A driver defensive driving ability training method and system based on line-of-sight tracking

Abstract

This invention provides a method for training a driver's defensive driving ability based on eye-tracking, comprising the following steps: Site data preparation step: establishing three-dimensional site data containing at least one defensive driving training scenario marker; Eye-tracking correlation analysis step: collecting the trainee's eye-tracking data in real time using an eye-tracking device, and correlating and matching the eye-tracking data with the three-dimensional site data to determine the positional relationship between the trainee's eye-tracking focus area in the real site and the defensive driving training scenario; Weakness assessment step: quantitatively assessing the trainee's defensive driving ability based on the analysis results of the eye-tracking correlation analysis step, and identifying at least one driving weakness; Scheme generation step: generating a personalized training scheme containing a target training scenario based on the driving weakness; Navigation training step: guiding the trainee to the target training scenario for intensive training using a navigation guidance module according to the personalized training scheme.

Description

Technical Field

[0001] This invention relates to the field of driver training technology, and more specifically, to a method and system for training drivers' defensive driving abilities based on eye tracking. Background Technology

[0002] Current driving school training generally suffers from the pain point of "emphasizing subject-based testing while neglecting the development of defensive driving skills," resulting in graduates lacking the ability to anticipate and avoid risks in complex road conditions after graduation. Traditional training relies heavily on the instructor's personal experience, which has the following technical shortcomings: First, the ability assessment is highly subjective, lacking quantitative data support, making it difficult to objectively evaluate key safe driving behavior indicators such as the distribution of the student's visual focus and the timing of risk prediction; second, the training program lacks specificity, failing to personalize adjustments based on different students' visual habits and cognitive weaknesses, leading to low training efficiency; third, the teaching guidance is delayed, unable to provide timely and accurate intervention and correction when students make visual deviations or operational errors.

[0003] Existing driver assistance training systems primarily focus on collecting and analyzing vehicle operation data related to subject examinations. They have not effectively incorporated eye-tracking technology for quantitative assessment of defensive driving abilities, nor have they been able to correlate trainees' visual behavior with real-time risk scenarios in driving school environments. This makes it difficult to meet the needs of precise and personalized defensive driving training. Therefore, there is an urgent need for a technological solution that can achieve objective assessment, precise positioning, targeted training, and support real-time guidance. Summary of the Invention

[0004] The present invention aims to provide a driver defensive driving ability training method and system based on eye tracking, so as to solve or improve the problems of subjective assessment of defensive driving ability, lack of targeted training and lagging guidance in the prior art.

[0005] To achieve the above objectives, the present invention provides a method for training a driver's defensive driving ability based on eye tracking, comprising the following steps: Field data preparation steps: Establish 3D field data containing at least one defensive driving training scenario marker; Line-of-sight correlation analysis steps: The line-of-sight data of trainees is collected in real time through line-of-sight tracking devices, and the line-of-sight data is correlated and matched with three-dimensional field data to determine the positional relationship between the focal area of ​​the trainee's line of sight in the real field and the defensive driving training scenario. Weakness assessment steps: Based on the analysis results of the line-of-sight correlation analysis step, quantitatively assess the trainee's defensive driving ability and identify at least one driving weakness; Solution generation steps: Based on driving weaknesses, generate personalized training solutions that include target training scenarios; Navigation training steps: Based on the personalized training plan, the navigation guidance module guides trainees to the target training scenario for intensive training.

[0006] Further steps in preparing site data include: The three-dimensional point cloud data of the driving school site was collected by laser scanning equipment, and a 1:1 scale three-dimensional site model was generated by combining the image with optimization. The location coordinates, scenario types, training objectives, and key observation areas of key defensive driving training scenarios are marked on a 3D site model to form a scenario database.

[0007] Further, the line-of-sight correlation analysis steps include: The collected gaze data is pre-processed to remove noise and eliminate abnormal data caused by blinking or head movement. The preprocessed line-of-sight data is fused with the real-time vehicle location data obtained through environmental sensors; Based on pre-completed coordinate calibration parameters, the fused data is mapped to the coordinate system of the 3D site data to determine whether the trainee's line of sight is a defensive critical area of ​​the current scene.

[0008] Furthermore, the line-of-sight correlation analysis steps also include: The percentage of coverage of key areas is the proportion of time that trainees spend looking at key defensive areas out of the total training time. The statistical risk target focusing delay time is the delay time from when the environmental sensor detects the risk target to when the trainee's line of sight first focuses on the target; Based on preset thresholds including a key area coverage percentage threshold and / or a risk target focusing delay time threshold, the rationality of the line of sight prediction is determined and specific shortcomings are identified.

[0009] Furthermore, the weakness assessment step also includes an attention analysis sub-step, which includes: Based on line-of-sight data and current driving scenario information, an attention assessment index is calculated. The attention assessment index includes at least one of the following: the frequency of line-of-sight focus switching per unit time, the continuity of line-of-sight dwelling in key areas, and the ratio of line-of-sight dwelling in key areas to non-key areas. By combining the differentiated qualification standards for current driving scenarios, the system assesses the matching degree between the student's attention and the driving scenario, and generates attention analysis results that include attention scores and weaknesses.

[0010] Furthermore, the weakness assessment step also includes a comprehensive assessment sub-step, including: Based on preset indicators and corresponding weights, combined with the results of gaze correlation analysis, attention analysis, driving operation data, and site environment data, a comprehensive score for defensive driving ability is calculated and classified into levels; an evaluation report is generated that includes the comprehensive score, individual indicator scores, descriptions of shortcomings, and suggestions for training scenarios.

[0011] Furthermore, the solution generation steps include: Extract the trainees' driving weaknesses and shortcomings from the assessment report; Select training scenarios and attention enhancement sub-scenarios from the scenario database that match the weaknesses; Configure training intensity, training duration, and passing standards according to the severity of the weaknesses; Generate personalized training plans that include training objectives, training scenarios, training focus, and real-time guidance rules.

[0012] Further, the navigation training steps include: The optimal route from the vehicle's current location to the target training scene is planned based on 3D site data, and navigation is provided through display and voice. During the student's driving process, line-of-sight data and vehicle driving data are continuously collected and compared in real time with preset passing standards; When a student's driving behavior is detected to deviate from the training focus of the personalized training program, real-time guidance instructions are issued via the in-vehicle interactive display module in the form of voice and / or visual means to correct it.

[0013] Furthermore, the navigation training step is followed by a loop iteration step, including: Repeat the line-of-sight correlation analysis, weakness assessment, solution generation, and navigation training steps until the trainee's assessment results meet the preset comprehensive assessment qualification standards. The overall assessment qualification criteria include: the overall score for defensive driving ability reaches the first threshold, and the scores for each individual indicator reach the second threshold.

[0014] The present invention also provides a driver defensive driving ability training system based on eye tracking, which is used for a driver defensive driving ability training method based on eye tracking. It includes a field scene construction module, which is used to establish a 1:1 scale three-dimensional field model through laser scanning and image optimization, and mark key scenes for defensive driving training to form a scene database. The data acquisition module includes an in-vehicle eye-tracking device, driving operation sensors, and environmental sensors, used to collect the trainee's eye-tracking data, driving operation data, and site environment data. The line-of-sight analysis module is used to correlate and match line-of-sight data with three-dimensional site data, calculate the coverage ratio of key areas and the focusing delay time of risk targets, and locate defensive weaknesses. The attention analysis module is used to calculate attention evaluation indicators based on gaze data and driving scenario information, determine the matching degree between attention and scenario, and generate attention analysis results. The defensive driving ability assessment module is used to comprehensively assess the learner's defensive driving ability based on a preset indicator system and generate an assessment report. The personalized training module is used to match the weak training scenarios with the evaluation report and generate personalized training plans. The vehicle navigation guidance module is used to plan the optimal route to the target training scene based on 3D site data and provide navigation guidance. The in-vehicle interactive display module is used to display information and provide guidance commands in a voice and visual manner.

[0015] The present invention also provides a driver defensive driving ability training system based on eye tracking, for performing the above method, including: a site scene construction module, a data acquisition module, an eye tracking analysis module, an attention analysis module, a defensive driving ability assessment module, a personalized training module, an in-vehicle navigation guidance module, and an in-vehicle interactive display module.

[0016] Compared with existing technologies, the beneficial effects of this invention are as follows: by constructing a "real-time correlation mechanism between line-of-sight data and three-dimensional defensive scenario" and a closed-loop training logic of "weakness location - scenario matching - navigation guidance - reinforcement training", it realizes objective assessment of defensive driving ability based on real driving school sites, accurate weakness location and targeted reinforcement training, effectively overcomes the technical problems of subjective assessment, generalized training and lagging guidance in traditional training, and significantly improves the scientificity and effectiveness of driver training. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a specific embodiment of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will be described in detail below with reference to specific embodiments. It should be understood that these embodiments are only used to explain the present invention and are not intended to limit the scope of the present invention.

[0019] An embodiment of the present invention provides a method for training a driver's defensive driving ability based on eye tracking, executed by a training system integrated into the onboard system of a driving school training vehicle. This system includes the following modules: a scenario construction module, a data acquisition module (including an onboard eye tracking device, driving operation sensors, and environmental sensors), an eye tracking analysis module, an attention analysis module, a defensive driving ability assessment module, a personalized training module, an onboard navigation guidance module, and an onboard interactive display module (using an onboard tablet computer). The modules communicate with each other via a CAN bus, constructing a closed-loop architecture of "data acquisition - analysis and evaluation - scheme generation - navigation training". This method achieves closed-loop training of defensive driving ability through the following steps: Step 1: Site Data Preparation This step is performed by the site scene construction module. First, laser scanning equipment and high-definition cameras are used to collect comprehensive data of the driving school site, generating high-precision 3D point cloud data. This data is then combined with image optimization to generate a 1:1 scale 3D site model, ensuring coordinate errors are ≤5cm. Next, various key defensive driving training scenarios, such as intersections, curves, and school zones, are marked on the model. For each scenario, detailed information is recorded, including boundary coordinates, core traffic element coordinates, risk prediction key area coordinates, and defensive observation requirements for each key area.

[0020] Taking the risk prediction scenario at intersections as an example, the marked content includes: ① Scene boundary localization: Mark the overall boundary coordinates of the intersection; ② Core traffic element marking: Mark the three-dimensional coordinates and types of elements such as road centerline, stop line, guide lane line, pedestrian crossing, traffic lights, and signs; ③ Risk prediction key area marking: highlight the observation areas for oncoming vehicles in each approach lane, the pedestrian entry and exit areas at both ends of the crosswalk, and the blind spots at intersection corners, and record the defensive observation requirements for each area; ④ Auxiliary reference markers: Mark the three-dimensional coordinates of auxiliary facilities such as guide lines and speed bumps. All marker information together constitutes a structured scene database.

[0021] Step 2: System Integration, Debugging, and Student Calibration The specific methods for system integration and debugging are as follows: ① Hardware integration and installation: Fix the vehicle-mounted eye-tracking device, driving operation sensor, GPS module, millimeter-wave radar, camera and vehicle-mounted interactive display terminal in the designated position of the training vehicle, and complete the power supply connection and data communication between each hardware and the vehicle main control unit through the CAN bus to ensure that the hardware is unobstructed and the data transmission is stable; ② Software deployment initialization: Burn the 3D site data and defensive scene database to the vehicle storage unit, initialize the sampling frequency of each sensor (120Hz for eye tracking device, 10Hz for GPS, 20Hz for millimeter-wave radar, and 50Hz for operation sensor), and load the eye-site scene association algorithm and attention evaluation model to the vehicle processor. ③ Joint debugging, verification and correction: The system is started to perform a full module self-test. The CRC cyclic redundancy check algorithm is used to ensure the integrity of CAN bus data transmission. The Kalman filter algorithm is used to fuse GPS and millimeter-wave radar position data to suppress sensor noise and drift error. The ICP iterative nearest point algorithm is used to complete the alignment of the vehicle local coordinate system and the site three-dimensional global coordinate system to ensure that the coordinate error is ≤5cm. ④ Threshold configuration verification: Write preset thresholds such as critical area coverage ≥60%, risk focus delay ≤1.5s, and curve line-of-sight switching ≤2 times / minute, and test navigation, voice, and interaction functions to ensure command response delay ≤100ms.

[0022] The eye-tracking calibration method is as follows: The trainee sits upright in a driving posture, and the in-vehicle interactive terminal displays a 9-point calibration dot matrix. The trainee gazes at the dot matrix sequentially as prompted. The in-vehicle eye-tracking device collects the relative deviation data between the pupil center and the corneal reflection point. Based on the collected data, the system constructs a personal eye feature model and establishes a mapping relationship between the pupil offset and the gaze direction vector. The personal gaze model is bound to the vehicle coordinate system to complete the spatial transformation from the gaze point to the three-dimensional field coordinates. Calibration points are randomly displayed for accuracy verification to ensure that the gaze point error is ≤2°. If it fails, recalibration is performed. The implementation principle is infrared eye feature localization combined with perspective projection mapping.

[0023] The vehicle position matching method is as follows: the GPS module obtains the initial latitude and longitude of the vehicle and matches it to the corresponding position in the three-dimensional field; the millimeter-wave radar detects fixed markers in the field and corrects the GPS positioning error; the Kalman filter algorithm is used to fuse multi-source position data and output the vehicle's three-dimensional coordinates and attitude in real time; the vehicle continuously matches scene markers while moving and dynamically calibrates its position to keep the error ≤5cm.

[0024] Step 3: Line-of-sight correlation analysis The trainee selects an initial training exercise (e.g., "maintaining a safe following distance") and begins driving. The data acquisition module collects real-time data on the trainee's line of sight, driving operation, and the surrounding environment. The workflow of the line-of-sight analysis module is as follows: ① Noise reduction: Median filtering algorithm is used to remove abnormal visual data caused by blinking or violent head shaking to ensure data validity; ②Association matching: Based on the calibration parameters of the trainee's eyes and the three-dimensional coordinate system of the field completed before training, the line of sight data (relative deviation between pupil center and corneal reflection point) is mapped into a spatial direction vector, and then combined with the real-time vehicle position data to match it into the three-dimensional field data coordinate system to determine whether the line of sight focus area is a defensive key area of ​​the current scene. ③Statistical analysis of indicators: Statistics on the coverage rate of key areas and the delay time in focusing on risk targets; ④ Shortcomings Identification: Based on preset thresholds (critical area coverage ≥ 60%, risk target focusing delay ≤ 1.5 seconds, which are determined based on statistical data from 500 groups of driving school students' defensive driving training), the rationality of vision prediction is judged, specific shortcomings are identified (such as failure to continuously pay attention to changes in the distance to the vehicle in front when following another vehicle), and the data is synchronized to the attention analysis module.

[0025] Step 4: Weakness Assessment The weakness assessment includes an attention analysis sub-step and a comprehensive assessment sub-step.

[0026] The attention analysis sub-step is executed by the attention analysis module: ① Receive the denoised gaze data transmitted by the gaze analysis module and synchronize the current scene type and timestamp information; ② Calculate core attention indicators, including the frequency of eye focus switching per unit time, the continuity of eye gaze in key areas (number of times a single gaze lasts ≥2 seconds / total number of observations), the proportion of gaze in key and non-key areas (duration of gaze in key areas / duration of gaze in non-key areas), and the proportion of qualified attention duration (qualified attention duration / total training time). ③ Apply differentiated qualification thresholds according to the scenario (e.g., for a curve scenario, the frequency of switching the focus of sight should be ≤2 times / minute, and for an intersection scenario, the ratio of dwell time in key and non-key areas should be ≥2) to determine whether the indicators meet the standards. ④ Calculate the weighted comprehensive attention score (switching frequency score × 25% + continuity score × 25% + area proportion score × 30% + qualified proportion score × 20%) to identify attention weaknesses; ⑤ Output attention analysis results including scores, weaknesses, and areas for improvement, synchronize them to the evaluation module, and display them in real time.

[0027] The comprehensive assessment sub-step is performed by the defensive driving capability assessment module: ① Load a preset weighting system (attention 20%, proactive observation 30%, safety margin maintenance 30%, risk avoidance operations 20%). ② Extract data from each module and calculate the scores of four individual indicators (0-100 points) according to preset thresholds. ③ Calculate the overall score for defensive driving using a weighted summation method; ④ Divide scores into grades (≥85 points is excellent, 70-84 points is satisfactory, 60-69 points is basically satisfactory, <60 points is unsatisfactory), and identify weak indicators and specific shortcomings; ⑤ Integrate scores, weaknesses, training scenario suggestions, and improvement directions to generate a visual evaluation report.

[0028] In one specific embodiment, the trainee chose to train on following distance safety. After the data collection and analysis were completed, the evaluation results showed an overall score of 57.9 points (unsatisfactory). The report clearly pointed out that the trainee's shortcomings were insufficient observation of following distance safety and distraction, and recommended a special training scenario for following distance safety.

[0029] Step 5: Solution Generation This step is performed by the personalized training module: ① Extract the trainees' defensive weaknesses and shortcomings from the assessment report; ② Filter training scenarios and attention enhancement sub-scenarios that closely match the weaknesses from the scenario database; ③ Configure training intensity, duration, and passing standards according to the severity of the weaknesses; ④ Generate an executable personalized training plan, including: basic information (student ID, training date, current overall score, type of weakness), training objectives (identify the defensive driving ability to be improved), training scenarios (match 1-3 defensive scenarios corresponding to the weaknesses), training focus (eye-tracking observation requirements, attention allocation rules, safe operating procedures), training duration (5-10 minutes per scenario, 20-30 minutes total), passing standards (thresholds that the key indicators of this training must reach), and guidance instructions (real-time voice and text prompt rules).

[0030] For example, if a trainee has a weakness in observing safe following distance, the system will match a "special training scenario for maintaining following distance" and generate a plan that requires the trainee to switch their line of sight reasonably between the vehicle in front, the left and right rearview mirrors and the dashboard during training.

[0031] Step 6: Navigation Training This step is performed jointly by the in-vehicle navigation guidance module and the in-vehicle interactive display module: ① The vehicle navigation guidance module plans the optimal route from the current location to the target training scene based on three-dimensional site data, and guides the trainees to the target scene by displaying and broadcasting navigation information; ② During the training process, the data acquisition module continuously collects line-of-sight data and vehicle driving data, and the system monitors them in real time and compares them with preset qualification standards; ③ When the trainee's driving behavior is detected to deviate from the training focus of the personalized training program (such as ignoring the observation of the vehicle in front for a long time again), the in-vehicle interactive display module immediately issues real-time guidance instructions through voice and screen text to correct the behavior (such as "Please observe the distance of the vehicle in front"), thus achieving immediate intervention.

[0032] Step 7: Iterate After one intensive training session, the system automatically returns to step 3 to re-collect, analyze, and evaluate gaze data. Steps 3-6 are repeated to form a closed loop of "training-evaluation-retraining" until the trainee's evaluation results meet the comprehensive evaluation passing standard.

[0033] The overall evaluation passing criteria are as follows: ① Overall Score Standard: A defensive driving score of ≥85 points is considered excellent, indicating completion of defensive driving training; 70-84 points is considered qualified, allowing graduation with recommendations for supplementary intensive training; 60-69 points is considered basically qualified, requiring specialized training and retesting; <60 points is considered unqualified, requiring complete retraining. ②Single mandatory standard: The scores of the four individual indicators of attention, anticipatory observation, safe space maintenance and risk avoidance operation must not be lower than 60 points. If any one of them is less than 60 points, it is considered as unqualified. ③ Final qualification criteria: A comprehensive score of ≥85 points, with each of the four individual indicators ≥60 points, and a 100% pass rate in key scenario training are required to be considered as having passed the comprehensive evaluation and completed the training.

[0034] Through the above methods and corresponding system architecture, this invention realizes a complete closed loop from data collection, objective evaluation, precise positioning to targeted training and real-time guidance, providing driving schools with a scientific and efficient means of defensive driving ability training. It effectively solves the industry's pain point of "emphasizing subjects and neglecting defense" and fills the gap in the existing technology of lacking a real-world scenario-based defensive training system.

[0035] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for training a driver's defensive driving ability based on eye-tracking, characterized in that, Includes the following steps: Field data preparation steps: Establish 3D field data containing at least one defensive driving training scenario marker; Line-of-sight correlation analysis steps: The line-of-sight data of the trainees is collected in real time through the line-of-sight tracking device, and the line-of-sight data is correlated and matched with the three-dimensional field data to determine the positional relationship between the focal area of ​​the trainees' line of sight in the real field and the defensive driving training scenario. Weakness assessment steps: Based on the analysis results of the line-of-sight correlation analysis steps, quantitatively assess the trainee's defensive driving ability and identify at least one driving weakness; Solution generation steps: Based on the driving shortcomings, generate a personalized training solution that includes the target training scenario; Navigation training steps: Based on the personalized training plan, guide trainees to the target training scenario for intensive training through the navigation guidance module.

2. The driver defensive driving ability training method based on eye tracking according to claim 1, characterized in that, The site data preparation steps include: The three-dimensional point cloud data of the driving school site was collected by laser scanning equipment, and a 1:1 scale three-dimensional site model was generated by combining the image with optimization. The location coordinates, scenario types, training objectives, and key observation areas of key defensive driving training scenarios are marked on the three-dimensional site model to form a scenario database.

3. The driver defensive driving ability training method based on eye tracking according to claim 1, characterized in that, The line-of-sight correlation analysis steps include: The collected gaze data is pre-processed to remove noise and eliminate abnormal data caused by blinking or head movement. The preprocessed line-of-sight data is fused with the real-time vehicle location data obtained through environmental sensors; Based on pre-completed coordinate calibration parameters, the fused data is mapped to the coordinate system of the three-dimensional site data to determine whether the trainee's line of sight is a defensive key area of ​​the current scene.

4. The driver defensive driving ability training method based on eye tracking according to claim 3, characterized in that, The line-of-sight correlation analysis step also includes: The percentage of coverage of key areas is the proportion of time that trainees spend looking at key defensive areas out of the total training time. The statistical risk target focusing delay time is the delay time from when the environmental sensor detects the risk target to when the trainee's line of sight first focuses on the target; Based on preset thresholds including a key area coverage percentage threshold and / or a risk target focusing delay time threshold, the rationality of the line of sight prediction is determined and specific shortcomings are identified.

5. The driver defensive driving ability training method based on eye tracking according to claim 1, characterized in that, The weakness assessment step also includes an attention analysis sub-step, which includes: Based on the gaze data and current driving scenario information, an attention assessment index is calculated. The attention assessment index includes at least one of the following: gaze focus switching frequency per unit time, gaze continuity in key areas, and gaze ratio between key and non-key areas. By combining the differentiated qualification standards for current driving scenarios, the system assesses the matching degree between the student's attention and the driving scenario, and generates attention analysis results that include attention scores and weaknesses.

6. The driver defensive driving ability training method based on eye tracking according to claim 1, characterized in that, The weakness assessment step also includes a comprehensive assessment sub-step, including: Based on preset indicators and corresponding weights, combined with the results of line-of-sight correlation analysis, attention analysis, driving operation data, and site environment data, a comprehensive score for defensive driving ability is calculated and classified into levels. Generate an evaluation report that includes an overall score, individual indicator scores, a description of weaknesses, and suggestions for training scenarios.

7. The driver defensive driving ability training method based on eye tracking according to claim 1, characterized in that, The scheme generation steps include: Extract the trainees' driving weaknesses and shortcomings from the assessment report; Select training scenarios and attention enhancement sub-scenarios that match the aforementioned shortcomings from the scenario database; Configure training intensity, training duration, and passing standards according to the severity of the weaknesses; Generate personalized training plans that include training objectives, training scenarios, training focus, and real-time guidance rules.

8. The method for training driver defensive driving ability based on eye tracking according to claim 1, characterized in that, The navigation training steps include: Based on the three-dimensional site data, the optimal route from the vehicle's current position to the target training scene is planned, and navigation is provided through display and voice. During the student's driving process, line-of-sight data and vehicle driving data are continuously collected and compared in real time with preset passing standards; When a student's driving behavior is detected to deviate from the training focus of the personalized training program, real-time guidance instructions are issued via the in-vehicle interactive display module in the form of voice and / or visual means to correct it.

9. The method for training a driver's defensive driving ability based on eye tracking according to claim 1, characterized in that, The navigation training step is followed by a loop iteration step, including: Repeat the line-of-sight correlation analysis step, the shortcoming assessment step, the solution generation step, and the navigation training step until the trainee's assessment results meet the preset comprehensive assessment qualification standard. The comprehensive evaluation qualification criteria include: the overall score of defensive driving ability reaches the first threshold, and the scores of each individual indicator reach the second threshold.

10. A driver defensive driving ability training system based on eye tracking, used to perform the method according to any one of claims 1-9, characterized in that, include: The site scene construction module is used to create a 1:1 scale 3D site model through laser scanning and image optimization, and to mark key defensive driving training scenarios to form a scene database; The data acquisition module includes an in-vehicle eye-tracking device, driving operation sensors, and environmental sensors, used to collect the trainee's eye-tracking data, driving operation data, and site environment data. The line-of-sight analysis module is used to correlate and match line-of-sight data with three-dimensional site data, calculate the coverage ratio of key areas and the focusing delay time of risk targets, and locate defensive weaknesses. The attention analysis module is used to calculate attention evaluation indicators based on gaze data and driving scenario information, determine the matching degree between attention and scenario, and generate attention analysis results. The defensive driving ability assessment module is used to comprehensively assess the learner's defensive driving ability based on a preset indicator system and generate an assessment report. The personalized training module is used to match the weak training scenarios with the evaluation report and generate personalized training plans. The vehicle navigation guidance module is used to plan the optimal route to the target training scene based on 3D site data and provide navigation guidance. The in-vehicle interactive display module is used to display information and provide guidance commands in a voice and visual manner.

Images