WO2019095887A1 - 通用的车内乘客防遗忘的传感装置的实现方法和*** - Google Patents

通用的车内乘客防遗忘的传感装置的实现方法和*** Download PDF

Info

Publication number
WO2019095887A1
WO2019095887A1 PCT/CN2018/109440 CN2018109440W WO2019095887A1 WO 2019095887 A1 WO2019095887 A1 WO 2019095887A1 CN 2018109440 W CN2018109440 W CN 2018109440W WO 2019095887 A1 WO2019095887 A1 WO 2019095887A1
Authority
WO
WIPO (PCT)
Prior art keywords
event
vehicle
passenger
recognition model
candidate
Prior art date
Application number
PCT/CN2018/109440
Other languages
English (en)
French (fr)
Inventor
孙宝石
Original Assignee
苏州数言信息技术有限公司
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 苏州数言信息技术有限公司 filed Critical 苏州数言信息技术有限公司
Publication of WO2019095887A1 publication Critical patent/WO2019095887A1/zh

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
    • B60R21/00—Arrangements or fittings on vehicles for protecting or preventing injuries to occupants or pedestrians in case of accidents or other traffic risks
    • B60R21/01—Electrical circuits for triggering passive safety arrangements, e.g. airbags, safety belt tighteners, in case of vehicle accidents or impending vehicle accidents
    • B60R21/015—Electrical circuits for triggering passive safety arrangements, e.g. airbags, safety belt tighteners, in case of vehicle accidents or impending vehicle accidents including means for detecting the presence or position of passengers, passenger seats or child seats, and the related safety parameters therefor, e.g. speed or timing of airbag inflation in relation to occupant position or seat belt use
    • B60R21/01512—Passenger detection systems
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
    • B60R21/00—Arrangements or fittings on vehicles for protecting or preventing injuries to occupants or pedestrians in case of accidents or other traffic risks
    • B60R21/01—Electrical circuits for triggering passive safety arrangements, e.g. airbags, safety belt tighteners, in case of vehicle accidents or impending vehicle accidents
    • B60R21/015—Electrical circuits for triggering passive safety arrangements, e.g. airbags, safety belt tighteners, in case of vehicle accidents or impending vehicle accidents including means for detecting the presence or position of passengers, passenger seats or child seats, and the related safety parameters therefor, e.g. speed or timing of airbag inflation in relation to occupant position or seat belt use
    • B60R21/01512—Passenger detection systems
    • B60R21/0153—Passenger detection systems using field detection presence sensors
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
    • B60R21/00—Arrangements or fittings on vehicles for protecting or preventing injuries to occupants or pedestrians in case of accidents or other traffic risks
    • B60R21/01—Electrical circuits for triggering passive safety arrangements, e.g. airbags, safety belt tighteners, in case of vehicle accidents or impending vehicle accidents
    • B60R21/015—Electrical circuits for triggering passive safety arrangements, e.g. airbags, safety belt tighteners, in case of vehicle accidents or impending vehicle accidents including means for detecting the presence or position of passengers, passenger seats or child seats, and the related safety parameters therefor, e.g. speed or timing of airbag inflation in relation to occupant position or seat belt use
    • B60R21/01512—Passenger detection systems
    • B60R21/01542—Passenger detection systems detecting passenger motion
    • G—PHYSICS
    • G08—SIGNALLING
    • G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
    • G08B21/02—Alarms for ensuring the safety of persons

Definitions

  • the present invention relates to a method and system for implementing a universal sensing device for in-vehicle passenger anti-forgetting.
  • the installation is cumbersome: the installation of sensors on the seat, by (the type of seat, the material varies widely, the installation of pressure sensors involves a lot of engineering work. For example, a school bus will be to modify dozens of seats.
  • High cost the system itself has high cost (more module components), high cost of vehicle modification and seat modification, and high maintenance cost.
  • an object of the present invention is to provide a sensor data acquired by a non-contact sensor, calculate various events occurring in the vehicle, and more accurately determine the car according to a logical relationship between various events.
  • the method for realizing the universal anti-forgotten sensing device for an in-vehicle passenger of the present invention includes:
  • the vehicle The related events include at least: vehicle engine starting, vehicle driving, vehicle waiting, starting and extinguishing, opening the door, closing the door;
  • passenger-related events include at least: human voice, human body movement, passengers getting on the bus, passengers getting off;
  • the event includes at least: at least one of a high or a low temperature, humidity, air pressure, temperature, humidity, and air pressure in the vehicle;
  • the event classifier is generated by supervising machine learning, and the process of identifying the event type by the event classifier includes:
  • the obtained sensor data is characterized to obtain characterization data, and the characterization data includes: performing fast Fourier transform, sliding time window, calculating maximum value, minimum value, median, and average value , sample number, standard variance 6 characteristic values;
  • the characterization data is sequentially substituted into m reserved candidate event recognition models for calculation, and the identified event types and credibility are output.
  • the screening method of the m highly correlated candidate event recognition model specifically includes:
  • the candidate event recognition model satisfying the activation condition is activated. If the candidate event recognition model satisfying the predetermined activation condition is greater than m, the m models with the smallest deviation score are selected as the screening result; the activation conditions of each candidate event recognition model are: (F x -F 0 )>2* ⁇ 0 , (F i -F x )/ ⁇ x is the deviation score of the candidate event recognition model;
  • calculating the average value of the sensor data acquired during the predetermined time period and the eigenvalue vector of the standard deviation are respectively recorded as F 0 and ⁇ 0 ; respectively, the sensing of the candidate event recognition model data during the training process
  • the eigenvalue vectors of the mean and standard deviation of the data are denoted as F x and ⁇ x , respectively; x is the number of the candidate event recognition model.
  • the predetermined time period includes a plurality of different event calculation cycles
  • the duration of the next event calculation cycle is determined based on the probability of a child forgetting event occurring within the next event calculation cycle.
  • the method further includes: acquiring vehicle information, loading the event classifier based on the vehicle information, and if the corresponding event classifier is stored locally, using a locally stored event classifier; if the corresponding event classifier is not stored locally And loading, by the wireless network, an event classifier of the remote service terminal, wherein the vehicle information includes one or a combination of a vehicle type, a vehicle brand, a number of passengers on the vehicle, and a model of the vehicle.
  • the method further includes determining the child The emergency priority of the forgotten event, different alarm modes and alarm frequencies are set according to the emergency priority.
  • the first level is urgent
  • the alarm information is sent to the bound private electronic terminal at the first frequency
  • the predetermined public receiving platform sends the alarm information.
  • Second-level emergency sending alarm information to the bound private electronic terminal with secondary frequency
  • third-level emergency sending alarm information to the bound private electronic terminal with three-level frequency
  • ...K-level emergency with K-level frequency
  • the bound private electronic terminal sends out the hug information, the emergency priority of the child forgetting event, level 1 > level 2 > level 3 > > level K, level 1 frequency > level 2 frequency > level 3 frequency > level K frequency.
  • the implementation system of the universal anti-forgotten sensing device for an in-vehicle passenger of the present invention comprises:
  • the sensors including an acceleration sensor, a gyroscope, a microphone, an electromagnetic induction, a temperature/humidity, a gas pressure, a Doppler radar, and/or a passive infrared PIR sensor;
  • the event classifier is configured to perform an analysis operation on the sensor data acquired by the plurality of non-contact sensor groups, and obtain a vehicle-related event, a passenger-related event, an in-vehicle environment-related event, and a user-defined event that occur within a predetermined time period;
  • Vehicle-related events include at least: vehicle engine start, vehicle travel, vehicle waiting, starting and extinguishing, opening the door, closing the door;
  • passenger-related events include at least: human voice, body movement, passengers getting on, passengers getting off; interior environment
  • the related events include at least: at least one of the temperature, humidity, normal air pressure, interior temperature, humidity, and air pressure of the vehicle is high or low;
  • the forgetting event determining unit is configured to determine whether the logical relationship between the vehicle related event, the passenger related event, the in-vehicle environment related event, and/or the user customized event satisfies a predetermined logical relationship of the child forgetting event on the vehicle,
  • the alarm unit sends an alarm message; if it is not satisfied, it will not process it.
  • the event classifier includes:
  • a characterization processing module configured to perform characterization processing on the acquired sensing data to obtain characterization data
  • a recognition model screening module configured to filter the selected event recognition model of the event classifier based on the sensing data, filter out a candidate event recognition model with low correlation, and retain m candidate events with high correlation model;
  • a result output module configured to substitute the characterization data into m reserved candidate event recognition models for calculation, and output the identified event type and credibility
  • Specific screening methods for identifying model screening modules include:
  • the candidate event recognition model satisfying the activation condition is activated. If the candidate event recognition model satisfying the predetermined activation condition is greater than m, the m models with the smallest deviation score are selected as the screening result; the activation conditions of each candidate event recognition model are: (F x -F 0 )>2* ⁇ 0 , (F i -F x )/ ⁇ x is the deviation score of the candidate event recognition model;
  • calculating the average value of the sensor data acquired during the predetermined time period and the eigenvalue vector of the standard deviation are respectively recorded as F 0 and ⁇ 0 ; respectively, the sensing of the candidate event recognition model data during the training process
  • the eigenvalue vectors of the mean and standard deviation of the data are denoted as F x and ⁇ x , respectively; x is the number of the candidate event recognition model.
  • the method further includes: an event calculation cycle duration determining module, configured to predict a next event based on a logical relationship between the vehicle related event and/or the in-vehicle environment related event occurring in the previous m event operation cycles Calculating a probability of a child forgetting event occurring in a next event calculation period based on the vehicle related event and/or the in-vehicle environment related event described in the calculation cycle;
  • the correspondence table between the vehicle-related event and/or the in-vehicle environment-related event and the occurrence probability of the child forgetting in-vehicle event is obtained by machine learning the historical data.
  • an event classifier loading module is further configured to acquire vehicle information, and load the event classifier based on the vehicle information. If a corresponding event classifier is stored locally, the locally stored event classifier is used; If the corresponding event classifier is not stored, the event classifier of the remote service terminal is loaded through the wireless network, wherein the vehicle information includes one or a combination of a vehicle type, a vehicle brand, a number of passengers on the vehicle, and a model of the vehicle. ;
  • the alarm unit includes an emergency priority determination module for determining an emergency priority of the child forgetting event, and setting different alarm modes and alarm frequencies according to the emergency priority, wherein the first level emergency is tied to the first frequency
  • the fixed private electronic terminal sends out the alarm information and the predetermined public receiving platform sends out the alarm information; the second-level emergency sends the alarm information to the bound private electronic terminal with the secondary frequency; the third-level emergency, the third-level frequency to the bound private
  • the electronic terminal issues an alarm message; ... K-level emergency, sending a tight message to the bound private electronic terminal at the K-level frequency, the emergency priority of the child forgetting event, level 1 > level 2 > level 3 > level K, Primary frequency > secondary frequency > tertiary frequency > K frequency.
  • the method and system for realizing the in-vehicle passenger anti-forgotten sensing device of the present invention have at least the following advantages:
  • the anti-forgetting integrated sensing device of the invention is composed of a 32-bit single chip microcomputer, a communication module, various physical sensors and a power supply module.
  • the communication module can be used for sending event notifications and alarm messages by using Bluetooth, wireless (Wi-Fi), 3G/4G, NB-IoT, eMTC, and the like.
  • Physical sensors include accelerometers, gyroscopes, microphones, electromagnetic induction, temperature and humidity, air pressure, and one or two sets of Doppler radar and passive infrared (PIR) sensors.
  • the power supply module is composed of a power management module, a battery, and a battery power detecting module, and supports power supply of the USB interface and power supply of the cigarette lighter and the power port (12V or 24V);
  • a set of event classifiers is integrated on the sensing device, corresponding to each in-vehicle event that needs to be identified, including: starting, starting, vehicle driving, waiting for the vehicle, starting and extinguishing, opening the door, closing the door, moving the human body, speaking, getting on and off Etc; the event classifier is generated by supervising machine learning;
  • the sensor raw data is characterized, including: fast Fourier transform, sliding time window to calculate six eigenvalues (maximum value, minimum value, median, average value, sample number, standard deviation);
  • the recognition speed is improved by the pre-screening algorithm of the classifier
  • Event notifications, raw data, and feature values are output as configurable items
  • the invention has wide application range, convenient installation, low cost and high reliability. Not only can accurate passenger forgetting detection be achieved, but a wide variety of in-vehicle events can be detected to support more applications.
  • FIG. 1 is a flow chart showing a method for realizing a general-purpose in-vehicle passenger anti-forgotten sensing device
  • FIG. 2 is a block diagram showing an implementation system of a general-purpose in-vehicle passenger anti-forgotten sensing device
  • Figure 3 is an anti-forgotten discrimination processing logic
  • Figure 4 is an event recognition model training process
  • Figure 5 is a production environment event identification process.
  • the method and system for realizing the in-vehicle passenger anti-forgotten sensing device of the present invention are based on the sensing data collected by a plurality of non-contact sensors, and the event classifier is obtained through data training, and the predetermined time period is obtained based on the event classifier.
  • the events occurring inside the vehicle determine the passenger forgetting event through the logical relationship between these events.
  • the invention utilizes multiple sets of non-contact sensors to acquire multiple sets of different sensor data, and performs comprehensive operations on multiple sensing data to obtain an event, and logical analysis of multiple events to obtain a final result.
  • a method for implementing an anti-forgotten sensing device for a passenger in a vehicle includes:
  • the vehicle The related events include at least: vehicle engine starting, vehicle driving, vehicle waiting, starting and extinguishing, opening the door, closing the door;
  • passenger-related events include at least: human voice, human body movement, passengers getting on the bus, passengers getting off;
  • the event includes at least: at least one of a high or a low temperature, humidity, air pressure, temperature, humidity, and air pressure in the vehicle;
  • the event classifier is generated by supervising machine learning, and the process of identifying the event type by the event classifier includes:
  • the characterization data is sequentially substituted into m reserved candidate event recognition models for calculation, and the identified event types and credibility are output.
  • the screening method of the m highly correlated candidate event recognition model specifically includes:
  • the candidate event recognition model satisfying the activation condition is activated. If the candidate event recognition model satisfying the predetermined activation condition is greater than m, the m models with the smallest deviation score are selected as the screening result; wherein the activation conditions of each candidate event recognition model Is: (F x -F 0 )>2* ⁇ 0 , (F i -F x )/ ⁇ x is the deviation score of the candidate event recognition model;
  • this background environment calculates the average value of the sensor data acquired in the predetermined time period and the eigenvalue vector of the standard deviation, respectively, F 0 and ⁇ 0 ; each candidate event recognition model data training process, the average value of the acquired sensor data and the eigenvalue vector of the standard deviation are respectively recorded as F x and ⁇ x ; x is the number of the candidate event recognition model.
  • the sensors are all non-contact sensors.
  • the acquisition of each event type is based on the results of the combined operation of the sensor data acquired by multiple sensors, rather than the result of a single sensor data.
  • the final conclusion of the passenger forgetting event is obtained by a logical relationship between multiple events, rather than pure sensory data.
  • the microphone detects the sound of someone in the car
  • the passive infrared (PIR) sensor detects that there is a human body in the car
  • the electromagnetic induction detects the change of the electromagnetic inside the car
  • the temperature sensor detects the temperature inside the car (the inside of the car and the inside of the car) There is no difference in temperature when there is no one.)
  • the event in the car is based on the data integration of the above sensors. After the three events of stopping the vehicle, some people in the car, and no one in the driver's seat, based on the logical relationship between the three events, it can be determined that the passenger forgetting event has occurred. For another example, if the vehicle is turned off for more than a preset time event, a door closing event, a sound event of a person in the vehicle, or a movement event in the vehicle, it can be determined that a passenger forgetting event has occurred. For another example, if the vehicle is in a normal range of speed driving state and there is an event in the vehicle, it can be judged that no passenger forgetting event has occurred.
  • a family car is taken as an example, and an anti-forgetting integrated sensing device is disposed at the front panel of the family car, wherein the physical sensor includes an acceleration sensor, a gyroscope, a microphone, electromagnetic induction, temperature and humidity, air pressure, and a group of Doppler. Radar and passive infrared (PIR) sensors.
  • the sensing device draws power through the USB port or the cigarette lighter.
  • the sensing device detects the door opening event and the boarding event; when the vehicle starts, the sensing device detects the engine starting event and sets the vehicle state to start; the vehicle starts to drive, and the sensing device sets the vehicle state For driving; the vehicle stops, the sensing device sets the vehicle state to stop; the driver turns off the fire, the sensing device sets the vehicle state to the engine stop state; the driver opens the door, gets off the vehicle, closes the door, and the sensing device detects the opening event and gets off the vehicle. Events, closing events. At this time, there are still passengers in the car. After 1 minute (assuming the anti-forgetting alarm delay is set to 1 minute), if the passenger makes a sound or moves the body, the sensing device will detect the human activity event, and an alarm message will be sent through the communication module.
  • the method for implementing the in-vehicle passenger anti-forgotten sensing device is obtained.
  • the vehicle information is acquired, and the event classifier is loaded based on the vehicle information, and the corresponding event classification is stored locally.
  • the event classifier is stored locally; if the corresponding event classifier is not stored locally, the event classifier of the remote service terminal is loaded through the wireless network, wherein the vehicle information includes a model, a vehicle brand, and a vehicle passenger A combination of one or several of the number of people and the model of the vehicle.
  • the predetermined time period includes a plurality of different event calculation periods, and the correspondence table between the vehicle related event and/or the vehicle environment related event and the occurrence probability of the child forgetting vehicle event is obtained based on the historical data training; A logical relationship between the vehicle-related event and/or the in-vehicle environment-related event occurring in an event computing cycle, predicting a vehicle-related event and/or an in-vehicle environment-related event within the next event computing cycle, based on The correspondence table obtains the probability of occurrence of the child forgetting event in the next event computing period; and determines the duration of the next event computing period based on the probability of the child forgetting event occurring in the next event computing period.
  • the probability of the passenger forgetting the event in the car is not large, and the time of the next event calculation cycle can be relatively long, that is, the time interval for acquiring the sensor data can be longer, which can reduce the data transmission frequency and Reduce the amount of calculations and save power.
  • a vehicle deceleration event (vehicle-related event) occurs, and the interior temperature is higher than a set threshold (in-vehicle environment-related event).
  • the next cycle is predicted to stop at this time. The possibility of coming down is relatively large.
  • the probability of the passenger forgetting event is higher than that of the vehicle.
  • the event of the next event calculation cycle is relatively short, so that it is convenient to find the passenger forgetting event in time.
  • the interior environment is also considered as a consideration. When the interior environment threatens human health, this is defined as a high probability of passenger forgetting events.
  • the generation of a specific probability correspondence table is generated by specific machine learning.
  • the method for realizing the in-vehicle passenger anti-forgotten sensing device in the embodiment is determined on the basis of the embodiment 1 or 2, and the vehicle-related event, the passenger-related event, the in-vehicle environment-related event, and/or the user is determined.
  • Defining the logical relationship of the event satisfies the predetermined logical relationship of the child forgetting the event on the vehicle and further includes determining the emergency priority of the child forgetting event, and setting different alarm modes and alarm frequencies according to the emergency priority, wherein Urgently, the alarm message is sent to the bound private electronic terminal at the first-level frequency and the alarm message is sent by the predetermined public receiving platform; the second-level emergency sends the alarm information to the bound private electronic terminal at the secondary frequency; The third-level frequency sends an alarm message to the bound private electronic terminal; ...
  • K-level emergency sending a tight message to the bound private electronic terminal with a K-level frequency, the emergency priority of the child forgetting event, level 1 > level 2 > Level 3...>K level, first order frequency>2nd frequency>3rd order frequency>K stage frequency.
  • the first-level frequency is bound to the private electronic terminal (mobile phone, notebook computer, PAD, smart).
  • the watch sends an alarm message and sends an alarm message to the public receiving platform (the server of the background operator and/or 110).
  • What is included in the alarm information can be set according to the program.
  • the vehicle has stopped, detecting that the driver's getting off the vehicle exceeds the predetermined time, and the passenger's voice and/or passenger movement in the rear seat of the vehicle may be defined as a secondary emergency, which is sent to the bound private electronic terminal.
  • Alarm information Specifically, the setting of the emergency situation and the setting of the alarm mode can be specifically trained according to the actual situation to generate a corresponding training model.
  • the specific binding quantity of the private electronic terminal is set by the user.
  • the output of the alarm signal also includes an audible and visual alarm signal.
  • the event classifier is configured to perform an analysis operation on the sensor data acquired by the plurality of non-contact sensor groups, and obtain a vehicle-related event, a passenger-related event, an in-vehicle environment-related event, and a user-defined event that occur within a predetermined time period;
  • Vehicle-related events include at least: vehicle engine start, vehicle travel, vehicle waiting, starting and extinguishing, opening the door, closing the door;
  • passenger-related events include at least: human voice, body movement, passengers getting on, passengers getting off; interior environment
  • the related events include at least: at least one of the temperature, humidity, normal air pressure, interior temperature, humidity, and air pressure of the vehicle is high or low;
  • the forgetting event determining unit is configured to determine whether the logical relationship between the vehicle related event, the passenger related event, the in-vehicle environment related event, and/or the user customized event satisfies a predetermined logical relationship of the child forgetting event on the vehicle,
  • the alarm unit sends an alarm message; if it is not satisfied, it will not process it.
  • the event classifier includes:
  • a characterization processing module configured to perform characterization processing on the acquired sensing data to obtain characterization data
  • a recognition model screening module configured to filter the selected event recognition model of the event classifier based on the sensing data, filter out a candidate event recognition model with low correlation, and retain m candidate events with high correlation model;
  • the result output module is configured to sequentially substitute the characterization data into the m reserved candidate event recognition models, and output the identified event type and credibility.
  • the screening method of the identification model screening module specifically includes:
  • the candidate event recognition model satisfying the activation condition is activated. If the candidate event recognition model satisfying the predetermined activation condition is greater than m, the m models with the smallest deviation score are selected as the screening result; wherein the activation conditions of each candidate event recognition model Is: (F x -F 0 )>2* ⁇ 0 , (F i -F x )/ ⁇ x is the deviation score of the candidate event recognition model;
  • the eigenvalue vectors of the average value and the standard deviation of the sensor data acquired in the predetermined time period are calculated as F 0 and ⁇ 0 respectively; in the training process of each candidate event recognition model data, the acquired sensor data
  • the eigenvalue vectors of the mean and standard deviation are denoted as F x and ⁇ x , respectively; x is the number of the candidate event recognition model.
  • a school bus is taken as an example, and an anti-forgetting integrated sensing device is deployed in the middle of the school bus.
  • the physical sensors include an acceleration sensor, a gyroscope, a microphone, an electromagnetic induction, a temperature and humidity, a gas pressure, and two sets of Doppler radar and passive infrared. (PIR) sensor.
  • the sensing device is powered by a 12V lighting circuit inside the vehicle.
  • the sensing device detects the door closing event, the door opening event, the getting off event and the closing event. But one student fell asleep in the car and was forgotten in the car. After 3 minutes (assuming the anti-forgetting alarm delay is set to 3 minutes), if the student makes a sound or moves the body, the sensing device detects the human activity event and sends an alarm message through the communication module.
  • the implementation system of the in-vehicle passenger anti-forgotten sensing device of the present embodiment further includes, on the basis of Embodiment 4, an event calculation cycle duration determining module, which is configured to be based on the occurrence of the previous m event calculation cycles.
  • an event calculation cycle duration determining module which is configured to be based on the occurrence of the previous m event calculation cycles.
  • the correspondence table between the vehicle-related event and/or the in-vehicle environment-related event and the occurrence probability of the child forgetting in-vehicle event is obtained by machine learning the historical data.
  • the implementation system of the in-vehicle passenger anti-forgotten sensing device of the present embodiment further includes an emergency priority determining module for determining the vehicle-related event and the passenger-related event on the basis of Embodiment 4 or 5.
  • the logical relationship between the environment related events in the vehicle and/or the user-defined event satisfies the predetermined logical relationship of the child forgetting event on the vehicle, determines the emergency priority of the child forgetting event, and sets different alarms according to the emergency priority.
  • Mode and alarm frequency wherein, the first level is urgent, the sound and light alarm is issued; the second level emergency sends an alarm message to the predetermined receiving terminal through the wireless network; the third level is urgent, and the alarm message is sent to the predetermined receiving terminal, and the predetermined public receiving is sent.
  • the platform sends an alarm message.
  • the method further includes: acquiring vehicle information, loading the event classifier based on the vehicle information, and if the corresponding event classifier is stored locally, using a locally stored event classifier; if the local does not store the corresponding The event classifier loads the event classifier of the remote service terminal through the wireless network, wherein the vehicle information includes one or a combination of a vehicle type, a vehicle brand, a number of passengers on the vehicle, and a model of the vehicle.
  • the senor includes an acceleration sensor, a gyroscope, a microphone, an electromagnetic induction, a temperature and humidity, a gas pressure, a Doppler radar, and a passive infrared (PIR) sensor.
  • the sensor data characterization is to extract the collected sensor raw data, which not only retains the key information, but also plays the role of data compression and confidentiality. Furthermore, for sensors with a sampling frequency lower than 1 kHz, such as Doppler radar, passive infrared, temperature and humidity, air pressure, etc., six characteristic values are calculated with a sliding time window of 1 to 2 seconds (maximum, minimum). Value, median, average, sample number, standard deviation); for sensors with sampling frequency higher than 1 kHz, such as: acceleration sensor, gyroscope, microphone, electromagnetic induction, etc. Leaf transformation (FFT, 256 sample sliding window), then calculate 6 eigenvalues.
  • FFT Fast Fourier transform
  • the characterization processed data packet (X matrix) is transmitted to the background machine learning tool, and the event recognition model is trained together with the event type tag (Y vector).
  • Machine learning tools are only used during the model training phase, support conventional machine learning algorithms (such as: SVM support vector machine, random forest, neural network), can be offline tools, or can be used online through automated scripts.
  • the m models with the smallest deviation score are selected, and the specific value of m may be 5 or may be specifically set according to a specific situation, and the specific data and the calculation amount that the device can support are specifically set.
  • the result of “passengers being forgotten in the car” is reflected in a series of related events. Therefore, by logically detecting the event for a period of time, it is possible to accurately recognize that a passenger is forgotten in the vehicle and issue an alarm message through the communication module.
  • the anti-forgotten related discriminating and processing logic is shown in Figure 3.
  • the vehicle status includes: starting from start, vehicle driving, vehicle waiting, starting and extinguishing, etc.; environmental events in the vehicle include: normal, over temperature, low temperature, humidity Excessively high, low humidity, high air pressure, low air pressure, etc.; other events include opening the door, closing the door, getting on the bus, getting off the bus, etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Mechanical Engineering (AREA)
  • Business, Economics & Management (AREA)
  • Emergency Management (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Emergency Alarm Devices (AREA)
  • Alarm Systems (AREA)

Abstract

一种通用的车内乘客防遗忘的传感装置的实现方法和***，为了提高儿童车内防遗忘识别准确率而设计。该通用的车内乘客防遗忘的传感装置的实现方法和***，通过判断车辆状态事件是否属于预定报警事件，若是，则判断当前车内环境是否满足人体移动、有人发出声音和/或当前车内环境数据超出预定阈值范围，若满足，则判定有人遗忘在车上，发出告警信息；否则不做处理。该***适用于所有类型、品牌的车辆，而且不需要对车辆进行任何改造，真正做到即插即用，成本也大大降低。

Description

通用的车内乘客防遗忘的传感装置的实现方法和*** 技术领域
本发明涉及一种通用的车内乘客防遗忘的传感装置的实现方法和***。
背景技术
每年都有多起儿童被遗忘在车内的悲剧发生，为了杜绝此类事件，各式各样的车辆防遗忘方案应运而生。目前的解决方案原理相似，这些方案通常需要在座椅上加装传感器(如：压力、温度等)，同时需要与车辆的电路或车载电脑(OBD)***连接。通过检测乘客座位是否有人、驾驶员座位是否有人，以及车辆发动起是否熄火等状态，作为是否有乘客被遗忘的判别标准。另外一类防遗忘***，则完全依靠人工巡检，即驾驶员停车熄火后必须及时走到车辆后端复位告警按钮，否则告警就会超时触发。
现有的儿童防遗忘方法及***具有如下缺陷：
1、不具备通用性：不同品牌、型号的汽车，其车载电子***差别很大，现有方案几乎都涉及到车辆电子***的对接，因此，不具有通用性。
2、安装繁琐：在座椅上加装传感器，由(座椅种类、材质千差万别压力传感器的安装都涉及大量的工程工作。例如，一辆校车就要改造几十个座椅。
3、成本高：***本身成本高(模块组件多)、车辆改造和座椅改造成本高，以及维护成本高。
4、可靠性差：通过座椅或者安全带检测，无法识别儿童离开座椅但仍留在车内、以及乘客不习惯系安全带等场合。
5、乘客遗忘的结论的得出，是基于一种传感器数据直接得到的，没有对传感数据进行深度的挖掘和开发。
鉴于上述的缺陷，本设计人积极加以研究创新，以期创设一种通用的车内乘客防遗忘的传感装置的实现方法和***，使其更具有产业上的利用价值。
发明内容
为解决上述技术问题，本发明的目的是提供一种利用非接触式传感器获取的传感数据，计算车内发生的多种事件，进而根据多种事件之间的逻辑关系，更精确地判定车内是否发生 儿童遗忘检测的通用的车内乘客防遗忘的传感装置的实现方法和***。
为达到上述发明目的，本发明通用的车内乘客防遗忘的传感装置的实现方法，包括：
获取设置在车内的多个非接触传感器组输出的各组传感数据；
利用预存储或预加载的事件分类器对所述的传感数据进行分析运算，得到预定时间周期内发生的车辆相关事件、乘客相关事件、车内环境相关事件以及用户自定义事件；其中，车辆相关事件至少包括：车辆发动机启动、车辆行驶、车辆等候、发动起熄火、打开车门、关闭车门；乘客相关事件至少包括：人发出声音、人体移动、乘客上车、乘客下车；车内环境相关事件至少包括：车内温度、湿度、气压正常、车内温度、湿度、气压中至少一项偏高或偏低；
判断所述的车辆相关事件、乘客相关事件、车内环境相关事件和/或用户自定义事件的逻辑关系是否满足预定的发生儿童遗忘在车上事件的逻辑关系，
若满足，则发出报警信息；若不满足，则不做处理。
进一步地，所述事件分类器通过监督机器学习方式生成，事件分类器对事件类型的识别过程包括：
对获取的所述传感数据进行特征化处理，得到特征化数据,所述的特征化数据包括：进行快速傅里叶变换、滑动时间窗、计算最大值、最小值、中位数、平均值、样本数、标准方差6个特征值；
基于所述传感数据，对所述事件分类器的获选事件识别模型进行筛选，过滤掉相关性低的候选事件识别模型，保留m个相关性高的候选事件识别模型；
特征化数据被依次代入m个保留的候选事件识别模型进行计算，输出识别到的事件类型及可信度。
进一步地，所述的m个相关性高的候选事件识别模型的筛选方法具体包括：
计算传感器数据的平均值特征值向量F i；
对满足激活条件的候选事件识别模型进行激活，若满足预定激活条件的候选事件识别模型大于m个，则选取偏离评分最小的m个模型，作为筛选结果；各候选事件识别模型的激活条件为：(F x-F 0)>2*σ 0，(F i-F x)/σ x为候选事件识别模型的偏离评分；
其中，背景环境下，计算预定时间周期内获取的传感数据的平均值和标准方差的特征值向量，分别记为F 0和σ 0；各候选事件识别模型数据训练过程中，获取的传感数据的平均值和标准方差的特征值向量，分别记为F x和σ x；x为候选事件识别模型的编号。
进一步地，预定时间周期包括多个不同的事件运算周期，
基于历史数据训练得到所述的车辆相关事件和/或车内环境相关事件与发生儿童遗忘车内事件概率的对应表；
基于前面m个事件运算周期中发生的所述的车辆相关事件和/或车内环境相关事件之间的逻辑关系，预测下一个事件运算周期内所述的车辆相关事件和/或车内环境相关事件，基于所述对应表得到下一个事件运算周期内发生儿童遗忘事件概率；
基于下一个事件运算周期内发生儿童遗忘事件概率，确定下一个事件运算周期的时长。
进一步地，还包括：获取车辆信息，基于车辆信息加载与所述的事件分类器，若本地存储有相应的事件分类器，则采用本地存储的事件分类器；若本地没有存储相应的事件分类器，则通过无线网络加载远程服务终端的事件分类器，其中，所述的车辆信息包括车型、车辆品牌、车辆载客人数、车辆型号中的一种或几种的组合。
进一步地，确定所述的车辆相关事件、乘客相关事件、车内环境相关事件和/或用户自定义事件的逻辑关系满足预定的发生儿童遗忘在车上事件的逻辑关系后，还包括确定该儿童遗忘事件的紧急优先级，根据紧急优先级设定不同的报警方式及报警频率，其中，一级紧急，以一级频率向绑定的私人电子终端发出报警信息以及预定的公共接收平台发出报警信息；二级紧急，以二级频率向绑定的私人电子终端发出报警信息；三级紧急，以三级频率向绑定的私人电子终端发出报警信息；……K级紧急，以K级频率向绑定的私人电子终端发出抱紧信息，儿童遗忘事件的紧急优先级，一级＞二级＞三级……＞K级，一级频率＞二级频率＞三级频率＞K级频率。
为达到上述发明目的，本发明通用的车内乘客防遗忘的传感装置的实现***，包括：
置在车内的多个非接触传感器组，所述的传感器包括加速度传感器、陀螺仪、麦克风、电磁感应、温/湿度、气压、多普勒雷达和/或被动红外PIR传感器；
事件分类器，用于对多个非接触传感器组获取的传感数据进行分析运算，得到预定时间周期内发生的车辆相关事件、乘客相关事件、车内环境相关事件以及用户自定义事件；其中，车辆相关事件至少包括：车辆发动机启动、车辆行驶、车辆等候、发动起熄火、打开车门、关闭车门；乘客相关事件至少包括：人发出声音、人体移动、乘客上车、乘客下车；车内环境相关事件至少包括：车内温度、湿度、气压正常、车内温度、湿度、气压中至少一项偏高或偏低；
遗忘事件判定单元，用于判断所述的车辆相关事件、乘客相关事件、车内环境相关事件 和/或用户自定义事件的逻辑关系是否满足预定的发生儿童遗忘在车上事件的逻辑关系，
若满足，则报警单元发出报警信息；若不满足，则不做处理。
进一步地，所述事件分类器包括：
特征化处理模块，用于对获取的所述传感数据进行特征化处理，得到特征化数据；
识别模型筛选模块，用于基于所述传感数据，对所述事件分类器的获选事件识别模型进行筛选，过滤掉相关性低的候选事件识别模型，保留m个相关性高的候选事件识别模型；
结果输出模块，用于将特征化数据被依次代入m个保留的候选事件识别模型进行计算，输出识别到的事件类型及可信度；
识别模型筛选模块的具体筛选方法包括：
计算传感器数据的平均值特征值向量F i；
对满足激活条件的候选事件识别模型进行激活，若满足预定激活条件的候选事件识别模型大于m个，则选取偏离评分最小的m个模型，作为筛选结果；各候选事件识别模型的激活条件为：(F x-F 0)>2*σ 0，(F i-F x)/σ x为候选事件识别模型的偏离评分；
其中，背景环境下，计算预定时间周期内获取的传感数据的平均值和标准方差的特征值向量，分别记为F 0和σ 0；各候选事件识别模型数据训练过程中，获取的传感数据的平均值和标准方差的特征值向量，分别记为F x和σ x；x为候选事件识别模型的编号。
进一步地，还包括：事件运算周期时长确定模块，用于基于前面m个事件运算周期中发生的所述的车辆相关事件和/或车内环境相关事件之间的逻辑关系关系，预测下一个事件运算周期内所述的车辆相关事件和/或车内环境相关事件，基于所述对应表得到下一个事件运算周期内发生儿童遗忘事件概率；
基于下一个事件运算周期内发生儿童遗忘事件概率，确定下一个事件运算周期的时长；
所述的车辆相关事件和/或车内环境相关事件与发生儿童遗忘车内事件概率的对应表通过对历史数据进行机器学习得到。
进一步地，还包括事件分类器加载模块，用于获取车辆信息，基于车辆信息加载与所述的事件分类器，若本地存储有相应的事件分类器，则采用本地存储的事件分类器；若本地没有存储相应的事件分类器，则通过无线网络加载远程服务终端的事件分类器，其中，所述的车辆信息包括车型、车辆品牌、车辆载客人数、车辆型号中的一种或几种的组合；
所述报警单元，包括紧急优先级确定模块，用于确定该儿童遗忘事件的紧急优先级，根据紧急优先级设定不同的报警方式及报警频率，其中，一级紧急，以一级频率向绑定的私人 电子终端发出报警信息以及预定的公共接收平台发出报警信息；二级紧急，以二级频率向绑定的私人电子终端发出报警信息；三级紧急，以三级频率向绑定的私人电子终端发出报警信息；……K级紧急，以K级频率向绑定的私人电子终端发出抱紧信息，儿童遗忘事件的紧急优先级，一级＞二级＞三级……＞K级，一级频率＞二级频率＞三级频率＞K级频率。
借由上述方案，本发明通用的车内乘客防遗忘的传感装置的实现方法和***至少具有以下优点：
本发明防遗忘集成传感装置由32位单片机、通信模块、各种物理传感器以及供电模块组成。通信模块可采用蓝牙、无线(Wi-Fi)、3G/4G、NB-IoT、eMTC等方式，用于发送事件通知和告警消息。物理传感器包括加速度传感器、陀螺仪、麦克风、电磁感应、温湿度、气压、以及一组或两组多普勒雷达和被动红外(PIR)传感器。供电模块由电源管理模块、电池、电池电量检测模块组成，支持USB接口供电和车载点烟口及电源口(12V或24V)供电；
传感装置上集成了一组事件分类器，分别对应每一个需要识别的车内事件，包括：发动起启动、车辆行驶、车辆等候、发动起熄火、开门、关门、人体移动、说话、上下车等；事件分类器通过监督机器学习方式生成；
对传感器原始数据进行特征化处理，包括：快速傅里叶变换，滑动时间窗计算6个特征值(最大值、最小值、中位数、平均值、样本数、标准方差)；
通过对一段时间内检测到事件的逻辑判断，识别出有乘客被遗忘在车内，并通过通讯模块发出告警消息；
通过分类器的预筛选算法提高识别速度；
事件通知、原始数据和特征值输出为可配置项；
本发明适用范围广、安装方便、成本低、可靠性高。不仅能够实现精确地乘客遗忘检测，而且能够检测到很多种车内事件，从而支持更多其他应用。
上述说明仅是本发明技术方案的概述，为了能够更清楚了解本发明的技术手段，并可依照说明书的内容予以实施，以下以本发明的较佳实施例并配合附图详细说明如后。
附图说明
图1是发明通用的车内乘客防遗忘的传感装置的实现方法的流程图；
图2是发明通用的车内乘客防遗忘的传感装置的实现***的框图；
图3是防遗忘判别处理逻辑；
图4是事件识别模型训练流程；
图5是生产环境事件识别流程。
具体实施方式
下面结合附图和实施例，对本发明的具体实施方式作进一步详细描述。以下实施例用于说明本发明，但不用来限制本发明的范围。
本发明通用的车内乘客防遗忘的传感装置的实现方法及***，以多个非接触式传感器采集的传感数据为基础，通过数据训练得到事件分类器，基于事件分类器得到预定时间周期内车内发生的事件，通过这些事件之间的逻辑关系确定是否发生了乘客遗忘事件。本发明利用多组非接触式传感器获取多组不同的传感器数据，对多做传感数据进行综合运算得到一个事件，对多个事件的逻辑分析得到最终的结果。
实施例1
如图1所示，本实施例通用的车内乘客防遗忘的传感装置的实现方法，包括：
获取设置在车内的多个非接触传感器组输出的各组传感数据；
利用预存储或预加载的事件分类器对所述的传感数据进行分析运算，得到预定时间周期内发生的车辆相关事件、乘客相关事件、车内环境相关事件以及用户自定义事件；其中，车辆相关事件至少包括：车辆发动机启动、车辆行驶、车辆等候、发动起熄火、打开车门、关闭车门；乘客相关事件至少包括：人发出声音、人体移动、乘客上车、乘客下车；车内环境相关事件至少包括：车内温度、湿度、气压正常、车内温度、湿度、气压中至少一项偏高或偏低；
判断所述的车辆相关事件、乘客相关事件、车内环境相关事件和/或用户自定义事件的逻辑关系是否满足预定的发生儿童遗忘在车上事件的逻辑关系，
若满足，则发出报警信息；若不满足，则不做处理。
本实施例中，所述事件分类器通过监督机器学习方式生成，事件分类器对事件类型的识别过程包括：
对获取的所述传感数据进行特征化处理，得到特征化数据；
基于所述传感数据，对所述事件分类器的获选事件识别模型进行筛选，过滤掉相关性低的候选事件识别模型，保留m个相关性高的候选事件识别模型；
特征化数据被依次代入m个保留的候选事件识别模型进行计算，输出识别到的事件类型及可信度。
所述的m个相关性高的候选事件识别模型的筛选方法具体包括：
计算传感器数据的平均值特征值向量F i；
对满足激活条件的候选事件识别模型进行激活，若满足预定激活条件的候选事件识别模型大于m个，则选取偏离评分最小的m个模型，作为筛选结果；其中，各候选事件识别模型的激活条件为：(F x-F 0)>2*σ 0，(F i-F x)/σ x为候选事件识别模型的偏离评分；
背景环境下(例如，没有任何事情发生的环境下，这个背景环境是用户定义的一个背景环境)，计算预定时间周期内获取的传感数据的平均值和标准方差的特征值向量，分别记为F 0和σ 0；各候选事件识别模型数据训练过程中，获取的传感数据的平均值和标准方差的特征值向量，分别记为F x和σ x；x为候选事件识别模型的编号。
本实施例中，第一，传感器均为非接触式传感器。第二，每一个事件类型的得到，是基于多个传感器获取的传感数据综合运算的结果，而不是单一的传感器数据的结果。第三，最终发生乘客遗忘事件的结论是由多个事件之间的逻辑关系运算得到的，而不是单纯的传感数据得到的。例如，麦克风检测到车内有人的声音，被动红外(PIR)传感器检测到车内有人体，电磁感应检测到车内电磁的变化，甚至温度传感器检测到车内的温度(车内有人和车内没人时温度有所差异)，这些数据都能作为车内有人这一事件的数据，车内有人这一事件是基于上述多个传感器的数据综合运算得到的。车辆停止行驶事件、车内有人事件、司机座位没有人这三个事件后，基于这三个事件之间的逻辑关系就可以判定发生了乘客遗忘事件。再例如，车辆熄火超过预设时间事件、车门关闭事件、车内有人的声音事件、车内有人移动事件，就可以判定发生了乘客遗忘事件。再例如，车辆处于正常范围内车速行驶状态，车内有人事件，此时就可以判断没有发生乘客遗忘事件。
本实施例以家用轿车为例，在家用轿车前部面板处部署一个防遗忘集成传感装置，其中物理传感器包括加速度传感器、陀螺仪、麦克风、电磁感应、温湿度、气压以及一组多普勒雷达和被动红外(PIR)传感器。传感装置通过USB口或点烟口取电。司机和乘客开门、上车，传感装置检测到开门事件和上车事件；车辆发动，传感装置检测到发动机启动事件，将车辆状态置为启动；车辆开始行驶，传感装置将车辆状态置为行驶中；车辆停止，传感装置将车辆状态置为停止；司机熄火，传感装置将车辆状态置为发动机停止状态；司机开门、下车、关门，传感装置检测到开门事件、下车事件、关门事件。此时，仍有乘客留着车内。1分钟后(假定防遗忘告警延时设置为1分钟)，如果乘客发出声音或者移动身体，传感装置就会检测到人体活动事件，则通过通信模块发出告警消息。
实施例2
本实施例通用的车内乘客防遗忘的传感装置的实现方法，在实施例1的基础上，获取车辆信息，基于车辆信息加载与所述的事件分类器，若本地存储有相应的事件分类器，则采用本地存储的事件分类器；若本地没有存储相应的事件分类器，则通过无线网络加载远程服务终端的事件分类器，其中，所述的车辆信息包括车型、车辆品牌、车辆载客人数、车辆型号中的一种或几种的组合。本实施例中预定时间周期包括多个不同的事件运算周期，基于历史数据训练得到所述的车辆相关事件和/或车内环境相关事件与发生儿童遗忘车内事件概率的对应表；基于前面m个事件运算周期中发生的所述的车辆相关事件和/或车内环境相关事件之间的逻辑关系，预测下一个事件运算周期内所述的车辆相关事件和/或车内环境相关事件，基于所述对应表得到下一个事件运算周期内发生儿童遗忘事件概率；基于下一个事件运算周期内发生儿童遗忘事件概率，确定下一个事件运算周期的时长。
例如，前m事件运算周期内，都是检测到车子在行驶状态，且没有持续减速，则可以预测出车辆时一直在行驶状态，此时预测下一周期车子在行驶状态的可能性比较大，因此车子在行驶状态下，发生乘客遗忘车内的事件的概率不大，下一个事件运算周期的时间可以相对长一些，也即获取传感器数据的时间间隔可以长一些，这样可以减少数据传输频率以及减少运算量，也省电。再例如，前m事件运算周期内，发生车辆减速事件(车辆相关事件)、车内温度高于设定阈值(车内环境相关事件)，在这种状态下，此时预测下一周期车子停下来的可能性比较大，车辆一旦停止，发生乘客遗忘事件的概率就比车辆行驶状态要高，此时，下一事件运算周期的事件要相对短一些，方便及时发现车内乘客遗忘事件。同时车内环境也是作为一个考虑项，当车内环境威胁到人体健康时，此时定义为乘客遗忘事件发生高概率情况。具体的概率对应表的生成，通过具体的机器学习生成。
实施例3
本实施例通用的车内乘客防遗忘的传感装置的实现方法，在实施例1或2的基础上，确定所述的车辆相关事件、乘客相关事件、车内环境相关事件和/或用户自定义事件的逻辑关系满足预定的发生儿童遗忘在车上事件的逻辑关系后，还包括确定该儿童遗忘事件的紧急优先级，根据紧急优先级设定不同的报警方式及报警频率，其中，一级紧急，以一级频率向绑定的私人电子终端发出报警信息以及预定的公共接收平台发出报警信息；二级紧急，以二级频 率向绑定的私人电子终端发出报警信息；三级紧急，以三级频率向绑定的私人电子终端发出报警信息；……K级紧急，以K级频率向绑定的私人电子终端发出抱紧信息，儿童遗忘事件的紧急优先级，一级＞二级＞三级……＞K级，一级频率＞二级频率＞三级频率＞K级频率。例如，车内温度过高，且此状况持续的时间超过预定时间，此时可以认定为一级紧急，此时以一级频率向绑定的私人电子终端(手机、笔记版电脑、PAD、智能手表)发出报警信息、向公共接收平台(后台运行商的服务器和/或110)发出报警信息。报警信息的具体包含哪些内容，例如本车位置、车内环境等，根据程序可以自行设定。再例如，车辆已经停止，检测到发生司机下车事件超出预定时间，且车内后座有乘客的声音和/或乘客移动，此时可以定义为二级紧急，向绑定的私人电子终端发出报警信息。具体地紧急情况的设定以及报警方式的设定可以根据实际情况进行具体地数据训练后生成相应的训练模型。同时，本实施例中，私人电子终端的具体绑定数量由用户自行设定。报警信号的输出还包括声光报警信号。
实施例4
本实施例通用的车内乘客防遗忘的传感装置的实现***，包括：
置在车内的多个非接触传感器组；
事件分类器，用于对多个非接触传感器组获取的传感数据进行分析运算，得到预定时间周期内发生的车辆相关事件、乘客相关事件、车内环境相关事件以及用户自定义事件；其中，车辆相关事件至少包括：车辆发动机启动、车辆行驶、车辆等候、发动起熄火、打开车门、关闭车门；乘客相关事件至少包括：人发出声音、人体移动、乘客上车、乘客下车；车内环境相关事件至少包括：车内温度、湿度、气压正常、车内温度、湿度、气压中至少一项偏高或偏低；
遗忘事件判定单元，用于判断所述的车辆相关事件、乘客相关事件、车内环境相关事件和/或用户自定义事件的逻辑关系是否满足预定的发生儿童遗忘在车上事件的逻辑关系，
若满足，则报警单元发出报警信息；若不满足，则不做处理。
所述事件分类器包括：
特征化处理模块，用于对获取的所述传感数据进行特征化处理，得到特征化数据；
识别模型筛选模块，用于基于所述传感数据，对所述事件分类器的获选事件识别模型进行筛选，过滤掉相关性低的候选事件识别模型，保留m个相关性高的候选事件识别模型；
结果输出模块，用于将特征化数据被依次代入m个保留的候选事件识别模型进行计算， 输出识别到的事件类型及可信度。
其中，识别模型筛选模块的筛选方法具体包括：
计算传感器数据的平均值特征值向量F i；
对满足激活条件的候选事件识别模型进行激活，若满足预定激活条件的候选事件识别模型大于m个，则选取偏离评分最小的m个模型，作为筛选结果；其中，各候选事件识别模型的激活条件为：(F x-F 0)>2*σ 0，(F i-F x)/σ x为候选事件识别模型的偏离评分；
背景环境下，计算预定时间周期内获取的传感数据的平均值和标准方差的特征值向量，分别记为F 0和σ 0；各候选事件识别模型数据训练过程中，获取的传感数据的平均值和标准方差的特征值向量，分别记为F x和σ x；x为候选事件识别模型的编号。
本实施例以校车为例，在校车前中部部署一个防遗忘集成传感装置，其中物理传感器包括加速度传感器、陀螺仪、麦克风、电磁感应、温湿度、气压以及两组多普勒雷达和被动红外(PIR)传感器。传感装置通过车内12V照明电路取电。司机停车、打开前车门，传感装置检测到停车事件和开门事件；学生陆续上车，传感装置检测到上车事件；司机关门、车辆启动，传感装置检测到关门事件、发动机启动事件和车辆行驶事件，并将车辆状态置为行驶中；车辆中途停车，传感装置将车辆状态置为车辆等候；车辆继续行驶，传感装置将车辆状态置为行驶中；车辆到达目的地、司机停车，传感装置将车辆状态置为停止；司机熄火，传感装置将车辆状态置为发动机停止状态；司机打开前后车门、学生陆续下车，传感装置检测到开门事件、下车事件；司机关闭前后车门，打开司机车门、下车、关闭司机车门，传感装置依次检测到关门事件、开门事件、下车事件和关门事件。但是有一个学生在车上睡着、被遗忘在车内。3分钟后(假定防遗忘告警延时设置为3分钟)，如果该学生发成声音或者移动身体，传感装置就会检测到人体活动事件，则通过通信模块发出告警消息。
实施例5
本实施例通用的车内乘客防遗忘的传感装置的实现***，在实施例4的基础上，还包括：事件运算周期时长确定模块，用于基于前面m个事件运算周期中发生的所述的车辆相关事件和/或车内环境相关事件之间的逻辑关系关系，预测下一个事件运算周期内所述的车辆相关事件和/或车内环境相关事件，基于所述对应表得到下一个事件运算周期内发生儿童遗忘事件概率；
基于下一个事件运算周期内发生儿童遗忘事件概率，确定下一个事件运算周期的时长；
所述的车辆相关事件和/或车内环境相关事件与发生儿童遗忘车内事件概率的对应表通过对历史数据进行机器学习得到。
实施例6
本实施例通用的车内乘客防遗忘的传感装置的实现***，在实施例4或5的基础上，还包括紧急优先级确定模块，用于在确定所述的车辆相关事件、乘客相关事件、车内环境相关事件和/或用户自定义事件的逻辑关系满足预定的发生儿童遗忘在车上事件的逻辑关系后，确定该儿童遗忘事件的紧急优先级，根据紧急优先级设定不同的报警方式及报警频率，其中，一级紧急，发出声光报警；二级紧急，通过无线网络向预定的接收终端发出报警信息；三级紧急，向预定的接收终端发出报警信息、向预定的公共接收平台发出报警信息。
上述各实施例中，还包括：获取车辆信息，基于车辆信息加载与所述的事件分类器，若本地存储有相应的事件分类器，则采用本地存储的事件分类器；若本地没有存储相应的事件分类器，则通过无线网络加载远程服务终端的事件分类器，其中，所述的车辆信息包括车型、车辆品牌、车辆载客人数、车辆型号中的一种或几种的组合。
上述各实施例中，所述的传感器包括加速度传感器、陀螺仪、麦克风、电磁感应、温湿度、气压、多普勒雷达和被动红外(PIR)传感器。
上述各实施例中，传感器数据特征化是将采集到的传感器原始数据进行特征提取，既保留了关键信息，也起了到数据压缩和保密的作用。更进一步，对采样频率低于1千赫兹的传感器，如：多普勒雷达、被动红外、温湿度、气压等，以1到2秒的滑动时间窗分别计算6个特征值(最大值、最小值、中位数、平均值、样本数、标准方差)；对采样频率高于1千赫兹的传感器，如：加速度传感器、陀螺仪、麦克风、电磁感应等，数据特征化时先进行快速傅里叶变换(FFT，256样本滑动窗口)，然后再计算6个特征值。
上述各实施例中，特征化处理后的数据封包(X矩阵)传送给后台机器学习工具，连同事件类型标签(Y向量)一起进行事件识别模型的训练。机器学习工具仅在模型训练阶段使用、支持常规的机器学习算法(如：SVM支持向量机、随机森林、神经网络)即可，可以为离线工具，也可以通过自动化脚本在线使用。
上述各实施例中，选取偏离评分最小的m个模型，m的具体数值可以为5也可以根据具体情况具体设置，这个具体数据与装置能够支撑的运算量具体设置。
“乘客被遗忘在车内”这一结果反映在一系列相关的事件上。因此，通过对一段时间内 检测到事件的逻辑判断，就能够准确识别出有乘客被遗忘在车内，并通过通讯模块发出告警消息。防遗忘相关的判别和处理逻辑如图3所示：车辆状态包括：发动起启动、车辆行驶、车辆等候、发动起熄火等；车内环境事件包括：正常、温度过高、温度过低、湿度过高、湿度过低、气压过高、气压过低等；其他事件包括开门、关门、上车、下车等。
以上所述仅是本发明的优选实施方式，并不用于限制本发明，应当指出，对于本技术领域的普通技术人员来说，在不脱离本发明技术原理的前提下，还可以做出若干改进和变型，这些改进和变型也应视为本发明的保护范围。

Claims (10)

  1. 一种通用的车内乘客防遗忘的传感装置的实现方法，其特征在于，包括：
    获取设置在车内的多个非接触传感器组输出的各组传感数据；
    利用预存储或预加载的事件分类器对所述的传感数据进行分析运算，得到预定时间周期内发生的车辆相关事件、乘客相关事件、车内环境相关事件以及用户自定义事件；其中，车辆相关事件至少包括：车辆发动机启动、车辆行驶、车辆等候、发动起熄火、打开车门、关闭车门；乘客相关事件至少包括：人发出声音、人体移动、乘客上车、乘客下车；车内环境相关事件至少包括：车内温度、湿度、气压正常、车内温度、湿度、气压中至少一项偏高或偏低；
    判断所述的车辆相关事件、乘客相关事件、车内环境相关事件和/或用户自定义事件的逻辑关系是否满足预定的发生儿童遗忘在车上事件的逻辑关系，
    若满足，则发出报警信息；若不满足，则不做处理。
  2. 根据权利要求1所述的通用的车内乘客防遗忘的传感装置的实现方法其特征在于，所述事件分类器通过监督机器学习方式生成，事件分类器对事件类型的识别过程包括：
    对获取的所述传感数据进行特征化处理，得到特征化数据，所述的特征化数据包括：进行快速傅里叶变换、滑动时间窗、计算最大值、最小值、中位数、平均值、样本数、标准方差6个特征值；
    基于所述传感数据，对所述事件分类器的获选事件识别模型进行筛选，过滤掉相关性低的候选事件识别模型，保留m个相关性高的候选事件识别模型；
    特征化数据被依次代入m个保留的候选事件识别模型进行计算，输出识别到的事件类型及可信度。
  3. 根据权利要求2所述的通用的车内乘客防遗忘的传感装置的实现方法，其特征在于，所述的m个相关性高的候选事件识别模型的筛选方法具体包括：
    计算传感器数据的平均值特征值向量F i；
    对满足激活条件的候选事件识别模型进行激活，若满足预定激活条件的候选事件识别模型大于m个，则选取偏离评分最小的m个模型，作为筛选结果；各候选事件识别模型的激活条件为：(F x-F 0)＞2*σ 0，(F i-F x)/σ x为候选事件识别模型的偏离评分；
    其中，背景环境下，计算预定时间周期内获取的传感数据的平均值和标准方差的特征值向量，分别记为F 0和σ 0；各候选事件识别模型数据训练过程中，获取的传感数据的平均值和标准方差的特征值向量，分别记为F x和σ x；x为候选事件识别模型的编号。
  4. 根据权利要求1所述的通用的车内乘客防遗忘的传感装置的实现方法，其特征在于，预定时间周期包括多个不同的事件运算周期，
    基于历史数据训练得到所述的车辆相关事件和/或车内环境相关事件与发生儿童遗忘车内事件概率的对应表；
    基于前面m个事件运算周期中发生的所述的车辆相关事件和/或车内环境相关事件之间的逻辑关系，预测下一个事件运算周期内所述的车辆相关事件和/或车内环境相关事件，基于所述对应表得到下一个事件运算周期内发生儿童遗忘事件概率；
    基于下一个事件运算周期内发生儿童遗忘事件概率，确定下一个事件运算周期的时长。
  5. 根据权利要求1所述的通用的车内乘客防遗忘的传感装置的实现方法，其特征在于，还包括：获取车辆信息，基于车辆信息加载与所述的事件分类器，若本地存储有相应的事件分类器，则采用本地存储的事件分类器；若本地没有存储相应的事件分类器，则通过无线网络加载远程服务终端的事件分类器，其中，所述的车辆信息包括车型、车辆品牌、车辆载客人数、车辆型号中的一种或几种的组合。
  6. 根据权利要求1所述的通用的车内乘客防遗忘的传感装置的实现方法，其特征在于，确定所述的车辆相关事件、乘客相关事件、车内环境相关事件和/或用户自定义事件的逻辑关系满足预定的发生儿童遗忘在车上事件的逻辑关系后，还包括确定该儿童遗忘事件的紧急优先级，根据紧急优先级设定不同的报警方式及报警频率，其中，一级紧急，以一级频率向绑定的私人电子终端发出报警信息以及预定的公共接收平台发出报警信息；二级紧急，以二级频率向绑定的私人电子终端发出报警信息；三级紧急，以三级频率向绑定的私人电子终端发出报警信息；......K级紧急，以K级频率向绑定的私人电子终端发出抱紧信息，儿童遗忘事件的紧急优先级，一级＞二级＞三级......＞K级，一级频率＞二级频率＞三级频率＞K级频率。
  7. 一种通用的车内乘客防遗忘的传感装置的实现***，其特征在于，包括：
    置在车内的多个非接触传感器组，所述的传感器包括加速度传感器、陀螺仪、麦克风、电磁感应、温/湿度、气压、多普勒雷达和/或被动红外PIR传感器；
    事件分类器，用于对多个非接触传感器组获取的传感数据进行分析运算，得到预定时间周期内发生的车辆相关事件、乘客相关事件、车内环境相关事件以及用户自定义事件；其中，车辆相关事件至少包括：车辆发动机启动、车辆行驶、车辆等候、发动起熄火、打开车门、关闭车门；乘客相关事件至少包括：人发出声音、人体移动、乘客上车、乘客下车；车内环 境相关事件至少包括：车内温度、湿度、气压正常、车内温度、湿度、气压中至少一项偏高或偏低；
    遗忘事件判定单元，用于判断所述的车辆相关事件、乘客相关事件、车内环境相关事件和/或用户自定义事件的逻辑关系是否满足预定的发生儿童遗忘在车上事件的逻辑关系，
    若满足，则报警单元发出报警信息；若不满足，则不做处理。
  8. 根据权利要求7所述的通用的车内乘客防遗忘的传感装置的实现***，其特征在于，所述事件分类器包括：
    特征化处理模块，用于对获取的所述传感数据进行特征化处理，得到特征化数据；
    识别模型筛选模块，用于基于所述传感数据，对所述事件分类器的获选事件识别模型进行筛选，过滤掉相关性低的候选事件识别模型，保留m个相关性高的候选事件识别模型；
    结果输出模块，用于将特征化数据被依次代入m个保留的候选事件识别模型进行计算，输出识别到的事件类型及可信度；
    识别模型筛选模块的具体筛选方法包括：
    计算传感器数据的平均值特征值向量Fi；
    对满足激活条件的候选事件识别模型进行激活，若满足预定激活条件的候选事件识别模型大于m个，则选取偏离评分最小的m个模型，作为筛选结果；各候选事件识别模型的激活条件为：(F x-F 0)＞2*σ 0，(F i-F x)/σ x为候选事件识别模型的偏离评分；
    其中，背景环境下，计算预定时间周期内获取的传感数据的平均值和标准方差的特征值向量，分别记为F 0和σ 0；各候选事件识别模型数据训练过程中，获取的传感数据的平均值和标准方差的特征值向量，分别记为F x和σ x；x为候选事件识别模型的编号。
  9. 根据权利要求7所述的通用的车内乘客防遗忘的传感装置的实现***，其特征在于，还包括：事件运算周期时长确定模块，用于基于前面m个事件运算周期中发生的所述的车辆相关事件和/或车内环境相关事件之间的逻辑关系关系，预测下一个事件运算周期内所述的车辆相关事件和/或车内环境相关事件，基于所述对应表得到下一个事件运算周期内发生儿童遗忘事件概率；
    基于下一个事件运算周期内发生儿童遗忘事件概率，确定下一个事件运算周期的时长；
    所述的车辆相关事件和/或车内环境相关事件与发生儿童遗忘车内事件概率的对应表通过对历史数据进行机器学习得到。
  10. 根据权利要求7所述的通用的车内乘客防遗忘的传感装置的实现***，其特征在于， 还包括事件分类器加载模块，用于获取车辆信息，基于车辆信息加载与所述的事件分类器，若本地存储有相应的事件分类器，则采用本地存储的事件分类器；若本地没有存储相应的事件分类器，则通过无线网络加载远程服务终端的事件分类器，其中，所述的车辆信息包括车型、车辆品牌、车辆载客人数、车辆型号中的一种或几种的组合；
    所述报警单元，包括紧急优先级确定模块，用于确定该儿童遗忘事件的紧急优先级，根据紧急优先级设定不同的报警方式及报警频率，其中，一级紧急，以一级频率向绑定的私人电子终端发出报警信息以及预定的公共接收平台发出报警信息；二级紧急，以二级频率向绑定的私人电子终端发出报警信息；三级紧急，以三级频率向绑定的私人电子终端发出报警信息；......K级紧急，以K级频率向绑定的私人电子终端发出抱紧信息，儿童遗忘事件的紧急优先级，一级＞二级＞三级......＞K级，一级频率＞二级频率＞三级频率＞K级频率。
PCT/CN2018/109440 2017-11-14 2018-10-09 通用的车内乘客防遗忘的传感装置的实现方法和*** WO2019095887A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201711124246.XA CN107933476B (zh) 2017-11-14 2017-11-14 通用的车内乘客防遗忘的传感装置的实现方法和***
CN201711124246.X 2017-11-14

Publications (1)

Publication Number Publication Date
WO2019095887A1 true WO2019095887A1 (zh) 2019-05-23

Family

ID=61932083

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/109440 WO2019095887A1 (zh) 2017-11-14 2018-10-09 通用的车内乘客防遗忘的传感装置的实现方法和***

Country Status (2)

Country Link
CN (1) CN107933476B (zh)
WO (1) WO2019095887A1 (zh)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107933476B (zh) * 2017-11-14 2019-08-13 苏州数言信息技术有限公司 通用的车内乘客防遗忘的传感装置的实现方法和***
CN108859944A (zh) * 2018-07-27 2018-11-23 马杰 基于人体传感器的车辆报警***及其控制方法
CN109177920A (zh) * 2018-09-03 2019-01-11 余利 一种智能安防的汽车门锁***及装置
CN109541498B (zh) * 2018-11-30 2021-10-15 苏州数言信息技术有限公司 一种通用的灯具故障智能检测方法和***
CN109615822B (zh) * 2019-01-17 2023-08-29 长安大学 一种基于儿童手表的校车防遗漏报警***及方法
CN110650246A (zh) * 2019-09-26 2020-01-03 陈学礼 一种手机防丢***
CN111483422A (zh) * 2020-04-24 2020-08-04 一汽奔腾轿车有限公司 一种生命体监测***
CN111882822A (zh) * 2020-08-07 2020-11-03 大陆投资(中国)有限公司 一种防止儿童被遗留于车内的提醒方法、装置以及服务器

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE102004044220A1 (de) * 2003-09-17 2005-04-14 Daimlerchrysler Ag Sicherheitsvorrichtung zur Erzeugung eines Warnsignals für ein Fahrzeug
CN202995965U (zh) * 2012-12-03 2013-06-12 浙江海洋学院 一种可用于校车的报警装置
CN106022235A (zh) * 2016-05-13 2016-10-12 中国人民解放军国防科学技术大学 基于人体探测的遗漏孩童检测方法
CN106627361A (zh) * 2016-12-02 2017-05-10 浙江吉利控股集团有限公司 防止儿童被遗忘车内的控制***及控制方法
CN107933476A (zh) * 2017-11-14 2018-04-20 苏州数言信息技术有限公司 通用的车内乘客防遗忘的传感装置的实现方法和***

Family Cites Families (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104637242A (zh) * 2013-11-12 2015-05-20 广州华久信息科技有限公司 一种基于多分类器集成的老人跌倒检测方法及***
CN104590172A (zh) * 2015-01-30 2015-05-06 无锡桑尼安科技有限公司 一种车辆停车时的内部监视控制方法
CN106328152B (zh) * 2015-06-30 2020-01-31 芋头科技（杭州）有限公司 一种室内噪声污染自动识别监测***
CN105059212A (zh) * 2015-08-07 2015-11-18 长沙高汇环保科技有限公司 车门遥控***及其控制方法
CN105261149B (zh) * 2015-11-02 2017-09-01 程宇笑 防止幼童被遗忘在汽车内的装置及方法
CN105590408B (zh) * 2016-02-06 2018-05-04 高强 一种人体跌倒检测方法及保护装置
CN105809890B (zh) * 2016-05-13 2017-10-13 中国人民解放军国防科学技术大学 面向校车安全的遗漏孩童检测方法
CN106228750A (zh) * 2016-09-20 2016-12-14 李国远 锁车后异常情况报警方法、装置及***
CN106427861B (zh) * 2016-11-16 2019-01-25 孙菁若 车载儿童遗漏提醒装置及其方法
CN106781236A (zh) * 2016-12-29 2017-05-31 东南大学 一种基于人物检测和再识别方法的远程报警装置及远程监控方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE102004044220A1 (de) * 2003-09-17 2005-04-14 Daimlerchrysler Ag Sicherheitsvorrichtung zur Erzeugung eines Warnsignals für ein Fahrzeug
CN202995965U (zh) * 2012-12-03 2013-06-12 浙江海洋学院 一种可用于校车的报警装置
CN106022235A (zh) * 2016-05-13 2016-10-12 中国人民解放军国防科学技术大学 基于人体探测的遗漏孩童检测方法
CN106627361A (zh) * 2016-12-02 2017-05-10 浙江吉利控股集团有限公司 防止儿童被遗忘车内的控制***及控制方法
CN107933476A (zh) * 2017-11-14 2018-04-20 苏州数言信息技术有限公司 通用的车内乘客防遗忘的传感装置的实现方法和***

Also Published As

Publication number Publication date
CN107933476B (zh) 2019-08-13
CN107933476A (zh) 2018-04-20

Similar Documents

Publication Publication Date Title
WO2019095887A1 (zh) 通用的车内乘客防遗忘的传感装置的实现方法和***
US11778144B2 (en) Vehicle monitoring
US11453363B2 (en) Vehicle cabin safety system and method
CN105247587B (zh) 利用响应***对私人设备的车辆追踪
US11230246B1 (en) Vehicle occupancy monitor
WO2021189641A1 (zh) 遗留对象检测
US6792339B2 (en) Artificial passenger with condition sensors
US9840166B2 (en) Determining the number of people in a vehicle
US20170043783A1 (en) Vehicle control system for improving occupant safety
US10446011B2 (en) System and method for providing rear seat monitoring within a vehicle
WO2016025680A1 (en) Integrated vehicle sensing and warning system
US11210540B2 (en) System and method for providing rear seat monitoring within a vehicle
CN108119003A (zh) 用于增强车库门开启器远程控制的***和方法
CN105898232A (zh) 车辆内部安全监控***及其监控方法
US11186244B2 (en) Smart child safety equipment in vehicles
CN110194117A (zh) 车辆的遗忘警报装置
US20230356719A1 (en) Methods and systems for detection and prevention of unattended vehicle deaths
CN107160967A (zh) 一种车内生命体监测应急方法
CN114463928B (zh) 一种智能报警方法和***
CN112614308B (zh) 车内生命侦测***
CN114789705A (zh) 一种车内生命遗留检测控制方法、装置以及车辆
Suryavanshi et al. In Cabin Driver Monitoring and Alerting System For Passenger cars using Machine Learning.
TWI816280B (zh) 車輛內環境監控方法及電子設備
CN213322875U (zh) 一种车内遗留物品探测装置
US20240059234A1 (en) Vehicle occupancy sensor systems and methods

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: 18878122

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18878122

Country of ref document: EP

Kind code of ref document: A1

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

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 29/10/2020)

122 Ep: pct application non-entry in european phase

Ref document number: 18878122

Country of ref document: EP

Kind code of ref document: A1