Abnormal driving behavior detection method and device, equipment and medium

Abstract

The invention discloses an abnormal driving behavior detection method and device, equipment and a medium, and relates to the technical field of artificial intelligence. The abnormal driving behavior detection method comprises the following steps: replacing a to-be-optimized convolutional layer in a target detection model through target convolution to obtain a to-be-pruned target model; pruning the target model to be pruned to obtain a pre-trained target model, and determining an abnormal driving behavior detection model of the cash truck driver according to the behavior sample data of the cash truck driver and the pre-trained target model; and detecting the driving behavior of the cash truck driver in real time based on the abnormal driving behavior detection model. According to the technical scheme of the embodiment of the invention, the limitation of driving state detection of the driver of the cash truck can be broken through, and the detection effect is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method, device, equipment and medium for detecting abnormal driving behavior. Background Art

[0002] With the booming financial industry and the growing demand for cash transport, the safety and stability of cash transport vehicles, as a critical means of transportation, have drawn considerable attention. In the daily operations of cash transport vehicles, the driver's behavior plays a crucial role in the safety of the vehicle and the goods being transported. However, due to factors such as long driving hours, complex road conditions, fatigue, or distraction, drivers may exhibit abnormal behaviors such as fatigued driving, distracted driving, speeding, and illegal lane changes, seriously threatening the safety of cash transport vehicles.

[0003] Traditional monitoring methods, such as manual monitoring and simple sensor detection, have limitations such as low efficiency, susceptibility to human interference, and inability to directly reflect the driver's specific behavior. Therefore, there is an urgent need for an efficient, accurate, and real-time method to detect abnormal driving behavior. Summary of the Invention

[0004] The present invention provides a method, device, equipment and medium for detecting abnormal driving behavior, so as to solve the problems of strong limitations and poor detection effect of existing cash transport vehicle driver's driving state detection.

[0005] According to one aspect of the present invention, a method for detecting abnormal driving behavior is provided, comprising:

[0006] Replace the convolutional layer to be optimized in the target detection model through target convolution to obtain the target model to be pruned;

[0007] The target model to be pruned is pruned to obtain a pre-trained target model, and an abnormal driving behavior detection model of the cash truck driver is determined based on the sample data of the cash truck driver's behavior and the pre-trained target model;

[0008] Real-time detection of the driving behavior of armored truck drivers based on the abnormal driving behavior detection model.

[0009] According to another aspect of the present invention, there is provided an abnormal driving behavior detection device, comprising:

[0010] The convolution layer replacement module is used to replace the convolution layer to be optimized in the target detection model through the target convolution to obtain the target model to be pruned;

[0011] The abnormal driving behavior detection model determination module is used to prune the target model to be pruned to obtain a pre-trained target model, and determine the abnormal driving behavior detection model of the cash truck driver based on the cash truck driver behavior sample data and the pre-trained target model;

[0012] The abnormal driving behavior detection module is used to detect the driving behavior of the armored car driver in real time based on the abnormal driving behavior detection model.

[0013] According to another aspect of the present invention, an electronic device is provided, comprising:

[0014] at least one processor; and

[0015] a memory communicatively connected to the at least one processor; wherein,

[0016] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the abnormal driving behavior detection method described in any embodiment of the present invention.

[0017] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the abnormal driving behavior detection method described in any embodiment of the present invention when executed.

[0018] According to another aspect of the present invention, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the abnormal driving behavior detection method according to any embodiment of the present invention.

[0019] The technical solution of the embodiment of the present invention replaces the convolution layer to be optimized in the target detection model through target convolution to obtain the target model to be pruned, thereby pruning the target model to be pruned to obtain a pre-trained target model, and determines the abnormal driving behavior detection model of the cash truck driver based on the sample data of the cash truck driver's behavior and the pre-trained target model, and then detects the driving behavior of the cash truck driver in real time based on the abnormal driving behavior detection model. In this solution, by replacing the convolution layer to be optimized in the target detection model, the accuracy of the model is improved, and the prediction efficiency of the model is effectively improved by the pruning operation, so that the final pre-trained target model takes into account both accuracy and calculation speed. After the pre-trained target model is trained based on the sample, the abnormal driving behavior of the cash truck driver can be quickly and accurately identified, solving the problem of strong limitations and poor detection effect of the existing cash truck driver's driving state detection, breaking the limitations of the cash truck driver's driving state detection, and effectively improving the detection effect.

[0020] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present invention, nor is it intended to limit the scope of the present invention. Other features of the present invention will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 This is a flow chart of a method for detecting abnormal driving behavior provided in Example 1 of the present invention;

[0023] Figure 2 This is a flow chart of a method for detecting abnormal driving behavior provided in Example 2 of the present invention;

[0024] Figure 3 A schematic structural diagram of an abnormal driving behavior detection device provided in a fourth embodiment of the present invention;

[0025] Figure 4 A schematic structural diagram of an electronic device that can be used to implement an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0027] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0028] Example 1

[0029] Figure 1 This is a flow chart of an abnormal driving behavior detection method provided in the first embodiment of the present invention. This embodiment is applicable to the situation of efficiently and accurately identifying the driving status of the driver of the cash transport vehicle. The method can be executed by an abnormal driving behavior detection device. The abnormal driving behavior detection device can be implemented in the form of hardware and / or software. The abnormal driving behavior detection device can be configured in an electronic device. Figure 1 As shown, the method includes:

[0030] Step 110: Replace the convolution layer to be optimized in the target detection model by target convolution to obtain the target model to be pruned.

[0031] The target convolution may be a preselected convolution method used to structurally replace a convolutional neural network to optimize its processing capabilities. The target detection model may be a model used for target detection. The target detection model may include an open-source convolutional neural network or an open-source extended convolutional neural network. The convolutional layer to be optimized may be a convolutional layer in the target detection model that requires structural optimization. The target model to be pruned may be the target detection model after the convolutional layer to be optimized has been replaced.

[0032] In an embodiment of the present invention, the target convolution for neural network structure optimization and the convolution layer to be optimized that needs structural optimization in the target detection model can be determined, and then the target convolution is used to replace the convolution layer to be optimized in the target detection model to obtain the target model to be pruned.

[0033] Step 120: Prune the target model to be pruned to obtain a pre-trained target model, and determine an abnormal driving behavior detection model for the cash truck driver based on the cash truck driver behavior sample data and the pre-trained target model.

[0034] The pre-trained target model can be a model obtained after pruning the target model to be pruned. The armored truck driver behavior sample data can be sample data describing the driving behavior of the armored truck driver. For example, the armored truck driver behavior sample data can include input features (such as driving-related behaviors of the armored truck driver) and output labels (such as whether the driver is driving attentively or fatigued). The abnormal driving behavior detection model can be used to detect abnormal driving behavior of armored truck drivers.

[0035] In an embodiment of the present invention, the target model to be pruned can be pruned while ensuring the accuracy of the model to obtain a pre-trained target model, and then the armored car driver behavior sample data can be obtained, and the pre-trained target model can be trained using the armored car driver behavior sample data to obtain an abnormal driving behavior detection model for the armored car driver.

[0036] Step 130: Detect the driving behavior of the cash transport vehicle driver in real time based on the abnormal driving behavior detection model.

[0037] In an embodiment of the present invention, based on the cash truck driver behavior collection camera, the cash truck driver's behavior image can be collected in real time, and the real-time collected image is input into the abnormal driving behavior detection model. According to the output result of the abnormal driving behavior detection model, it is determined whether the driving behavior of the cash truck driver is abnormal, so as to intervene in time when it is identified that the driving behavior of the cash truck driver poses a safety accident risk.

[0038] The technical solution of the embodiment of the present invention replaces the convolution layer to be optimized in the target detection model through target convolution to obtain the target model to be pruned, thereby pruning the target model to be pruned to obtain a pre-trained target model, and determines the abnormal driving behavior detection model of the cash truck driver based on the sample data of the cash truck driver's behavior and the pre-trained target model, and then detects the driving behavior of the cash truck driver in real time based on the abnormal driving behavior detection model. In this solution, by replacing the convolution layer to be optimized in the target detection model, the accuracy of the model is improved, and the prediction efficiency of the model is effectively improved by the pruning operation, so that the final pre-trained target model takes into account both accuracy and calculation speed. After the pre-trained target model is trained based on the sample, the abnormal driving behavior of the cash truck driver can be quickly and accurately identified, solving the problem of strong limitations and poor detection effect of the existing cash truck driver's driving state detection, breaking the limitations of the cash truck driver's driving state detection, and effectively improving the detection effect.

[0039] Example 2

[0040] Figure 2 This is a flowchart of a method for detecting abnormal driving behavior provided by the second embodiment of the present invention. This embodiment is specific based on the above embodiment and provides a specific optional implementation method for pruning the target model to be pruned to obtain a pre-trained target model. Figure 2 As shown, the method includes:

[0041] Step 210: Replace the convolution layer to be optimized in the target detection model by target convolution to obtain the target model to be pruned.

[0042] In an optional embodiment of the present invention, the convolution layer to be optimized in the target detection model is replaced by target convolution to obtain the target model to be pruned, which may include: when the convolution layer in the backbone network and the neck network in the target detection model is the convolution layer to be optimized, the spatial depth conversion convolution is used as the target convolution; and the convolution layer in the backbone network and the neck network in the target detection model is replaced by spatial depth conversion convolution to obtain the target model to be pruned.

[0043] In an embodiment of the present invention, the convolutional layers in the backbone network and the neck network in the target detection model can be selected as the convolutional layers to be optimized, and the spatial depth conversion convolution can be used as the target convolution. Then, the convolutional layers in the backbone network and the neck network in the target detection model can be replaced by the spatial depth conversion convolution to optimize the model structure, improve the model accuracy, and finally obtain the target model to be pruned.

[0044] In an optional embodiment of the present invention, the convolutional layer to be optimized in the target detection model is replaced by target convolution to obtain the target model to be pruned, which may include: when the convolutional layers in the backbone network, the neck network and the C2f module in the target detection model are the convolutional layers to be optimized, replacing the backbone network and the convolutional layers in the neck network in the target detection model by spatial depth conversion convolution to obtain the initial replacement target model; replacing the convolutional layer in the C2f module by ghost convolution to obtain the target model to be pruned.

[0045] The C2f (Cross Stage Partial Network with 2Convolutions Fusion) module is a network structure used for feature fusion in object detection or image processing. The initial replacement target model can be a model that replaces the convolutional layers in the backbone network and the neck network of the object detection model based on spatial depth conversion convolution. Ghost Conv is used for ghost convolution.

[0046] In an embodiment of the present invention, the convolutional layers in the backbone network, neck network and C2f module in the target detection model can be used as convolutional layers to be optimized, and the spatial depth conversion convolution and ghost convolution can be used as target convolutions. Specifically, the convolutional layers in the backbone network and the neck network in the target detection model are replaced by the spatial depth conversion convolution to obtain the initial replacement target model, and the convolutional layers in the C2f module are replaced by the ghost convolution to further optimize the model structure, effectively improve the model accuracy, and obtain the target model to be pruned.

[0047] Step 220: Obtain model accuracy and rate constraints.

[0048] Among them, model accuracy and rate constraints can be used to constrain the prediction accuracy and prediction efficiency of the pre-trained target model.

[0049] In the embodiment of the present invention, the model accuracy and rate constraint conditions that need to be set by technicians based on the identification of abnormal driving behavior of armored trucks can be obtained.

[0050] Step 230: Prune the target model to be pruned according to the model accuracy and rate constraints to obtain a pre-trained target model, and determine the abnormal driving behavior detection model of the cash truck driver based on the cash truck driver behavior sample data and the pre-trained target model.

[0051] In an embodiment of the present invention, the target model to be pruned can be pruned, and it can be determined whether the pruned model meets the model accuracy and rate constraints. The pruned model that meets the model accuracy and rate constraints can then be used as a pre-trained target model, thereby training the pre-trained target model based on the sample behavior data of the armored car driver to obtain an abnormal driving behavior detection model for the armored car driver.

[0052] In an optional embodiment of the present invention, after pruning the target model to be pruned to obtain a pre-trained target model, the method may also include: obtaining historical monitoring data from a driver monitoring camera in an armored truck; performing image preprocessing on the historical monitoring data to obtain a first preprocessed image, and performing format conversion on the first preprocessed image according to the data input format of the target detection model to obtain a second preprocessed image; and labeling the second preprocessed image based on key recognition features to obtain armored truck driver behavior sample data.

[0053] The driver monitoring camera in the armored truck is a camera installed inside the armored truck that captures the driver's driving behavior. The historical monitoring data may be monitoring data of the driver's historical driving behavior in the armored truck. The first preprocessed image may be an image obtained by preprocessing the historical monitoring data. The second preprocessed image may be an image obtained by converting the first preprocessed image into a data format according to the data input format of the target detection model. Key recognition features may be features used to identify abnormal driving behavior of the armored truck driver. Exemplary key recognition features may include, but are not limited to, blinking, yawning, turning the head while talking, and holding an object in the hand.

[0054] In an embodiment of the present invention, historical monitoring data of the driver monitoring camera in the armored truck can be read, and then the historical monitoring data can be subjected to image preprocessing such as image cleaning and enhancement to obtain a first preprocessed image. Then, the first preprocessed image can be format converted according to the data input format of the target detection model to obtain a second preprocessed image. Then, the second preprocessed image can be labeled according to key recognition features to obtain armored truck driver behavior sample data. Data preprocessing can improve the readability of the data and the accuracy of subsequent detection, and can be labeled from recognition features of different dimensions (such as focus dimension and fatigue dimension), providing a basis for multi-dimensional abnormal driving behavior analysis.

[0055] Step 240: Detect the driving behavior of the cash transport vehicle driver in real time based on the abnormal driving behavior detection model.

[0056] In an optional embodiment of the present invention, real-time detection of the driving behavior of the armored car driver based on the abnormal driving behavior detection model can include: inputting the real-time monitoring data of the armored car driver into the abnormal driving behavior detection model to obtain concentration recognition data and fatigue recognition data; generating abnormal driving behavior analysis data based on the concentration recognition data, fatigue recognition data, driving concentration state judgment conditions and driving fatigue state judgment conditions.

[0057] The real-time monitoring data may be monitoring data of the current driving behavior of the cash truck driver. The concentration identification data may be the identification result of the cash truck driver's driving concentration as determined by the abnormal driving behavior detection model. The fatigue identification data may be the identification result of the cash truck driver's driving fatigue as determined by the abnormal driving behavior detection model. The driving concentration state discrimination condition may be the discrimination condition for determining whether the cash truck driver is driving attentively. The driving fatigue state discrimination condition may be the discrimination condition for determining whether the cash truck driver is driving fatigued. The abnormal driving behavior analysis data may be the analysis result of abnormal driving behavior of the cash truck driver as determined by the real-time monitoring data.

[0058] In an embodiment of the present invention, real-time monitoring data of an armored car driver can be input into an abnormal driving behavior detection model to be identified according to key identification features through the abnormal driving behavior detection model, thereby using the driver's head twisting conversation identification data and the hand-held object identification data as concentration identification data, and the driver's blinking identification data and yawning identification data as fatigue identification data, thereby using the driving concentration state discrimination conditions and the concentration identification data to determine whether the driver is driving attentively, and using the driving fatigue state discrimination conditions and the fatigue identification data to determine whether the driver is driving fatigued, thereby generating abnormal driving behavior analysis data based on the concentration driving discrimination results and the fatigue driving discrimination results, so as to automatically analyze the abnormal driving behavior of the armored car driver from multiple dimensions without consuming a lot of manpower.

[0059] In an optional embodiment of the present invention, after real-time detection of the driving behavior of the armored truck driver based on the abnormal driving behavior detection model, it may also include: when it is determined based on the abnormal driving behavior analysis data that the driver has abnormal driving behavior, obtaining the real-time positioning data of the armored truck and the device status data of the driver monitoring camera in the armored truck; generating early warning information based on the real-time positioning data of the armored truck, the device status data of the driver monitoring camera in the armored truck and the abnormal driving behavior analysis data.

[0060] The real-time positioning data of the cash transport vehicle can be provided by the vehicle's positioning device in real time. The device status data can describe the operating status of the driver monitoring camera. The early warning information can be data describing abnormal driving behavior of the cash transport vehicle driver.

[0061] In an embodiment of the present invention, if the abnormal driving behavior analysis data indicates that the driver has abnormal driving behavior, the real-time positioning data of the armored car and the device status data of the driver monitoring camera in the armored car are further obtained, so as to organize the real-time positioning data of the armored car, the device status data of the driver monitoring camera in the armored car and the abnormal driving behavior analysis data, and generate early warning information, so that the management personnel can grasp the relevant situation of the abnormally driven armored car in real time and manage and deal with the abnormally driven armored car in time.

[0062] The technical solution of the embodiment of the present invention replaces the convolutional layer to be optimized in the target detection model through target convolution to obtain a target model to be pruned, thereby obtaining model accuracy and rate constraints. Then, based on the model accuracy and rate constraints, the target model to be pruned is pruned to obtain a pre-trained target model. Then, based on the cash truck driver behavior sample data and the pre-trained target model, a cash truck driver's abnormal driving behavior detection model is determined, so that the cash truck driver's driving behavior can be detected in real time based on the abnormal driving behavior detection model. In this solution, by replacing the convolutional layer to be optimized in the target detection model, the model accuracy is improved. The model accuracy and rate constraints are used to constrain the pruning operation from two dimensions, model accuracy and computational rate, so that the final pre-trained target model takes into account both accuracy and computational speed. After the pre-trained target model is trained based on samples, it can quickly and accurately identify the abnormal driving behavior of cash truck drivers. This solves the problems of strong limitations and poor detection effect of existing cash truck driver driving state detection, can break the limitations of cash truck driver driving state detection, and effectively improve the detection effect.

[0063] The real-time monitoring data of drivers collected are information and data authorized by users or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of relevant data comply with the relevant laws, regulations and standards of relevant countries and regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0064] Example 3

[0065] The third embodiment of the present invention provides an optional embodiment of a method for detecting abnormal driving behavior, and its specific implementation can be found in the following embodiments. Technical terms that are the same as or corresponding to the above embodiments are not repeated here.

[0066] The cash truck's driver monitoring camera is activated to capture real-time video streams of the driver's facial, eye, and body movements, generating real-time monitoring data. This data is then transmitted to the data processing module via the vehicle's onboard network. The cash truck's driver monitoring camera uses a high-resolution, wide-angle lens to ensure clear capture of the truck driver's real-time monitoring data, providing a high-quality data source for subsequent abnormal behavior detection.

[0067] The real-time monitoring data is preliminarily screened, denoised, enhanced, and format converted to improve the readability of the data and the accuracy of subsequent detection. The specific steps are as follows: denoise the real-time monitoring data to remove noise and interference in the image to ensure image clarity, and further perform image enhancement processing on the image, such as contrast adjustment and brightness adjustment, to improve image recognizability. The data after the above processing is converted into a data format, and the pre-trained target model is trained based on the format-converted data. Among them, the denoising process may include Gaussian filtering or median filtering algorithms. The image enhancement process uses histogram equalization or contrast stretching algorithms. Format conversion ensures that the data meets the input requirements of the pre-trained target model, such as resolution, number of channels, etc.

[0068] The pre-trained abnormal driving behavior detection model is used to perform real-time analysis on the pre-processed real-time monitoring data. The abnormal driving behavior detection model can accurately identify abnormal driving behaviors of the driver, such as fatigue driving (such as frequent blinking and yawning), distracted driving (such as looking at the phone, turning the head to talk), etc., and output the detection results. Based on the detection results (including the driver's mouth movement, eye movement, body movement, etc.), the model outputs the detection results to determine whether abnormal driving behavior exists. Among them, the abnormal driving behavior detection model uses a convolutional neural network to extract features and detect targets on the input data. The model output results include target category, confidence level, and bounding box information. Based on the preset driving concentration state judgment conditions and driving fatigue state judgment conditions, it is judged whether abnormal behavior exists, such as frequent blinking, yawning, looking at the phone, etc.

[0069] Abnormal driving behavior analysis data identifies abnormal driving behavior and immediately triggers an early warning mechanism. Warning methods include but are not limited to audible alarms, light prompts, screen displays, and alerts sent to a remote management center to prompt drivers and management to take timely action. The early warning mechanism is triggered based on the confidence level of the detection results and preset thresholds. Warning information is transmitted to the remote management center in real time via the vehicle network to ensure a timely response.

[0070] The monitoring data collected by driver monitoring cameras in armored vehicles can be stored and statistically analyzed based on detection results and warning information to generate driver behavior reports for management review and evaluation, providing financial institutions with a scientific basis for driver behavior management. Furthermore, based on feedback, the performance of the abnormal driving behavior detection model can be continuously optimized and redeployed to improve detection accuracy and efficiency. Data storage utilizes a distributed database to ensure data security and scalability. Statistical analysis utilizes machine learning algorithms, such as cluster analysis and regression analysis, to generate driver behavior reports. Model optimization utilizes transfer learning or incremental learning algorithms to improve the model's detection accuracy and efficiency.

[0071] In order to achieve comprehensive monitoring and management of abnormal driving behavior of armored truck drivers, role-based access control can be used to manage user login, user permissions and role assignment. The parameter settings of on-board equipment, equipment status data and real-time positioning data of armored trucks can be managed based on Internet of Things technology. The deployment of abnormal driving behavior detection models and the analysis of driver behavior data can be managed based on big data analysis technology.

[0072] Utilizing an abnormal driving behavior detection model, the system processes the video stream captured by the driver monitoring camera in the armored vehicle in real time, rapidly identifying the driver's facial and body movements, enabling instant detection of abnormal driving behavior. Through optimized model structure and computational efficiency, detection can be completed within milliseconds, ensuring an immediate response to abnormal driving behavior. This effectively prevents traffic accidents caused by driver misoperation and safeguards the security of the armored vehicle and its contents. Compared to traditional post-event analysis, this method enables proactive monitoring and early warning of driving behavior, enabling intervention before risks occur, significantly reducing the probability of safety incidents and enhancing transportation safety management capabilities.

[0073] Administrators can use the cash truck management system to monitor the truck's driving status and driver behavior in real time. If any unusual driving behavior is detected, the system immediately issues an alert, enabling administrators to quickly respond and take appropriate measures, such as contacting the driver, adjusting the route, or dispatching a support team. The system automatically generates driver behavior analysis reports, including detailed information such as the number, type, and timing of unusual behaviors. This significantly reduces administrators' data collation and analysis workload and improves management efficiency.

[0074] Example 4

[0075] Figure 3 This is a schematic diagram of the structure of an abnormal driving behavior detection device provided by the fourth embodiment of the present invention. Figure 3 As shown, the device includes:

[0076] A convolutional layer replacement module 310 is configured to replace the convolutional layer to be optimized in the target detection model by performing target convolution to obtain a target model to be pruned.

[0077] The abnormal driving behavior detection model determination module 320 is used to prune the target model to be pruned to obtain a pre-trained target model, and determine the abnormal driving behavior detection model of the cash truck driver based on the cash truck driver behavior sample data and the pre-trained target model;

[0078] The abnormal driving behavior detection module 330 is used to detect the driving behavior of the cash transport vehicle driver in real time based on the abnormal driving behavior detection model.

[0079] The technical solution of the embodiment of the present invention replaces the convolution layer to be optimized in the target detection model through target convolution to obtain the target model to be pruned, thereby pruning the target model to be pruned to obtain a pre-trained target model, and determines the abnormal driving behavior detection model of the cash truck driver based on the sample data of the cash truck driver's behavior and the pre-trained target model, and then detects the driving behavior of the cash truck driver in real time based on the abnormal driving behavior detection model. In this solution, by replacing the convolution layer to be optimized in the target detection model, the accuracy of the model is improved, and the prediction efficiency of the model is effectively improved by the pruning operation, so that the final pre-trained target model takes into account both accuracy and calculation speed. After the pre-trained target model is trained based on the sample, the abnormal driving behavior of the cash truck driver can be quickly and accurately identified, solving the problem of strong limitations and poor detection effect of the existing cash truck driver's driving state detection, breaking the limitations of the cash truck driver's driving state detection, and effectively improving the detection effect.

[0080] Optionally, a convolutional layer replacement module 310 is specifically used to use the spatial depth conversion convolution as the target convolution when the convolutional layers in the backbone network and the neck network in the target detection model are the convolutional layers to be optimized; the convolutional layers in the backbone network and the neck network in the target detection model are replaced by the spatial depth conversion convolution to obtain the target model to be pruned.

[0081] Optionally, the convolution layer replacement module 310 is specifically used to replace the backbone network, the neck network and the convolution layer in the cross-stage dual convolution feature fusion C2f module in the target detection model by spatial depth conversion convolution to obtain the initial replacement target model; and replace the convolution layer in the C2f module by ghost convolution to obtain the target model to be pruned.

[0082] Optionally, the abnormal driving behavior detection model determination module 320 is used to obtain model accuracy and rate constraints; according to the model accuracy and rate constraints, the target model to be pruned is pruned to obtain a pre-trained target model.

[0083] Optionally, the abnormal driving behavior detection device also includes a sample data determination module, which is used to obtain historical monitoring data of the driver monitoring camera in the armored truck; perform image preprocessing on the historical monitoring data to obtain a first preprocessed image, and perform format conversion on the first preprocessed image according to the data input format of the target detection model to obtain a second preprocessed image; and label the second preprocessed image based on key recognition features to obtain the armored truck driver behavior sample data.

[0084] Optionally, the abnormal driving behavior detection module 330 is used to input the real-time monitoring data of the armored car driver into the abnormal driving behavior detection model to obtain concentration identification data and fatigue identification data; and generate abnormal driving behavior analysis data based on the concentration identification data, fatigue identification data, driving concentration state judgment conditions and driving fatigue state judgment conditions.

[0085] Optionally, the abnormal driving behavior detection device also includes an early warning module for obtaining real-time positioning data of the armored car and device status data of the driver monitoring camera in the armored car when it is determined based on the abnormal driving behavior analysis data that the driver has abnormal driving behavior; and generating early warning information based on the real-time positioning data of the armored car, the device status data of the driver monitoring camera in the armored car and the abnormal driving behavior analysis data.

[0086] The abnormal driving behavior detection device provided in the embodiment of the present invention can execute the abnormal driving behavior detection method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0087] Example 5

[0088] Figure 4 A schematic diagram of the structure of an electronic device that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0089] like Figure 4 As shown, electronic device 10 includes at least one processor 11 and memory, such as ROM 12 and RAM 13, communicatively connected to at least one processor 11. The memory stores computer programs executable by the at least one processor, and processor 11 can perform various appropriate actions and processes based on the computer programs stored in ROM 12 or loaded from storage unit 18 into RAM 13. RAM 13 can also store various programs and data required for the operation of electronic device 10. Processor 11, ROM 12, and RAM 13 are interconnected via bus 14. An I / O interface 15 is also connected to bus 14. ROM 12 is a read-only memory, RAM 13 is a random access memory, and I / O interface 15 is an input / output interface.

[0090] Multiple components in the electronic device 10 are connected to the I / O interface 15, including an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0091] The processor 11 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any other suitable processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as the abnormal driving behavior detection method.

[0092] In some embodiments, the abnormal driving behavior detection method can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the abnormal driving behavior detection method described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to perform the abnormal driving behavior detection method in any other appropriate manner (e.g., by means of firmware).

[0093] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0094] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the computer program is executed by the processor, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer program may be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0095] In the context of the present invention, computer-readable storage medium can be a tangible medium that can contain or store a computer program for use with an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. Computer-readable storage medium can include but is not limited to electronic, magnetic, optical, electromagnetic, infrared or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, computer-readable storage medium can be a machine-readable signal medium. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, RAM, ROM, an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device or any suitable combination of the foregoing.

[0096] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0097] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0098] A computing system may include clients and servers. The clients and servers are generally remote from each other and typically interact via a communication network. This client-server relationship arises through computer programs running on the respective computers, creating a client-server relationship. The server may be a cloud server, also known as a cloud computing server or cloud host. This server is a hosting product within the cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosts and VPS servers.

[0099] The present application also discloses a computer program product comprising a computer program that, when executed by a processor, implements the abnormal driving behavior detection method provided in any of the embodiments of the present application. This program product shares the same inventive concept as the abnormal driving behavior detection method disclosed in each embodiment of the present application and is therefore not further described here.

[0100] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present invention can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved. This is not limited herein.

[0101] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A method for detecting abnormal driving behavior, characterized in that: include: Replace the convolutional layer to be optimized in the target detection model through target convolution to obtain the target model to be pruned; Pruning the target model to be pruned to obtain a pre-trained target model, and determining an abnormal driving behavior detection model for the cash truck driver based on the cash truck driver behavior sample data and the pre-trained target model; The driving behavior of the cash transport vehicle driver is detected in real time based on the abnormal driving behavior detection model.

2. The abnormal driving behavior detection method according to claim 1, characterized in that: The target convolution layer to be optimized in the target detection model is replaced by the target convolution to obtain the target model to be pruned, including: In the target detection model, when the convolutional layers in the backbone network and the neck network are the convolutional layers to be optimized, spatial depth conversion convolution is used as the target convolution; The backbone network in the target detection model and the convolution layer in the neck network are replaced by the spatial depth conversion convolution to obtain the target model to be pruned.

3. The abnormal driving behavior detection method according to claim 1, characterized in that: The target convolution layer to be optimized in the target detection model is replaced by the target convolution to obtain the target model to be pruned, including: When the convolutional layers in the backbone network, the neck network, and the cross-stage dual convolutional feature fusion C2f module in the target detection model are the convolutional layers to be optimized, the backbone network and the convolutional layers in the neck network in the target detection model are replaced by spatial depth conversion convolution to obtain an initial replacement target model; The convolution layer in the C2f module is replaced by ghost convolution to obtain the target model to be pruned.

4. The abnormal driving behavior detection method according to claim 1, characterized in that: Pruning the target model to be pruned to obtain a pre-trained target model includes: Obtain model accuracy and rate constraints; According to the model accuracy and rate constraints, the target model to be pruned is pruned to obtain a pre-trained target model.

5. The abnormal driving behavior detection method according to claim 1, characterized in that: After pruning the target model to be pruned to obtain a pre-trained target model, the method further includes: Obtain historical monitoring data from the driver monitoring camera in the armored truck; Performing image preprocessing on the historical monitoring data to obtain a first preprocessed image, and performing format conversion on the first preprocessed image according to a data input format of the target detection model to obtain a second preprocessed image; The second pre-processed image is labeled based on key recognition features to obtain the sample data of the cash truck driver's behavior.

6. The abnormal driving behavior detection method according to claim 1, characterized in that: Real-time detection of the driving behavior of the cash transport vehicle driver based on the abnormal driving behavior detection model includes: Input the real-time monitoring data of the cash transport vehicle driver into the abnormal driving behavior detection model to obtain concentration recognition data and fatigue recognition data; Abnormal driving behavior analysis data is generated based on the concentration recognition data, the fatigue recognition data, the driving concentration state discrimination condition, and the driving fatigue state discrimination condition.

7. The abnormal driving behavior detection method according to claim 6, characterized in that: After detecting the driving behavior of the cash transport vehicle driver in real time based on the abnormal driving behavior detection model, the method further includes: When it is determined that the driver has abnormal driving behavior based on the abnormal driving behavior analysis data, real-time positioning data of the armored car and device status data of the driver monitoring camera in the armored car are obtained; An early warning message is generated based on the real-time positioning data of the armored car, the device status data of the driver monitoring camera in the armored car, and the abnormal driving behavior analysis data.

8. An abnormal driving behavior detection device, characterized in that: include: The convolution layer replacement module is used to replace the convolution layer to be optimized in the target detection model through the target convolution to obtain the target model to be pruned; An abnormal driving behavior detection model determination module is used to prune the target model to be pruned to obtain a pre-trained target model, and determine the abnormal driving behavior detection model of the cash truck driver based on the cash truck driver behavior sample data and the pre-trained target model; The abnormal driving behavior detection module is used to detect the driving behavior of the cash transport vehicle driver in real time based on the abnormal driving behavior detection model.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the abnormal driving behavior detection method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the abnormal driving behavior detection method according to any one of claims 1 to 7 when executed.

11. A computer program product, characterized in that The computer program product comprises a computer program, which, when executed by a processor, implements the abnormal driving behavior detection method according to any one of claims 1 to 7.

Images