertia, this step identifies time periods that deviate significantly from normal behavior patterns by establishing individualized inertial behavior models, serving as candidate regions for subsequent scenario determination.

[0037] Furthermore, each dump truck is equipped with a three-axis IMU sensor to collect real-time acceleration and angular velocity data during acceleration, braking, and steering. Specifically, the sensor records three-dimensional acceleration (α) per second. t ) and angular velocity (ω t This constitutes the vehicle's inertial behavior sequence.

[0038] The data at each time point t includes:

[0039] These represent the accelerations along the x, y, and z axes, respectively.

[0040] These represent the angular velocities along the x, y, and z axes, respectively.

[0041] These acceleration and angular velocity data will be used for subsequent behavior analysis. For ease of processing, a sliding window method is employed. Assuming a window length of τ, the data from time point t-τ to t constitutes a complete inertial behavior segment X. t ={a t ,ω t}

[0042] Furthermore, for each dump truck, this invention trains an individualized behavior model using its historical normal operation data (excluding violations). This model employs a Gaussian Mixture Model (GMM), where the normal behavior pattern of each vehicle is represented as a weighted Gaussian distribution.

[0043] For each new behavioral period X t This invention compares the deviation of the vehicle with its historical normal behavior distribution to calculate its "deviation" within that distribution. This invention uses the negative log-likelihood value as the deviation score s. t The formula is as follows:

[0044] s t =-logf θ (X t (1)

[0045] Among them, f θ (X t) represents the current behavior segment X calculated by the GMM model. t The likelihood value is θ, which is the model parameter and is adaptively adjusted according to the normal behavior of each vehicle after training.

[0046] If s t The value is significantly higher than the mean score for normal behavior, indicating that the behavior deviates significantly from the norm and is therefore marked as a possible abnormal behavior.

[0047] Furthermore, in order to identify potential violations, the present invention uses the current deviation score s t Average of historical ratings Compare the data to determine if it falls within an abnormal time period. Set a threshold δ, when... When the current time period is identified as a candidate for abnormal behavior, it is determined that the current time period is an abnormal behavior candidate.

[0048] Specifically, δ is 1.5 times the historical standard deviation of the ratings, i.e., δ = 1.5 × σ s , where σ s It is the standard deviation of historical scores. If s t If the behavior exceeds this threshold, the time period is marked as abnormal, and X t This abnormal period was recorded and passed on to subsequent steps.

[0049] The output of this step includes: Abnormal behavior period X t That is, a time series containing acceleration and angular velocity data; abnormal behavior score s t This is used for subsequent compliance assessments.

[0050] S2. When abnormal behavior occurs, the vehicle-mounted camera is triggered to collect environmental images, and the legality of the scene where the dump truck is located is determined by the visual semantic recognition model.

[0051] Specifically, the purpose of this step is to further determine whether these behaviors occurred within a legal operating environment by combining the abnormal behavior periods detected in the first step. Dump trucks may exhibit certain abnormal behaviors under normal circumstances, such as sudden braking or acceleration, but whether these behaviors violate regulations depends on the environment in which they occur. Therefore, this step uses semantic scene understanding by combining inertial sensor data and image data collected by the onboard camera to determine whether the behavior occurred in a permitted area (such as a construction site) or a prohibited unloading area (such as a no-dumping zone). This not only helps to further confirm whether the behavior is in violation of regulations but also enhances the accuracy and environmental adaptability of the system.

[0052] Furthermore, in step one, the present invention identifies an abnormal behavior period X. tThe acceleration and angular velocity sequences during this period indicate the abnormal behavior of the dump truck. Using the timestamps of these abnormal behavior periods, the onboard camera automatically captures corresponding images. t This ensures that the images are synchronized with the time period of abnormal behavior. The image acquisition module is typically an in-vehicle 360-degree panoramic camera, which can capture the actual driving status of the vehicle in different environments.

[0053] Here, X t As candidate time periods for abnormal behavior, they identify the time periods during which vehicles may commit violations, while I t These are captured environmental image data related to these time periods. The image data will be used for scene analysis by a subsequent visual recognition module.

[0054] Semantic scene analysis and environmental legality judgment

[0055] Through image I t It is then input into the trained visual semantic recognition model g. φ (I t In this model, the scene label c corresponding to the image can be output. t This indicates the environment type in the image. Possible environment type labels include:

[0056] c t =Construction site: This indicates that dump trucks are unloading goods normally at the construction site;

[0057] c t =No-dumping zone: This indicates that the dump truck has entered an area where unloading is prohibited;

[0058] c t = Road: Indicates that dump trucks travel on ordinary roads.

[0059] Through image model g φ (I t This invention can accurately identify the environment in which dump trucks are located. The model has been trained on a large scale and can handle images in complex scenes, such as image classification of urban construction sites, no-dumping zones, and road areas.

[0060] After determining the environmental label c t The system then performs further environmental compliance checks. This invention defines a compliance flag l. t If c t =Forbidden zone, then l t =Illegal, indicating that the dump truck's actions occurred in a prohibited dumping area, violating relevant regulations; if c t =Construction site, then l t =Legal, indicating that the area is a legal unloading site.

[0061] This compliance assessment process not only improves the system's accurate understanding of vehicle behavior, but also closely integrates visual scene and behavioral data, eliminating the misjudgment problem that may be caused by data from a single sensor.

[0062] Furthermore, in step one, the present invention obtains a behavioral anomaly score s. t In this step, the present invention obtains the legality mark l t To comprehensively assess the compliance of behavior and environment, this invention introduces a joint judgment model h. ψ (s t ,l t This model will combine s t and l t Generate the final violation determination result y t The formula is as follows:

[0063]

[0064] In this formula: s t The abnormal behavior score obtained from step one indicates whether the behavior deviates from the normal behavior pattern; t This indicates whether the current environment allows the behavior to occur; w1 and w2 are weighting coefficients used to balance the impact between abnormal behavior and scene legality; λ is a regularization factor, especially when l t = When it is illegal, λ enhances the impact of the illegal environment on the final violation judgment by adding penalty items.

[0065] This approach not only considers the anomalousness of the behavior itself but also enhances the role of environmental legitimacy in determining violations. For example, when the dump truck's behavior is rated as s... t Very high, but if its environment is designated as an "illegal" area (l t =Illegal), then the judgment of violation will be increased. This addition of regularization improves the system's sensitivity to the environment.

[0066] Finally, the environmental scene label c is obtained. t This indicates the type of environment in which the dump truck is located (e.g., construction site, no-dumping zone, etc.); legality marker l t This indicates whether the environment permits the dump truck's current behavior; the final violation determination is y. t This serves as input for subsequent steps (such as regulatory response). The output will be passed to the next step of processing (such as collaborative monitoring and enforcement response) to determine whether further enforcement action is needed.

[0067] This step directly addresses a key issue in monitoring dump truck violations—the difficulty of determining environmental compliance based solely on behavioral data. By introducing semantic scene recognition and environmental legitimacy assessment, this invention effectively eliminates the limitations of single-sensor data and combines vehicle behavior with actual environmental information for joint judgment. This approach not only improves the accuracy of violation determination but also enhances the system's adaptability to complex scenarios, enabling more accurate detection of dump truck violations in more realistic environments.

[0068] S3. Integrate the abnormal score and the scene legality mark, and generate the final violation judgment result through the joint judgment model.

[0069] Specifically, step three aims to further integrate the data obtained in the previous steps, using joint modeling of behavior and environment to accurately determine the illegal behavior of dump trucks. This is achieved by combining the behavioral anomaly score extracted in step one. t And the scene legality label obtained in step two t (and scene tag c) t The final output of this step is the judgment result y. t To determine whether any violations have occurred.

[0070] Combining the outputs of these two steps, this invention performs joint modeling in this step, fusing s t and l t (and c) t The final determination is made by:

[0071] To comprehensively assess the severity of abnormal behavior and the legality of the scenario, this invention constructs a joint judgment model h. ψ (s t ,l t ,c t The goal of this model is to output the final violation judgment y. t .

[0072] The joint judgment model uses behavioral anomaly scoring s t Environmental legality marker t and environmental label c t A comprehensive assessment will be conducted. The combination of these three factors will more accurately determine whether a dump truck has engaged in any illegal activities.

[0073] The calculation formula for the joint model is as follows:

[0074]

[0075] Among them, y t For the final determination of the violation, y t =1 indicates a violation, yt =0 indicates normal behavior. t This is the abnormal behavior score obtained from step one, reflecting the degree of deviation between the current behavior and normal behavior. t The older they are, the more abnormal their behavior becomes. t This is the legality flag obtained from step two, indicating whether the current environment allows the dump truck's behavior. If l t =Illegal, meaning the act occurred in a no-dumping zone or other illegal area. c t The scene label, obtained from step two, indicates the specific type of the current environment. For example, the dump truck is at a construction site or in a no-dumping zone. w1, w2, and w3 are weighting coefficients used to control the relative importance of behavior scoring, environmental legality, and scene label. λ is a regularization factor, especially in l t = When illegal, λ enhances the response to illegal regions, ensuring that violations in illegal regions receive sufficient attention. α is a threshold weighting term; when s t When a certain critical value is exceeded, A value of 1 indicates that the behavior is highly abnormal, and the system gives it additional weight.

[0076] By introducing a regularization factor λ, this invention can strengthen the determination of illegal areas, especially when dump trucks enter no-dumping zones or other illegal areas, even if the behavior score s t Even at relatively low levels, the system can still determine a violation based on the scene's legality. This overcomes the limitations of traditional methods in complex scenarios.

[0077] To improve the system's adaptability and accuracy, this step further considers the impact of different scenarios on behavior determination. This is achieved through scenario labeling. t By introducing this feature, the model can fully understand the actual operating scenarios of dump trucks. For example, certain behaviors may be permitted on construction sites and in non-restricted dumping areas, but in restricted dumping areas, even if the behavior score is low, it should still be judged as a violation. In this way, this step solves the problem of misjudgment when "the behavior itself is abnormal but occurs in a legal area".

[0078] Finally, model h ψ y will be generated t This serves as the final determination of the violation. t =1 indicates that the behavior during that period was a violation, y t =0 indicates that the behavior during that period was normal.

[0079] S4. Based on the violation determination result, trigger a multi-level linkage response mechanism, including real-time vehicle feedback, linkage of regional monitoring equipment, and uploading and storage of violation data.

[0080] Specifically, the goal of this step is to ensure timely response and subsequent tracking of the illegal behaviors of dump trucks through a multi-level linkage response and collaborative monitoring mechanism. Through the processing of the final determination of illegal behavior y t and combined with a multi-level response system, it quickly responds and ensures that illegal behaviors are handled, while enhancing regional monitoring.

[0081] Once the illegal behavior y t = 1 is determined, the system activates the following multi-level response mechanism:

[0082] Local response mechanism (in-vehicle feedback): At the in-vehicle terminal, the system reminds the driver that the behavior is non-compliant by means of voice warnings and warning messages displayed on the dashboard. At the same time, the system can restrict the vehicle's driving route through in-vehicle intelligent devices, such as prohibiting entry into specific areas or restricting the vehicle speed, to ensure that non-compliant behaviors are immediately corrected.

[0083] Regional linkage response mechanism: When an illegal behavior occurs in a sensitive area (such as a prohibited dumping area), the system will trigger a regional-level monitoring response through linkage with surrounding intelligent monitoring devices. For example, roadside cameras, drones, and other sensing devices will synchronously collect and analyze data to ensure that other vehicles in the area can also receive warnings and make corresponding adjustments.

[0084] Data upload and record keeping: All data related to illegal behaviors (such as acceleration data, environmental image data, illegal timestamps, illegal areas, etc.) will be uploaded to the central platform in real time. These data will be used for subsequent audits, penalty decisions, and further supervision. The platform will generate a detailed illegal report and provide it to the traffic management department for further review.

[0085] The system also has a calculation mechanism for the regional risk index R t . By monitoring illegal behaviors in a specific area, the risk index of the area is calculated to reflect the degree of illegal risk in the area. The risk index R t is calculated by the following formula:

[0086]

[0087] where: N is the number of vehicles participating in the monitoring of the area; is the illegal determination of the i-th vehicle at time t, and y i,t = 1 indicates illegal; is whether the i-th vehicle is in the illegal area D.

[0088] When R t exceeds the preset threshold, the system will automatically increase the monitoring frequency of the area and deploy more sensors or drones in the area for dynamic monitoring.

[0089] The intelligent feedback mechanism dynamically adjusts the monitoring intensity of a region based on real-time monitoring data and historical violation records. For areas with frequent violations, the system will automatically increase the monitoring density in that area and reduce the monitoring frequency in other areas. This dynamic adjustment mechanism ensures efficient resource allocation, enabling the monitoring system to more flexibly respond to the monitoring needs of different regions.

[0090] Automated enforcement feedback mechanism: When the frequency of violations in a region reaches a certain threshold, the system will automatically generate a violation report and trigger an alarm in the enforcement system. Enforcement officers can quickly obtain specific data on the violations through this report and initiate manual intervention when necessary.

[0091] This invention also provides an AI-driven collaborative monitoring system for illegal dump truck behavior, the system comprising:

[0092] The vehicle-mounted terminal module is used to collect three-dimensional acceleration and three-dimensional angular velocity data of dump trucks through vehicle-mounted inertial sensors, construct an inertial behavior sequence, establish an individualized behavior model based on historical normal behavior data, identify abnormal behavior periods that deviate from the norm, and generate anomaly scores.

[0093] The cloud-based analytics platform is used to trigger the vehicle-mounted camera to capture environmental images when abnormal behavior is detected, and to determine the legality of the scene in which the dump truck is located through a visual semantic recognition model.

[0094] The multi-level response module generates the final violation judgment result of the violation behavior by integrating the anomaly score and the scene legality mark through a joint judgment model;

[0095] The dynamic monitoring and scheduling module triggers a multi-level linkage response mechanism based on the violation determination result, including real-time vehicle feedback, linkage with regional monitoring equipment, and uploading and storing violation data.

[0096] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0097] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of apparatuses or units may be electrical, mechanical, or other forms.

[0098] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0099] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. An AI-driven collaborative monitoring method for illegal activities of dump trucks, characterized in that, The method includes the following steps: S1. Collect three-dimensional acceleration and three-dimensional angular velocity data of dump trucks through vehicle-mounted inertial sensors to construct an inertial behavior sequence; establish an individualized behavior model based on historical normal behavior data, identify abnormal behavior periods that deviate from the norm, and generate anomaly scores. S2. When abnormal behavior occurs, the vehicle-mounted camera is triggered to collect environmental images, and the legality of the scene where the dump truck is located is determined by the visual semantic recognition model. S3. Integrate the aforementioned anomaly score and scene legality marker, and generate the final violation judgment result through a joint judgment model; S4. Based on the violation determination result, trigger a multi-level linkage response mechanism, including real-time vehicle feedback, linkage of regional monitoring equipment, and uploading and storage of violation data; The S4 further includes: Calculate the regional risk index, and dynamically adjust the allocation of monitoring resources based on the number of illegal vehicles and the distribution of illegal areas within the region. When the risk index of a region exceeds a preset threshold, the deployment density of monitoring equipment in that region is automatically increased, and an enforcement report is generated and pushed to the management platform. Among them, the risk index Calculated using the following formula: ； in: The number of vehicles participating in the monitoring of this area; For the first The car at any time The violation determination This indicates a violation; For the first Is the vehicle in a prohibited area? ; When the risk index When the preset threshold is exceeded, the system will automatically increase the monitoring frequency of the area and deploy more sensors or drones in the area for dynamic monitoring.

2. The AI-driven collaborative monitoring method for illegal dump truck behavior according to claim 1, characterized in that, The individualized behavior model is trained using a Gaussian mixture model, and the anomaly score is determined by calculating the deviation of the current behavior sequence from the historical distribution; the deviation is determined based on the negative log-likelihood value. If the deviation is higher than the average score of normal behavior, it indicates that the behavior deviates significantly from the norm and is therefore marked as a possible abnormal behavior. Compare the deviation with the mean of historical scores to determine if it is an abnormal period: If the deviation is greater than the sum of the historical average score and the preset threshold, the current period is marked as an abnormal behavior period.

3. The AI-driven collaborative monitoring method for illegal dump truck behavior according to claim 1, characterized in that, The visual semantic recognition model takes the environmental image as input and outputs scene labels including construction site, no-dumping zone and road, and the environmental legality mark is dynamically generated based on the scene labels.

4. The AI-driven collaborative monitoring method for illegal dump truck behavior according to claim 3, characterized in that, The environment legality marker is dynamically generated based on the scene label, specifically including: Define a scene legality marker. If the construction site is a no-dumping zone, the scene legality marker is illegal, indicating that the dump truck's behavior occurred in a no-dumping zone and violated relevant regulations. If the construction site is a construction site, the scene legality marker is legal, indicating that the area is a legal unloading site.

5. The AI-driven collaborative monitoring method for illegal dump truck behavior according to claim 4, characterized in that, The joint judgment model generates a final judgment result by weighting the abnormal behavior score and the legality of the scene, wherein the illegal scene mark triggers additional penalty weights; the final judgment result is either a violation or normal behavior.

6. The AI-driven collaborative monitoring method for illegal dump truck behavior according to claim 1, characterized in that, The vehicle-mounted real-time feedback includes at least one of voice warnings and driving route restrictions, and the regional monitoring equipment linkage includes triggering roadside cameras and drones to conduct collaborative monitoring of the violation area.

7. The AI-driven collaborative monitoring method for illegal dump truck behavior according to claim 1, characterized in that, The inertial sensor includes a three-axis accelerometer and a gyroscope, with a data acquisition frequency of no less than once per second; the vehicle-mounted camera is a 360-degree panoramic camera, and the image acquisition is strictly synchronized with the timestamp of the abnormal behavior period.

8. A system for implementing the AI-driven collaborative monitoring method for illegal dump truck behavior as described in claim 1, characterized in that, The system includes: The vehicle-mounted terminal module is used to collect three-dimensional acceleration and three-dimensional angular velocity data of dump trucks through vehicle-mounted inertial sensors, construct an inertial behavior sequence, establish an individualized behavior model based on historical normal behavior data, identify abnormal behavior periods that deviate from the norm, and generate anomaly scores. The cloud-based analytics platform is used to trigger the vehicle-mounted camera to capture environmental images when abnormal behavior is detected, and to determine the legality of the scene in which the dump truck is located through a visual semantic recognition model. The multi-level response module is used to generate the final violation judgment result of the violation behavior by combining the anomaly score and the scene legality mark through a joint judgment model; The dynamic monitoring and scheduling module is used to trigger a multi-level linkage response mechanism based on the violation judgment result, including real-time vehicle feedback, linkage of regional monitoring equipment, and uploading and storage of violation data. The dynamic monitoring and scheduling module also performs the following: Calculate the regional risk index, and dynamically adjust the allocation of monitoring resources based on the number of illegal vehicles and the distribution of illegal areas within the region. When the risk index of a region exceeds a preset threshold, the deployment density of monitoring equipment in that region is automatically increased, and an enforcement report is generated and pushed to the management platform.

Images