WO2022016583A1 - 一种教练车智能辅助训练方法及系统 - Google Patents

一种教练车智能辅助训练方法及系统 Download PDF

Info

Publication number
WO2022016583A1
WO2022016583A1 PCT/CN2020/105268 CN2020105268W WO2022016583A1 WO 2022016583 A1 WO2022016583 A1 WO 2022016583A1 CN 2020105268 W CN2020105268 W CN 2020105268W WO 2022016583 A1 WO2022016583 A1 WO 2022016583A1
Authority
WO
WIPO (PCT)
Prior art keywords
coach
training
vehicle
path
machine learning
Prior art date
Application number
PCT/CN2020/105268
Other languages
English (en)
French (fr)
Inventor
高青
陈小兵
高岩
Original Assignee
南京智金科技创新服务中心
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 南京智金科技创新服务中心 filed Critical 南京智金科技创新服务中心
Priority to SE2051569A priority Critical patent/SE2051569A1/en
Publication of WO2022016583A1 publication Critical patent/WO2022016583A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G09—EDUCATION; CRYPTOGRAPHY; DISPLAY; ADVERTISING; SEALS
    • G09B—EDUCATIONAL OR DEMONSTRATION APPLIANCES; APPLIANCES FOR TEACHING, OR COMMUNICATING WITH, THE BLIND, DEAF OR MUTE; MODELS; PLANETARIA; GLOBES; MAPS; DIAGRAMS
    • G09B9/00—Simulators for teaching or training purposes
    • G09B9/02—Simulators for teaching or training purposes for teaching control of vehicles or other craft
    • G09B9/04—Simulators for teaching or training purposes for teaching control of vehicles or other craft for teaching control of land vehicles
    • G09B9/042—Simulators for teaching or training purposes for teaching control of vehicles or other craft for teaching control of land vehicles providing simulation in a real vehicle
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G06V20/54—Surveillance or monitoring of activities, e.g. for recognising suspicious objects of traffic, e.g. cars on the road, trains or boats
    • G—PHYSICS
    • G09—EDUCATION; CRYPTOGRAPHY; DISPLAY; ADVERTISING; SEALS
    • G09B—EDUCATIONAL OR DEMONSTRATION APPLIANCES; APPLIANCES FOR TEACHING, OR COMMUNICATING WITH, THE BLIND, DEAF OR MUTE; MODELS; PLANETARIA; GLOBES; MAPS; DIAGRAMS
    • G09B19/00—Teaching not covered by other main groups of this subclass
    • G09B19/16—Control of vehicles or other craft
    • G09B19/167—Control of land vehicles

Definitions

  • Embodiments of the present invention relate to the technical field of intelligent assistance in driving school teaching, and in particular, to a method and system for intelligent auxiliary training of coach vehicles.
  • the smart products provided by existing driving schools are basically virtual products, such as driving training simulators.
  • driving training simulators because the operational follow-up of driving training simulators is quite different from that of real vehicles, and the practicability of supporting teaching materials is poor, trainees cannot The effect of practice is not obvious; and the driving teaching in different regions is different, and such products cannot be changed according to the changes of driving schools, so the training simulators of various driving schools have almost become decorations.
  • subject 2 In the learning of driving skills, subject 2 requires students to spend a lot of time in training, especially the reversing and storage project in subject 2. Because of its complex practice behavior, students need to put more energy into practice to increase their proficiency. , and in the process of reversing into the garage, students often need to fine-tune the direction of the vehicle to make the vehicle enter the parking space, but it is difficult for students to realize the direction of fine-tuning and the angle of fine-tuning at the beginning, and it is difficult for the coach to remind in real time. , If the coach is in a bad mood, it will also increase the tension of the students, which is not conducive to the improvement of the students' driving level. Therefore, it is necessary to improve one or more problems existing in the above-mentioned related technical solutions.
  • the purpose of the embodiments of the present invention is to provide an intelligent auxiliary training method and system for a coach vehicle, thereby at least to a certain extent overcoming one or more problems caused by the limitations and defects of the related art.
  • an intelligent auxiliary training method for a coach vehicle comprising:
  • Detecting the image of the training ground by using the target detection model pre-established in the machine learning model, to determine the type of the training ground, and to determine whether there is a coach car in the image of the training ground;
  • the machine learning model detects the image of the training site by using a pre-established deep learning-based coach vehicle target detection model
  • the machine learning model calculates and analyzes the driving path to determine whether the driving path deviates from the preset path;
  • the machine learning model is used by learning the driving path of the coach vehicle for many times and consolidating the parameters of the machine learning model after training;
  • the receiving module sends the received correction information to the control module in the coaching vehicle, and the control module controls the correction information to be displayed and played in the coaching vehicle, so that the students in the coaching vehicle can adjust their own driving actions according to the correction information. , so that the driving path is adjusted by the student to approach the preset path;
  • the correction information includes voice prompt information and video demonstration information, and the video demonstration information is demonstration animation information of the pre-rotation direction and angle of the steering wheel in the coach;
  • the machine learning model stores the actual driving path data of the coach vehicle this time;
  • the method further includes:
  • the traveling speed of the coach vehicle is monitored in real time, and it is judged whether the traveling speed value is within a preset range;
  • a prompt message is generated, and the prompt message is sent to the coach vehicle.
  • the method further includes:
  • a braking signal is sent to the coach vehicle to control the coach The car stopped in time.
  • the route comparison information is the comparison video information of the actual driving video of the coach vehicle and the preset virtual coach vehicle driving video.
  • the types of training venues include: reversing parking training items, side parking training items, S-bend training items, and right-angle bend training items.
  • an intelligent auxiliary training system for a coach vehicle comprising:
  • a camera module arranged above the training ground, for acquiring image information of the training ground in real time, and uploading the image information of the training ground to the machine learning model;
  • a machine learning model used to detect the image of the training site through a target detection model pre-established in the machine learning model, to determine the type of the training site, and to determine whether there is a coach car in the image of the training site; wherein , the machine learning model detects the image of the training site by using a pre-established deep learning-based coaching vehicle target detection model; if there is a coaching vehicle, multiple moving positions of the coaching vehicle are detected by the machine learning model performing calculation and analysis processing to generate a one-vehicle path from the plurality of moving positions;
  • the machine learning model is used to determine whether the driving path deviates from the preset path by calculating and analyzing the driving path; wherein, the machine learning model solidifies the machine after multiple times of learning and training of the driving path of the coach vehicle.
  • the parameters of the learning model are used; if the driving path deviates, a correction information is generated according to the degree of deviation of the driving path through the machine learning model, and the correction information is sent to the receiving module on the coach; when After the coach vehicle completes a single training project, the machine learning model is used to store the actual driving path data of the coach vehicle;
  • the coaching vehicle module includes a receiving module and a control module, the receiving module is used to send the received correction information to the control module in the coaching vehicle; the control module is used to control the correction information to be displayed and displayed in the coaching vehicle. Play, so that the students in the coach car adjust their own driving actions according to the correction information, so that the driving path is adjusted by the students to approach the preset path;
  • the correction information includes voice prompt information and video demonstration information, and the video demonstration information is demonstration animation information of the pre-rotation direction and angle of the steering wheel in the coach;
  • the comparison module is used for comparing the stored actual driving path with the preset path, generating path comparison information, and sending the path comparison information to the terminal for students to correct errors in time.
  • the machine learning module is used to monitor the travel speed of the coach vehicle in real time during the process of the coach vehicle completing the training program, and determine whether the travel speed value is within the within the preset range; if the traveling speed exceeds the preset range, a prompt message is generated, and the prompt message is sent to the coach vehicle.
  • the machine learning model is further used for when the driving path of the coach vehicle deviates from the preset path by more than a preset offset, and the machine learning model identifies that the coach vehicle is in danger In the state, the brake signal is sent to the coach vehicle to control the coach vehicle to stop in time.
  • the route comparison information is the comparison video information of the actual driving video of the coach vehicle and the preset virtual coach vehicle driving video.
  • the types of training venues include: reversing parking training items, side parking training items, S-bend training items, and right-angle bend training items.
  • the image information is captured in real time by a camera arranged above the training ground, and the image information is uploaded to the machine learning model, and then the machine learning model is used to analyze the image information in real time.
  • the calculation and analysis of the actual driving path of the coach vehicle, the offset between the actual driving path and the preset path is obtained, and correction information is generated according to the offset, and the correction information can be displayed and played in the coach vehicle in real time. Students can intuitively adjust the steering and angle of the steering wheel according to the video demonstration in the correction information, and at the same time, voice prompts will be given in the coach car.
  • This intelligent auxiliary training method can replace the coach to a certain extent, provide the students with intelligent driving assistance in the process of driving, and can improve the training level of the students to a certain extent.
  • Fig. 1 shows the flow chart of the intelligent auxiliary training method of the coach vehicle in the exemplary embodiment of the present invention
  • Fig. 2 shows the module schematic diagram of the intelligent auxiliary training system of the coach vehicle in the exemplary embodiment of the present invention
  • FIG. 3 shows a schematic diagram of the steering wheel steering and angle in the display screen of the coach car in an exemplary embodiment of the present invention
  • FIG. 4 shows a schematic diagram of a preset travel path of a coach vehicle in a training field in an exemplary embodiment of the present invention
  • Fig. 5 shows the schematic diagram of the actual travel path of the coach vehicle in the training field in the exemplary embodiment of the present invention
  • FIG. 6 is a schematic diagram showing a path comparison between the actual travel path of the coach vehicle and the preset path in the training field in the exemplary embodiment of the present invention.
  • Example embodiments will now be described more fully with reference to the accompanying drawings.
  • Example embodiments can be embodied in various forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art.
  • the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
  • This exemplary embodiment first provides an intelligent auxiliary training method for a coach vehicle.
  • the method may include:
  • step S101 the image information of the training field is acquired in real time through a camera disposed above the training field, and the image information of the training field is uploaded to the machine learning model.
  • Step S102 the image of the training field is detected by the target detection model pre-established in the machine learning model, to determine the type of the training field, and to determine whether there is a coach car in the image of the training field;
  • the machine learning model detects the training field image by using a pre-established deep learning-based coach vehicle target detection model.
  • Step S103 if there is a coach vehicle, calculate and analyze multiple moving positions of the coach vehicle through the machine learning model, so as to generate a one-vehicle path from the multiple moving positions.
  • Step S104 the machine learning model calculates and analyzes the driving path to determine whether the driving path deviates from the preset path; wherein, the machine learning model solidifies the driving path after multiple times of learning and training of the driving path of the coach vehicle.
  • the parameters of the machine learning model are used.
  • Step S105 if the driving path deviates, generate correction information according to the degree of deviation of the driving path through the machine learning model, and send the correction information to the receiving module on the coach vehicle.
  • Step S106 the receiving module sends the received correction information to the control module in the coach car, and the control module controls the correction information to be displayed and played in the coach car, so that the students in the coach car adjust themselves according to the correction information.
  • driving action so that the driving path is adjusted by the trainee to be close to the preset path;
  • the correction information includes voice prompt information and video demonstration information, and the video demonstration information is the pre-rotation direction and angle of the steering wheel in the coach car. Demonstrate animation information.
  • Step S107 after the coach vehicle completes a single training item, the machine learning model stores the actual driving path data of the coach vehicle this time.
  • Step S108 compare the stored actual driving route with the preset route, generate route comparison information, and send the route comparison information to the terminal for students to correct errors in time.
  • the image information is captured in real time by the camera set above the training ground, and the image information is uploaded to the machine learning model, and then the actual driving path of the coach vehicle is calculated and analyzed by the machine learning model. Analyze, get the offset between the actual driving path and the preset path, and generate correction information according to the offset.
  • the correction information can be displayed and played in the coach car in real time, and students can demonstrate according to the video in the correction information. Intuitively adjust the steering and angle of the steering wheel, and at the same time, there will be voice prompts in the coach car. Through the joint action of the video demonstration and the voice prompts, the students can better correct their own driving movements.
  • This intelligent auxiliary training method can be used to a certain extent.
  • the substitute coach provides intelligent driving assistance for the trainees in the process of driving, and can improve the training level of the trainees to a certain extent.
  • step S101 the image information of the training field is acquired in real time through a camera disposed above the training field, and the image information of the training field is uploaded to the machine learning model.
  • the camera is fixed on the top of the training ground.
  • the camera above each training ground is turned on, and the camera will obtain the image information of the training ground in real time, and the obtained training ground image information Uploaded to the machine learning model
  • the image information here can be understood as picture frame data, but not limited to this.
  • the machine learning model first reads the data, that is, collects the data; and processes and corrects the data in advance, such as feature extraction, feature dimension reduction, feature null value processing, feature transformation, feature normalization, target value null value processing, Target value conversion, etc.; in the model testing stage, the cross-validation method can be selected first, and the data can be divided in advance; when the data is processed, use the data to build a training model.
  • There are many parameters for the evaluation of the model such as score, precision rate , recall rate, etc. The details can be understood with reference to the prior art, and details are not repeated here.
  • step S102 the image of the training field is detected by the target detection model pre-established in the machine learning model to determine the type of the training field, and to determine whether there is a coach car in the image of the training field; wherein, The machine learning model detects the image of the training field by using a pre-established deep learning-based coach vehicle target detection model.
  • the steps of establishing a deep learning-based target detection model for coaching vehicles may be as follows: building a deep learning network for coaching vehicle target detection; the deep learning network for coaching vehicle target detection includes: 8 convolutional layers, 8 ReLU activation layers and 3 A coach car feature extractor composed of two pooling layers, and a first convolution layer and a second convolution layer connected to the coach car feature extractor; obtain training site image samples with different angles, illumination and image quality; adopt A rectangular frame marks the position of the training area of the coach car in the training ground image sample, and records the coordinate information of the rectangular frame; with the coordinate information of each of the training ground image samples and the rectangular frame in the training ground image sample For a group of training data, generate a coach vehicle target detection training data set; use the generated coach vehicle target detection training data set to train the coach vehicle target detection deep learning network, and obtain a coach vehicle target detection based on deep learning after parameter updating Model.
  • the target detection model After the target detection model completes the image detection of the training site, it will obtain the type of the training site.
  • the type of the training site includes: reversing parking training items, side parking training items, S-curves Training items and right-angle bending training items, each of the above items can be trained by the intelligent auxiliary training method provided in this embodiment.
  • step S103 if there is a coaching vehicle, the machine learning model is used to calculate and analyze multiple moving positions of the coaching vehicle, so as to generate a one-vehicle path from the multiple moving positions.
  • the machine learning model can receive the image information of the training site captured by the camera in real time. Because there are coach vehicles in the training site, and the machine learning model can be trained by a large number of positive and negative samples of the coach vehicles, so as to solidify its parameters for use.
  • the positive samples are: The sample corresponding to the category that you want to classify correctly, in this embodiment, the positive sample is the coach car in the image information, and the negative sample can be the training ground other than the coach car, such as cement floor, lawn, yellow line, etc.
  • the machine learning model calculates and analyzes the received multiple moving positions of the coach vehicle, so as to generate a one-vehicle path from the multiple moving positions.
  • the camera uploads the frame of images captured by the camera to the machine learning module in turn, and the machine learning model calculates and processes each frame of the image received, so that it is consistent with the previous frame and
  • the previous image forms the driving path, which shows that the driving path is generated when the coach vehicle starts to reverse, and changes in real time as the coach vehicle moves.
  • step S104 the machine learning model calculates and analyzes the driving path to determine whether the driving path deviates from the preset path; wherein, the machine learning model is based on the learning and training of the driving path of the coach vehicle for many times It is used after curing the parameters of the machine learning model.
  • the machine learning model is used after learning and training a large number of coaching vehicle driving paths to solidify its parameters.
  • the library is successful, or how to adjust the steering and angle of the steering wheel when it deviates from the original route so that the coach car is put into the garage, and this adjustment method is converted into voice, which is used to assist the trainee in the training process when the trainee is practicing.
  • the machine learning model will calculate and analyze the driving path of the coach vehicle in real time through the captured image information, and determine whether the driving path deviates from the preset path. , the machine learning model at this time is equivalent to the driving school coach staring at the student to practice the car.
  • the coach will timely remind the student to adjust the direction so that the coach car is close to the conventional path.
  • step S105 if the driving path deviates, the machine learning model generates correction information according to the degree of deviation of the driving path, and sends the correction information to the receiving module on the coach vehicle.
  • the machine learning model finds that the driving path of the coach deviates from the driving path, or continues to drive on this driving path, the training subject will fail.
  • the degree of deviation generates a correction information, and sends the correction information to the coach.
  • the machine learning model will generate a correction information through calculation and analysis. , the correction information is sent to the receiving module on the coach car to prompt the trainee to adjust the direction.
  • step S106 the receiving module sends the received correction information to the control module in the coaching vehicle, and the control module controls the correction information to be displayed and played in the coaching vehicle, so that the students in the coaching vehicle can display and play the correction information according to the correction information.
  • Adjust your own driving action so that the driving path is close to the preset path through the adjustment of the trainee;
  • the correction information includes voice prompt information and video demonstration information, and the video demonstration information is the pre-rotation direction of the steering wheel in the coach car and Demo animation information for angles.
  • a control module is provided in the coach car, which is used to control the display screen in the coach car to display correction information, and to control the horn in the car to give voice prompts.
  • the display screen in the coach car to display correction information
  • the horn in the car to give voice prompts.
  • the correction information is sent to the receiving module on the coach, the receiving module sends the correction information to the control module in the car, and the control module will control the display screen to display the correction information, as shown in Figure 3 , the displayed content includes the video demonstration information in the correction information.
  • the video demonstration information is specifically the demonstration animation information of the pre-rotation direction and angle of the steering wheel in the coach car, and the direction of the steering wheel rotation can be marked by arrows.
  • the coach car will issue a voice reminder according to the voice prompt information in the correction information, such as "return to the wheel for half a circle", after driving for a certain distance according to the path , the machine learning model will generate another correction information through calculation and analysis and send it to the coach car.
  • the display screen will display the steering wheel video animation of different angles and rotation directions, and at the same time, it will be accompanied by voice reminders, such as "kill to the right” , through the video prompts and voice prompts on the display screen, until the trainee correctly parks the coach car in the garage.
  • the content of the voice prompt information in the correction information can be selected and set according to the actual situation, and is not limited to what is described in this embodiment.
  • the machine learning model stores the actual driving path data of the coach vehicle; compares the stored actual driving path with the preset path. For comparison, a path comparison information is generated, and the path comparison information is sent to the terminal for students to correct errors in time.
  • the storage module in the machine learning model will store the actual driving path of the training vehicle, and compare the actual driving path with the preset path.
  • the comparison is performed to generate a path comparison information.
  • the path comparison information is the comparison video information of the actual driving video of the coach vehicle and the preset virtual coach vehicle driving video. Specifically, it is set in the training venue.
  • the camera above can shoot the actual path of the coach vehicle training project, and store the video. This video is the actual driving video of the coach vehicle.
  • the actual driving video can be compared with the preset virtual driving video through the comparison module. , and generate a comparison video, and send the comparison video information to the terminal, for example, to the student's mobile phone. Rapid progress of students.
  • the training method further includes:
  • Step S1031 in the process of the coach vehicle completing the training item, monitor the travel speed of the coach vehicle in real time, and determine whether the travel speed value is within a preset range.
  • Step S1032 if the traveling speed exceeds the preset range, generate a prompt message, and send the prompt message to the coach vehicle.
  • the slower the vehicle speed the easier it is for students to adjust the state of the vehicle.
  • Monitor the travel speed of the coach and judge whether the travel speed value is within the preset range. For example, when the travel speed of the coach vehicle is greater than 5km/h, the coach vehicle will voice prompt the students to slow down.
  • the specific preset range can be based on actual conditions. It is set according to the situation, and there is no restriction here.
  • the training method further includes:
  • Step S1041 when the driving path of the coach vehicle deviates from the preset path by more than a preset offset, and the machine learning model recognizes that the coach vehicle is in a dangerous state, then send a brake signal to the coach vehicle to Control the coach to stop in time.
  • the trainee when the trainee is practicing, in order to prevent accidents, such as the trainee accidentally stepping on the brakes and causing the trainer to lose control, that is, the trainer deviates too much from the preset path, or the deviation is abnormal.
  • the machine learning model judges that the coach vehicle is in a dangerous state, it will send a braking signal to the receiving module in the coaching vehicle, and the braking module in the coaching vehicle will receive the braking signal sent by the receiving module, thereby controlling the The coach car is stopped to ensure the safety of the students during the practice.
  • This exemplary embodiment also provides an intelligent auxiliary training system for a coach car.
  • the system may include a camera module, a machine learning model, a coach module, and an alignment module.
  • the camera module is arranged above the training ground, and is used for acquiring image information of the training ground in real time, and uploading the image information of the training ground to the machine learning model.
  • the machine learning model is used to detect the image of the training field through the target detection model pre-established in the machine learning model, to determine the type of the training field, and to determine whether there is a coaching vehicle in the image of the training field; Wherein, the machine learning model detects the image of the training site by using a pre-established deep learning-based coach vehicle target detection model; if there is a coach vehicle, the machine learning model detects multiple movements of the coach vehicle. Calculate and analyze the position to generate a one-vehicle path from the plurality of moving positions;
  • the machine learning model is used to determine whether the driving path deviates from the preset path by calculating and analyzing the driving path; wherein, the machine learning model solidifies the machine after multiple times of learning and training of the driving path of the coach vehicle.
  • the parameters of the learning model are used; if the driving path deviates, a correction information is generated according to the degree of deviation of the driving path through the machine learning model, and the correction information is sent to the receiving module on the coach; when After the coach vehicle completes a single training item, the machine learning model is used to store the actual driving path data of the coach vehicle.
  • the coaching car module is used to send the received correction information to the control module in the coaching vehicle;
  • the control module is used to control the correction information to be displayed and played in the coaching vehicle, so that the students in the coaching vehicle can be corrected according to the correction.
  • information to adjust their own driving actions so that the driving path is adjusted to the preset path by the students;
  • the correction information includes voice prompt information and video demonstration information, and the video demonstration information is the pre-rotation direction of the steering wheel in the coach car and angle demonstration animation information.
  • the comparison module is used for comparing the stored actual driving path with the preset path, generating path comparison information, and sending the path comparison information to the terminal for students to correct errors in time.
  • the machine learning module is configured to monitor the travel speed of the coach vehicle in real time during the process of the coach vehicle completing the training program, and determine whether the travel speed value is within a preset range If the traveling speed exceeds the preset range, a prompt message is generated, and the prompt message is sent to the coach vehicle.
  • the machine learning model is further configured to, when the driving path of the coach vehicle deviates from the preset path by more than a preset offset, and the machine learning model recognizes that the coach vehicle is in a dangerous state, Then, a braking signal is sent to the coach vehicle to control the coach vehicle to stop in time.
  • the route comparison information is the comparison video information of the actual driving video of the coach vehicle and the preset virtual coach vehicle driving video.
  • the types of training venues include: reversing parking training items, side parking training items, S-bend training items, and right-angle bend training items.
  • the image information is captured in real time by the camera set above the training ground, and the image information is uploaded to the machine learning model, and then the actual driving path of the coach vehicle is analyzed by the machine learning model. Calculate and analyze to obtain the offset between the actual driving path and the preset path, and generate correction information according to the offset.
  • the correction information can be displayed and played in the coach car in real time, and students can The video demonstration intuitively adjusts the steering and angle of the steering wheel, and at the same time, there will be voice prompts in the coach car. Through the joint action of the video demonstration and the voice prompts, the students can better correct their own driving movements.
  • This intelligent auxiliary training method can be used in To a certain extent, it replaces the coach, provides intelligent driving assistance for students in the process of driving, and can improve the training level of students to a certain extent.
  • first and second are only used for descriptive purposes, and should not be construed as indicating or implying relative importance or implying the number of indicated technical features. Thus, a feature defined as “first” or “second” may expressly or implicitly include one or more of that feature.
  • “plurality” means two or more, unless otherwise expressly and specifically defined.
  • the first feature "on” or “under” the second feature may include the first and second features in direct contact, or may include the first and second features The two features are not in direct contact but through another feature between them.
  • the first feature being “above”, “over” and “above” the second feature includes the first feature being directly above and obliquely above the second feature, or simply means that the first feature is level higher than the second feature.
  • the first feature is “below”, “below” and “below” the second feature includes the first feature being directly below and diagonally below the second feature, or simply means that the first feature has a lower level than the second feature.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Educational Administration (AREA)
  • Educational Technology (AREA)
  • Computing Systems (AREA)
  • Multimedia (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Databases & Information Systems (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Electrically Operated Instructional Devices (AREA)

Abstract

一种教练车智能辅助训练方法及系统。该方法包括：通过摄像头实时获取训练场地图像信息，并上传至机器学习模型；通过模型对训练场地图像进行检测，以判断训练场地所属类型，以及判断训练场地图像中是否存在教练车；若存在教练车，则通过模型计算及分析处理，生成一行车路径；模型通过计算及分析行车路径，判断行车路径是否偏离预设路径；若偏离，则通过模型生成一矫正信息，并将该信息发送至教练车；将矫正信息在车内进行显示及播放，使得车内的学员根据矫正信息调整自身的驾驶动作，从而使行车路径通过调整而靠近预设路径。可以在一定程度上替代教练，为学员在练车过程中提供智能的练车辅助，并且提高了学员的训练水平。

Description

一种教练车智能辅助训练方法及系统 技术领域
本发明实施例涉及驾校教学智能辅助技术领域，尤其涉及一种教练车智能辅助训练方法及系统。
背景技术
随着社会的进步发展，人们的生活水平逐渐提高，越来越多的家庭拥有小汽车，从而越来越多的人开始学习开车技术，面对众多的考生，驾校教练人数相对来说比较少，从而导致驾校教练的工作量非常大，同样的技术动作同一个教练可能在一天内要不断地重复，这或许会导致教练不耐烦的心理变化，从而引起学生与驾校教练之间的矛盾。另外，由于学车人数众多，传统面授教学在学车时间和次数上受到较多限制，学员不能灵活选择，从而在一定程度上影响了学员驾驶技能水平的提高。
现有驾校所配备的智能产品基本上都是虚拟产品，例如驾驶训练模拟器，但因驾驶训练模拟器的操作随动性方面与实车差异较大、配套教材实用性较差等，因此学员练习的效果并不明显；并且不同地区的驾驶教学是由差别的，而这类产品不能根据驾校的改变而改变，所以各个驾校的训练模拟器几乎成了摆设。
在驾驶技能的学习中，科目二则需要学员通过大量的时间进行训练，尤其是科目二中的倒车入库项目，因其练习行为较为复杂，需要学员投入更多的精力进行练习以增加熟练度，而且在倒车入库的过程中往往需要学员对车辆方向进行微调以使车辆进入车位中，但学员在初学时很难意识到微调的方向以及微调的角度，而教练也很难实时的进行提醒，若教练情绪不佳，也会加剧学员的紧张感，不利于学员驾驶水平的提高。因此，有必要改善上述相关技术方案中存在的一个或者多个问题。
需要注意的是，本部分旨在为权利要求书中陈述的本发明的实施方式提供背景或上下文。此处的描述不因为包括在本部分中就承认是现有技术。
发明内容
本发明实施例的目的在于提供一种教练车智能辅助训练方法及系统，进而至少在一定程度上克服由于相关技术的限制和缺陷而导致的一个或者多个问题。
根据本发明实施例的第一方面，提供一种教练车智能辅助训练方法，该方法包括：
通过设置于训练场地上方的摄像头实时获取该训练场地图像信息，并将该训练场地图像信息上传至机器学习模型中；
通过所述机器学习模型中预建立的目标检测模型对所述训练场地图像进行检测，以判断所述训练场地所属类型，以及判断该训练场地图像中是否存在教练车；
其中，所述机器学习模型通过采用预建立的基于深度学习的教练车目标检测模型对所述训练场地图像进行检测；
若存在教练车，则通过所述机器学习模型对该教练车的多个移动位置进行计算及分析处理，以将该多个移动位置生成一行车路径；
所述机器学习模型通过计算及分析所述行车路径，以判断所述行车路径是否偏离预设路径；
其中，所述机器学习模型是通过多次教练车行车路径的学习及训练后固化该机器学习模型的参数而进行使用的；
若所述行车路径偏离，则通过所述机器学习模型根据所述行车路径偏离程度生成一矫正信息，并将该矫正信息发送至教练车上的接收模块；
所述接收模块将接收到的矫正信息发送至教练车中的控制模块，该控制模块控制所述矫正信息在教练车内进行显示及播放，使得教练车内的学员根据矫正信息调整自身的驾驶动作，从而使行车路径通过学员调整而靠近所述预设路径；
所述矫正信息包括语音提示信息和视频演示信息，所述视频演示信息为所述教练车内方向盘预转动方向及角度的演示动画信息；
当所述教练车完成单次训练项目后，所述机器学习模型将该次教练 车的实际行车路径数据进行存储；
将存储的所述实际行车路径与预设路径进行比对，生成一路径比对信息，并将该路径比对信息发送至终端，以供学员及时纠错。
本发明的一实施例中，该方法还包括：
在所述教练车完成所述训练项目的过程中，实时对所述教练车的行进速度进行监控，并判断所述行进速度值是否位于预设范围内；
若所述行进速度超过所述预设范围，则生成一提示信息，并将该提示信息发送至所述教练车。
本发明的一实施例中，该方法还包括：
当所述教练车行车路径偏移所述预设路径超过预设偏移量，且经所述机器学习模型识别该教练车处于危险状态时，则向该教练车发送刹车信号，以控制该教练车及时停车。
本发明的一实施例中，所述路经比对信息为教练车实际行车视频与预设虚拟教练车行车视频的比对视频信息。
本发明的一实施例中，所述训练场地所属类型包括：倒车入库训练项目、侧方停车训练项目、S弯训练项目和直角弯训练项目。
根据本发明实施例的第二方面，提供一种教练车智能辅助训练系统，该系统包括：
摄像头模块，设置于训练场地上方，用于实时获取训练场地图像信息，并将该训练场地图像信息上传至机器学习模型中；
机器学习模型，用于通过所述机器学习模型中预建立的目标检测模型对所述训练场地图像进行检测，以判断所述训练场地所属类型，以及判断该训练场地图像中是否存在教练车；其中，所述机器学习模型通过采用预建立的基于深度学习的教练车目标检测模型对所述训练场地图像进行检测；若存在教练车，则通过所述机器学习模型对该教练车的多个移动位置进行计算及分析处理，以将该多个移动位置生成一行车路径；
所述机器学习模型用于通过计算及分析所述行车路径，以判断所述行车路径是否偏离预设路径；其中，所述机器学习模型通过多次教练车行车路径的学习及训练后固化该机器学习模型的参数而进行使用的；若 所述行车路径偏离，则通过所述机器学习模型根据所述行车路径偏离程度生成一矫正信息，并将该矫正信息发送至教练车上的接收模块；当所述教练车完成单次训练项目后，所述机器学习模型用于将该次教练车的实际行车路径数据进行存储；
教练车模块，包括接收模块和控制模块，所述接收模块用于将接收到的矫正信息发送至教练车中的控制模块；所述控制模块用于控制所述矫正信息在教练车内进行显示及播放，使得教练车内的学员根据矫正信息调整自身的驾驶动作，从而使行车路径通过学员调整而靠近所述预设路径；
所述矫正信息包括语音提示信息和视频演示信息，所述视频演示信息为所述教练车内方向盘预转动方向及角度的演示动画信息；
比对模块，用于将存储的所述实际行车路径与预设路径进行比对，生成一路径比对信息，并将该路径比对信息发送至终端，以供学员及时纠错。
本发明的一实施例中，所述机器学习模块用于在所述教练车完成所述训练项目的过程中，实时对所述教练车的行进速度进行监控，并判断所述行进速度值是否位于预设范围内；若所述行进速度超过所述预设范围，则生成一提示信息，并将该提示信息发送至所述教练车。
本发明的一实施例中，所述机器学习模型还用于当所述教练车行车路径偏移所述预设路径超过预设偏移量，且经所述机器学习模型识别该教练车处于危险状态时，则向该教练车发送刹车信号，以控制该教练车及时停车。
本发明的一实施例中，所述路经比对信息为教练车实际行车视频与预设虚拟教练车行车视频的比对视频信息。
本发明的一实施例中，所述训练场地所属类型包括：倒车入库训练项目、侧方停车训练项目、S弯训练项目和直角弯训练项目。
本发明的实施例提供的技术方案可以包括以下有益效果：
本发明的实施例中，根据上述提供的教练车智能辅助训练方法及系统，通过设置于训练场地上方的摄像头实时捕捉图像信息，并将图像信息上传至机器学习模型中，然后通过机器学习模型对教练车实际的行车 路径的计算及分析，得出实际行车路径与预设路径之间的偏移量，并且根据偏移量生成矫正信息，该矫正信息能够实时的在教练车内显示及播放，学员能够根据矫正信息内的视频演示直观的调整方向盘的转向及角度，同时教练车内还会进行语音提示，通过视频演示配合语音提示的共同作用，能够使得学员较好的矫正自身驾驶动作，该种智能辅助训练方法能够在一定程度上替代教练，为学员在练车过程中提供了智能的练车辅助，并且能够在一定程度上提高学员的训练水平。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分，示出了符合本公开的实施例，并与说明书一起用于解释本公开的原理。显而易见地，下面描述中的附图仅仅是本公开的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1示出本发明示例性实施例中教练车智能辅助训练方法流程图；
图2示出本发明示例性实施例中教练车智能辅助训练系统的模块示意图；
图3示出本发明示例性实施例中教练车显示屏中方向盘转向及角度示意图；
图4示出本发明示例性实施例中训练场地中教练车预设行进路径示意图；
图5示出本发明示例性实施例中训练场地中教练车实际行进路径示意图；
图6示出本发明示例性实施例中训练场地中教练车实际行进路径与预设路径比对路径示意图。
具体实施方式
现在将参考附图更全面地描述示例实施方式。然而，示例实施方式能够以多种形式实施，且不应被理解为限于在此阐述的范例；相反，提供这些实施方式使得本发明将更加全面和完整，并将示例实施方式的构 思全面地传达给本领域的技术人员。所描述的特征、结构或特性可以以任何合适的方式结合在一个或更多实施方式中。
此外，附图仅为本发明实施例的示意性图解，并非一定是按比例绘制。图中相同的附图标记表示相同或类似的部分，因而将省略对它们的重复描述。附图中所示的一些方框图是功能实体，不一定必须与物理或逻辑上独立的实体相对应。
本示例实施方式中首先提供了一种教练车智能辅助训练方法。参考图1中所示，该方法可以包括：
步骤S101，通过设置于训练场地上方的摄像头实时获取该训练场地图像信息，并将该训练场地图像信息上传至机器学习模型中。
步骤S102，通过所述机器学习模型中预建立的目标检测模型对所述训练场地图像进行检测，以判断所述训练场地所属类型，以及判断该训练场地图像中是否存在教练车；其中，所述机器学习模型通过采用预建立的基于深度学习的教练车目标检测模型对所述训练场地图像进行检测。
步骤S103，若存在教练车，则通过所述机器学习模型对该教练车的多个移动位置进行计算及分析处理，以将该多个移动位置生成一行车路径。
步骤S104，所述机器学习模型通过计算及分析所述行车路径，以判断所述行车路径是否偏离预设路径；其中，所述机器学习模型通过多次教练车行车路径的学习及训练后固化该机器学习模型的参数而进行使用的。
步骤S105，若所述行车路径偏离，则通过所述机器学习模型根据所述行车路径偏离程度生成一矫正信息，并将该矫正信息发送至教练车上的接收模块。
步骤S106，所述接收模块将接收到的矫正信息发送至教练车中的控制模块，该控制模块控制所述矫正信息在教练车内进行显示及播放，使得教练车内的学员根据矫正信息调整自身的驾驶动作，从而使行车路径通过学员调整而靠近所述预设路径；所述矫正信息包括语音提示信息和视频演示信息，所述视频演示信息为所述教练车内方向盘预转动方向 及角度的演示动画信息。
步骤S107，当所述教练车完成单次训练项目后，所述机器学习模型将该次教练车的实际行车路径数据进行存储。
步骤S108，将存储的所述实际行车路径与预设路径进行比对，生成一路径比对信息，并将该路径比对信息发送至终端，以供学员及时纠错。
根据上述提供的教练车智能辅助训练方法，通过设置于训练场地上方的摄像头实时捕捉图像信息，并将图像信息上传至机器学习模型中，然后通过机器学习模型对教练车实际的行车路径的计算及分析，得出实际行车路径与预设路径之间的偏移量，并且根据偏移量生成矫正信息，该矫正信息能够实时的在教练车内显示及播放，学员能够根据矫正信息内的视频演示直观的调整方向盘的转向及角度，同时教练车内还会进行语音提示，通过视频演示配合语音提示的共同作用，能够使得学员较好的矫正自身驾驶动作，该种智能辅助训练方法能够在一定程度上替代教练，为学员在练车过程中提供了智能的练车辅助，并且能够在一定程度上提高学员的训练水平。
下面，将参考图1至图6对本示例实施方式中的上述教练车智能辅助训练方法的各个部分进行更详细的说明。
在步骤S101中，通过设置于训练场地上方的摄像头实时获取该训练场地图像信息，并将该训练场地图像信息上传至机器学习模型中。
示例的，摄像头固定安装于训练场地的上方，当学员进行科目二项目训练时，将各训练场地上方的摄像头打开，摄像头便会实时获取训练场地的图像信息，并将获取到的训练场地图像信息上传至机器学习模型中，该处图像信息可理解为图片帧数据，但不限于此。机器学习模型首先读取数据，即对数据进行收集；并对数据提前进行处理和修正，如特征提取、特征降维、特征空值处理、特征转换、特征归一化、目标值空值处理、目标值转换等；在模型测试阶段，可先选择交叉验证方法，提前划分好数据；当数据处理好后，用该数据建立训练模型，对模型的评价有很多参数，例如，得分、查准率、查全率等。具体可参考现有技术进行理解，在此不再赘述。
在步骤S102中，通过所述机器学习模型中预建立的目标检测模型对所述训练场地图像进行检测，以判断所述训练场地所属类型，以及判断该训练场地图像中是否存在教练车；其中，所述机器学习模型通过采用预建立的基于深度学习的教练车目标检测模型对所述训练场地图像进行检测。
示例的，建立基于深度学习的教练车目标检测模型的步骤可以为：搭建教练车目标检测深度学习网络；教练车目标检测深度学习网络包括：由8个卷积层、8个ReLU激活层和3个池化层组成的教练车特征提取器，以及与所述教练车特征提取器连接的第一卷积层和第二卷积层；获取不同角度、光照及图像质量的训练场地图像样本；采用矩形框标记所述训练场地图像样本中的教练车开始训练区域位置，并记录所述矩形框的坐标信息；以每张所述训练场地图像样本和该训练场地图像样本中的矩形框的坐标信息为一组训练数据，生成教练车目标检测训练数据集；使用生成的所述教练车目标检测训练数据集训练所述教练车目标检测深度学习网络，参数更新后得到基于深度学习的教练车目标检测模型。具体可参考现有技术进行理解。该目标检测模型对训练场地图像检测完成后，将会得出该训练场地所属的类型，在一个示例中，所述训练场地所属类型包括：倒车入库训练项目、侧方停车训练项目、S弯训练项目和直角弯训练项目，上述每个项目都可通过本实施例提供的智能辅助训练方法进行训练。
在步骤S103中，若存在教练车，则通过所述机器学习模型对该教练车的多个移动位置进行计算及分析处理，以将该多个移动位置生成一行车路径。
示例的，机器学习模型可接收通过摄像头实时拍摄的训练场地图像信息，因训练场地存在教练车，且机器学习模型能够通过大量教练车正负样本的训练，从而固化其参数进行使用，正样本为想要正确分类出的类别所对应的样本，在本实施例中，正样本为图像信息中的教练车，负样本可以为除教练车外的训练场地，例如水泥地、草坪、黄线等等，此时机器学习模型对接收到的教练车的多个移动位置进行计算及分析处理，从而将多个移动位置生成一行车路径。例如，当学员在倒车入库场 地进行训练时，摄像头将拍摄的一帧帧图像依次上传至机器学习模块中，机器学习模型对接收的每一帧图像进行计算及处理，使得与上一帧及之前的图像形成行车路径，由此可看出，该行车路径从教练车开始倒车时便会生成，且随着教练车的移动而实时变化。
在步骤S104中，所述机器学习模型通过计算及分析所述行车路径，以判断所述行车路径是否偏离预设路径；其中，所述机器学习模型是通过多次教练车行车路径的学习及训练后固化该机器学习模型的参数而进行使用的。
示例的，机器学习模型是通过大量的教练车行车路径学习及训练后从而固化其参数进行使用的，可以理解为机器学习模型在经过大量样本的训练后，能够计算出教练车如何行驶才能够倒库成功，或者是在偏离原路线的情况下，如何调整方向盘的转向及角度使得教练车倒入车库，并将该种调整方式转化为语音，用于当学员在练车过程中辅助学员进行训练。如图4所示，当学员驾驶车辆进行如倒车入库训练时，机器学习模型将会实时的通过捕捉获得的图像信息计算和分析教练车的行车路径，并判断该行车路径是否偏离预设路径，此时的机器学习模型相当于驾校教练盯着学员练车一样，当学员的行车路径与常规路径不同时，教练将会进行适时的提醒学员进行方向的调整，以使教练车靠近常规路径。
在步骤S105中，若所述行车路径偏离，则通过所述机器学习模型根据所述行车路径偏离程度生成一矫正信息，并将该矫正信息发送至教练车上的接收模块。
示例的，若机器学习模型计算发现教练车行车路径偏离，或者是以该种行车路径继续行驶下去，该项训练科目将会失败时，该机器学习模型将会根据行车路径相对于预设路径的偏离程度生成一矫正信息，并将该矫正信息发送至教练车。例如，当学员在练习倒车入库项目时，由于过早打方向盘，若是以该种路径继续行驶下去，教练车肯定会压车库线，此时机器学习模型将会通过计算及分析生成一矫正信息，该矫正信息被发送至教练车上的接收模块中，用于提示学员进行方向的调整。
在步骤S106中，所述接收模块将接收到的矫正信息发送至教练车 中的控制模块，该控制模块控制所述矫正信息在教练车内进行显示及播放，使得教练车内的学员根据矫正信息调整自身的驾驶动作，从而使行车路径通过学员调整而靠近所述预设路径；所述矫正信息包括语音提示信息和视频演示信息，所述视频演示信息为所述教练车内方向盘预转动方向及角度的演示动画信息。
示例的，教练车中设置有控制模块，用于控制教练车内的显示屏进行矫正信息的显示，以及控制车内的喇叭进行语音提示。例如，当学员在练习倒车入库项目时，由于过早打方向盘，若是以该种路径继续行驶下去，车肯定会压车库线，从而导致科目训练失败，此时机器学习模型将会通过计算及分析生成一矫正信息，该矫正信息被发送至教练车上的接收模块，接收模块将矫正信息发送至车内的控制模块，控制模块将控制显示屏将该矫正信息进行显示，如图3所示，显示的内容包括矫正信息中的视频演示信息，该视频演示信息具体为教练车内方向盘预转动方向及角度的演示动画信息，并且可用箭头标识方向盘转动的方向，通过显示屏中视频的显示，能够更直观的向学员显示如何操作方向盘，提高学员的训练效率；同时教练车将会根据矫正信息中的语音提示信息发出语音提醒，如“回盘半圈”，当按照该路径行驶一段距离后，机器学习模型又会通过计算及分析生成另一矫正信息并发送至教练车，此时显示屏将会显示不同角度及转动方向的方向盘视频动画，同时配合语音提醒，如“往右打死”，通过显示屏中视频的提示以及语音的提醒，直至学员将教练车正确停到车库中。需要说明的是，矫正信息中的语音提示信息内容可根据实际情况进行选择设置，不限于本实施例中所述。
在步骤S107、S108中，当所述教练车完成单次训练项目后，所述机器学习模型将该次教练车的实际行车路径数据进行存储；将存储的所述实际行车路径与预设路径进行比对，生成一路径比对信息，并将该路径比对信息发送至终端，以供学员及时纠错。
示例的，如图5、6所示，当学员每完成一次训练项目后，机器学习模型中的存储模块便会将该次训练车子的实际行车路径进行存储，并且将实际行车路径与预设路径进行比对，生成一路径比对信息，在一个示例中，所述路经比对信息为教练车实际行车视频与预设虚拟教练车行 车视频的比对视频信息，具体的，设置于训练场地上方的摄像头能够对教练车训练项目的实际路径进行拍摄，并将该视频进行存储，该视频即为教练车实际行车视频，可通过比对模块将实际行车视频与预设虚拟行车视频进行比对，并生成一比对视频，并将该比对视频信息发送至终端，例如发送至学员的手机，学员可通过观看路径比对视频更加直观的看出自己在训练中所存在的问题，以利于学员的快速进步。
在步骤S103后，该训练方法还包括：
步骤S1031，在所述教练车完成所述训练项目的过程中，实时对所述教练车的行进速度进行监控，并判断所述行进速度值是否位于预设范围内。
步骤S1032，若所述行进速度超过所述预设范围，则生成一提示信息，并将该提示信息发送至所述教练车。
示例的，在科目二的考试中，尤其是倒车入库及侧方停车，车速越慢，学员越容易对车子状态进行调整，因此，在完成训练项目的过程中，可实时对所述教练车的行进速度进行监控，并判断所述行进速度值是否位于预设范围内，如教练车的行进速度大于5km/h时，教练车将会语音提示学员减慢速度，具体预设范围可根据实际情况进行设置，在此不作限制。
在步骤S104后，该训练方法还包括：
步骤S1041，当所述教练车行车路径偏移所述预设路径超过预设偏移量，且经所述机器学习模型识别该教练车处于危险状态时，则向该教练车发送刹车信号，以控制该教练车及时停车。
示例的，当学员在练车的过程中，为防止意外发生，如学员不小心踩到刹车使教练车失控等情况的发生，即教练车偏移预设路径过多，或是偏移不正常，且通过机器学习模型的判断认为该教练车处于危险状态时，将会向教练车中的接收模块发送刹车信号，教练车内的刹车模块将会接收到接收模块发送的刹车信号，从而控制该教练车停下，以保证学员在练车过程中的安全。
本示例实施方式中还提供了一种教练车智能辅助训练系统。参考图2中所示，该系统可以包括摄像头模块、机器学习模型、教练车模块和 比对模块。
所述摄像头模块设置于训练场地上方，用于实时获取训练场地图像信息，并将该训练场地图像信息上传至机器学习模型中。
所述机器学习模型用于通过所述机器学习模型中预建立的目标检测模型对所述训练场地图像进行检测，以判断所述训练场地所属类型，以及判断该训练场地图像中是否存在教练车；其中，所述机器学习模型通过采用预建立的基于深度学习的教练车目标检测模型对所述训练场地图像进行检测；若存在教练车，则通过所述机器学习模型对该教练车的多个移动位置进行计算及分析处理，以将该多个移动位置生成一行车路径；
所述机器学习模型用于通过计算及分析所述行车路径，以判断所述行车路径是否偏离预设路径；其中，所述机器学习模型通过多次教练车行车路径的学习及训练后固化该机器学习模型的参数而进行使用的；若所述行车路径偏离，则通过所述机器学习模型根据所述行车路径偏离程度生成一矫正信息，并将该矫正信息发送至教练车上的接收模块；当所述教练车完成单次训练项目后，所述机器学习模型用于将该次教练车的实际行车路径数据进行存储。
所述教练车模块用于将接收到的矫正信息发送至教练车中的控制模块；所述控制模块用于控制所述矫正信息在教练车内进行显示及播放，使得教练车内的学员根据矫正信息调整自身的驾驶动作，从而使行车路径通过学员调整而靠近所述预设路径；所述矫正信息包括语音提示信息和视频演示信息，所述视频演示信息为所述教练车内方向盘预转动方向及角度的演示动画信息。
所述比对模块用于将存储的所述实际行车路径与预设路径进行比对，生成一路径比对信息，并将该路径比对信息发送至终端，以供学员及时纠错。
具体实施方式可参考上述实施例进行理解，在此不再赘述。
在一个示例中，所述机器学习模块用于在所述教练车完成所述训练项目的过程中，实时对所述教练车的行进速度进行监控，并判断所述行进速度值是否位于预设范围内；若所述行进速度超过所述预设范围，则 生成一提示信息，并将该提示信息发送至所述教练车。
在一个示例中，所述机器学习模型还用于当所述教练车行车路径偏移所述预设路径超过预设偏移量，且经所述机器学习模型识别该教练车处于危险状态时，则向该教练车发送刹车信号，以控制该教练车及时停车。
在一个示例中，所述路经比对信息为教练车实际行车视频与预设虚拟教练车行车视频的比对视频信息。
在一个示例中，所述训练场地所属类型包括：倒车入库训练项目、侧方停车训练项目、S弯训练项目和直角弯训练项目。
根据上述提供的教练车智能辅助训练方法及系统，通过设置于训练场地上方的摄像头实时捕捉图像信息，并将图像信息上传至机器学习模型中，然后通过机器学习模型对教练车实际的行车路径的计算及分析，得出实际行车路径与预设路径之间的偏移量，并且根据偏移量生成矫正信息，该矫正信息能够实时的在教练车内显示及播放，学员能够根据矫正信息内的视频演示直观的调整方向盘的转向及角度，同时教练车内还会进行语音提示，通过视频演示配合语音提示的共同作用，能够使得学员较好的矫正自身驾驶动作，该种智能辅助训练方法能够在一定程度上替代教练，为学员在练车过程中提供了智能的练车辅助，并且能够在一定程度上提高学员的训练水平。
需要理解的是，上述描述中的术语“中心”、“纵向”、“横向”、“长度”、“宽度”、“厚度”、“上”、“下”、“前”、“后”、“左”、“右”、“竖直”、“水平”、“顶”、“底”“内”、“外”、“顺时针”、“逆时针”等指示的方位或位置关系为基于附图所示的方位或位置关系，仅是为了便于描述本发明实施例和简化描述，而不是指示或暗示所指的装置或元件必须具有特定的方位、以特定的方位构造和操作，因此不能理解为对本发明实施例的限制。
此外，术语“第一”、“第二”仅用于描述目的，而不能理解为指示或暗示相对重要性或者隐含指明所指示的技术特征的数量。由此，限定有“第一”、“第二”的特征可以明示或者隐含地包括一个或者更多个该特征。在本发明实施例的描述中，“多个”的含义是两个或两个以上， 除非另有明确具体的限定。
在本发明实施例中，除非另有明确的规定和限定，术语“安装”、“相连”、“连接”、“固定”等术语应做广义理解，例如，可以是固定连接，也可以是可拆卸连接，或成一体；可以是机械连接，也可以是电连接；可以是直接相连，也可以通过中间媒介间接相连，可以是两个元件内部的连通或两个元件的相互作用关系。对于本领域的普通技术人员而言，可以根据具体情况理解上述术语在本发明中的具体含义。
在本发明实施例中，除非另有明确的规定和限定，第一特征在第二特征之“上”或之“下”可以包括第一和第二特征直接接触，也可以包括第一和第二特征不是直接接触而是通过它们之间的另外的特征接触。而且，第一特征在第二特征“之上”、“上方”和“上面”包括第一特征在第二特征正上方和斜上方，或仅仅表示第一特征水平高度高于第二特征。第一特征在第二特征“之下”、“下方”和“下面”包括第一特征在第二特征正下方和斜下方，或仅仅表示第一特征水平高度小于第二特征。
在本说明书的描述中，参考术语“一个实施例”、“一些实施例”、“示例”、“具体示例”、或“一些示例”等的描述意指结合该实施例或示例描述的具体特征、结构、材料或者特点包含于本发明的至少一个实施例或示例中。在本说明书中，对上述术语的示意性表述不必须针对的是相同的实施例或示例。而且，描述的具体特征、结构、材料或者特点可以在任何的一个或多个实施例或示例中以合适的方式结合。此外，本领域的技术人员可以将本说明书中描述的不同实施例或示例进行接合和组合。
本领域技术人员在考虑说明书及实践这里公开的发明后，将容易想到本发明的其它实施方案。本申请旨在涵盖本发明的任何变型、用途或者适应性变化，这些变型、用途或者适应性变化遵循本发明的一般性原理并包括本发明未公开的本技术领域中的公知常识或惯用技术手段。说明书和实施例仅被视为示例性的，本发明的真正范围和精神由所附的权利要求指出。

Claims (12)

  1. 一种教练车智能辅助训练方法，其特征在于，该方法包括：
    通过设置于训练场地上方的摄像头实时获取该训练场地图像信息，并将该训练场地图像信息上传至机器学习模型中；
    通过所述机器学习模型中预建立的目标检测模型对所述训练场地图像进行检测，以判断所述训练场地所属类型，以及判断该训练场地图像中是否存在教练车；
    其中，所述机器学习模型通过采用预建立的基于深度学习的教练车目标检测模型对所述训练场地图像进行检测；
    若存在教练车，则通过所述机器学习模型对该教练车的多个移动位置进行计算及分析处理，以将该多个移动位置生成一行车路径；
    所述机器学习模型通过计算及分析所述行车路径，以判断所述行车路径是否偏离预设路径；
    其中，所述机器学习模型是通过多次教练车行车路径的学习及训练后固化该机器学习模型的参数而进行使用的；
    若所述行车路径偏离，则通过所述机器学习模型根据所述行车路径偏离程度生成一矫正信息，并将该矫正信息发送至教练车上的接收模块；
    所述接收模块将接收到的矫正信息发送至教练车中的控制模块，该控制模块控制所述矫正信息在教练车内进行显示及播放，使得教练车内的学员根据矫正信息调整自身的驾驶动作，从而使行车路径通过学员调整而靠近所述预设路径；
    所述矫正信息包括语音提示信息和视频演示信息，所述视频演示信息为所述教练车内方向盘预转动方向及角度的演示动画信息。
  2. 根据权利要求1所述教练车智能辅助训练方法，其特征在于，该方法还包括：
    在所述教练车完成所述训练项目的过程中，实时对所述教练车的行进速度进行监控，并判断所述行进速度值是否位于预设范围内；
    若所述行进速度超过所述预设范围，则生成一提示信息，并将该提 示信息发送至所述教练车。
  3. 根据权利要求2所述教练车智能辅助训练方法，其特征在于，该方法还包括：
    当所述教练车行车路径偏移所述预设路径超过预设偏移量，且经所述机器学习模型识别该教练车处于危险状态时，则向该教练车发送刹车信号，以控制该教练车及时停车。
  4. 根据权利要求1所述教练车智能辅助训练方法，其特征在于，该方法还包括：
    当所述教练车完成单次训练项目后，所述机器学习模型将该次教练车的实际行车路径数据进行存储；
    将存储的所述实际行车路径与预设路径进行比对，生成一路径比对信息，并将该路径比对信息发送至终端，以供学员及时纠错。
  5. 根据权利要求4所述教练车智能辅助训练方法，其特征在于，所述路经比对信息为教练车实际行车视频与预设虚拟教练车行车视频的比对视频信息。
  6. 根据权利要求1所述教练车智能辅助训练方法，其特征在于，所述训练场地所属类型包括：倒车入库训练项目、侧方停车训练项目、S弯训练项目和直角弯训练项目。
  7. 一种教练车智能辅助训练系统，其特征在于，该系统包括：
    摄像头模块，设置于训练场地上方，用于实时获取训练场地图像信息，并将该训练场地图像信息上传至机器学习模型中；
    机器学习模型，用于通过所述机器学习模型中预建立的目标检测模型对所述训练场地图像进行检测，以判断所述训练场地所属类型，以及判断该训练场地图像中是否存在教练车；其中，所述机器学习模型通过采用预建立的基于深度学习的教练车目标检测模型对所述训练场地图像进行检测；若存在教练车，则通过所述机器学习模型对该教练车的多个移动位置进行计算及分析处理，以将该多个移动位置生成一行车路径；
    所述机器学习模型用于通过计算及分析所述行车路径，以判断所述行车路径是否偏离预设路径；其中，所述机器学习模型通过多次教练车 行车路径的学习及训练后固化该机器学习模型的参数而进行使用的；若所述行车路径偏离，则通过所述机器学习模型根据所述行车路径偏离程度生成一矫正信息，并将该矫正信息发送至教练车上的接收模块；当所述教练车完成单次训练项目后，所述机器学习模型用于将该次教练车的实际行车路径数据进行存储；
    教练车模块，包括接收模块和控制模块，所述接收模块用于将接收到的矫正信息发送至教练车中的控制模块；所述控制模块用于控制所述矫正信息在教练车内进行显示及播放，使得教练车内的学员根据矫正信息调整自身的驾驶动作，从而使行车路径通过学员调整而靠近所述预设路径；
    所述矫正信息包括语音提示信息和视频演示信息，所述视频演示信息为所述教练车内方向盘预转动方向及角度的演示动画信息。
  8. 根据权利要求7所述教练车智能辅助训练方法，其特征在于，所述机器学习模块用于在所述教练车完成所述训练项目的过程中，实时对所述教练车的行进速度进行监控，并判断所述行进速度值是否位于预设范围内；若所述行进速度超过所述预设范围，则生成一提示信息，并将该提示信息发送至所述教练车。
  9. 根据权利要求8所述教练车智能辅助训练方法，其特征在于，所述机器学习模型还用于当所述教练车行车路径偏移所述预设路径超过预设偏移量，且经所述机器学习模型识别该教练车处于危险状态时，则向该教练车发送刹车信号，以控制该教练车及时停车。
  10. 根据权利要求7所述教练车智能辅助训练方法，其特征在于，该系统还包括：
    比对模块，用于将存储的实际行车路径与预设路径进行比对，生成一路径比对信息，并将该路径比对信息发送至终端，以供学员及时纠错
  11. 根据权利要求10所述教练车智能辅助训练方法，其特征在于，所述路经比对信息为教练车实际行车视频与预设虚拟教练车行车视频的比对视频信息。
  12. 根据权利要求7所述教练车智能辅助训练方法，其特征在于，所述训练场地所属类型包括：倒车入库训练项目、侧方停车训练项目、 S弯训练项目和直角弯训练项目。
PCT/CN2020/105268 2020-07-20 2020-07-28 一种教练车智能辅助训练方法及系统 WO2022016583A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
SE2051569A SE2051569A1 (en) 2020-07-20 2020-07-28 Intelligent training aiding method and system for instructional cars

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010696315.XA CN111667742A (zh) 2020-07-20 2020-07-20 一种教练车智能辅助训练方法及系统
CN202010696315.X 2020-07-20

Publications (1)

Publication Number Publication Date
WO2022016583A1 true WO2022016583A1 (zh) 2022-01-27

Family

ID=72392693

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/105268 WO2022016583A1 (zh) 2020-07-20 2020-07-28 一种教练车智能辅助训练方法及系统

Country Status (5)

Country Link
CN (1) CN111667742A (zh)
AU (1) AU2020104282A4 (zh)
SE (1) SE2051569A1 (zh)
WO (1) WO2022016583A1 (zh)
ZA (1) ZA202007882B (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114822169A (zh) * 2022-05-06 2022-07-29 辽宁科技大学 一种教练车驾驶辅助练习方法及装置
GB2620025A (en) * 2022-06-23 2023-12-27 Motional Ad Llc Generating corrected future maneuver parameters in a planner

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112201113A (zh) * 2020-10-21 2021-01-08 广东星唯信息技术有限公司 一种采用rtk测绘技术监控教练车倒车入库的方法
CN114399932A (zh) * 2021-12-28 2022-04-26 南京财经大学 一种基于智能学习的人才培养仿真实训系统
CN115578914B (zh) * 2022-11-23 2023-03-14 湖南视觉伟业智能科技有限公司 一种驾车训练时倒车入位用训练系统及方法
CN116645848A (zh) * 2023-04-17 2023-08-25 武汉未来幻影科技有限公司 一种车辆运行的控制方法以及相关设备
CN116729422B (zh) * 2023-06-07 2024-03-08 广州市德赛西威智慧交通技术有限公司 车辆轨迹的偏离修正方法、车辆驾驶辅助方法及设备

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170193842A1 (en) * 2012-03-06 2017-07-06 State Farm Mutual Automobile Insurance Company Online System For Training Novice Drivers And Rating Insurance Products
CN108109481A (zh) * 2017-12-21 2018-06-01 云南冲浪科技有限公司 一种机动车驾驶人培训智能教练系统
CN108389477A (zh) * 2018-03-05 2018-08-10 广州星唯信息科技有限公司 一种驾培场地训练的纠偏指导方法
CN108806371A (zh) * 2018-08-31 2018-11-13 成都杏仁网络科技有限公司 一种基于驾考科目训练的智能判定方法及系统
CN208376741U (zh) * 2017-03-24 2019-01-15 多伦科技股份有限公司 一种机器人智能教练辅助驾驶系统
CN109272821A (zh) * 2018-10-22 2019-01-25 广州星唯信息科技有限公司 一种基于高精度视觉定位的场地驾驶评判方法
CN111047948A (zh) * 2019-11-27 2020-04-21 康忠文 一种学车辅助方法及系统

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105590449B (zh) * 2015-12-29 2017-12-12 南京邮电大学 一种教练车行驶监控系统及监控方法
CN110599853B (zh) * 2019-08-05 2022-07-05 深圳华桥智能设备科技有限公司 驾校智能教学系统及方法

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170193842A1 (en) * 2012-03-06 2017-07-06 State Farm Mutual Automobile Insurance Company Online System For Training Novice Drivers And Rating Insurance Products
CN208376741U (zh) * 2017-03-24 2019-01-15 多伦科技股份有限公司 一种机器人智能教练辅助驾驶系统
CN108109481A (zh) * 2017-12-21 2018-06-01 云南冲浪科技有限公司 一种机动车驾驶人培训智能教练系统
CN108389477A (zh) * 2018-03-05 2018-08-10 广州星唯信息科技有限公司 一种驾培场地训练的纠偏指导方法
CN108806371A (zh) * 2018-08-31 2018-11-13 成都杏仁网络科技有限公司 一种基于驾考科目训练的智能判定方法及系统
CN109272821A (zh) * 2018-10-22 2019-01-25 广州星唯信息科技有限公司 一种基于高精度视觉定位的场地驾驶评判方法
CN111047948A (zh) * 2019-11-27 2020-04-21 康忠文 一种学车辅助方法及系统

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114822169A (zh) * 2022-05-06 2022-07-29 辽宁科技大学 一种教练车驾驶辅助练习方法及装置
CN114822169B (zh) * 2022-05-06 2023-06-09 辽宁科技大学 一种教练车驾驶辅助练习方法及装置
GB2620025A (en) * 2022-06-23 2023-12-27 Motional Ad Llc Generating corrected future maneuver parameters in a planner

Also Published As

Publication number Publication date
SE2051569A1 (en) 2022-01-27
ZA202007882B (en) 2021-09-29
CN111667742A (zh) 2020-09-15
AU2020104282A4 (en) 2021-04-15

Similar Documents

Publication Publication Date Title
WO2022016583A1 (zh) 一种教练车智能辅助训练方法及系统
US10255821B2 (en) Driver training
CN106710360B (zh) 基于增强虚拟现实人机交互的智能驾驶训练系统及方法
US20230306866A1 (en) Pre-License Development Tool
CN105741643B (zh) 用于培训驾驶技术的自动化实现方法、装置和系统
CN108806371B (zh) 一种基于驾考科目训练的智能判定方法及系统
CN111402671A (zh) 一种驾校车语音视频教学方法及装置
CN102096930A (zh) 一种教学智能录播系统学生起坐检测方法
CN111798718A (zh) 一种基于虚拟现实的驾驶适应性训练方法、主机及装置
JP6086515B1 (ja) 運転技量評価装置、サーバ装置、運転技量評価システム、プログラムおよび運転技量評価方法
CN110619767A (zh) 智能教育机器人及计算机可读介质
CN109637289A (zh) 曲线行驶科目的自动评分装置及方法
CN109215419A (zh) 教育机器人及体验式教学系统
JP2012155285A (ja) 運転危険予測学習支援システム
CN112801526B (zh) 基于生物识别交叉认证的驾培作弊监管系统和方法
KR101408325B1 (ko) 열차 시뮬레이터 장치, 열차 시뮬레이터 제공방법 및 그 기록매체
CN113506196A (zh) 一种基于单向视频交互的模拟驾驶训练课程制作方法
JP5971611B2 (ja) 教習支援装置
CN113556507A (zh) 一种培训用车载智能监控系统及其监控方法
CN113052041A (zh) 一种具备智能化教学系统和监测方法
CN114093219B (zh) 小型汽车驾驶证考试科目二的辅助训练系统及方法
TWI731577B (zh) 學習狀態改善管理系統
US20230368698A1 (en) Auto Bot Driver Training System, Vehicle, and Method of Use
CN114694394B (zh) 一种特种车辆智能提示方法及装置
CN113539008B (zh) 基于ar的便携式城市轨道交通车辆驾驶模拟系统及方法

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: 20946175

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: 20946175

Country of ref document: EP

Kind code of ref document: A1