Open environment automatic driving danger early warning method and system based on Internet of Things

Abstract

The invention provides an open environment automatic driving danger early warning method and system based on the Internet of Things, and the method comprises the steps: a first sensor group disposed on a vehicle continuously monitors the condition of a driving path, and transmits the monitored condition information of the driving path to an intelligent auxiliary driving center; a second sensor group deployed on the vehicle continuously monitors the state of the driver and sends the monitored state information of the driver to the intelligent auxiliary driving center; and the advanced driving assistance system fuses the road condition information and the driver state information to form a danger evaluation coefficient, and performs graded early warning according to the danger evaluation coefficient. By means of the method and the corresponding system, the attention index of the driver can be monitored in real time, the early warning threshold value and strategy are dynamically adjusted according to the fatigue degree and the concentration degree of the driver, and personalized risk prompts are provided; the risk level is subdivided, early warning measures are changed from light to heavy, safety is ensured, excessive interference is avoided, and the driving experience is improved.

Description

Technical Field

[0001] The present invention proposes an open environment autonomous driving hazard warning method and system based on the Internet of Things, relating to the field of autonomous driving warning technology. Background Art

[0002] Most existing methods rely on a single or limited number of sensors (e.g., radar / cameras), making it difficult to timely perceive complex road conditions (slippery, icy, potholes). They also fail to fully consider external factors such as weather, traffic flow, and road type, resulting in insufficient accuracy and robustness in identifying hazardous situations. While some systems have begun to integrate camera and radar data, their utilization of vehicle dynamics (e.g., tire-road adhesion coefficient) or external IoT data such as weather, road network events, and other vehicle information remains insufficiently systematic or comprehensive. This lack of a comprehensive understanding of the real-world road environment makes it difficult for algorithms to make precise assessments when sudden hazards arise. Most warnings rely solely on a few single metrics, such as time to collision (TTC), following distance, or relative speed, using fixed thresholds or linear models. These systems fail to fully consider the dynamic coupling of factors such as tire adhesion coefficient, driver status, road type, weather conditions, and obstacle motion in complex scenarios, resulting in insufficient warning sensitivity and accuracy. Summary of the Invention

[0003] The present invention provides an open environment autonomous driving hazard warning method and system based on the Internet of Things to solve the above-mentioned problems:

[0004] The present invention proposes a method for early warning of dangers in an open environment autonomous driving based on the Internet of Things, the method comprising:

[0005] The first sensor group deployed on the vehicle continuously monitors the driving path condition and sends the monitored driving path condition information to the intelligent assisted driving center;

[0006] A second sensor group deployed on the vehicle continuously monitors the driver's status and sends the monitored driver status information to the intelligent assisted driving center;

[0007] The advanced driver assistance system integrates road condition information and driver status information to form a risk assessment coefficient, and performs graded warnings based on the risk assessment coefficient.

[0008] Furthermore, the first sensor group deployed on the vehicle continuously monitors the driving path condition and sends the monitored driving path condition information to the intelligent assisted driving center, including:

[0009] The six-axis force sensor deployed in the vehicle's wheel hub motor measures the tire vertical load F in real time. z , longitudinal force F x, lateral force F y , get the estimated tire-road adhesion coefficient:

[0010] The ABS sensor installed on the inner side of the wheel hub monitors the vehicle speed in real time;

[0011] The millimeter-wave radar and binocular cameras installed at both ends of the vehicle's front bumper obtain the status of obstacles in front of the vehicle, including:

[0012] Obtain the obstacle status in front of the vehicle through millimeter-wave radar;

[0013] When the obstacle ahead is within a first safety preset distance from the vehicle, the motion state of the obstacle ahead within the first safety preset distance is determined, the video frames of the binocular camera are read, the video frames are cropped into left and right monocular photos, and the BM matching algorithm is used for stereo matching to obtain a disparity map;

[0014] Using a median filter algorithm to filter out noise points whose errors exceed a first preset threshold value from the disparity map generated by stereo matching;

[0015] A depth map is generated based on the disparity map after filtering out noise points. The mean value processing method is used to select the distance points of the target object in the depth map, including:

[0016] After obtaining the rectangular bounding box of the target object in the video frame, calculate the center pixel point P of the rectangular bounding box. Then, with P as the center, select the intersection points (a, b, c, d) of the rectangular box with a width and height of 1 / N of the length of the rectangular bounding box and the diagonal of the rectangular bounding box. Calculate the corresponding three-dimensional coordinates of the five points a, b, c, d, and P using the coordinates in the depth map, and then calculate the average of the Z-axis values as the target object distance.

[0017] Furthermore, determining the motion state of a front obstacle within a first preset safety distance from the vehicle includes:

[0018] The absolute speed v of the vehicle is obtained through the on-board inertial measurement unit and GPS car and direction of travel;

[0019] The millimeter-wave radar is used to obtain the angle θ between the front obstacle and the vehicle's driving direction and the relative speed of the front obstacle to the vehicle. The velocity component of the vehicle in the direction of the radar ray is: v car,radial =v car ·cosθ;

[0020] Get the absolute radial velocity v of the obstacle ahead object,radial =v relative +v car,radialIf the absolute radial velocity of the obstacle ahead is less than the preset threshold, it is determined to be a stationary object; if the absolute radial velocity of the obstacle ahead is greater than the preset threshold, it is determined to be a moving object;

[0021] When the obstacle ahead is detected as a moving object, the time margin between the vehicle and the obstacle ahead is calculated. The time margin is expressed as:

[0022]

[0023] Where |ΔV| represents the relative speed between the vehicle and the obstacle along the collision direction, ε represents a positive number, TTC represents the time margin, and ΔD represents the current distance between the vehicle and the obstacle.

[0024] Furthermore, a second sensor group deployed on the vehicle continuously monitors the driver's status and sends the monitored driver status information to the intelligent assisted driving center; including:

[0025] The second sensor group on the vehicle is used to monitor the driver's status, and the driver's status information includes: driver's blink frequency, driver's head orientation, heart rate, skin resistance and steering wheel grip;

[0026] The driver's attention index is obtained by integrating the driver's status information.

[0027] Furthermore, the advanced driver assistance system integrates road condition information and driver status information to form a risk assessment coefficient, and performs graded warnings based on the risk assessment coefficient, including:

[0028] The risk assessment coefficient is calculated using a risk assessment coefficient model based on the road condition information and the driver status information. Specifically, the risk assessment coefficient model is:

[0029]

[0030] Where R(t) represents the risk assessment coefficient at time t, a1 and a2 represent the first weight coefficients, b0 and b1 represent the second weight coefficients, w1, w2, w3, and w4 represent the exponential weight coefficients, and I driver (t) represents the driver attention index, μ(t) represents the road adhesion coefficient, TTC(t) represents the time margin at time t, f(v obj,radar (t)) represents the relative motion risk of obstacles;

[0031]

[0032] Among them, v obj,radar (t) represents the absolute radial velocity of the obstacle ahead at time t, and k represents the relative velocity coefficient;

[0033] If 0 ≤ R(t) < r1, it is determined that the vehicle is in a safe state, and only normal driving is maintained;

[0034] If r1 ≤ R total (t) < r2, it is determined that the vehicle is in a low-risk state, and the driver is prompted by voice to increase attention;

[0035] If r2 ≤ R total (t) < r3, it is determined that the vehicle is in a medium-risk state, a red warning appears on the vehicle's central control large screen, and steering wheel and seat vibration reminders are applied;

[0036] If r3 ≤ R total (t) < 1.0, it is determined that the vehicle is in a high-risk state, and the ADAS system actively performs the operation of pre-tightening the seat belt. Among them, r1 represents the first risk threshold, r2 represents the second risk threshold, and r3 represents the third risk threshold.

[0037] The open environment automatic driving danger warning system based on the Internet of Things proposed by the present invention, the system includes:

[0038] The module for monitoring the driving path condition, the first sensor group deployed on the vehicle continuously monitors the driving path condition and sends the monitored driving path condition information to the intelligent assisted driving center;

[0039] The module for monitoring the driver's state, the second sensor group deployed on the vehicle continuously monitors the driver's state and sends the monitored driver's state information to the intelligent assisted driving center;

[0040] The danger assessment module, the advanced driving assistance system combines the road condition information and the driver's state information to form a danger assessment coefficient, and performs hierarchical warning according to the danger assessment coefficient.

[0041] Further, the module for monitoring the driving path condition includes:

[0042] The module for obtaining the adhesion estimation coefficient, which is used for the six-axis force sensor deployed on the vehicle hub motor to measure the tire vertical load F z , longitudinal force F x , lateral force F y , and obtains the adhesion estimation coefficient of the tire-road surface:

[0043] The module for monitoring the vehicle speed, the ABS sensor installed on the inner side of the wheel hub continuously monitors the vehicle speed;

[0044] The module for obtaining the front obstacle, the millimeter wave radar and the binocular camera installed at both ends of the vehicle front bumper obtain the condition of the front obstacle of the vehicle, including:

[0045] Obtaining the condition of the front obstacle of the vehicle through the millimeter wave radar;

[0046] When the obstacle ahead is within a first safety preset distance from the vehicle, the motion state of the obstacle ahead within the first safety preset distance is determined, the video frames of the binocular camera are read, the video frames are cropped into left and right monocular photos, and the BM matching algorithm is used for stereo matching to obtain a disparity map;

[0047] Using a median filter algorithm to filter out noise points whose errors exceed a first preset threshold value from the disparity map generated by stereo matching;

[0048] A depth map is generated based on the disparity map after filtering out noise points. The mean value processing method is used to select the distance points of the target object in the depth map, including:

[0049] After obtaining the rectangular bounding box of the target object in the video frame, calculate the center pixel point P of the rectangular bounding box. Then, with P as the center, select the intersection points (a, b, c, d) of the rectangular box with a width and height of 1 / N of the length of the rectangular bounding box and the diagonal of the rectangular bounding box. Calculate the corresponding three-dimensional coordinates of the five points a, b, c, d, and P using the coordinates in the depth map, and then calculate the average of the Z-axis values as the target object distance.

[0050] Furthermore, the module for obtaining obstacles ahead includes:

[0051] Obtain absolute speed and direction module, obtain the absolute speed v of the vehicle through the vehicle inertial measurement unit and GPS car and direction of travel;

[0052] The relative speed module obtains the angle θ between the front obstacle and the vehicle's driving direction and the relative speed of the front obstacle relative to the vehicle through the millimeter wave radar. The velocity component of the vehicle in the direction of the radar ray is: v car,radial =v car ·cosθ;

[0053] The stationary motion determination module is used to obtain the absolute radial velocity v of the obstacle in front object,radial =v relative +

[0054] v car,radial If the absolute radial velocity of the obstacle ahead is less than the preset threshold, it is determined to be a stationary object; if the absolute radial velocity of the obstacle ahead is greater than the preset threshold, it is determined to be a moving object;

[0055] The time margin calculation module is used to calculate the time margin between the vehicle and the obstacle in front when the obstacle in front is detected to be a moving object. The time margin is expressed as:

[0056]

[0057] Where, |ΔV| represents the relative speed of the vehicle and the obstacle along the collision direction, ε represents a positive number, TTC represents the time margin, and ΔD represents the current distance between the vehicle and the obstacle.

[0058] Further, the driver state monitoring module includes:

[0059] A monitoring module, where the second sensor group on the vehicle is used to monitor the driver state, and the driver state information includes: driver blink frequency, driver head orientation, heart rate, skin resistance, and steering wheel grip force;

[0060] An attention index acquisition module, which is used to obtain the driver attention index by synthesizing the driver state information.

[0061] Further, the risk assessment module includes:

[0062] A risk assessment coefficient calculation module, which is used to calculate the risk assessment coefficient based on the road condition information and the driver state information by using the risk assessment coefficient model. Specifically, the risk assessment coefficient model is:

[0063]

[0064] Where, R(t) represents the risk assessment coefficient at time t, a1 and a2 represent the first weight coefficients, b0 and b1 represent the second weight coefficients, w1, w2, w3, w4 represent the exponential weight coefficients, I driver (t) represents the driver attention index, μ(t) represents the road surface adhesion coefficient, TTC(t) represents the time margin at time t, f(v obj,radar (t)0 represents the relative motion risk of the obstacle;

[0065]

[0066] Where, v obj,radar (t) represents the absolute radial speed of the front obstacle at time t, and k represents the relative speed coefficient;

[0067] A hierarchical early warning module, which is used to determine that the vehicle is in a safe state and only maintain normal driving if 0 ≤ R(t) < r1;

[0068] If r1 ≤ R total (t) < r2, it is determined that the vehicle is in a low risk, and the driver is prompted by voice to improve attention;

[0069] If r2 ≤ R total (t) < r3, it is determined that the vehicle is in a medium risk, a red warning appears on the vehicle's central control large screen, and steering wheel and seat vibration reminders are applied;

[0070] If r3 ≤ R totalIf (t)<1.0, the vehicle is judged to be at high risk, and the ADAS system actively performs the seat belt pre-tightening operation, where r1 represents the first risk threshold, r2 represents the second risk threshold, and r3 represents the third risk threshold.

[0071] The present invention has the following beneficial effects: comprehensively improving driving safety. Through real-time monitoring and data fusion from multiple sensors, the system can quickly and accurately identify potential dangers ahead, reducing the accident rate. It comprehensively considers the driver's status, vehicle operating status, road environment, and obstacle information, and utilizes a nonlinear risk assessment model to accurately reflect the actual risk level, avoiding missed and false alarms. It also takes into account the driver's status, monitors the driver's attention index in real time, and dynamically adjusts the warning threshold and strategy based on their fatigue and concentration, providing personalized risk warnings. It also subdivides risk levels and implements warning measures from light to heavy, ensuring safety while avoiding excessive interference and improving the driving experience. By calculating the road adhesion coefficient in real time, the system can adjust the warning threshold in complex road conditions such as slippery and icy conditions, enhancing its adaptability to special road conditions. It distinguishes between stationary and moving obstacles, assigning higher risk weights to moving targets approaching at high speed, improving the accuracy of warnings. In high-risk situations, the system can proactively implement safety measures such as pre-tightening seatbelts and, when necessary, intervene in vehicle control (such as emergency braking and steering assistance), providing more comprehensive safety protection for drivers and passengers. This solution provides a comprehensive hazard warning mechanism for Advanced Driver Assistance Systems (ADAS) and autonomous driving technologies, enhancing the system's decision-making capabilities and safety in complex traffic environments, and accelerating the development and application of autonomous driving technology. Leveraging the Internet of Things and multi-sensor fusion, this technical solution significantly improves vehicle safety and reliability in open environments. Furthermore, personalized warning strategies and adaptability to complex road conditions enhance the driving experience, laying a solid foundation for the further development of autonomous driving technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 This is a schematic diagram of the IoT-based open environment autonomous driving hazard warning method described in the present invention. DETAILED DESCRIPTION

[0073] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features therein may be combined with each other.

[0074] The following description sets forth numerous specific details to facilitate a thorough understanding of the present invention. The embodiments described are merely a portion of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are intended to fall within the scope of protection of the present invention.

[0075] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0076] One embodiment of the present invention provides a method for hazard warning of autonomous driving in an open environment based on the Internet of Things, the method comprising:

[0077] The first sensor group deployed on the vehicle continuously monitors the driving path condition and sends the monitored driving path condition information to the intelligent assisted driving center;

[0078] A second sensor group deployed on the vehicle continuously monitors the driver's status and sends the monitored driver status information to the intelligent assisted driving center;

[0079] The advanced driver assistance system integrates road condition information and driver status information to form a risk assessment coefficient, and performs graded warnings based on the risk assessment coefficient.

[0080] The working principle and effect of the above technical solution are as follows: Multi-source sensors enable multi-dimensional information collection. The first sensor group on the vehicle continuously collects environmental data on the road ahead, such as traffic flow, obstacles, and road markings, to comprehensively monitor the driving path. In addition, the second sensor group focuses on real-time monitoring of the driver's state, collecting data such as the driver's gaze, drowsiness, distraction, and operating behavior. The collected driving path conditions (external environmental information) and driver status (internal behavioral information) are deeply integrated through the intelligent assisted driving center. The AD system, based on preset algorithms and models, combines external risk factors with the driver's current state to generate a more objective and dynamic risk profile. Ultimately, a quantifiable "risk assessment coefficient" is generated. The system then issues graded warnings to the driver based on the different levels of the risk assessment coefficient. When the risk is low, only a friendly reminder or interface warning is issued. When the risk is moderate, a voice or vibration warning may be issued. When the risk is extremely high, the system can automatically take intervention measures such as active braking and forced deceleration to ensure driving safety. Dynamic and personalized risk identification: through real-time coupling of environmental perception and driver behavior, the system can dynamically judge risk probability, avoid misjudgment that may be caused by a single information source, and better adapt to changing road conditions and individual differences; the warning strategy is scientific and accurate. Danger warning is based on multi-dimensional data fusion and classification mechanism. Regardless of the harsh external environment or poor driver condition, it can accurately match the corresponding warning and protection measures, effectively improving the timeliness and acceptability of the warning; overall driving safety is improved. This solution enables the vehicle to actively intervene and make comprehensive judgments, helping drivers to promptly perceive and avoid potential risks, significantly reducing traffic accidents caused by driver negligence, fatigue or external emergencies, and helping to achieve more intelligent and humane advanced assisted driving.

[0081] In one embodiment of the present invention, a first sensor group deployed on a vehicle continuously monitors a driving path condition and transmits the monitored driving path condition information to an intelligent assisted driving center, including:

[0082] The six-axis force sensor deployed in the vehicle's wheel hub motor measures the tire vertical load F in real time. z , longitudinal force F x , lateral force F y , get the estimated tire-road adhesion coefficient:

[0083] The ABS sensor installed on the inner side of the wheel hub monitors the vehicle speed in real time;

[0084] The millimeter-wave radar and binocular cameras installed at both ends of the vehicle's front bumper obtain the status of obstacles in front of the vehicle, including:

[0085] Obtain the obstacle status in front of the vehicle through millimeter-wave radar;

[0086] When the obstacle ahead is within a first safety preset distance from the vehicle, the motion state of the obstacle ahead within the first safety preset distance is determined, the video frames of the binocular camera are read, the video frames are cropped into left and right monocular photos, and the BM matching algorithm is used for stereo matching to obtain a disparity map;

[0087] Using a median filter algorithm to filter out noise points whose errors exceed a first preset threshold value from the disparity map generated by stereo matching;

[0088] A depth map is generated based on the disparity map after filtering out noise points. The mean value processing method is used to select the distance points of the target object in the depth map, including:

[0089] After obtaining the rectangular bounding box of the target object in the video frame, calculate the center pixel point P of the rectangular bounding box. Then, with P as the center, select the intersection points (a, b, c, d) of the rectangular box with a width and height of 1 / N of the length of the rectangular bounding box and the diagonal of the rectangular bounding box. Calculate the corresponding three-dimensional coordinates of the five points a, b, c, d, and P using the coordinates in the depth map, and then calculate the average of the Z-axis values as the target object distance.

[0090] The working principle and effect of the above technical solution are: real-time measurement of tire vertical load F z , longitudinal force F x , lateral force F y , get the estimated tire-road adhesion coefficient: This adhesion coefficient can reflect whether the road surface is wet, loose, or the risk of tire-to-ground slippage, facilitating real-time warning adjustments. The vehicle's motion status is captured in real time, and ABS sensors mounted on the inside of the wheel hubs continuously monitor and provide feedback on the vehicle's precise speed, providing foundational data for subsequent obstacle detection and collision risk prediction. Millimeter-wave radar and binocular cameras are positioned at either end of the vehicle's front bumper to achieve a combined sense of the forward environment. The millimeter-wave radar can quickly detect obstacles ahead and their relative positions, and is particularly adept at detecting moving or stationary obstacles in adverse weather and lighting conditions. The binocular camera system is responsible for acquiring high-resolution video information for precise target detection. When the millimeter-wave radar detects an obstacle within the first safety preset distance, the video frame processing process is triggered. Stereo vision and depth estimation uses a Block Matching (BM) algorithm to perform stereo matching on the synchronized left and right image frames captured by the binocular cameras to generate a disparity map. A median filter is then used to remove noise, ensuring stable and reliable data. After processing the disparity map, a depth map is further generated to achieve accurate distance extraction of obstacles. For each obstacle, a detection algorithm is first used to obtain its rectangular bounding box in the video frame and determine the center pixel point P. Then, around point P, through the bounding box and its zoom area, the intersection points with the diagonal lines (a, b, c, d) are found to form the core sampling area, and the three-dimensional coordinates of the five points a, b, c, d, and P are extracted from the depth map. By averaging the values of the Z axis (i.e., the front-to-back direction) of these five points, a more robust and noise-resistant distance estimate is obtained, which improves the reliability of obstacle ranging. Through the fusion of multi-dimensional physical data and information, high-precision dynamic perception of the driving environment state and the distance to the obstacle ahead is performed, providing rich, reliable, and structured data support for the intelligent assisted driving center. Significantly improves the accuracy of road environment perception. Through a combination of multi-source sensors (force sensors, ABS, millimeter-wave radar, binocular cameras), it achieves full coverage and multi-dimensional acquisition of key factors such as road adhesion conditions, vehicle speed, obstacle distance and dynamics, improving the accuracy of environmental perception, especially in complex road conditions and changing environments. It can still maintain stable perception capabilities; dynamic risk control and auxiliary safety decision-making, real-time tire-road adhesion coefficient can help the drive system adjust output in time, reducing the risk of loss of control caused by wet, muddy, and slippery conditions. Accurate distance judgment and dynamic analysis of obstacles provide a solid data foundation for ADAS functions such as automatic emergency braking and adaptive cruise control, making the system's warning response earlier and more scientific; strong robustness and error resistance. In the process of obstacle distance extraction, through algorithms such as multi-point averaging and median filtering, it effectively suppresses noise and occasional misjudgments in the image processing process, improves the reliability and consistency of recognition results, and reduces safety hazards caused by single-point ranging anomalies.Laying the foundation for the deep integration of subsequent intelligent driving systems, the data collected by this system provides rich and structured input for driving risk assessment, path planning, early warning decision-making and even advanced intelligent algorithms for autonomous driving, facilitating the further development of more complex scene recognition and active safety control strategies.

[0091] In one embodiment of the present invention, determining the motion state of a front obstacle within a first predetermined safety distance from a vehicle includes:

[0092] The absolute speed v of the vehicle is obtained through the on-board inertial measurement unit and GPS car and direction of travel;

[0093] The millimeter-wave radar is used to obtain the angle θ between the front obstacle and the vehicle's driving direction and the relative speed of the front obstacle to the vehicle. The velocity component of the vehicle in the direction of the radar ray is: v car,radial =v car ·cosθ;

[0094] Get the absolute radial velocity v of the obstacle ahead object,radial =v relative +v car,radial If the absolute radial velocity of the obstacle ahead is less than the preset threshold, it is determined to be a stationary object; if the absolute radial velocity of the obstacle ahead is greater than the preset threshold, it is determined to be a moving object;

[0095] When the obstacle ahead is detected as a moving object, the time margin between the vehicle and the obstacle ahead is calculated. The time margin is expressed as:

[0096]

[0097] Where |ΔV| represents the relative speed between the vehicle and the obstacle along the collision direction, ε represents a positive number, TTC represents the time margin, and ΔD represents the current distance between the vehicle and the obstacle.

[0098] The working principle and effect of the above technical solution are: the absolute motion parameters of the vehicle are obtained, and the absolute speed v of the vehicle is accurately obtained through the vehicle-mounted inertial measurement unit (IMU) and GPS. car and current heading, providing basic data for all spatial calculations;

[0099] The forward millimeter-wave radar not only detects the position of the obstacle, but also provides the angle θ between the obstacle and the vehicle's direction of travel and the relative speed of the obstacle. The velocity component of the vehicle along the radar ray can be calculated using the trigonometric function relationship v car,radial =v carcosθ, which decomposes the vehicle's overall velocity into the radar's line of sight, ensuring that the velocity component is consistent with the projection of the obstacle's orientation. The obstacle's absolute radial velocity is calculated. If this value is less than a preset threshold, the obstacle is considered stationary (such as a road sign or a stationary vehicle). Otherwise, it is considered dynamic (such as a pedestrian or other moving vehicle). This distinction is crucial for the autonomous driving system to understand the nature of the environment and safely avoid obstacles. TTC (Time-to-Collision) is dynamically calculated. When a dynamic obstacle is detected ahead, the intelligent hub collects the distance ΔD and relative velocity |ΔV| between the vehicle and the obstacle in real time.

[0100] (i.e., the velocity difference along the possible collision path), and substitute into the formula TTC represents the estimated time to collision if both parties maintain constant speed and direction, and is directly used to assist driving decisions and graded warnings. Dynamic target detection and risk level determination are efficient and accurate. By integrating radar with the vehicle's own status data, it can accurately distinguish between stationary and moving obstacles, reducing misjudgments. For example, only dynamic traffic participants that pose a real collision threat are tracked and warned, optimizing resource allocation and system response speed. Real-time collision risk quantification improves the scientific nature of safety warnings. The dynamic calculation of TTC time margin can provide continuous and forward-looking risk warnings to drivers and autonomous driving systems. It can not only be used to trigger emergency braking and deceleration to avoid risks, but also support multi-level warning designs (such as early yellow light prompts and red light mandatory intervention), significantly improving active driving safety capabilities.

[0101] In one embodiment of the present invention, a second sensor group deployed on a vehicle continuously monitors the driver's status and sends the monitored driver status information to an intelligent assisted driving center; the system includes:

[0102] The second sensor group on the vehicle is used to monitor the driver's status, and the driver's status information includes: driver's blink frequency, driver's head orientation, heart rate, skin resistance and steering wheel grip;

[0103] The driver's attention index is obtained by integrating the driver's status information.

[0104] The working principle and effect of the above technical solution are as follows: multimodal driver status monitoring, blinking frequency: the driver's eye movements are detected through cameras and facial recognition algorithms, and the blinking frequency is collected at high speed and automatically identified; the blinking frequency is closely related to the driver's fatigue state, and an abnormal increase or decrease in frequency may indicate that the driver is not concentrating or is about to enter a fatigue state; head orientation is also based on the in-car camera to capture and analyze the driver's head posture to determine whether he is looking at the road, or turning his head, lowering his head, or other distracting behaviors; heart rate uses the bioelectric sensor in the steering wheel or wearable devices to collect the driver's heart rate data in real time. Abnormal heart rate may indicate driver stress, drowsiness, or sudden health risks. Physiological sensors detect changes in the driver's skin resistance, which reflects emotional changes and alertness levels. Abnormal changes can help identify tension, anxiety, and fatigue. A force sensor built into the steering wheel continuously measures grip force distribution and strength, reflecting the driver's hand movement habits. Insufficient grip strength can also indicate a lack of attention. A driver attention index is calculated by inputting these multiple signals into a driving monitoring algorithm to generate a numerical "attention index" reflecting the driver's alertness and focus. This index numerically describes the driver's alertness level and provides an intuitive indicator for driving safety assessment. Status information is transmitted to the intelligent driving assistance center (such as the ADAS system and domain controller) to serve as a key reference for safety warnings, automatic interventions, and risk assessments, ensuring real-time, personalized safety strategy formulation while driving. Accurately monitoring driver alertness and fatigue, integrating behavioral and physiological data such as eye movements, head orientation, heart rate, skin resistance, and grip strength, enables the system to promptly and accurately identify driver distraction, fatigue, and abnormal physiological states. This helps to intervene at the very beginning of dangerous trends and reduce accident risks; it realizes active warning and safety assistance. According to the real-time attention index, the system can dynamically adjust the warning level or take proactive measures (such as alarms, vibrations, vehicle deceleration, etc.), which greatly improves safety during driving and reduces traffic accidents caused by poor driver condition; it improves driving experience and personalized safety strategies. Due to the rich collection parameters, the system can optimize safety measures according to different driver habits and status, realize personalized driving monitoring solutions, and enhance user experience and trust; it lays the foundation for advanced assisted driving and autonomous driving. High-precision driver status monitoring provides necessary safety redundancy and judgment support for ADAS and future autonomous driving systems, which helps smart vehicles make scientific decisions in key scenarios such as switching between different automation levels and monitoring the driver's takeover ability.

[0105] In one embodiment of the present invention, the advanced driver assistance system integrates road condition information and driver status information to form a risk assessment coefficient, and performs graded warnings based on the risk assessment coefficient, including:

[0106] Calculate the risk assessment coefficient using the risk assessment coefficient model based on road condition information and driver status information. Specifically, the risk assessment coefficient model is as follows:

[0107]

[0108] where \(R(t)\) represents the risk assessment coefficient at time \(t\), \(a_1\) and \(a_2\) represent the first weight coefficients, \(b_0\) and \(b_1\) represent the second weight coefficients, \(w_1\), \(w_2\), \(w_3\), \(w_4\) represent the exponential weight coefficients, \(I\) driver (t) represents the driver attention index, \(\mu(t)\) represents the road surface adhesion coefficient, \(TTC(t)\) represents the time margin at time \(t\), \(f(v\) obj,radar (t)) represents the relative motion risk of the obstacle;

[0109]

[0110] where \(v\) obj,radar (t) represents the absolute radial velocity of the front obstacle at time \(t\), and \(k\) represents the relative velocity coefficient;

[0111] If \(0\leq R(t)\lt r_1\), it is determined that the vehicle is in a safe state and only maintains normal driving;

[0112] If \(r_1\leq R\) total (t)\lt r_2, it is determined that the vehicle is in a low-risk state, and the driver is prompted by voice to improve attention;

[0113] If \(r_2\leq R\) total (t)\lt r_3, it is determined that the vehicle is in a medium-risk state, a red warning appears on the vehicle's central control large screen, and steering wheel and seat vibration reminders are applied;

[0114] If \(r_3\leq R\) total (t)\lt 1.0, it is determined that the vehicle is in a high-risk state, and the ADAS system actively performs the operation of pre-tightening the seat belt. Here, \(r_1\) represents the first risk threshold, \(r_2\) represents the second risk threshold, and \(r_3\) represents the third risk threshold.

[0115] The working principle and effect of the above technical solution are as follows: multi-source information fusion input, the driver attention index reflects the driver's concentration and alertness level; the road adhesion coefficient reflects the road condition indicators related to road safety through tire-road force, slippage, etc.; the time margin TTC(t) measures the expected remaining time for the vehicle to collide (or close contact) with the obstacle; the absolute radial velocity of the obstacle represents the speed relationship and motion risk between the obstacle and the vehicle on the road; the driver's attention value is mapped to a risk score through a nonlinear function; the lower the road adhesion coefficient (μ decreases), the higher the risk score, reflecting the increased risk of low-adhesion roads such as wet / icy roads; the time margin TTC The shorter the distance (the higher the possibility of collision), the higher the risk score; it is determined by the movement state of the obstacle, with a static value of 0 and the risk of a dynamic obstacle increasing exponentially with increasing speed. R(t) is divided into four levels, corresponding to safety, low risk, medium risk and high risk, and each level automatically corresponds to different reminders or even active intervention plans: Safe zone: [0, r1) - just maintain normal driving; Low risk: [r1, r2) - voice prompts the driver to pay more attention; Medium risk: [r2, r3) - red warning on the central control screen + steering wheel and seat vibration reminder; High risk: [r3, 1.0) - ADAS actively performs seat belt pre-tightening, pre-emergency braking and other measures. In-depth integration of multi-dimensional risk information, more scientific and accurate assessment, comprehensive superposition of the environment, driver behavior, obstacle dynamic information and vehicle-road conditions, dynamic and quantitative characterization of risks, effective avoidance of single factor misjudgment, and providing a solid foundation for the vehicle safety strategy; graded prompts and progressive intervention to enhance the early warning experience, gradually upgrade warnings and intervention measures according to risk levels, ensure that the driver is not overly fatigued by warnings, and does not miss key high-risk states, to achieve human-machine friendly and scientific safety interaction; significantly improve the active safety level of autonomous driving and ADAS; automatically respond to potential dangers within a very short delay, and immediately trigger physical interventions such as seat belt pre-tightening and vibration, effectively reducing the probability of accidents caused by driver inattention, distraction, and sudden road conditions. As long as one item is very high (close to 1), the corresponding item is close to 0, the product quickly tends to 0, and R(t) tends to 1 (indicating extremely high risk), which can be adjusted by wi

[0116] Let the key sub-items (such as TTC or attention) have a greater weight than other items, the formula is extremely risk-dominated (high-risk factors are not masked by the average), and multiple high-risk aggregation is enhanced, which is close to "high-risk dominance" and retains the ability of "multiple high risks synergistically increasing".

[0117] One embodiment of the present invention is an open environment autonomous driving hazard warning system based on the Internet of Things, the system comprising:

[0118] A driving path condition monitoring module, in which a first sensor group deployed on the vehicle continuously monitors the driving path condition and sends the monitored driving path condition information to the intelligent assisted driving center;

[0119] The driver status monitoring module, a second sensor group deployed on the vehicle, continuously monitors the driver's status and sends the monitored driver status information to the intelligent assisted driving center;

[0120] The advanced driver assistance system integrates road condition information and driver status information to form a risk assessment coefficient, and performs graded warnings based on the risk assessment coefficient.

[0121] In one embodiment of the present invention, the module for monitoring the driving path condition includes:

[0122] Obtain adhesion estimation coefficient module, which is used to measure the tire vertical load F in real time using the six-axis force sensor deployed in the vehicle's wheel hub motor z , longitudinal force F x , lateral force F y , get the estimated tire-road adhesion coefficient:

[0123] Vehicle speed monitoring module: The ABS sensor installed on the inner side of the wheel hub monitors the vehicle speed in real time;

[0124] Obtaining the front obstacle module: The millimeter-wave radar and binocular camera installed at both ends of the vehicle's front bumper obtain the obstacle status in front of the vehicle, including:

[0125] Obtain the obstacle status in front of the vehicle through millimeter-wave radar;

[0126] When the obstacle ahead is within a first safety preset distance from the vehicle, the motion state of the obstacle ahead within the first safety preset distance is determined, the video frames of the binocular camera are read, the video frames are cropped into left and right monocular photos, and the BM matching algorithm is used for stereo matching to obtain a disparity map;

[0127] Using a median filter algorithm to filter out noise points whose errors exceed a first preset threshold value from the disparity map generated by stereo matching;

[0128] A depth map is generated based on the disparity map after filtering out noise points. The mean value processing method is used to select the distance points of the target object in the depth map, including:

[0129] After obtaining the rectangular bounding box of the target object in the video frame, calculate the center pixel point P of the rectangular bounding box. Then, with P as the center, select the intersection points (a, b, c, d) of the rectangular box with a width and height of 1 / N of the length of the rectangular bounding box and the diagonal of the rectangular bounding box. Calculate the corresponding three-dimensional coordinates of the five points a, b, c, d, and P using the coordinates in the depth map, and then calculate the average of the Z-axis values as the target object distance.

[0130] In one embodiment of the present invention, the module for obtaining obstacles ahead includes:

[0131] Obtain absolute speed and direction module, obtain the absolute speed v of the vehicle through the vehicle inertial measurement unit and GPS car and direction of travel;

[0132] The relative speed module obtains the angle θ between the front obstacle and the vehicle's driving direction and the relative speed of the front obstacle relative to the vehicle through the millimeter wave radar. The velocity component of the vehicle in the direction of the radar ray is: v car,radia1 =v car ·cosθ;

[0133] The stationary motion determination module is used to obtain the absolute radial velocity v of the obstacle in front object,radial =v relative +

[0134] v car,radial If the absolute radial velocity of the obstacle ahead is less than the preset threshold, it is determined to be a stationary object; if the absolute radial velocity of the obstacle ahead is greater than the preset threshold, it is determined to be a moving object;

[0135] The time margin calculation module is used to calculate the time margin between the vehicle and the obstacle in front when the obstacle in front is detected to be a moving object. The time margin is expressed as:

[0136]

[0137] Where |ΔV| represents the relative speed between the vehicle and the obstacle along the collision direction, ε represents a positive number, TTC represents the time margin, and ΔD represents the current distance between the vehicle and the obstacle.

[0138] In one embodiment of the present invention, the driver status monitoring module includes:

[0139] A monitoring module, wherein the second sensor group on the vehicle is used to monitor the driver's status, wherein the driver's status information includes: the driver's blink frequency, the driver's head orientation, the heart rate, the skin resistance and the steering wheel grip;

[0140] The attention index acquisition module is used to obtain the driver's attention index based on the driver's status information.

[0141] In one embodiment of the present invention, the risk assessment module includes:

[0142] The risk assessment coefficient calculation module is used to calculate the risk assessment coefficient based on the road condition information and the driver status information using the risk assessment coefficient model. Specifically, the risk assessment coefficient model is:

[0143]

[0144] Among them, R(t) represents the risk assessment coefficient at time t, a1 and a2 represent the first weight coefficients, b0 and b1 represent the second weight coefficients, w1, w2, w3, w4 represent exponential weight coefficients, and I driver (t) represents the driver attention index, μ(t) represents the road surface adhesion coefficient, TTC(t) represents the time margin at time t, and f(v obj,radar (t)) represents the relative motion risk of the obstacle;

[0145]

[0146] Among them, v obj,radar (t) represents the absolute radial velocity of the front obstacle at time t, and k represents the relative velocity coefficient;

[0147] The grading early warning module is used to determine that the vehicle is in a safe state if 0 ≤ R(t) < r1, and only maintain normal driving;

[0148] If r1 ≤ R total (t) < r2, it is determined that the vehicle is in a low risk, and a voice prompt is given to the driver to improve attention;

[0149] If r2 ≤ R total (t) < r3, it is determined that the vehicle is in a medium risk, a red warning appears on the vehicle's central control large screen, and steering wheel and seat vibration reminders are applied;

[0150] If r3 ≤ R total (t) < 1.0, it is determined that the vehicle is in a high risk, and the ADAS system actively performs the operation of pre-tightening the seat belt. Among them, r1 represents the first risk threshold, r2 represents the second risk threshold, and r3 represents the third risk threshold.

[0151] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these changes and modifications.

Claims

1. An open environment autonomous driving hazard warning method based on the Internet of Things, characterized by: The method comprises: The first sensor group deployed on the vehicle continuously monitors the driving path condition and sends the monitored driving path condition information to the intelligent assisted driving center; A second sensor group deployed on the vehicle continuously monitors the driver's status and sends the monitored driver status information to the intelligent assisted driving center; The advanced driver assistance system integrates road condition information and driver status information to form a risk assessment coefficient, and performs graded warnings based on the risk assessment coefficient.

2. The method for early warning of dangers in an open environment autonomous driving based on the Internet of Things according to claim 1, characterized in that: The first sensor group deployed on the vehicle continuously monitors the driving path conditions and sends the monitored driving path condition information to the intelligent assisted driving center, including: The six-axis force sensor deployed in the vehicle's wheel hub motor measures the tire vertical load F in real time. z , longitudinal force F x , lateral force F y , get the estimated tire-road adhesion coefficient: The ABS sensor installed on the inner side of the wheel hub monitors the vehicle speed in real time; The millimeter-wave radar and binocular cameras installed at both ends of the vehicle's front bumper obtain the status of obstacles in front of the vehicle, including: Obtain the obstacle status in front of the vehicle through millimeter-wave radar; When the obstacle ahead is within a first safety preset distance from the vehicle, the motion state of the obstacle ahead within the first safety preset distance is determined, the video frames of the binocular camera are read, the video frames are cropped into left and right monocular photos, and the BM matching algorithm is used for stereo matching to obtain a disparity map; Using a median filter algorithm to filter out noise points whose errors exceed a first preset threshold value from the disparity map generated by stereo matching; A depth map is generated based on the disparity map after filtering out noise points. The mean value processing method is used to select the distance points of the target object in the depth map, including: After obtaining the rectangular bounding box of the target object in the video frame, calculate the center pixel point P of the rectangular bounding box. Then, with P as the center, select the intersection points (a, b, c, d) of the rectangular box with a width and height of 1 / N of the length of the rectangular bounding box and the diagonal of the rectangular bounding box. Calculate the corresponding three-dimensional coordinates of the five points a, b, c, d, and P using the coordinates in the depth map, and then calculate the average of the Z-axis values as the target object distance.

3. The method for early warning of dangers of autonomous driving in an open environment based on the Internet of Things according to claim 1, characterized in that: Determining the motion state of a front obstacle within a first safety preset distance from the vehicle includes: The absolute speed v of the vehicle is obtained through the on-board inertial measurement unit and GPS car and direction of travel; The millimeter-wave radar is used to obtain the angle θ between the front obstacle and the vehicle's driving direction and the relative speed of the front obstacle to the vehicle. The velocity component of the vehicle in the direction of the radar ray is: v car,radial =v car ·cosθ; Get the absolute radial velocity v of the obstacle ahead object,radial =v relative +v car,radial If the absolute radial velocity of the obstacle ahead is less than the preset threshold, it is determined to be a stationary object; if the absolute radial velocity of the obstacle ahead is greater than the preset threshold, it is determined to be a moving object; When the obstacle ahead is detected as a moving object, the time margin between the vehicle and the obstacle ahead is calculated. The time margin is expressed as: Where |ΔV| represents the relative speed between the vehicle and the obstacle along the collision direction, ε represents a positive number, TTC represents the time margin, and ΔD represents the current distance between the vehicle and the obstacle.

4. The method for early warning of dangers in an open environment autonomous driving based on the Internet of Things according to claim 1 is characterized in that ， The second sensor group deployed on the vehicle continuously monitors the driver's status and sends the monitored driver status information to the intelligent assisted driving center; it includes: The second sensor group on the vehicle is used to monitor the driver's status, and the driver's status information includes: driver's blink frequency, driver's head orientation, heart rate, skin resistance and steering wheel grip; The driver's attention index is obtained by integrating the driver's status information.

5. The method for early warning of dangers in an open environment autonomous driving based on the Internet of Things according to claim 1, characterized in that: The advanced driver assistance system integrates road condition information and driver status information to form a risk assessment coefficient, and performs graded warnings based on the risk assessment coefficient, including: The risk assessment coefficient is calculated using a risk assessment coefficient model based on the road condition information and the driver status information. Specifically, the risk assessment coefficient model is: Where R(t) represents the risk assessment coefficient at time t, a1 and a2 represent the first weight coefficients, b0 and b1 represent the second weight coefficients, w1, w2, w3, and w4 represent the exponential weight coefficients, and I driver (t) represents the driver attention index, μ(t) represents the road adhesion coefficient, TTC(t) represents the time margin at time t, f(v obj,radar (t)) represents the relative motion risk of obstacles; Among them, v obj,radar (t) represents the absolute radial velocity of the obstacle ahead at time t, and k represents the relative velocity coefficient; If 0 ≤ R(t) < r1, it is determined that the vehicle is in a safe state, and only normal driving is maintained; If r1 ≤ R total (t) < r2, it is determined that the vehicle is in a low-risk state, and a voice prompt is given to the driver to increase attention; If r2 ≤ R total (t) < r3, it is determined that the vehicle is at medium risk, a red warning appears on the vehicle's central control large screen, and steering wheel and seat vibrations are applied for reminder; If r3≤R total If (t)<1.0, the vehicle is judged to be at high risk, and the ADAS system actively performs the seat belt pre-tightening operation, where r1 represents the first risk threshold, r2 represents the second risk threshold, and r3 represents the third risk threshold.

6. The open environment autonomous driving hazard warning system based on the Internet of Things is characterized by: The system includes: A module for monitoring the driving path condition. The first sensor group deployed on the vehicle continuously monitors the driving path condition and sends the monitored driving path condition information to the intelligent assisted driving center; A module for monitoring the driver's state. The second sensor group deployed on the vehicle continuously monitors the driver's state and sends the monitored driver's state information to the intelligent assisted driving center; A risk assessment module. The advanced driving assistance system combines road condition information and driver's state information to form a risk assessment coefficient, and performs hierarchical warning according to the risk assessment coefficient.

7. The IoT-based open environment autonomous driving hazard warning system according to claim 6 is characterized in that: The module for monitoring the driving path condition includes: Obtain adhesion estimation coefficient module, which is used to measure the tire vertical load F in real time using the six-axis force sensor deployed in the vehicle's wheel hub motor z , longitudinal force F x , lateral force F y , get the estimated tire-road adhesion coefficient: A module for monitoring the vehicle speed. The ABS sensor installed on the inner side of the wheel hub continuously monitors the vehicle speed; A module for obtaining the obstacles ahead. The millimeter-wave radar and binocular camera installed at both ends of the front bumper of the vehicle obtain the condition of the obstacles ahead of the vehicle, including: Obtaining the condition of the obstacles ahead of the vehicle through the millimeter-wave radar; When the obstacle ahead is within the first safety preset distance from the vehicle, judge the motion state of the obstacle ahead within the first safety preset distance from the vehicle, read the video frame of the binocular camera, crop the video frame into left and right monocular photos, and use the BM matching algorithm for stereo matching to obtain a disparity map; Filter out the noise points with errors exceeding the first preset threshold from the disparity map generated by stereo matching through the median filtering algorithm; Generate a depth map based on the disparity map after filtering out the noise points. For the selection of the target object distance points in the depth map, the mean processing method is adopted, including: After obtaining the rectangular bounding box of the target object in the video frame, calculate the center pixel point P of the rectangular bounding box. Then, with P as the center, select the intersection points (a, b, c, d) of the rectangle with a width and height of 1 / N of the length of the rectangular bounding box and the diagonal of the rectangular bounding box. Calculate the corresponding three-dimensional coordinates of the five points a, b, c, d, and P through the coordinates in the depth map, and then find the average value of the Z-axis values as the target object distance.

8. The IoT-based open environment autonomous driving hazard warning system according to claim 6 is characterized in that: The module for obtaining the obstacles ahead includes: Obtain absolute speed and direction module, obtain the absolute speed v of the vehicle through the vehicle inertial measurement unit and GPS car and direction of travel; The relative speed module obtains the angle θ between the front obstacle and the vehicle's driving direction and the relative speed of the front obstacle relative to the vehicle through the millimeter wave radar. The velocity component of the vehicle in the direction of the radar ray is: v car,radial =v car ·cosθ; The stationary motion determination module is used to obtain the absolute radial velocity v of the obstacle in front object,radial =v relative +v car,radial If the absolute radial velocity of the obstacle ahead is less than the preset threshold, it is determined to be a stationary object; if the absolute radial velocity of the obstacle ahead is greater than the preset threshold, it is determined to be a moving object; A module for calculating the time margin. When detecting that the obstacle ahead is a moving object, it is used to calculate the time margin between the vehicle and the obstacle ahead. The time margin is expressed as: where |ΔV| represents the relative speed between the vehicle and the obstacle along the collision direction, ε represents a positive number, TTC represents the time margin, and ΔD represents the current distance between the vehicle and the obstacle.

9. The IoT-based open environment autonomous driving hazard warning system according to claim 6, characterized in that: The module for monitoring the driver's state includes: A monitoring module. The second sensor group on the vehicle is used to monitor the driver's state. The driver's state information includes: driver's blink frequency, driver's head orientation, heart rate, skin resistance, and steering wheel grip force; A module for obtaining the attention index. It is used to obtain the driver's attention index by synthesizing the driver's state information.

10. The open environment autonomous driving hazard warning system based on the Internet of Things according to claim 6 is characterized in that: The risk assessment module includes: A module for calculating the risk assessment coefficient. It is used to calculate the risk assessment coefficient based on the road condition information and the driver's state information by using the risk assessment coefficient model. Specifically, the risk assessment coefficient model is: Where R(t) represents the risk assessment coefficient at time t, a1 and a2 represent the first weight coefficients, b0 and b1 represent the second weight coefficients, w1, w2, w3, and w4 represent the exponential weight coefficients, and I driver (t) represents the driver attention index, μ(t) represents the road adhesion coefficient, TTC(t) represents the time margin at time t, f(v obj,radar (t)) represents the relative motion risk of obstacles; Among them, v obj,radar (t) represents the absolute radial velocity of the obstacle ahead at time t, and k represents the relative velocity coefficient; The hierarchical warning module is used to determine that the vehicle is in a safe state and only maintain normal driving if 0 ≤ R(t) < r1; If r1 ≤ R total (t) < r2, it is determined that the vehicle is in a low-risk state, and a voice prompt is given to the driver to improve attention; If r2 ≤ R total (t) < r3, it is determined that the vehicle is at medium risk, a red warning appears on the vehicle's central control large screen, and steering wheel and seat vibration reminders are applied; If r3≤R t0tal If (t)<1.0, the vehicle is judged to be at high risk, and the ADAS system actively performs the seat belt pre-tightening operation, where r1 represents the first risk threshold, r2 represents the second risk threshold, and r3 represents the third risk threshold.

Images