WO2020151339A1 - 基于无人驾驶车辆的异常处理方法、装置及相关设备 - Google Patents

基于无人驾驶车辆的异常处理方法、装置及相关设备 Download PDF

Info

Publication number
WO2020151339A1
WO2020151339A1 PCT/CN2019/118606 CN2019118606W WO2020151339A1 WO 2020151339 A1 WO2020151339 A1 WO 2020151339A1 CN 2019118606 W CN2019118606 W CN 2019118606W WO 2020151339 A1 WO2020151339 A1 WO 2020151339A1
Authority
WO
WIPO (PCT)
Prior art keywords
passenger
facial image
unmanned vehicle
abnormal behavior
trained
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2019/118606
Other languages
English (en)
French (fr)
Inventor
黄秋凤
罗霄
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Ping An Technology Shenzhen Co Ltd
Original Assignee
Ping An Technology Shenzhen Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Ping An Technology Shenzhen Co Ltd filed Critical Ping An Technology Shenzhen Co Ltd
Publication of WO2020151339A1 publication Critical patent/WO2020151339A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition

Definitions

  • This application relates to the field of unmanned driving technology, and in particular to an abnormal handling method and device based on an unmanned vehicle, an unmanned vehicle, and a storage medium.
  • the search for abducted persons is mainly through the family members of the abducted persons and the public security organs. Specifically, it can be spread through channels such as posting tracing notices, television broadcast tracing notices, and the Internet, so that more and more people Find out the characteristics of trafficked persons.
  • the above-mentioned tracing method requires a lot of manpower.
  • relying on people’s description of the appearance of the person they saw and the photo and appearance of the abducted person to determine whether they are abducted persons, in this way may cause misjudgment and cause unnecessary disputes.
  • unmanned vehicles are also equipped with related facilities to determine whether passengers are trafficked persons, that is, by acquiring facial images of passengers in unmanned vehicles, and comparing them with the facial images of known abducted persons. The similarity determines whether the person is abducted.
  • this method can only identify known abductees. For abductees who have been abducted but are in an unknown state, they cannot be identified even if they ride in an unmanned vehicle, so they cannot play a preventive role. The effect of trafficking is poor.
  • the first aspect of the present application provides an abnormal handling method based on an unmanned vehicle, the method including:
  • the real-time geographic location information of the unmanned vehicle and the facial image of the passenger who has the abnormal behavior are sent to a remote server for storage.
  • the second aspect of the present application provides an abnormality processing device based on an unmanned vehicle, which runs in an unmanned vehicle, and the device includes:
  • the image acquisition module is used to acquire the facial image of the passenger when it is detected that there is a passenger on the board;
  • the first recognition module is configured to calculate the similarity between the facial image of the passenger and the facial image of each known abducted person according to the pre-trained recognition model of the abducted person;
  • the first sending module is configured to send the real-time geographic location information of the unmanned vehicle and the facial image of the passenger to the public security department for rescue when the calculated similarity is greater than or equal to a preset similarity threshold;
  • the second recognition module is used to determine whether there is any abnormal behavior among the passengers according to the pre-trained abnormal behavior recognition model when the calculated similarity is less than the preset similarity threshold;
  • the second sending module is used to send the real-time geographic location information of the unmanned vehicle and the facial image of the passenger who has the abnormal behavior to a remote server for storage when it is determined that an abnormal behavior occurs.
  • a third aspect of the present application provides an unmanned vehicle.
  • the unmanned vehicle includes a processor and a memory, and the processor is configured to implement the unmanned driving-based vehicle when executing computer-readable instructions stored in the memory.
  • the abnormal handling method of the vehicle is configured to implement the unmanned driving-based vehicle when executing computer-readable instructions stored in the memory.
  • a fourth aspect of the present application provides a non-volatile readable storage medium having computer readable instructions stored on the non-volatile readable storage medium, and when the computer readable instructions are executed by a processor, the An exception handling method based on unmanned vehicles.
  • the abnormal handling method, device, unmanned vehicle, and storage medium based on the unmanned vehicle described in this application can determine that the passenger in the unmanned vehicle is a known passenger based on the identification model of the abducted person when there are passengers.
  • the geographic location information of the driverless vehicle and the facial image of the passenger can be sent to the public security department for rescue in real time.
  • the abnormal behavior recognition model is used to determine whether the passenger has experienced abnormal behavior.
  • the geographical location information of the person driving the vehicle and the The passenger’s facial image is sent to the remote server for storage, so that when the remote server subsequently receives the same passenger’s facial image from other unmanned vehicles, it can be determined that there has been a trafficking behavior, and then rescued by the public security department .
  • FIG. 1 is a schematic diagram of the application environment of the abnormal handling method based on the unmanned vehicle provided by the present application.
  • Fig. 2 is a flowchart of an abnormality processing method based on an unmanned vehicle provided in Embodiment 1 of the present application.
  • Fig. 3 is a functional block diagram of an abnormality processing device based on an unmanned vehicle provided in Embodiment 2 of the present application.
  • Fig. 4 is a schematic diagram of an unmanned vehicle provided in Embodiment 3 of the present application.
  • FIG. 1 is a schematic diagram of the application environment of the abnormal handling method based on the unmanned vehicle provided by this application.
  • the abnormal handling method based on unmanned vehicles can be applied in an application environment composed of unmanned vehicles 1, network 2, remote server 3, terminal equipment 4 and public security department 5.
  • the driverless vehicle 1 may be various types of driverless vehicles, such as driverless buses, driverless cars, and so on.
  • a high-definition digital image acquisition device is installed in the unmanned vehicle 1, and the high-definition digital image acquisition device may be a pinhole camera, which can be hidden in the unmanned vehicle to avoid being discovered by passengers.
  • the high-definition digital image acquisition device collects facial images of passengers riding in the unmanned vehicle 1 and sends the facial images of the passengers to the remote server 3 through the network 2.
  • the unmanned vehicle 1 determines whether the passenger is the abducted person according to the facial image of the passenger.
  • the network 2 is used to provide a communication connection medium between the unmanned vehicle 1 and the remote server 3.
  • the network 2 may include various connection types, such as wired, wireless communication links, or fiber optic cables.
  • the remote server 3 may be a remote server that provides various services.
  • the network 2 provides the abducted person identification model to a plurality of unmanned vehicles 1 and receives facial images of passengers sent by the unmanned vehicle 1 Cloud remote server.
  • the remote server 3 may also send prompt information to the terminal device 4 via the network 2.
  • the terminal device 4 may be a terminal device held by a related organization or related person looking for a trafficked person.
  • the terminal device 4 may be various unmanned vehicles with a display screen and supporting wireless communication, including but not limited to a smart phone, a tablet computer, a laptop portable computer, and so on.
  • the abnormality processing method based on the unmanned vehicle in the embodiment of the present application can be executed by the unmanned vehicle 1. Accordingly, the abnormality processing device based on the unmanned vehicle is generally installed in the unmanned vehicle. 1 in.
  • the abnormality processing method based on the unmanned vehicle in the embodiment of the present application can also be executed by the remote server 3. Accordingly, the abnormality processing device based on the unmanned vehicle is generally set in the remote server 3.
  • the numbers of unmanned vehicles, networks, remote servers, and terminal devices in FIG. 1 are merely illustrative. According to implementation needs, there can be any number of mobile terminals, networks, remote servers, and terminal devices. In other embodiments, the terminal device may not be included in the application environment of the method.
  • Fig. 2 is a flowchart of an abnormality processing method based on an unmanned vehicle provided in Embodiment 1 of the present application.
  • the abnormal handling method based on unmanned vehicles is applied to unmanned vehicles. According to different requirements, the execution sequence in the flowchart can be changed, and some steps can be omitted.
  • the unmanned vehicle When the unmanned vehicle detects that there are passengers in the unmanned vehicle, it acquires the facial images of the passengers in the unmanned vehicle through the high-definition digital image acquisition device installed in the unmanned vehicle.
  • the method includes:
  • the high-definition digital image acquisition device is controlled to shut down.
  • the high-definition digital image acquisition device when it is detected that a passenger is riding, after the high-definition digital image acquisition device is controlled to be turned on, the high-definition digital image acquisition device acquires the facial image of the passenger in the unmanned vehicle.
  • the unmanned vehicle does not detect that there is a passenger, it is not necessary to control the high-definition digital image acquisition device to turn on, or when the unmanned vehicle detects that the passenger gets off the vehicle, it controls the high-definition digital image acquisition device to turn off.
  • the high-definition digital image acquisition device By controlling the high-definition digital image acquisition device to turn on when there are passengers in the car, and controlling the high-definition digital image acquisition device to turn off after no one is riding or the passenger gets off the bus, it can prevent the high-definition digital image acquisition device from working and reduce the high-definition digital image Collect the power consumption of the equipment, thereby saving the power of unmanned vehicles and extending the battery life of unmanned vehicles.
  • S22 Calculate the similarity between the facial image of the passenger and each known facial image of the abducted person according to the pre-trained person identification model.
  • the calculation of the similarity between the passenger's facial image and each known abductee's facial image according to a pre-trained abductee recognition model includes:
  • the unmanned vehicle can download a pre-trained person identification model from a remote server (for example, a cloud server) via the network every predetermined time (for example, 24 hours), and based on the pre-trained abductee
  • the person recognition model calculates the similarity between the facial image of each passenger and the facial image of each known abducted person for each passenger in the ride.
  • the remote server may be connected to a remote server of a related department, such as the Ministry of Public Security, or a remote server of a tracing platform for public welfare through a network.
  • the remote server may obtain multiple facial images of abducted persons from the remote server of the relevant department or from the remote server of the non-profit tracing platform at predetermined intervals.
  • the facial image of the abducted person may include facial images taken from multiple angles.
  • the abductee corresponding to the facial images of multiple abducted persons obtained from the remote server of the relevant department or from the remote server of the non-profit tracing platform is a known abducted person, that is, a person who has been known with his name and other identity information. Abducting people.
  • the remote server When the remote server obtains multiple known facial images of the abducted person, it trains the abducted person identification model according to the facial images. Specifically, the known facial images of each abducted person and other reference facial images obtained from the remote server of the relevant department or the remote server of the non-profit tracing platform can be used to train the abducted person recognition model. After training the abducted person recognition model, the abducted person recognition model can recognize the similarity between the newly input face image and the known facial images of the abducted person. It is understandable that as newly added known abducted persons are continuously acquired, the abducted person identification model can be continuously trained and updated, thereby obtaining a highly accurate abducted person identification model.
  • the specific process of the remote server pre-training the abducted person identification model may include:
  • the parameters of the convolutional neural network model are trained with default parameters, and the parameters are continuously adjusted during the training process.
  • the verified sample image verifies the abducted person identification model. If the verification pass rate is greater than or equal to the preset first threshold, for example, the pass rate is greater than or equal to 98%, the training ends, and the abducted person is identified by the training.
  • the model recognizes; if the verification pass rate is less than a preset threshold, for example, less than 98%, increase the number of face images participating in training, and re-execute the above steps until the verification pass rate is greater than or equal to the preset first threshold.
  • the trained abductee recognition model is used to recognize the face pictures in the test set to evaluate the recognition effect of the trained abductee recognition model.
  • the unmanned vehicle can input the acquired facial image of each passenger into the pre-trained person identification model, and calculate the difference between the facial image of each passenger and the facial image of the abducted person. Similarity. When it is determined that the similarity between the facial images of a passenger is greater than or equal to the preset similarity threshold, it is determined that abduction has occurred, and there is a known abducted person among the passengers.
  • the unmanned vehicle can send the facial images and real-time geographic location information of all the passengers currently riding to the public security department through the network, or it can only determine the facial image and real-time geographic location of the known abductee who is currently riding. The information is sent to the public security department, and the public security department performs rescue based on the geographic location information and the facial image of the passenger.
  • the unmanned vehicle determines that the similarity between the facial images of any one of the passengers is less than the preset similarity threshold, it can be determined that there is no known abducted person among the passengers, but it is impossible to determine whether there has been abduction. Unknown abducted persons cannot be calculated and judged through the abducted person identification model, and S24 is required for further judgment.
  • the training process of the abnormal behavior recognition model includes:
  • the abnormal behavior may include, but is not limited to: nervousness, fear, irritability, crying, pain, hiding one's face, dementia, giggles, etc.
  • the abnormal behavior recognition model is used to identify whether the passenger's behavior is unnatural or abnormal. For example, the passenger's nervous expression, the crying of children, and the passenger hiding their face are all unnatural or abnormal behaviors.
  • the face pictures of the second proportion are used as the training set, and the remaining face pictures are used as the test set.
  • the number of face pictures in the training set is greater than the number of face pictures in the test set. For example, faces with abnormal behaviors are included 80% of the images and face pictures that do not contain abnormal behavior are used as the training set, and the remaining 20% of the face pictures are used as the test set.
  • the parameters of the residual neural network model are trained with the default parameters, and the parameters are continuously adjusted during the training process.
  • the sample pictures to be verified are used Verify the generated abnormal behavior recognition model. If the verification pass rate is greater than or equal to the preset threshold, for example, the pass rate is greater than or equal to 98%, then the training ends, and the abnormal behavior recognition model obtained by the training is used to identify unmanned vehicles Whether there is any abnormal behavior among the passengers currently in the middle of the school; if the verification pass rate is less than the preset threshold, for example, less than 98%, increase the number of face images and re-execute the above steps until the verification pass rate is greater than or equal to the preset The second threshold. During the test, the abnormal behavior recognition model obtained by training is used to recognize the abnormal behavior of the face pictures in the test set to evaluate the recognition effect of the trained convolutional neural network model.
  • the passenger's facial image calculated by the pre-trained abductee recognition model is less than the preset similarity threshold, then the passenger's facial image is input into the pre-trained abnormal behavior recognition model to determine Whether any abnormal behavior occurs.
  • step S25 is executed.
  • S25 Send the real-time geographic location information of the unmanned vehicle and the facial image of the passenger who has the abnormal behavior to a remote server for storage.
  • the unmanned vehicle can send the real-time geographic location information and the facial image of the passenger who has the abnormal behavior to a remote server for storage.
  • the remote server subsequently receives the abnormal behavior information sent by other unmanned vehicles, it determines whether the currently determined facial image of the passenger who has the abnormal behavior is the same as the historically determined facial image of the passenger who has the abnormal behavior. If the remote server determines that the facial image of the passenger who is currently determined to have abnormal behavior is different from the facial image of the passenger who is determined to have abnormal behavior in history, it is considered that the passenger who is currently riding is the first time that the abnormal behavior has occurred. If the remote server determines that the facial image of the passenger who is currently determined to have the abnormal behavior is the same as the facial image of the passenger who is determined to have the abnormal behavior in history, it can notify the public security for rescue.
  • the remote server stores the real-time geographic location information of the unmanned vehicle and the facial image of the passenger who has the abnormal behavior.
  • the passenger who has the abnormal behavior is a known abducted person, it can be based on the Real-time geographic location information assists public security rescue.
  • the method further includes:
  • the facial image of the passenger and the preset first warning information are sent to other unmanned vehicles, so that the other unmanned vehicles will monitor the real-time geographic location when the passenger is riding.
  • the information and the facial image of the passenger are sent to the public security department.
  • the preset first warning information may be preset text information. For example, if a trafficked person gets off the vehicle, please pay attention to whether to transfer to another unmanned vehicle.
  • the other unmanned vehicle detects that a passenger is riding, it is determined whether the passenger is a passenger sent by the unmanned vehicle.
  • the real-time geographic location information and the facial image of the passenger are sent to the public security department. In this way, it helps the public security department to rescue.
  • the method further includes:
  • the face of the passenger can be detected by the high-definition digital image acquisition device of the unmanned vehicle. It should be understood that the number of human faces of the passenger detected by the high-definition digital image acquisition device is either one or multiple.
  • a face When a face is detected, it is considered that there is only one passenger in the unmanned vehicle, and when multiple faces are detected, it is considered that there are multiple passengers in the unmanned vehicle.
  • the facial image of each passenger is acquired. In the case where it is determined that the number of passengers is 1, it may not be necessary to obtain facial images of passengers. This can also save the workload of high-definition digital image acquisition equipment, save the network resources required when the high-definition digital image acquisition equipment uploads the facial images of passengers to the remote server, and is beneficial to increase the rate of uploading facial images of passengers by other unmanned vehicles.
  • the method may further include:
  • a preset second warning message is sent to pedestrians within a preset distance of the passenger's drop-off location.
  • the unmanned vehicle may send the preset second warning message to the terminal device of the pedestrian within the preset distance from the passenger alighting point through the network.
  • the alarm information can be sent in the form of SMS.
  • the preset second alarm information may be preset text information, for example, please take care of your belongings or children, and there are unknown persons nearby.
  • a preset warning message is sent to pedestrians near the passenger getting off the bus to remind pedestrians to pay attention to personal and property safety, which can be effective Eliminate potential safety hazards, improve pedestrian safety, and effectively reduce the occurrence of trafficking or criminal acts.
  • the identification model of criminals in the embodiments of this application is obtained by obtaining the facial image training of the public security system's online pursuit, habitual offenders, etc., and is similar to the training process of the abducted person identification model. , I won’t go into details here.
  • the abnormal handling method based on the unmanned vehicle obtains the facial image of the passenger when it is detected that there is a passenger in the passenger;
  • the similarity between the facial image of the passenger and the facial image of each known abducted person when the calculated similarity is greater than or equal to the preset similarity threshold, the real-time geographic location information of the unmanned vehicle and the The facial image of the passenger is sent to the public security department for rescue; when the calculated similarity is less than the preset similarity threshold, judge whether there is any abnormal behavior in the passenger according to the pre-trained abnormal behavior recognition model; when it is determined that abnormal behavior has occurred
  • the real-time geographic location information of the unmanned vehicle and the facial image of the passenger who has the abnormal behavior are sent to a remote server for storage.
  • the geographical location information of the unmanned vehicle and the facial image of the passenger can be sent to The public security department rescues, and when it is impossible to determine whether the passenger in the unmanned vehicle is an unknown abducted person according to the abducted person identification model, it further uses the abnormal behavior recognition model to determine whether the passenger has abnormal behavior, and recognizes that the passenger is abnormal.
  • Behaviors to prevent the situation of being abducted but in an unknown state and send the geographic location information of the vehicle driven by the person and the facial image of the passenger to the remote server for storage in real time, so that the remote server will subsequently receive other unmanned driving
  • the same facial image of the passenger occurs in the vehicle, it can be deemed that there has been abduction, and the public security department can rescue it.
  • the following describes the functional modules and hardware structure of the unmanned vehicle that implement the above-mentioned abnormal handling method based on the unmanned vehicle in conjunction with FIGS. 3 to 4.
  • FIG. 3 is a diagram of functional modules in a preferred embodiment of an abnormality processing device based on an unmanned vehicle in this application.
  • the unmanned vehicle-based abnormality processing device 30 runs in an unmanned vehicle.
  • the abnormality processing device 30 based on the unmanned vehicle may include a plurality of functional modules composed of program code segments.
  • the program code of each program segment in the abnormal handling device 30 based on the unmanned vehicle can be stored in the memory and executed by at least one processor for execution (see Figure 2 and its related description for details).
  • the abnormal handling method of driving vehicle can be stored in the memory and executed by at least one processor for execution (see Figure 2 and its related description for details).
  • the abnormality processing device 30 based on the unmanned vehicle can be divided into multiple functional modules according to the functions it performs.
  • the functional modules may include: an image acquisition module 301, an opening control module 302, a first identification module 303, a first sending module 304, a second identification module 305, a second sending module 306, a third sending module 307, and a quantity judgment module 308, a third identification module 309, and a fourth sending module 310.
  • the module referred to in this application refers to a series of computer program segments that can be executed by at least one processor and can complete fixed functions, and are stored in a memory. In some embodiments, the functions of each module will be detailed in subsequent embodiments.
  • the image acquisition module 301 is used to acquire the facial image of the passenger when it is detected that there is a passenger taking the ride.
  • the unmanned vehicle When the unmanned vehicle detects that there are passengers in the unmanned vehicle, it acquires the facial images of the passengers in the unmanned vehicle through the high-definition digital image acquisition device installed in the unmanned vehicle.
  • the turn-on control module 302 is used to control the high-definition digital image acquisition device installed in the unmanned vehicle to turn on and acquire the facial image of the passenger when it is detected that a passenger is riding; when it is detected that the passenger gets off the car, control The high-definition digital image acquisition device is turned off.
  • the high-definition digital image acquisition device when it is detected that a passenger is riding, after the high-definition digital image acquisition device is controlled to be turned on, the high-definition digital image acquisition device acquires the facial image of the passenger in the unmanned vehicle.
  • the unmanned vehicle does not detect that there is a passenger, it is not necessary to control the high-definition digital image acquisition device to turn on, or when the unmanned vehicle detects that the passenger gets off the vehicle, it controls the high-definition digital image acquisition device to turn off.
  • the high-definition digital image acquisition device By controlling the high-definition digital image acquisition device to turn on when there are passengers in the car, and controlling the high-definition digital image acquisition device to turn off after no one is riding or the passenger gets off the bus, it can prevent the high-definition digital image acquisition device from working and reduce the high-definition digital image Collect the power consumption of the equipment, thereby saving the power of unmanned vehicles and extending the battery life of unmanned vehicles.
  • the first recognition module 303 is configured to calculate the similarity between the facial image of the passenger and each known facial image of the abducted person according to the pre-trained recognition model of the abducted person.
  • the first recognition module 303 calculates the similarity between the passenger's facial image and each known abductee's facial image according to a pre-trained abductee recognition model, including:
  • the unmanned vehicle can download a pre-trained person identification model from a remote server (for example, a cloud server) via the network every predetermined time (for example, 24 hours), and based on the pre-trained abductee
  • the person recognition model calculates the similarity between the facial image of each passenger and the facial image of each known abducted person for each passenger in the ride.
  • the remote server may be connected to a remote server of a related department, such as the Ministry of Public Security, or a remote server of a tracing platform for public welfare through a network.
  • the remote server may obtain multiple facial images of abducted persons from the remote server of the relevant department or from the remote server of the non-profit tracing platform at predetermined intervals.
  • the facial image of the abducted person may include facial images taken from multiple angles.
  • the abductee corresponding to the facial images of multiple abducted persons obtained from the remote server of the relevant department or from the remote server of the non-profit tracing platform is a known abducted person, that is, a person who has been known with his name and other identity information Abducting people.
  • the remote server When the remote server obtains multiple known facial images of the abducted person, it trains the abducted person identification model according to the facial images. Specifically, the known facial images of each abducted person and other reference facial images obtained from the remote server of the relevant department or the remote server of the non-profit tracing platform can be used to train the abducted person recognition model. After training the abducted person recognition model, the abducted person recognition model can recognize the similarity between the newly input face image and the known facial images of the abducted person. It is understandable that as newly added known abducted persons are continuously acquired, the abducted person identification model can be continuously trained and updated, thereby obtaining a highly accurate abducted person identification model.
  • the specific process of the remote server pre-training the abducted person identification model may include:
  • the parameters of the convolutional neural network model are trained with default parameters, and the parameters are continuously adjusted during the training process.
  • the verified sample image verifies the abducted person identification model. If the verification pass rate is greater than or equal to the preset first threshold, for example, the pass rate is greater than or equal to 98%, the training ends, and the abducted person is identified by the training.
  • the model recognizes; if the verification pass rate is less than a preset threshold, for example, less than 98%, increase the number of face images participating in training, and re-execute the above steps until the verification pass rate is greater than or equal to the preset first threshold.
  • the trained abductee recognition model is used to recognize the face pictures in the test set to evaluate the recognition effect of the trained abductee recognition model.
  • the first sending module 304 is configured to send the real-time geographic location information of the unmanned vehicle and the facial image of the passenger to the public security department for rescue when the calculated similarity is greater than or equal to a preset similarity threshold.
  • the unmanned vehicle can input the acquired facial image of each passenger into the pre-trained person identification model, and calculate the difference between the facial image of each passenger and the facial image of the abducted person. Similarity. When it is determined that the similarity between the facial images of a passenger is greater than or equal to the preset similarity threshold, it is determined that abduction has occurred, and there is a known abducted person among the passengers.
  • the unmanned vehicle can send the facial images and real-time geographic location information of all the passengers currently riding to the public security department through the network, or it can only determine the facial image and real-time geographic location of the known abductee who is currently riding. The information is sent to the public security department, and the public security department performs rescue based on the geographic location information and the facial image of the passenger.
  • the unmanned vehicle determines that the similarity between the facial images of any one of the passengers is less than the preset similarity threshold, it can be determined that there is no known abducted person among the passengers, but it is impossible to determine whether there has been abduction.
  • the unknown abducted person cannot calculate and make a judgment through the abducted person identification model, and the second identification module 305 needs to perform further judgment.
  • the second recognition module 305 is configured to determine whether there is any abnormal behavior among the passengers according to the pre-trained abnormal behavior recognition model when the calculated similarity is less than the preset similarity threshold.
  • the training process of the abnormal behavior recognition model includes:
  • the abnormal behavior may include, but is not limited to: nervousness, fear, irritability, crying, pain, hiding one's face, dementia, giggles, etc.
  • the abnormal behavior recognition model is used to identify whether the passenger's behavior is unnatural or abnormal. For example, the passenger's nervous expression, the crying of children, and the passenger hiding their face are all unnatural or abnormal behaviors.
  • the face pictures of the second proportion are used as the training set, and the remaining face pictures are used as the test set.
  • the number of face pictures in the training set is greater than the number of face pictures in the test set. For example, faces with abnormal behaviors are included 80% of the images and face pictures that do not contain abnormal behavior are used as the training set, and the remaining 20% of the face pictures are used as the test set.
  • the parameters of the residual neural network model are trained with the default parameters, and the parameters are continuously adjusted during the training process.
  • the sample pictures to be verified are used Verify the generated abnormal behavior recognition model. If the verification pass rate is greater than or equal to the preset threshold, for example, the pass rate is greater than or equal to 98%, then the training ends, and the abnormal behavior recognition model obtained by the training is used to identify unmanned vehicles Whether there is any abnormal behavior among the passengers currently in the middle of the school; if the verification pass rate is less than the preset threshold, for example, less than 98%, increase the number of face images and re-execute the above steps until the verification pass rate is greater than or equal to the preset The second threshold. During the test, the abnormal behavior recognition model obtained by training is used to recognize the abnormal behavior of the face pictures in the test set to evaluate the recognition effect of the trained convolutional neural network model.
  • the passenger's facial image calculated by the pre-trained abductee recognition model is less than the preset similarity threshold, then the passenger's facial image is input into the pre-trained abnormal behavior recognition model to determine Whether any abnormal behavior occurs.
  • the second sending module 306 is configured to send the real-time geographic location information of the unmanned vehicle and the facial image of the passenger who has the abnormal behavior to a remote server for storage when it is determined that an abnormal behavior occurs.
  • the unmanned vehicle can send the real-time geographic location information and the facial image of the passenger who has the abnormal behavior to a remote server for storage.
  • the remote server subsequently receives the abnormal behavior information sent by other unmanned vehicles, it determines whether the currently determined facial image of the passenger who has the abnormal behavior is the same as the historically determined facial image of the passenger who has the abnormal behavior. If the remote server determines that the facial image of the passenger who is currently determined to have abnormal behavior is different from the facial image of the passenger who is determined to have abnormal behavior in history, it is considered that the passenger who is currently riding is the first time that the abnormal behavior has occurred. If the remote server determines that the facial image of the passenger who is currently determined to have the abnormal behavior is the same as the facial image of the passenger who is determined to have the abnormal behavior in history, it can notify the public security for rescue.
  • the remote server stores the real-time geographic location information of the unmanned vehicle and the facial image of the passenger who has the abnormal behavior.
  • the passenger who has the abnormal behavior is a known abducted person, it can be based on the Real-time geographic location information assists public security rescue.
  • the abnormality processing device 30 based on unmanned vehicles further includes:
  • the third sending module 307 is configured to send the facial image of the passenger and the preset first alarm information to the other when the passenger gets off the bus after the calculated similarity is greater than or equal to the preset similarity threshold.
  • the unmanned vehicle enables the other unmanned vehicle to send real-time geographic location information and the facial image of the passenger to the public security department when the other unmanned vehicle detects that the passenger is riding.
  • the preset first warning information may be preset text information. For example, if a trafficked person gets off the vehicle, please pay attention to whether to transfer to another unmanned vehicle.
  • the other unmanned vehicle detects that a passenger is riding, it is determined whether the passenger is a passenger sent by the unmanned vehicle.
  • the real-time geographic location information and the facial image of the passenger are sent to the public security department. In this way, it helps the public security department to rescue.
  • the number determining module 308 is configured to determine whether the number of passengers is 1 after the passenger is detected and before the facial image of the passenger is obtained.
  • the image acquisition module 301 is further configured to acquire a facial image of each passenger when the number judgment module 308 determines that the number of passengers is not one.
  • the face of the passenger can be detected by the high-definition digital image acquisition device of the unmanned vehicle. It should be understood that the number of human faces of the passenger detected by the high-definition digital image acquisition device is either one or multiple.
  • a face When a face is detected, it is considered that there is only one passenger in the unmanned vehicle, and when multiple faces are detected, it is considered that there are multiple passengers in the unmanned vehicle.
  • the facial image of each passenger is acquired. In the case where it is determined that the number of passengers is 1, it may not be necessary to obtain facial images of passengers. This can also save the workload of high-definition digital image acquisition equipment, save the network resources required when the high-definition digital image acquisition equipment uploads the facial images of passengers to the remote server, and is beneficial to increase the rate of uploading facial images of passengers by other unmanned vehicles.
  • the image acquisition module 301 is further configured to acquire the facial image of the passenger when it is determined that the number of the passenger is one.
  • the third identification module 309 is used to determine whether the passenger is a criminal or not according to a pre-trained identification model for criminals.
  • the fourth sending module 310 is configured to send a preset second warning message to pedestrians within a preset distance of the passenger's alighting location when it is determined that the passenger is a criminal offender and/or when the passenger gets off the bus.
  • the unmanned vehicle can send the preset second warning message via the network to the terminal device of the pedestrian within the preset distance from the passenger alighting point.
  • the alarm information can be sent in the form of SMS.
  • the preset second alarm information may be preset text information, for example, please take care of your belongings or children, and there are unknown persons nearby.
  • a preset warning message is sent to pedestrians near the passenger getting off the bus to remind pedestrians to pay attention to personal and property safety, which can be effective Eliminate potential safety hazards, improve pedestrian safety, and effectively reduce the occurrence of trafficking or criminal acts.
  • the identification model of criminals in the embodiments of this application is obtained by obtaining the facial image training of the public security system's online pursuit, habitual offenders, etc., and is similar to the training process of the abducted person identification model. , I won’t go into details here.
  • the anomaly processing device based on an unmanned vehicle acquires a facial image of the passenger when it is detected that a passenger is riding; and calculates the position of the passenger based on the pre-trained abductee recognition model.
  • the similarity between the facial image of the passenger and the facial image of each known abducted person when the calculated similarity is greater than or equal to the preset similarity threshold, the real-time geographic location information of the unmanned vehicle and the The facial image of the passenger is sent to the public security department for rescue; when the calculated similarity is less than the preset similarity threshold, judge whether there is any abnormal behavior in the passenger according to the pre-trained abnormal behavior recognition model; when it is determined that abnormal behavior has occurred
  • the real-time geographic location information of the unmanned vehicle and the facial image of the passenger who has the abnormal behavior are sent to a remote server for storage.
  • the geographical location information of the unmanned vehicle and the facial image of the passenger can be sent to The public security department rescues, and when it is impossible to determine whether the passenger in the unmanned vehicle is an unknown abducted person according to the abducted person identification model, it further uses the abnormal behavior recognition model to determine whether the passenger has abnormal behavior, and recognizes that the passenger is abnormal.
  • Behaviors to prevent the situation of being abducted but in an unknown state and send the geographic location information of the vehicle driven by the person and the facial image of the passenger to the remote server for storage in real time, so that the remote server will subsequently receive other unmanned driving
  • the same facial image of the passenger occurs in the vehicle, it can be deemed that there has been abduction, and the public security department can rescue it.
  • the above-mentioned integrated unit implemented in the form of a software function module may be stored in a computer readable storage medium.
  • the above-mentioned software function module is stored in a storage medium, and includes several instructions to make a computer device (may be a personal computer, a dual-screen device, or a network device, etc.) or a processor to execute the various embodiments of the present application Method part.
  • FIG. 4 is a schematic diagram of an unmanned vehicle provided in Embodiment 3 of the application.
  • the unmanned vehicle 4 includes: a vehicle body 40, a memory 41, at least one processor 42, a computer program 43 stored in the memory 41 and running on the at least one processor 42, and at least one communication bus 44.
  • the at least one processor 42 executes the computer program 43 to implement the steps in the foregoing method embodiment.
  • the computer program 43 may be divided into one or more modules/units, and the one or more modules/units are stored in the memory 41 and executed by the at least one processor 42.
  • the one or more modules/units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program 43 in the unmanned vehicle 4.
  • the unmanned vehicle 4 may be a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud remote server.
  • a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud remote server.
  • the schematic diagram 4 is only an example of the unmanned vehicle 4, and does not constitute a limitation on the unmanned vehicle 4. It may include more or less components than those shown in the figure, or combine certain components. Components, or different components, for example, the unmanned vehicle 4 may also include input and output devices, network access devices, buses, and so on.
  • the at least one processor 42 may be a central processing unit (Central Processing Unit, CPU), or other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), and application specific integrated circuits (ASICs). ), Field-Programmable Gate Array (FPGA) or other programmable logic devices or transistor logic devices, discrete hardware components, etc.
  • the processor 42 can be a microprocessor or the processor 42 can also be any conventional processor, etc.
  • the processor 42 is the control center of the unmanned vehicle 4, which uses various interfaces and lines to connect the entire vehicle. People drive various parts of the vehicle 4.
  • the memory 41 may be used to store the computer program 43 and/or modules/units, the processor 42 runs or executes a series of computer-readable instructions and/or modules/units stored in the memory 41, and The data stored in the memory 41 is called to realize various functions of the unmanned vehicle 4.
  • the memory 41 may mainly include a storage program area and a storage data area.
  • the storage program area may store an operating system, an application program required by at least one function (such as a sound playback function, an image playback function, etc.), etc.; the storage data area may Data (such as audio data) and the like created according to the use of the unmanned vehicle 4 are stored.
  • the memory 41 may include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a Secure Digital (SD) card, a flash memory card (Flash Card), At least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device.
  • non-volatile memory such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a Secure Digital (SD) card, a flash memory card (Flash Card), At least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device.
  • the integrated module/unit of the unmanned vehicle 4 is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a non-volatile readable storage medium.
  • this application implements all or part of the processes in the above-mentioned embodiments and methods, and can also be completed by instructing relevant hardware through a computer program.
  • the computer program can be stored in a non-volatile readable storage medium.
  • the computer program includes a series of computer-readable instruction codes, and the computer-readable instruction codes may be in the form of source code, object code, executable file, or some intermediate forms.
  • the non-volatile readable medium may include: any entity or device capable of carrying the computer readable instruction code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read only memory (ROM, Read-Only Memory).
  • the unmanned vehicle and method disclosed may be implemented in other ways.
  • the embodiment of the unmanned vehicle described above is only illustrative, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Alarm Systems (AREA)
  • Image Analysis (AREA)
  • Emergency Alarm Devices (AREA)

Abstract

一种基于无人驾驶车辆的异常处理方法、处理装置、无人驾驶车辆及存储介质，该方法包括：监测到有乘客乘坐时，获取乘客的面部图像（S21）；根据预先训练好的被拐人员识别模型计算乘客与已知的被拐人员的相似度（S22）；当相似度大于或等于预设相似度阈值时，将实时地理位置信息及乘客的面部图像发送至公安部门进行援救（S23）；当相似度小于预设相似度阈值时，根据预先训练好的异常行为识别模型判断乘客中是否有异常行为发生（S24）；当确定有异常行为发生时，将实时地理位置信息及所述发生异常行为的乘客的面部图像发送至远程服务器存储（S25）。该方法不仅能够识别已知的被拐人员，还能够获取未知的被拐人员的实时地理位置，便于后续的援救。

Description

基于无人驾驶车辆的异常处理方法、装置及相关设备
本申请要求于2019年01月24日提交中国专利局，申请号为201910070260.9申请名称为“基于无人驾驶车辆的防拐卖方法、装置及相关设备”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及无人驾驶技术领域，具体涉及一种基于无人驾驶车辆的异常处理方法、装置、无人驾驶车辆及存储介质。
背景技术
当前对于被拐卖人员的寻找主要是通过被拐卖人员的家属、公安机关共同寻找，具体地，可以通过张贴寻人启事、电视广播寻人通知、互联网等渠道进行扩散，让越来越多的人了解被拐卖人员的特征来寻找。上述寻人方法需要比较多的人力。此外，依靠人们将所见到的人员的相貌与被拐卖人员的照片及相貌描述来判断是否为被拐人员，在这种方式下，有可能引起误判，带来不必要的纷争。
另外，随着无人驾驶车辆的发展，乘坐无人驾驶车辆的乘客将会越来越多。目前，虽然无人驾驶车辆上也有配备判断乘客是否为被拐卖人员的相关设施，即通过获取乘坐无人驾驶车辆的乘客的面部图像，并与各个已知的被拐人员的面部图像之间的相似度确定是否为被拐人员。然而，此种方法只能识别已知的被拐人员，对于已发生被拐但处于未知状态的被拐人员，即使乘坐无人驾驶车辆，也无法进行识别，因而起不到预防的作用，防拐卖效果较差。
发明内容
鉴于以上内容，有必要提出一种基于无人驾驶车辆的异常处理方法、装置、无人驾驶车辆及存储介质，不仅能够识别已知的被拐人员，还能够获取未知的被拐人员的实时地理位置，便于后续的援救。
本申请的第一方面提供一种基于无人驾驶车辆的异常处理方法，所述方法包括：
监测到有乘客乘坐时，获取所述乘客的面部图像；
根据预先训练好的被拐人员识别模型计算所述乘客的面部图像与各个已知的被拐人员的面部图像之间的相似度；
当所计算的相似度大于或者等于预设相似度阈值时，将所述无人驾驶车辆的实时地理位置信息及所述乘客的面部图像发送至公安部门进行援救；
当所计算的相似度小于所述预设相似度阈值时，根据预先训练好的异常行为识别模型判断所述乘客中是否有异常行为发生；
当确定有异常行为发生时，将所述无人驾驶车辆的实时地理位置信息及所述发生异常行为的乘客的面部图像发送至远程服务器进行存储。
本申请的第二方面提供一种基于无人驾驶车辆的异常处理装置，运行于无人驾驶车辆中，所述装置包括：
图像获取模块，用于监测到有乘客乘坐时，获取所述乘客的面部图像；
第一识别模块，用于根据预先训练好的被拐人员识别模型计算所述乘客的面部图像与各个已知的被拐人员的面部图像之间的相似度；
第一发送模块，用于当所计算的相似度大于或者等于预设相似度阈值时，将所述无人驾驶车辆的实时地理位置信息及所述乘客的面部图像发送至公安部门进行援救；
第二识别模块，用于当所计算的相似度小于所述预设相似度阈值时，根据预先训练好的异常行为识别模型判断所述乘客中是否有异常行为发生；
第二发送模块，用于当确定有异常行为发生时，将所述无人驾驶车辆的实时地理位置信息及所述发生异常行为的乘客的面部图像发送至远程服务器进行存储。
本申请的第三方面提供一种无人驾驶车辆，所述无人驾驶车辆包括处理器和存储器，所述处理器用于执行所述存储器中存储的计算机可读指令时实现所述基于无人驾驶车辆的异常处理方法。
本申请的第四方面提供一种非易失性可读存储介质，所述非易失性可读存储介质上存储有计算机可读指令，所述计算机可读指令被处理器执行时实现所述基于无人驾驶车辆的异常处理方法。
本申请所述的基于无人驾驶车辆的异常处理方法、装置、无人驾驶车辆及存储介质，能够在有乘客乘坐时，根据被拐人员识别模型确定乘坐无人驾驶车辆的乘客为已知被拐人员时，可实时将无人驾驶车辆的地理位置信息及所述乘客的面部图像发送至公安部门进行援救，而在根据被拐人员识别模型无法确定乘坐无人驾驶车辆的乘客是否为未知的被拐人员时，进一步通过异常行为识别模型判断乘客是否发生了异常行为，通过识别乘客发生异常行为来预防已发生被拐但处于未知状态的情况，并实时的将人驾驶车辆的地理位置信息及所述乘客的面部图像发送至远程服务器进行存储，以便远程服务器后续再次接收到其他无人驾驶车辆发生的相同的乘客的面部图像时，即可认定为有拐卖行为发生，进而通过公安部门进行援救。
附图说明
图1是本申请提供的基于无人驾驶车辆的异常处理方法的应用环境示意图。
图2是本申请实施例一提供的基于无人驾驶车辆的异常处理方法的流程图。
图3是本申请实施例二提供的基于无人驾驶车辆的异常处理装置的功能模块图。
图4是本申请实施例三提供的无人驾驶车辆的示意图。
如下具体实施方式将结合上述附图进一步说明本申请。
具体实施方式
为了能够更清楚地理解本申请的上述目的、特征和优点，下面结合附图和具体实施例对本申请进行详细描述。需要说明的是，在不冲突的情况下，本申请的实施例及实施例中的特征可以相互组合。
参阅图1所示，为本申请提供的基于无人驾驶车辆的异常处理方法的应用环境示意图。
所述基于无人驾驶车辆的异常处理方法可以应用在由无人驾驶车辆1、网络2、远程服务器3、终端设备4及公安部门5构成的应用环境中。
所述无人驾驶车辆1可以是各种类型的无人驾驶车辆，例如无人驾驶公交车、无人驾驶轿车等等。本实施例中，所述无人驾驶车辆1中安装有高清数字图像采集设备，所述高清数字图像采集设备可以是针孔摄像机，能够隐藏在无人驾驶车辆中，避免被乘客发现。所述高清数字图像采集设备采集乘坐所述无人驾驶车辆1的乘客的面部图像，并通过所述网络2向所述远程服务器3发送所述乘客的面部图像。同时，所述无人驾驶车辆1根据所述乘客的面部图像判断所述乘客是否为被拐人员。
所述网络2用以在所述无人驾驶车辆1和所述远程服务器3之间提供通信连接的介质。所述网络2可以包括各种连接类型，例如有线、无线通信链路或者光纤电缆等等。
所述远程服务器3可以是提供各种服务的远程服务器，例如通过所述网络2向多个无人驾驶车辆1提供被拐人员识别模型，以及接收无人驾驶车辆1发送的乘客的面部图像的云端远程服务器。所述远程服务器3还可以通过所述网络2向所述终端设备4发送提示信息。
所述终端设备4上可以安装有各种通讯客户端应用，例如社交类应用等。所述终端设备4可以是寻找被拐卖人员的相关组织或相关人员所持的终端设备。终端设备4可以是具有显示屏并且支持无线通信的各种无人驾驶车辆，包括但不限于智能手机、平板电脑、 膝上型便携计算机等等。
需要说明的是，本申请实施例中的基于无人驾驶车辆的异常处理方法可以由所述无人驾驶车辆1执行，相应地，基于无人驾驶车辆的异常处理装置一般设置于无人驾驶车辆1中。本申请实施例中的基于无人驾驶车辆的异常处理方法也可以由所述远程服务器3执行，相应地，基于无人驾驶车辆的异常处理装置一般设置于远程服务器3中。
应该理解，图1中的无人驾驶车辆、网络、远程服务器和终端设备的数目仅仅是示意性的。根据实现需要，可以具有任意数目的移动终端、网络、远程服务器和终端设备。在其他实施例中，所述方法的应用环境中还可以不包括所述终端设备。
实施例一
图2是本申请实施例一提供的基于无人驾驶车辆的异常处理方法的流程图。所述基于无人驾驶车辆的异常处理方法应用于无人驾驶车辆中，根据不同的需求，该流程图中的执行顺序可以改变，某些步骤可以省略。
S21：监测到有乘客乘坐时，获取所述乘客的面部图像。
无人驾驶车辆监测到有乘客乘坐时，通过安装在无人驾驶车辆内部的高清数字图像采集设备获取乘坐该无人驾驶车辆的乘客的面部图像。
优选的，所述方法包括：
当监测到有乘客乘坐时，控制安装在所述无人驾驶车辆内部的高清数字图像采集设备开启并获取所述乘客的面部图像；
当监测到乘客下车时，控制所述高清数字图像采集设备关闭。
本实施例中，当监测到有乘客乘坐时，控制所述高清数字图像采集设备开启后，所述高清数字图像采集设备获取乘坐该无人驾驶车辆的乘客的面部图像。当无人驾驶车辆没有监测到有乘客乘坐时，不需控制所述高清数字图像采集设备开启，或者当无人驾驶车辆监测到乘客下车时，控制所述高清数字图像采集设备关闭。通过在有乘客乘坐时，控制高清数字图像采集设备开启，在无人乘坐或者乘客下车后，控制高清数字图像采集设备关闭，可以避免高清数字图像采集设备一种处于工作状态，减少高清数字图像采集设备的耗电量，从而节约无人驾驶车辆的电量，延长无人驾驶车辆的续航时间。
S22：根据预先训练好的被拐人员识别模型计算所述乘客的面部图像与各个已知的被拐人员的面部图像之间的相似度。
具体的，所述根据预先训练好的被拐人员识别模型计算所述乘客的面部图像与各个已知的被拐人员的面部图像的相似度包括：
每隔预定时间从所述远程服务器上下载所述预先训练好的被拐人员识别模型；
基于最新下载的所述预先训练好的被拐人员识别模型，计算每一个乘客的面部图像与各个已知的被拐人员的面部图像之间的相似度。
本实施例中，无人驾驶车辆可以每隔预定时间(例如24小时)通过网络从远程服务器(例如，云端服务器)上下载预先训练好的被拐人员识别模型，并基于预先训练好的被拐人员识别模型，对于乘坐的每一个乘客，计算每一个乘客的面部图像与各个已知的被拐人员的面部图像之间的相似度。
在一些应用场景中，所述远程服务器可以通过网络与相关部门，例如公安部的远程服务器连接或者和公益寻人平台的远程服务器连接。所述远程服务器可以每隔预定时间从相关部门的远程服务器中或者从公益寻人平台的远程服务器中获取多个被拐人员的面部图像。所述被拐人员的面部图像可以包括从多个角度拍摄的面部图像。从相关部门的远程服务器中或者从公益寻人平台的远程服务器中获取的多个被拐人员的面部图像对应的被拐人员为已知的被拐人员，即已被知晓姓名等身份信息的被拐人员。
所述远程服务器在获取到多个已知的被拐人员的面部图像时，根据所述面部图像训练被拐人员识别模型。具体地，可以使用从相关部门的远程服务器或者从公益寻人平台的远程服务器中获取的已知的各个被拐人员的面部图像以及其他参考人脸图像对所述被拐人员识别模型进行训练。对被拐人员识别模型进行训练之后，所述被拐人员识别模型可以识别出新输入的人脸图像与所述已知的各个被拐人员面部图像之间的相似度。可以理解的是，随着不断地获取新增加的已知的被拐人员，可以对被拐人员识别模型持续地进行训练和更新，从而得到精确度较高的被拐人员识别模型。
所述远程服务器预先训练被拐人员识别模型的具体过程可以包括：
1)获取预设第一数量的已知的被拐人员的人脸图片，将所述预设数量的人脸图片划分为第一图片集和第二图片集；
2)从所述第一图片集和第二图片集中分别提取出预设第一比例的人脸图片作为待训练的样本图片，并将第一图片集和第二图片集中剩余的人脸图片作为待验证的样本图片；
3)将各待训练的样本图片输入至卷积神经网络模型中进行训练得到被拐人员识别模型，并利用各待验证的样本图片对所训练得到的被拐人员识别模型进行验证；
4)若验证通过率大于或者等于预设第一阈值，则训练完成，否则增加待训练的样本图片的数量，以重新进行训练及验证。
在第一次训练卷积神经网络模型时，所述卷积神经网络模型的参数采用默认的参数进行训练，在训练过程不断调整参数，在训练得到所述被拐人员识别模型后，利用各待 验证的样本图片对所述被拐人员识别模型进行验证，如果验证通过率大于或者等于预设第一阈值，例如通过率大于或者等于98％，则训练结束，以该训练得到的被拐人员识别模型进行识别；如果验证通过率小于预设阈值，例如小于98％，则增加参与训练的人脸图片的数量，并重新执行上述的步骤，直至验证通过率大于或者等于预设第一阈值。在测试时，使用训练得到的被拐人员识别模型对测试集中的人脸图片进行识别，以评估所训练的被拐人员识别模型的识别效果。
S23：当所计算的相似度大于或者等于预设相似度阈值时，将所述无人驾驶车辆的实时地理位置信息及所述乘客的面部图像发送至公安部门进行援救。
本实施例中，无人驾驶车辆可以将获取的每一个乘客的面部图像输入至预先训练好的被拐人员识别模型中，并计算每一个乘客的面部图像与被拐人员的面部图像之间的相似度。当确定有一个乘客的面部图像之间的相似度大于或者等于预设相似度阈值时，即确定有拐卖行为发生，所述乘客中有已知的被拐人员。无人驾驶车辆可以通过网络将当前乘坐的所有乘客的面部图像及实时的地理位置信息发送至公安部门，也可以仅将当前乘坐的确定为已知的被拐人员的面部图像及实时的地理位置信息发送至公安部门，公安部门根据所述地理位置信息及所述乘客的面部图像进行援救。
当无人驾驶车辆确定任何一个乘客的面部图像之间的相似度小于预设相似度阈值时，则可以确定所述乘客中没有已知的被拐人员，但无法确定是否有拐卖行为发生，对于未知的被拐人员则无法通过被拐人员识别模型进行计算并做出判断，需执行S24进一步判断。
S24：当所计算的相似度小于预设相似度阈值时，根据预先训练好的异常行为识别模型判断所述乘客中是否有异常行为发生。
所述异常行为识别模型的训练过程包括：
1)获取预设第二数量的人脸图片，将包含异常行为的人脸图片作为正样本图片集，并将不包含异常行为的人脸图片作为负样本图片集；
所述异常行为可以包括，但不限于：紧张、恐惧、烦躁、哭泣、痛苦、掩面、痴呆、傻笑等。异常行为识别模型用于识别乘客的行为是否不符合自然或者是否异常，比如，乘客神情紧张、儿童不停的哭泣、乘客掩面等均属于不自然或不正常的行为。
2)从所述正样本图片集和负样本图片集中分别提取出预设第二比例的人脸图片作为待训练的样本图片，并将正样本图片集和负样本图片集中剩余的人脸图片作为待验证的样本图片；
3)将各待训练的样本图片输入至残差神经网络模型中进行训练得到异常行为识别模 型，并利用各待验证的样本图片对所训练得到的异常行为识别模型进行验证；
4)若验证通过率大于或者等于预设第二阈值，则训练完成，否则增加待训练的样本图片的数量，以重新进行训练及验证。
示例性的，假设获取1万张包含异常行为的人脸图片和1万张不包含异常行为的人脸图片，分别提取包含异常行为的人脸图片和不包含异常行为的人脸图片中的预设第二比例的人脸图片作为训练集，并将剩余的人脸图片作为测试集，训练集中的人脸图片的数量大于测试集中的人脸图片的数量，例如分别将包含异常行为的人脸图像和不包含异常行为的人脸图片中的80％的人脸图片作为训练集，将剩余的20％的人脸图片作为测试集。
在第一次训练残差神经网络模型时，该残差神经网络模型的参数采用默认的参数进行训练，在训练过程不断调整参数，在训练得到异常行为识别模型后，利用各待验证的样本图片对所生成的异常行为识别模型进行验证，如果验证通过率大于或者等于预设阈值，例如通过率大于或者等于98％，则训练结束，以该训练得到的异常行为识别模型进行识别无人驾驶车辆中当前乘坐的乘客中是否有异常行为发生；如果验证通过率小于预设阈值，例如小于98％，则增加人脸图片的数量，并重新执行上述的步骤，直至验证通过率大于或者等于预设第二阈值。在测试时，使用训练得到的异常行为识别模型对测试集中的人脸图片进行异常行为识别，以评估所训练的卷积神经网络模型的识别效果。
本实施例中，当通过预先训练好的被拐人员识别模型计算出的乘客面部图像小于预设相似度阈值时，再将所述乘客的面部图像输入至预先训练好的异常行为识别模型中确定是否有异常行为发生。
当确定有异常行为发生时，执行步骤S25。
S25：将所述无人驾驶车辆的实时地理位置信息及所述发生异常行为的乘客的面部图像发送至远程服务器进行存储。
如果确定有异常行为发生，则无人驾驶车辆可以将实时地理位置信息及发生异常行为的乘客的面部图像发送至远程服务器进行存储。后续远程服务器再接收到其他无人驾驶车辆发送的异常行为信息时，判断当前确定的发生异常行为的乘客的面部图像与历史确定的发生异常行为的乘客的面部图像是否相同。若远程服务器判断当前确定的发生异常行为的乘客的面部图像与历史确定的发生异常行为的乘客的面部图像不相同时，认为当前乘坐的乘客为首次发生了异常行为。若远程服务器判断当前确定的发生异常行为的乘客的面部图像与历史确定的发生异常行为的乘客的面部图像相同时，则可通知公安进行援救。
在其他实施例中，远程服务器存储无人驾驶车辆的实时地理位置信息及所述发生异 常行为的乘客的面部图像，后续在确定发生异常行为的乘客为已知被拐人员时，可根据所述实时地理位置信息辅助公安援救。
优选的，当所述所计算的相似度大于或者等于预设相似度阈值时，所述方法还包括：
在所述乘客下车时，将所述乘客的面部图像及预设第一告警信息发送至其他无人驾驶车辆，使得所述其他无人驾驶车辆在监测到所述乘客乘坐时将实时地理位置信息及所述乘客的面部图像发送至所述公安部门。
无人驾驶车辆在乘客下车时，将所述乘客的面部图像及预设第一告警信息发送至其他无人驾驶车辆，以便乘客通过不断的转乘其他无人驾驶车辆而避开搜索。所述预设第一告警信息可以是预设文字信息，例如，有被拐人员下车，请注意是否换乘其他无人驾驶车辆。
所述其他无人驾驶车辆在监测到乘客乘坐时，判断所述乘客是否为所述无人驾驶车辆发送的乘客。当确定所述乘客是否为所述无人驾驶车辆发送的乘客时，将实时地理位置信息及所述乘客的面部图像发送至所述公安部门。如此，有助于公安部门进行援救。
优选的，在所述监测到有乘客乘坐之后，在所述获取所述乘客的面部图像之前，所述方法还包括：
判断所述乘客的数量是否为1；
当确定所述乘客的数量不为1时，获取每一个乘客的面部图像。
本实施例中，在通过无人驾驶车辆监测到有乘客乘坐之后，可以通过无人驾驶车辆的高清数字图像采集设备检测所述乘客的人脸。应当理解的是，高清数字图像采集设备检测所述乘客的人脸的数量要么为1，要么为多个。
当检测到一张人脸时，认为只有一个乘客乘坐无人驾驶车辆，当检测到有多张人脸时，认为有多个乘客乘坐无人驾驶车辆。当确定所述乘客的数量不为1时，即至少有两个及两个以上的乘客乘坐无人驾驶车辆，则获取每一个乘客的面部图像。在确定所述乘客的数量为1的情况下，可以不必获取乘客的面部图像。如此还可以节约高清数字图像采集设备的工作量，节省高清数字图像采集设备上传乘客的面部图像至远程服务器时所需的网络资源，有利于提高其他无人驾驶车辆上传乘客的面部图像的速率。
优选的，在确定所述乘客的数量为1时，所述方法还可以包括：
获取乘客的面部图像；
根据预先训练的违法犯罪人员识别模型，确定所述乘客是否为违法犯罪人员；
在确定所述乘客为违法犯罪人员及/或当所述乘客下车时，发送预设第二告警信息至乘客下车地点预设距离内的行人。
无人驾驶车辆可以通过网络发送预设第二告警信息至乘客下车地点预设距离内的行人的终端设备。可以以短信的形式发送告警信息。预设第二告警信息可以是预设文字信息，例如，请看管好自己的随身物品或者小孩，有不明人员在附近。
本实施例中，通过预先训练违法犯罪人员识别模型，在确定所述乘客为违法犯罪人员时，发送预设告警信息至乘客下车附近的行人，以提示行人注意人身和财产安全，可有效的消除安全隐患，提高行人安全，有效的减少拐卖或者犯罪行为的发生。
需要说明的是，本申请实施例中的违法犯罪人员识别模型是通过获取公安系统的网上追逃、惯犯等的数据库中的人脸面部图像训练得到的，与被拐人员识别模型的训练过程类似，在此不再详细赘述。
综上所述，本申请实施例提供的所述基于无人驾驶车辆的异常处理方法，监测到有乘客乘坐时，获取所述乘客的面部图像；根据预先训练好的被拐人员识别模型计算所述乘客的面部图像与各个已知的被拐人员的面部图像的相似度；当所计算的相似度大于或者等于预设相似度阈值时，将所述无人驾驶车辆的实时地理位置信息及所述乘客的面部图像发送至公安部门进行援救；当所计算的相似度小于预设相似度阈值时，根据预先训练好的异常行为识别模型判断所述乘客中是否有异常行为发生；当确定有异常行为发生时，将所述无人驾驶车辆的实时地理位置信息及所述发生异常行为的乘客的面部图像发送至远程服务器进行存储。本申请在有乘客乘坐时，根据被拐人员识别模型确定乘坐无人驾驶车辆的乘客为已知被拐人员时，可实时将无人驾驶车辆的地理位置信息及所述乘客的面部图像发送至公安部门进行援救，而在根据被拐人员识别模型无法确定乘坐无人驾驶车辆的乘客是否为未知的被拐人员时，进一步通过异常行为识别模型判断乘客是否发生了异常行为，通过识别乘客发生异常行为来预防已发生被拐但处于未知状态的情况，并实时的将人驾驶车辆的地理位置信息及所述乘客的面部图像发送至远程服务器进行存储，以便远程服务器后续再次接收到其他无人驾驶车辆发生的相同的乘客的面部图像时，即可认定为有拐卖行为发生，进而通过公安部门进行援救。
以上所述，仅是本申请的具体实施方式，但本申请的保护范围并不局限于此，对于本领域的普通技术人员来说，在不脱离本申请创造构思的前提下，还可以做出改进，但这些均属于本申请的保护范围。
下面结合第3至4图，分别对实现上述基于无人驾驶车辆的异常处理方法的无人驾驶车辆的功能模块及硬件结构进行介绍。
实施例二
图3为本申请基于无人驾驶车辆的异常处理装置较佳实施例中的功能模块图。
在一些实施例中，所述基于无人驾驶车辆的异常处理装置30运行于无人驾驶车辆中。所述基于无人驾驶车辆的异常处理装置30可以包括多个由程序代码段所组成的功能模块。所述基于无人驾驶车辆的异常处理装置30中的各个程序段的程序代码可以存储于存储器中，并由至少一个处理器所执行，以执行(详见图2及其相关描述)基于无人驾驶车辆的异常处理方法。
本实施例中，所述基于无人驾驶车辆的异常处理装置30根据其所执行的功能，可以被划分为多个功能模块。所述功能模块可以包括：图像获取模块301、开启控制模块302、第一识别模块303、第一发送模块304、第二识别模块305、第二发送模块306、第三发送模块307、数量判断模块308、第三识别模块309及第四发送模块310。本申请所称的模块是指一种能够被至少一个处理器所执行并且能够完成固定功能的一系列计算机程序段，其存储在存储器中。在一些实施例中，关于各模块的功能将在后续的实施例中详述。
图像获取模块301，用于监测到有乘客乘坐时，获取所述乘客的面部图像。
无人驾驶车辆监测到有乘客乘坐时，通过安装在无人驾驶车辆内部的高清数字图像采集设备获取乘坐该无人驾驶车辆的乘客的面部图像。
开启控制模块302，用于当监测到有乘客乘坐时，控制安装在所述无人驾驶车辆内部的高清数字图像采集设备开启并获取所述乘客的面部图像；当监测到乘客下车时，控制所述高清数字图像采集设备关闭。
本实施例中，当监测到有乘客乘坐时，控制所述高清数字图像采集设备开启后，所述高清数字图像采集设备获取乘坐该无人驾驶车辆的乘客的面部图像。当无人驾驶车辆没有监测到有乘客乘坐时，不需控制所述高清数字图像采集设备开启，或者当无人驾驶车辆监测到乘客下车时，控制所述高清数字图像采集设备关闭。通过在有乘客乘坐时，控制高清数字图像采集设备开启，在无人乘坐或者乘客下车后，控制高清数字图像采集设备关闭，可以避免高清数字图像采集设备一种处于工作状态，减少高清数字图像采集设备的耗电量，从而节约无人驾驶车辆的电量，延长无人驾驶车辆的续航时间。
第一识别模块303，用于根据预先训练好的被拐人员识别模型计算所述乘客的面部图像与各个已知的被拐人员的面部图像之间的相似度。
具体的，所述第一识别模块303根据预先训练好的被拐人员识别模型计算所述乘客的面部图像与各个已知的被拐人员的面部图像的相似度包括：
每隔预定时间从所述远程服务器上下载所述预先训练好的被拐人员识别模型；
基于最新下载的所述预先训练好的被拐人员识别模型，计算每一个乘客的面部图像与各个已知的被拐人员的面部图像之间的相似度。
本实施例中，无人驾驶车辆可以每隔预定时间(例如24小时)通过网络从远程服务器(例如，云端服务器)上下载预先训练好的被拐人员识别模型，并基于预先训练好的被拐人员识别模型，对于乘坐的每一个乘客，计算每一个乘客的面部图像与各个已知的被拐人员的面部图像之间的相似度。
在一些应用场景中，所述远程服务器可以通过网络与相关部门，例如公安部的远程服务器连接或者和公益寻人平台的远程服务器连接。所述远程服务器可以每隔预定时间从相关部门的远程服务器中或者从公益寻人平台的远程服务器中获取多个被拐人员的面部图像。所述被拐人员的面部图像可以包括从多个角度拍摄的面部图像。从相关部门的远程服务器中或者从公益寻人平台的远程服务器中获取的多个被拐人员的面部图像对应的被拐人员为已知的被拐人员，即已被知晓姓名等身份信息的被拐人员。
所述远程服务器在获取到多个已知的被拐人员的面部图像时，根据所述面部图像训练被拐人员识别模型。具体地，可以使用从相关部门的远程服务器或者从公益寻人平台的远程服务器中获取的已知的各个被拐人员的面部图像以及其他参考人脸图像对所述被拐人员识别模型进行训练。对被拐人员识别模型进行训练之后，所述被拐人员识别模型可以识别出新输入的人脸图像与所述已知的各个被拐人员面部图像之间的相似度。可以理解的是，随着不断地获取新增加的已知的被拐人员，可以对被拐人员识别模型持续地进行训练和更新，从而得到精确度较高的被拐人员识别模型。
所述远程服务器预先训练被拐人员识别模型的具体过程可以包括：
1)获取预设第一数量的已知的被拐人员的人脸图片，将所述预设数量的人脸图片划分为第一图片集和第二图片集；
2)从所述第一图片集和第二图片集中分别提取出预设第一比例的人脸图片作为待训练的样本图片，并将第一图片集和第二图片集中剩余的人脸图片作为待验证的样本图片；
3)将各待训练的样本图片输入至卷积神经网络模型中进行训练得到被拐人员识别模型，并利用各待验证的样本图片对所训练得到的被拐人员识别模型进行验证；
4)若验证通过率大于或者等于预设第一阈值，则训练完成，否则增加待训练的样本图片的数量，以重新进行训练及验证。
在第一次训练卷积神经网络模型时，所述卷积神经网络模型的参数采用默认的参数进行训练，在训练过程不断调整参数，在训练得到所述被拐人员识别模型后，利用各待验证的样本图片对所述被拐人员识别模型进行验证，如果验证通过率大于或者等于预设第一阈值，例如通过率大于或者等于98％，则训练结束，以该训练得到的被拐人员识别模型进行识别；如果验证通过率小于预设阈值，例如小于98％，则增加参与训练的人脸 图片的数量，并重新执行上述的步骤，直至验证通过率大于或者等于预设第一阈值。在测试时，使用训练得到的被拐人员识别模型对测试集中的人脸图片进行识别，以评估所训练的被拐人员识别模型的识别效果。
第一发送模块304，用于当所计算的相似度大于或者等于预设相似度阈值时，将所述无人驾驶车辆的实时地理位置信息及所述乘客的面部图像发送至公安部门进行援救。
本实施例中，无人驾驶车辆可以将获取的每一个乘客的面部图像输入至预先训练好的被拐人员识别模型中，并计算每一个乘客的面部图像与被拐人员的面部图像之间的相似度。当确定有一个乘客的面部图像之间的相似度大于或者等于预设相似度阈值时，即确定有拐卖行为发生，所述乘客中有已知的被拐人员。无人驾驶车辆可以通过网络将当前乘坐的所有乘客的面部图像及实时的地理位置信息发送至公安部门，也可以仅将当前乘坐的确定为已知的被拐人员的面部图像及实时的地理位置信息发送至公安部门，公安部门根据所述地理位置信息及所述乘客的面部图像进行援救。
当无人驾驶车辆确定任何一个乘客的面部图像之间的相似度小于预设相似度阈值时，则可以确定所述乘客中没有已知的被拐人员，但无法确定是否有拐卖行为发生，对于未知的被拐人员则无法通过被拐人员识别模型进行计算并做出判断，需执行第二识别模块305进一步判断。
第二识别模块305，用于当所计算的相似度小于预设相似度阈值时，根据预先训练好的异常行为识别模型判断所述乘客中是否有异常行为发生。
所述异常行为识别模型的训练过程包括：
1)获取预设第二数量的人脸图片，将包含异常行为的人脸图片作为正样本图片集，并将不包含异常行为的人脸图片作为负样本图片集；
所述异常行为可以包括，但不限于：紧张、恐惧、烦躁、哭泣、痛苦、掩面、痴呆、傻笑等。异常行为识别模型用于识别乘客的行为是否不符合自然或者是否异常，比如，乘客神情紧张、儿童不停的哭泣、乘客掩面等均属于不自然或不正常的行为。
2)从所述正样本图片集和负样本图片集中分别提取出预设第二比例的人脸图片作为待训练的样本图片，并将正样本图片集和负样本图片集中剩余的人脸图片作为待验证的样本图片；
3)将各待训练的样本图片输入至残差神经网络模型中进行训练得到异常行为识别模型，并利用各待验证的样本图片对所训练得到的异常行为识别模型进行验证；
4)若验证通过率大于或者等于预设第二阈值，则训练完成，否则增加待训练的样本图片的数量，以重新进行训练及验证。
示例性的，假设获取1万张包含异常行为的人脸图片和1万张不包含异常行为的人脸图片，分别提取包含异常行为的人脸图片和不包含异常行为的人脸图片中的预设第二比例的人脸图片作为训练集，并将剩余的人脸图片作为测试集，训练集中的人脸图片的数量大于测试集中的人脸图片的数量，例如分别将包含异常行为的人脸图像和不包含异常行为的人脸图片中的80％的人脸图片作为训练集，将剩余的20％的人脸图片作为测试集。
在第一次训练残差神经网络模型时，该残差神经网络模型的参数采用默认的参数进行训练，在训练过程不断调整参数，在训练得到异常行为识别模型后，利用各待验证的样本图片对所生成的异常行为识别模型进行验证，如果验证通过率大于或者等于预设阈值，例如通过率大于或者等于98％，则训练结束，以该训练得到的异常行为识别模型进行识别无人驾驶车辆中当前乘坐的乘客中是否有异常行为发生；如果验证通过率小于预设阈值，例如小于98％，则增加人脸图片的数量，并重新执行上述的步骤，直至验证通过率大于或者等于预设第二阈值。在测试时，使用训练得到的异常行为识别模型对测试集中的人脸图片进行异常行为识别，以评估所训练的卷积神经网络模型的识别效果。
本实施例中，当通过预先训练好的被拐人员识别模型计算出的乘客面部图像小于预设相似度阈值时，再将所述乘客的面部图像输入至预先训练好的异常行为识别模型中确定是否有异常行为发生。
第二发送模块306，用于当确定有异常行为发生时，将所述无人驾驶车辆的实时地理位置信息及所述发生异常行为的乘客的面部图像发送至远程服务器进行存储。
如果确定有异常行为发生，则无人驾驶车辆可以将实时地理位置信息及发生异常行为的乘客的面部图像发送至远程服务器进行存储。后续远程服务器再接收到其他无人驾驶车辆发送的异常行为信息时，判断当前确定的发生异常行为的乘客的面部图像与历史确定的发生异常行为的乘客的面部图像是否相同。若远程服务器判断当前确定的发生异常行为的乘客的面部图像与历史确定的发生异常行为的乘客的面部图像不相同时，认为当前乘坐的乘客为首次发生了异常行为。若远程服务器判断当前确定的发生异常行为的乘客的面部图像与历史确定的发生异常行为的乘客的面部图像相同时，则可通知公安进行援救。
在其他实施例中，远程服务器存储无人驾驶车辆的实时地理位置信息及所述发生异常行为的乘客的面部图像，后续在确定发生异常行为的乘客为已知被拐人员时，可根据所述实时地理位置信息辅助公安援救。
优选的，所述基于无人驾驶车辆的异常处理装置30还包括：
第三发送模块307，用于当所计算的相似度大于或者等于所述预设相似度阈值之后，在所述乘客下车时，将所述乘客的面部图像及预设第一告警信息发送至其他无人驾驶车辆，使得所述其他无人驾驶车辆在监测到所述乘客乘坐时将实时地理位置信息及所述乘客的面部图像发送至所述公安部门。
无人驾驶车辆在乘客下车时，将所述乘客的面部图像及预设第一告警信息发送至其他无人驾驶车辆，以便乘客通过不断的转乘其他无人驾驶车辆而避开搜索。所述预设第一告警信息可以是预设文字信息，例如，有被拐人员下车，请注意是否换乘其他无人驾驶车辆。
所述其他无人驾驶车辆在监测到乘客乘坐时，判断所述乘客是否为所述无人驾驶车辆发送的乘客。当确定所述乘客是否为所述无人驾驶车辆发送的乘客时，将实时地理位置信息及所述乘客的面部图像发送至所述公安部门。如此，有助于公安部门进行援救。
数量判断模块308，用于在所述监测到有乘客乘坐之后，在所述获取所述乘客的面部图像之前，判断所述乘客的数量是否为1。
所述图像获取模块301，还用于当所述数量判断模块308确定所述乘客的数量不为1时，获取每一个乘客的面部图像。
本实施例中，在通过无人驾驶车辆监测到有乘客乘坐之后，可以通过无人驾驶车辆的高清数字图像采集设备检测所述乘客的人脸。应当理解的是，高清数字图像采集设备检测所述乘客的人脸的数量要么为1，要么为多个。
当检测到一张人脸时，认为只有一个乘客乘坐无人驾驶车辆，当检测到有多张人脸时，认为有多个乘客乘坐无人驾驶车辆。当确定所述乘客的数量不为1时，即至少有两个及两个以上的乘客乘坐无人驾驶车辆，则获取每一个乘客的面部图像。在确定所述乘客的数量为1的情况下，可以不必获取乘客的面部图像。如此还可以节约高清数字图像采集设备的工作量，节省高清数字图像采集设备上传乘客的面部图像至远程服务器时所需的网络资源，有利于提高其他无人驾驶车辆上传乘客的面部图像的速率。
优选的，所述图像获取模块301，还用于在确定所述乘客的数量为1时，获取乘客的面部图像。
第三识别模块309，用于根据预先训练的违法犯罪人员识别模型，确定所述乘客是否为违法犯罪人员。
第四发送模块310，用于在确定所述乘客为违法犯罪人员及/或当所述乘客下车时，发送预设第二告警信息至乘客下车地点预设距离内的行人。
无人驾驶车辆可以通过网络发送预设第二告警信息至乘客下车地点预设距离内的行 人的终端设备。可以以短信的形式发送告警信息。预设第二告警信息可以是预设文字信息，例如，请看管好自己的随身物品或者小孩，有不明人员在附近。
本实施例中，通过预先训练违法犯罪人员识别模型，在确定所述乘客为违法犯罪人员时，发送预设告警信息至乘客下车附近的行人，以提示行人注意人身和财产安全，可有效的消除安全隐患，提高行人安全，有效的减少拐卖或者犯罪行为的发生。
需要说明的是，本申请实施例中的违法犯罪人员识别模型是通过获取公安系统的网上追逃、惯犯等的数据库中的人脸面部图像训练得到的，与被拐人员识别模型的训练过程类似，在此不再详细赘述。
综上所述，本申请实施例提供的所述基于无人驾驶车辆的异常处理装置，监测到有乘客乘坐时，获取所述乘客的面部图像；根据预先训练好的被拐人员识别模型计算所述乘客的面部图像与各个已知的被拐人员的面部图像的相似度；当所计算的相似度大于或者等于预设相似度阈值时，将所述无人驾驶车辆的实时地理位置信息及所述乘客的面部图像发送至公安部门进行援救；当所计算的相似度小于预设相似度阈值时，根据预先训练好的异常行为识别模型判断所述乘客中是否有异常行为发生；当确定有异常行为发生时，将所述无人驾驶车辆的实时地理位置信息及所述发生异常行为的乘客的面部图像发送至远程服务器进行存储。本申请在有乘客乘坐时，根据被拐人员识别模型确定乘坐无人驾驶车辆的乘客为已知被拐人员时，可实时将无人驾驶车辆的地理位置信息及所述乘客的面部图像发送至公安部门进行援救，而在根据被拐人员识别模型无法确定乘坐无人驾驶车辆的乘客是否为未知的被拐人员时，进一步通过异常行为识别模型判断乘客是否发生了异常行为，通过识别乘客发生异常行为来预防已发生被拐但处于未知状态的情况，并实时的将人驾驶车辆的地理位置信息及所述乘客的面部图像发送至远程服务器进行存储，以便远程服务器后续再次接收到其他无人驾驶车辆发生的相同的乘客的面部图像时，即可认定为有拐卖行为发生，进而通过公安部门进行援救。
上述以软件功能模块的形式实现的集成的单元，可以存储在一个计算机可读取存储介质中。上述软件功能模块存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，双屏设备，或者网络设备等)或处理器(processor)执行本申请各个实施例所述方法的部分。
实施例三
图4为本申请实施例三提供的无人驾驶车辆的示意图。
所述无人驾驶车辆4包括：车辆本体40、存储器41、至少一个处理器42、存储在 所述存储器41中并可在所述至少一个处理器42上运行的计算机程序43及至少一条通讯总线44。
所述至少一个处理器42执行所述计算机程序43时实现上述方法实施例中的步骤。
示例性的，所述计算机程序43可以被分割成一个或多个模块/单元，所述一个或者多个模块/单元被存储在所述存储器41中，并由所述至少一个处理器42执行，以完成本申请上述方法实施例中的步骤。所述一个或多个模块/单元可以是能够完成特定功能的一系列计算机程序指令段，该指令段用于描述所述计算机程序43在所述无人驾驶车辆4中的执行过程。
所述无人驾驶车辆4可以是桌上型计算机、笔记本、掌上电脑及云端远程服务器等计算设备。本领域技术人员可以理解，所述示意图4仅仅是无人驾驶车辆4的示例，并不构成对无人驾驶车辆4的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件，例如所述无人驾驶车辆4还可以包括输入输出设备、网络接入设备、总线等。
所述至少一个处理器42可以是中央处理单元(Central Processing Unit，CPU)，还可以是其他通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现场可编程门阵列(Field-Programmable Gate Array，FPGA)或者其他可编程逻辑器件或者晶体管逻辑器件、分立硬件组件等。该处理器42可以是微处理器或者该处理器42也可以是任何常规的处理器等，所述处理器42是所述无人驾驶车辆4的控制中心，利用各种接口和线路连接整个无人驾驶车辆4的各个部分。
所述存储器41可用于存储所述计算机程序43和/或模块/单元，所述处理器42通过运行或执行存储在所述存储器41内的一系列计算机可读指令和/或模块/单元，以及调用存储在存储器41内的数据，实现所述无人驾驶车辆4的各种功能。所述存储器41可主要包括存储程序区和存储数据区，其中，存储程序区可存储操作系统、至少一个功能所需的应用程序(比如声音播放功能、图像播放功能等)等；存储数据区可存储根据无人驾驶车辆4的使用所创建的数据(比如音频数据)等。此外，存储器41可以包括非易失性存储器，例如硬盘、内存、插接式硬盘，智能存储卡(Smart Media Card，SMC)，安全数字(Secure Digital，SD)卡，闪存卡(Flash Card)、至少一个磁盘存储器件、闪存器件、或其他非易失性固态存储器件。
所述无人驾驶车辆4集成的模块/单元如果以软件功能模块的形式实现并作为独立的产品销售或使用时，可以存储在一个非易失性可读取存储介质中。基于这样的理解，本 申请实现上述实施例方法中的全部或部分流程，也可以通过计算机程序来指令相关的硬件来完成，所述的计算机程序可存储于一非易失性可读存储介质中，该计算机程序在被处理器执行时，可实现上述各个方法实施例的步骤。其中，所述计算机程序包括一系列计算机可读指令代码，所述计算机可读指令代码可以为源代码形式、对象代码形式、可执行文件或某些中间形式等。所述非易失性可读介质可以包括：能够携带所述计算机可读指令代码的任何实体或装置、记录介质、U盘、移动硬盘、磁碟、光盘、计算机存储器、只读存储器(ROM，Read-Only Memory)。
在本申请所提供的几个实施例中，应该理解到，所揭露的无人驾驶车辆和方法，可以通过其它的方式实现。例如，以上所描述的无人驾驶车辆实施例仅仅是示意性的，例如，所述模块的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式。
最后应说明的是，以上实施例仅用以说明本申请的技术方案而非限制，尽管参照较佳实施例对本申请进行了详细说明，本领域的普通技术人员应当理解，可以对本申请的技术方案进行修改或等同替换，而不脱离本申请技术方案的精神范围。

Claims (20)

  1. 一种基于无人驾驶车辆的异常处理方法，应用于无人驾驶车辆中，其特征在于，所述方法包括：
    监测到有乘客乘坐时，获取所述乘客的面部图像；
    根据预先训练好的被拐人员识别模型计算所述乘客的面部图像与各个已知的被拐人员的面部图像之间的相似度；
    当所计算的相似度大于或者等于预设相似度阈值时，将所述无人驾驶车辆的实时地理位置信息及所述乘客的面部图像发送至公安部门进行援救；
    当所计算的相似度小于所述预设相似度阈值时，根据预先训练好的异常行为识别模型判断所述乘客中是否有异常行为发生；
    当确定有异常行为发生时，将所述无人驾驶车辆的实时地理位置信息及所述发生异常行为的乘客的面部图像发送至远程服务器进行存储。
  2. 如权利要求1所述的方法，其特征在于，当所述所计算的相似度大于或者等于预设相似度阈值时，所述方法还包括：
    在所述乘客下车时，将所述乘客的面部图像及预设第一告警信息发送至其他无人驾驶车辆，使得所述其他无人驾驶车辆在监测到所述乘客乘坐时将实时地理位置信息及所述乘客的面部图像发送至所述公安部门。
  3. 如权利要求1所述的方法，其特征在于，在所述监测到有乘客乘坐之后，在所述获取所述乘客的面部图像之前，所述方法还包括：
    判断所述乘客的数量是否为1；
    当确定所述乘客的数量不为1时，获取每一个乘客的面部图像。
  4. 如权利要求3所述的方法，其特征在于，在确定所述乘客的数量为1时，所述方法还包括：
    获取所述乘客的面部图像；
    根据预先训练的违法犯罪人员识别模型，确定所述乘客是否为违法犯罪人员；
    在确定所述乘客为违法犯罪人员及/或当所述乘客下车时，发送预设第二告警信息至所述乘客下车地点预设距离内的行人。
  5. 如权利要求1所述的方法，其特征在于，所述根据预先训练好的被拐人员识别模型计算所述乘客的面部图像与各个已知的被拐人员的面部图像之间的相似度包括：
    每隔预定时间从所述远程服务器上下载所述预先训练好的被拐人员识别模型；
    基于最新下载的所述预先训练好的被拐人员识别模型，计算每一个乘客的面部图像与各个已知的被拐人员的面部图像之间的相似度。
  6. 如权利要求1所述的方法，其特征在于，所述异常行为识别模型的训练过程包括：
    获取预设数量的人脸图片，将包含异常行为的人脸图片作为正样本图片集，并将不包含异常行为的人脸图片作为负样本图片集；
    从所述正样本图片集和所述负样本图片集中分别提取出预设比例的人脸图片作为待训练的样本图片，并将所述正样本图片集和所述负样本图片集中剩余的人脸图片作为待验证的样本图片；
    将各待训练的样本图片输入至残差神经网络模型中进行训练得到异常行为识别模型，并利用各待验证的样本图片对所训练得到的异常行为识别模型进行验证；
    若验证通过率大于或者等于预设阈值时，则训练完成；否则，若验证通过率大于或者等于所述预设阈值时，增加待训练的样本图片的数量，以重新进行训练及验证。
  7. 如权利要求1所述的方法，其特征在于，所述方法还包括：
    当监测到有乘客乘坐时，控制安装在所述无人驾驶车辆内部的高清数字图像采集设备开启并获取所述乘客的面部图像；
    当监测到乘客下车时，控制所述高清数字图像采集设备关闭。
  8. 一种基于无人驾驶车辆的异常处理装置，运行于无人驾驶车辆中，其特征在于，所述装置包括：
    图像获取模块，用于监测到有乘客乘坐时，获取所述乘客的面部图像；
    第一识别模块，用于根据预先训练好的被拐人员识别模型计算所述乘客的面部图像与各个已知的被拐人员的面部图像之间的相似度；
    第一发送模块，用于当所计算的相似度大于或者等于预设相似度阈值时，将所述无人驾驶车辆的实时地理位置信息及所述乘客的面部图像发送至公安部门进行援救；
    第二识别模块，用于当所计算的相似度小于所述预设相似度阈值时，根据预先训练好的异常行为识别模型判断所述乘客中是否有异常行为发生；
    第二发送模块，用于当确定有异常行为发生时，将所述无人驾驶车辆的实时地理位置信息及所述发生异常行为的乘客的面部图像发送至远程服务器进行存储。
  9. 一种无人驾驶车辆，其特征在于，所述无人驾驶车辆包括处理器和存储器，所述处理器用于执行所述存储器中存储的至少一个计算机可读指令以实现以下步骤：
    监测到有乘客乘坐时，获取所述乘客的面部图像；
    根据预先训练好的被拐人员识别模型计算所述乘客的面部图像与各个已知的被拐人员的 面部图像之间的相似度；
    当所计算的相似度大于或者等于预设相似度阈值时，将所述无人驾驶车辆的实时地理位置信息及所述乘客的面部图像发送至公安部门进行援救；
    当所计算的相似度小于所述预设相似度阈值时，根据预先训练好的异常行为识别模型判断所述乘客中是否有异常行为发生；
    当确定有异常行为发生时，将所述无人驾驶车辆的实时地理位置信息及所述发生异常行为的乘客的面部图像发送至远程服务器进行存储。
  10. 如权利要求9所述的无人驾驶车辆，其特征在于，当所述所计算的相似度大于或者等于预设相似度阈值时，所述处理器执行所述至少一个计算机可读指令时，还用以实现以下步骤：
    在所述乘客下车时，将所述乘客的面部图像及预设第一告警信息发送至其他无人驾驶车辆，使得所述其他无人驾驶车辆在监测到所述乘客乘坐时将实时地理位置信息及所述乘客的面部图像发送至所述公安部门。
  11. 如权利要求9所述的无人驾驶车辆，其特征在于，在所述监测到有乘客乘坐之后，在所述获取所述乘客的面部图像之前，所述处理器执行所述至少一个计算机可读指令时，还用以实现以下步骤：
    判断所述乘客的数量是否为1；
    当确定所述乘客的数量不为1时，获取每一个乘客的面部图像。
  12. 如权利要求11所述的无人驾驶车辆，其特征在于，在确定所述乘客的数量为1时，所述处理器执行所述至少一个计算机可读指令时，还用以实现以下步骤：
    获取所述乘客的面部图像；
    根据预先训练的违法犯罪人员识别模型，确定所述乘客是否为违法犯罪人员；
    在确定所述乘客为违法犯罪人员及/或当所述乘客下车时，发送预设第二告警信息至所述乘客下车地点预设距离内的行人。
  13. 如权利要求9所述的无人驾驶车辆，其特征在于，所述根据预先训练好的被拐人员识别模型计算所述乘客的面部图像与各个已知的被拐人员的面部图像之间的相似度时，所述处理器执行所述至少一个计算机可读指令以实现以下步骤：
    每隔预定时间从所述远程服务器上下载所述预先训练好的被拐人员识别模型；
    基于最新下载的所述预先训练好的被拐人员识别模型，计算每一个乘客的面部图像与各个已知的被拐人员的面部图像之间的相似度。
  14. 如权利要求9所述的无人驾驶车辆，其特征在于，所述处理器执行所述至少一个计算 机可读指令还用以实现以下步骤：
    当监测到有乘客乘坐时，控制安装在所述无人驾驶车辆内部的高清数字图像采集设备开启并获取所述乘客的面部图像；
    当监测到乘客下车时，控制所述高清数字图像采集设备关闭。
  15. 一种非易失性可读存储介质，所述非易失性可读存储介质上存储有计算机可读指令，其特征在于，所述计算机可读指令被处理器执行时实现以下步骤：
    监测到有乘客乘坐时，获取所述乘客的面部图像；
    根据预先训练好的被拐人员识别模型计算所述乘客的面部图像与各个已知的被拐人员的面部图像之间的相似度；
    当所计算的相似度大于或者等于预设相似度阈值时，将所述无人驾驶车辆的实时地理位置信息及所述乘客的面部图像发送至公安部门进行援救；
    当所计算的相似度小于所述预设相似度阈值时，根据预先训练好的异常行为识别模型判断所述乘客中是否有异常行为发生；
    当确定有异常行为发生时，将所述无人驾驶车辆的实时地理位置信息及所述发生异常行为的乘客的面部图像发送至远程服务器进行存储。
  16. 如权利要求15所述的存储介质，其特征在于，当所述所计算的相似度大于或者等于预设相似度阈值时，所述至少一个计算机可读指令被处理器执行时，还用以实现以下步骤：
    在所述乘客下车时，将所述乘客的面部图像及预设第一告警信息发送至其他无人驾驶车辆，使得所述其他无人驾驶车辆在监测到所述乘客乘坐时将实时地理位置信息及所述乘客的面部图像发送至所述公安部门。
  17. 如权利要求15所述的存储介质，其特征在于，在所述监测到有乘客乘坐之后，在所述获取所述乘客的面部图像之前，所述至少一个计算机可读指令被处理器执行时，还用以实现以下步骤：
    判断所述乘客的数量是否为1；
    当确定所述乘客的数量不为1时，获取每一个乘客的面部图像。
  18. 如权利要求17所述的存储介质，其特征在于，在确定所述乘客的数量为1时，所述至少一个计算机可读指令被处理器执行时，还用以实现以下步骤：
    获取所述乘客的面部图像；
    根据预先训练的违法犯罪人员识别模型，确定所述乘客是否为违法犯罪人员；
    在确定所述乘客为违法犯罪人员及/或当所述乘客下车时，发送预设第二告警信息至所述乘客下车地点预设距离内的行人。
  19. 如权利要求15所述的存储介质，其特征在于，所述根据预先训练好的被拐人员识别模型计算所述乘客的面部图像与各个已知的被拐人员的面部图像之间的相似度时，所述至少一个计算机可读指令被处理器执行以实现以下步骤：
    每隔预定时间从所述远程服务器上下载所述预先训练好的被拐人员识别模型；
    基于最新下载的所述预先训练好的被拐人员识别模型，计算每一个乘客的面部图像与各个已知的被拐人员的面部图像之间的相似度。
  20. 如权利要求15所述的存储介质，其特征在于，所述至少一个计算机可读指令被处理器执行还用以实现以下步骤：
    当监测到有乘客乘坐时，控制安装在所述无人驾驶车辆内部的高清数字图像采集设备开启并获取所述乘客的面部图像；
    当监测到乘客下车时，控制所述高清数字图像采集设备关闭。
PCT/CN2019/118606 2019-01-24 2019-11-14 基于无人驾驶车辆的异常处理方法、装置及相关设备 Ceased WO2020151339A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910070260.9A CN109902575B (zh) 2019-01-24 2019-01-24 基于无人驾驶车辆的防拐卖方法、装置及相关设备
CN201910070260.9 2019-01-24

Publications (1)

Publication Number Publication Date
WO2020151339A1 true WO2020151339A1 (zh) 2020-07-30

Family

ID=66944236

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/118606 Ceased WO2020151339A1 (zh) 2019-01-24 2019-11-14 基于无人驾驶车辆的异常处理方法、装置及相关设备

Country Status (2)

Country Link
CN (1) CN109902575B (zh)
WO (1) WO2020151339A1 (zh)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113507593A (zh) * 2021-06-24 2021-10-15 中汽创智科技有限公司 一种用于车辆座舱的监测方法、装置、系统及终端
CN114360230A (zh) * 2021-09-24 2022-04-15 深圳市卡联科技股份有限公司 一种公交车安全预警方法、系统、装置及存储介质
CN114475623A (zh) * 2021-12-28 2022-05-13 阿波罗智联(北京)科技有限公司 车辆的控制方法、装置、电子设备及存储介质
CN114550076A (zh) * 2021-10-19 2022-05-27 深圳市中博科创信息技术有限公司 区域异常行为监控方法、装置、设备及存储介质
CN114677745A (zh) * 2022-04-13 2022-06-28 安徽华昇能源互联网研究院有限公司 一种智能电缆井安全管理系统
CN114692757A (zh) * 2022-03-30 2022-07-01 北京汇通天下物联科技有限公司 一种异常驾驶行为的监测方法、装置及电子设备

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109902575B (zh) * 2019-01-24 2024-03-15 平安科技（深圳）有限公司 基于无人驾驶车辆的防拐卖方法、装置及相关设备
CN112208475B (zh) * 2019-07-09 2023-02-03 奥迪股份公司 用于车辆乘员的安全保护系统、车辆及相应的方法和介质
CN110956143A (zh) * 2019-12-03 2020-04-03 交控科技股份有限公司 一种异常行为检测方法、装置、电子设备及存储介质
CN111798356B (zh) * 2020-07-09 2023-11-10 北京交通大学 一种基于大数据的轨道交通客流异常模式识别方法
CN112070011A (zh) * 2020-09-08 2020-12-11 安徽兰臣信息科技有限公司 一种可用于走失儿童找寻的无感人脸识别摄像抓拍机
CN113157833A (zh) * 2021-01-07 2021-07-23 北京码牛科技有限公司 一种一标三实信息采集方法、装置及电子设备
CN112906486B (zh) * 2021-01-26 2023-09-12 吉利汽车研究院（宁波）有限公司 无人驾驶出租车的乘客状况检测方法、控制方法及系统
CN113190372B (zh) * 2021-05-24 2023-02-07 深圳赛安特技术服务有限公司 多源数据的故障处理方法、装置、电子设备及存储介质
CN113432644A (zh) * 2021-06-16 2021-09-24 苏州艾美睿智能系统有限公司 一种无人搬运车异常检测系统及检测方法
CN113313087B (zh) * 2021-07-28 2021-11-02 江西科技学院 无人驾驶汽车的乘客行为监督方法与装置

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2013000300A (ja) * 2011-06-15 2013-01-07 Nissan Motor Co Ltd 気分判定装置及び気分判定方法
CN107742106A (zh) * 2017-10-18 2018-02-27 百度在线网络技术（北京）有限公司 基于无人驾驶车辆的面部匹配方法和装置
CN108090446A (zh) * 2017-12-18 2018-05-29 大陆汽车投资（上海）有限公司 基于人脸识别的车辆智能响应方法
DE102017201328A1 (de) * 2017-01-27 2018-08-02 Bayerische Motoren Werke Aktiengesellschaft Verfahren zur Bestimmung eines Gesichtsausdrucks
CN108369645A (zh) * 2018-02-08 2018-08-03 深圳前海达闼云端智能科技有限公司 一种出租车运营监测方法、设备、存储介质和系统
DE102018004479A1 (de) * 2018-06-06 2018-11-08 Daimler Ag Verfahren zum Betrieb eines Assistenzsystems eines Fahrzeuges
CN109902575A (zh) * 2019-01-24 2019-06-18 平安科技（深圳）有限公司 基于无人驾驶车辆的防拐卖方法、装置及相关设备

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102831442A (zh) * 2011-06-13 2012-12-19 索尼公司 异常行为检测设备和方法及生成该检测设备的设备和方法
CN103246869B (zh) * 2013-04-19 2016-07-06 福建亿榕信息技术有限公司 基于人脸识别和行为语音识别的犯罪监控方法
WO2017173640A1 (zh) * 2016-04-08 2017-10-12 厦门中控智慧信息技术有限公司 一种基于多模式生物识别信息的个人识别方法和装置
CN106355829A (zh) * 2016-08-29 2017-01-25 佛山市海科知识产权交易有限公司 一种安全监测方法和装置
CN107730873A (zh) * 2017-09-14 2018-02-23 王淑芳 一种重点营运车辆异常行为监控方法和系统

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2013000300A (ja) * 2011-06-15 2013-01-07 Nissan Motor Co Ltd 気分判定装置及び気分判定方法
DE102017201328A1 (de) * 2017-01-27 2018-08-02 Bayerische Motoren Werke Aktiengesellschaft Verfahren zur Bestimmung eines Gesichtsausdrucks
CN107742106A (zh) * 2017-10-18 2018-02-27 百度在线网络技术（北京）有限公司 基于无人驾驶车辆的面部匹配方法和装置
CN108090446A (zh) * 2017-12-18 2018-05-29 大陆汽车投资（上海）有限公司 基于人脸识别的车辆智能响应方法
CN108369645A (zh) * 2018-02-08 2018-08-03 深圳前海达闼云端智能科技有限公司 一种出租车运营监测方法、设备、存储介质和系统
DE102018004479A1 (de) * 2018-06-06 2018-11-08 Daimler Ag Verfahren zum Betrieb eines Assistenzsystems eines Fahrzeuges
CN109902575A (zh) * 2019-01-24 2019-06-18 平安科技（深圳）有限公司 基于无人驾驶车辆的防拐卖方法、装置及相关设备

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113507593A (zh) * 2021-06-24 2021-10-15 中汽创智科技有限公司 一种用于车辆座舱的监测方法、装置、系统及终端
CN114360230A (zh) * 2021-09-24 2022-04-15 深圳市卡联科技股份有限公司 一种公交车安全预警方法、系统、装置及存储介质
CN114360230B (zh) * 2021-09-24 2024-02-02 深圳市卡联科技股份有限公司 一种公交车安全预警方法、系统、装置及存储介质
CN114550076A (zh) * 2021-10-19 2022-05-27 深圳市中博科创信息技术有限公司 区域异常行为监控方法、装置、设备及存储介质
CN114475623A (zh) * 2021-12-28 2022-05-13 阿波罗智联(北京)科技有限公司 车辆的控制方法、装置、电子设备及存储介质
CN114692757A (zh) * 2022-03-30 2022-07-01 北京汇通天下物联科技有限公司 一种异常驾驶行为的监测方法、装置及电子设备
CN114677745A (zh) * 2022-04-13 2022-06-28 安徽华昇能源互联网研究院有限公司 一种智能电缆井安全管理系统
CN114677745B (zh) * 2022-04-13 2024-05-07 安徽华昇能源互联网研究院有限公司 一种智能电缆井安全管理系统

Also Published As

Publication number Publication date
CN109902575B (zh) 2024-03-15
CN109902575A (zh) 2019-06-18

Similar Documents

Publication Publication Date Title
WO2020151339A1 (zh) 基于无人驾驶车辆的异常处理方法、装置及相关设备
US20200349666A1 (en) Enhanced vehicle sharing system
US10373280B2 (en) Systems and methods for location fencing within a controlled environment
US20210024080A1 (en) Method for data analysis, electronic device, and computer readable medium
US11551117B1 (en) Policy based artificial intelligence engine
US9843611B2 (en) Incident data collection for public protection agencies
US10896598B1 (en) Ambient situational abnormality detection and response
CN109859437A (zh) 一种报警方法及设备、系统
CN111125382A (zh) 人员轨迹实时监测方法及终端设备
US11089112B2 (en) Low latency interactive media distribution using shared user hardware
US10943117B2 (en) Translation to braille
KR102157074B1 (ko) 안면 인식 기술이 적용된 차량 내 방치어린이 예방 안전시스템
CN105512478A (zh) 基于云计算技术的老年人险情报警方法
Kim et al. Development of unmanned remote smart rescue platform applying Internet of Things technology
US20240219583A1 (en) Location determination using a plurality of geo-location techniques
US20230230190A1 (en) Personal protector platform
US11463841B1 (en) System and method for monitoring and ensuring student safety
Saad et al. Monitoring and complaining web-based face recognition using Haar-MATLAB
CN106340183B (zh) 一种基于gnss下的智能交通通信系统
US10800289B1 (en) Addressing risk associated with a vehicular seat component
Sharma et al. Design and Implementation of IoT Based Vehicle Emergency Relieve System
US12483634B2 (en) Cloud-based, geospatially-enabled data recording, notification, and rendering system and method
ES2941893T3 (es) Activación de cámara de respuesta basada en despacho
JP2026028084A (ja) システム
US20250246063A1 (en) System and method for monitoring and ensuring student safety

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19911724

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS (EPO FORM 1205A DATED 14.09.2021)

122 Ep: pct application non-entry in european phase

Ref document number: 19911724

Country of ref document: EP

Kind code of ref document: A1