Artificial intelligence auxiliary intelligent driving system based on driving blind area

Abstract

The invention discloses an artificial intelligence auxiliary intelligent driving system based on a driving blind area, and relates to the technical field of artificial intelligence, the artificial intelligence auxiliary intelligent driving system is composed of an environment perception core module, a data acquisition core module and a decision early warning core module, and closed-loop management of driving safety is realized through a multi-level response mechanism; the environment sensing module analyzes the visibility of the environment where the vehicle is located by using an edge computing node, dynamically adjusts the acquisition frequency of a sensor, and ensures the optimal solution of data acquisition under different visibility conditions; the data acquisition module acquires a mobile entity in a driving blind area, and the monitoring precision is improved through fusion of a multi-source heterogeneous sensor; the decision early warning module adopts a trajectory prediction algorithm to analyze the relative motion relationship between the mobile entity and the vehicle, and when the collision risk exceeds a threshold value, a three-level early warning mechanism is started: a visual warning is firstly sent to a driver, if no response is made within the specified time, a sound-light alarm is triggered, and if no response is made continuously, an intelligent driving system takes over vehicle control.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence, and specifically relates to an artificial intelligence-assisted intelligent driving system based on a driving blind spot. Background Art

[0002] In recent years, with the rapid development of artificial intelligence, sensor technology, and computing power, intelligent driving technology has gradually become the core development direction of the automotive industry.

[0003] The existing intelligent driving systems have limited perception capabilities in complex environments. Especially in the driving blind spot, the sensors cannot adaptively adjust the acquisition frequency to accurately capture the dynamics of moving entities, resulting in inaccurate collision prediction. Secondly, most of the existing collision prediction algorithms are trained based on known scenarios and have poor adaptability to long-tail scenarios, prone to misjudgment. Some intelligent driving systems have deficiencies in sensor fusion and cannot fully utilize the advantages of cameras, millimeter-wave radars, and ultrasonic sensors, resulting in low accuracy of collision prediction. In view of the above problems, the present invention proposes an artificial intelligence-assisted intelligent driving system based on a driving blind spot. Summary of the Invention

[0004] Aiming at the deficiencies of the existing technology, the present invention provides an artificial intelligence-assisted intelligent driving system based on a driving blind spot, which solves the problem of difficult early prediction of vehicle collisions in the existing technology.

[0005] The object of the present invention can be achieved by the following technical solutions:

[0006] An artificial intelligence-assisted intelligent driving system based on a driving blind spot, the system includes the following:

[0007] An environment perception module, which obtains the visibility of the vehicle's surrounding environment in real time, analyzes and locks the visibility state of the current vehicle's surrounding environment through the marginalized computing nodes equipped in the environment perception module, and adaptively adjusts the acquisition frequency of the data acquisition sensors based on the visibility state;

[0008] A data acquisition module, which receives the visibility state determined by the environment perception module and the acquisition frequency of the data acquisition sensors adapted to this visibility state;

[0009] Real-time obtain the blind spot-related data in the driving blind spot through the data acquisition sensors;

[0010] A decision-making and warning module, which analyzes the blind spot-related data to determine whether a collision occurs between the running trajectories of the moving entity and the vehicle, and generates a warning signal or a safety signal based on the collision result;

[0011] If it is a warning signal, a warning reminder is sent to the driver, and the adjustment actions made by the driver are monitored in real time;

[0012] If the driver does not make any adjustment actions within the specified time, the decision-making early warning module issues an audible alarm to the driver. If the driver still does not make any adjustment actions within the specified time, the intelligent driving system automatically makes adjustment actions;

[0013] If it is detected that the driver makes corresponding adjustment actions, it is determined again whether a collision will occur in the running trajectories between the moving entity and the vehicle, and corresponding safety signals and early warning signals are generated;

[0014] If a safety signal is generated, no processing is performed;

[0015] If the moving entity leaves the driving blind area, the monitoring is stopped.

[0016] As a further solution of the present invention, the data acquisition sensors in the data acquisition module include: cameras, ultrasonic sensors, and millimeter-wave radars;

[0017] The blind area-related data includes the blind area images of the vehicle obtained in real time by the camera and the moving entities identified in the blind area images, as well as the real-time position relationships between the moving entities and the vehicle in the driving blind area determined by the ultrasonic sensors and millimeter-wave radars.

[0018] As a further solution of the present invention, in the environment perception module, the specific method for analyzing and locking the visibility state of the current vehicle's environment is as follows:

[0019] Collect the visibility of the environment where the current vehicle is located. Taking the current moment as the start moment, continuously collect the visibility within the time t, and summarize it as the visibility set Q, and transmit it to the marginalization calculation node, where t is a preset value set by the operator;

[0020] Process the visibility set Q using the sliding window method and the three-standard-deviation principle;

[0021] Extract the processed visibility set Q, perform normalization, and take the mean to obtain the mean visibility within the time t

[0022] Repeat the above method to obtain the mean visibility within consecutive j time intervals of t Denote it as the mean visibility set: where j is a counting index starting from 1;

[0023] Taking the mean visibility as the vertical axis and the time line as the horizontal axis, construct a two-dimensional coordinate system, and mark the mean visibility set in the two-dimensional coordinate system. Starting from connect adjacent mean visibilities successively with short lines backward until to obtain the visibility change broken line Q ～ ;

[0024] For Q ～ Further analyze to lock the visibility status associated with different time points.

[0025] As a further solution of the present invention, in the environmental perception module, the specific method for analyzing and locking the visibility status of the environment where the current vehicle is located further includes:

[0026] For Q ～ , after Construct a straight line perpendicular to the horizontal axis, denoted as the first moment line L1, copy L1 and translate it in the positive direction of the horizontal axis by a time t to obtain a straight line passing through , denoted as the second moment line L2;

[0027] Obtain the area S1 of the closed area formed by the horizontal axis, Q ～ , L1, and L2;

[0028] And so on, obtain the areas of j - 1 closed areas, denoted as the closed area sequence: S1, S2,..., S j-1 , where S i is any one in the closed area sequence, and i does not exceed j;

[0029] Extract the area threshold S 阈 preset by the operator, and compare the area S i of any closed area in the closed area sequence with the area threshold S 阈 preset by the operator;

[0030] If the area S i of any closed area ≥ S 阈 , then calibrate the visibility of the environment where the vehicle is located as normal visibility;

[0031] If S i < S 阈 , then calibrate the visibility of the environment where the vehicle is located as abnormal visibility.

[0032] As a further solution of the present invention, in the environmental perception module, the specific method for adaptively adjusting the acquisition frequency of the data acquisition sensor based on the visibility status is:

[0033] Lock the acquisition frequencies R1, R2, R3 of the camera, ultrasonic sensor, and millimeter - wave radar in their respective normal states, and the combination is denoted as the normal frequency feature R = R1, R2, R3;

[0034] Obtain the acquisition frequencies of each data acquisition sensor in the non - normal state preset by the operator, and the combination is denoted as the non - normal frequency feature R ′ = R ′ 1, R ′ 2, R′ 3, where R ′ 1> is R1, R ′ 2> is R2, R ′ 3> is R3;

[0035] Extract the visibility status of the current vehicle's environment. If it is normal visibility, enable R as the acquisition frequency;

[0036] If it is abnormal visibility, enable R ′ as the acquisition frequency.

[0037] As a further solution of the present invention, in the decision-making and warning module, the specific method for determining whether a collision occurs between the running trajectories of the moving entity and the vehicle and generating a warning signal or a safety signal based on the collision result is:

[0038] Extract the blind area image of the current vehicle that is monitored and captured in real time by the camera, and identify the moving entity through infrared technology;

[0039] If there is a moving entity in the vehicle's blind area, record the moment t1 when the moving entity is detected, determine the positional relationship between the moving entity and the current vehicle, and continuously monitor until the next moment t2. The time interval between the moment t1 and the moment t2 is a preset value set by the operator;

[0040] Taking the position of the vehicle at the moment t1 as the coordinate origin A(0, 0), taking the positive direction of the vehicle's movement as the vertical axis, and constructing a horizontal axis perpendicular to the positive direction of the vehicle's movement and in the same plane as the vertical axis, and combining the above to form a two-dimensional plane coordinate system;

[0041] Extract the positional relationship between the moving entity and the vehicle, mark it in the two-dimensional plane coordinate system, and record the position coordinates of the moving entity at the moment t1 as B(o1, p1);

[0042] Obtain the position coordinates of the vehicle at the moment t2, which are recorded as A(x1, y1), and the position coordinates of the moving entity at the moment t2, which are recorded as B(o2, p2);

[0043] For the vehicle, taking A(0, 0) as the starting point, connecting A(x1, y1) with a ray and continuing to extend in this direction, obtaining a ray V1, and calculating the straight-line distance from A(0, 0) to A(x1, y1), which is recorded as the moving trajectory G1 of the vehicle from the moment t1 to the moment t2;

[0044] For the moving entity, taking B(o1, p1) as the starting point, connecting B(o2, p2) with a ray and continuing to extend in this direction, obtaining a ray V2. Among them, the straight-line distance between B(o1, p1) and B(o2, p2) is recorded as the moving trajectory D1 of the moving entity from the moment t1 to the moment t2;

[0045] Along the V1 direction, continuously obtain n distances identical to G1 as the n moving trajectories when the vehicle moves continuously for n + 1 moments in the current moving state, and together with G1, denote them as the vehicle moving trajectory sequence: G1, G2,..., G n , where n is a value preset by the operator;

[0046] Similarly, obtain the moving entity moving trajectory sequence: D1, D2,..., D n ;

[0047] Further analyze the vehicle moving trajectory sequence and the moving entity moving trajectory sequence, obtain the collision result, and generate a safety signal and a warning signal associated with the collision result.

[0048] As a further solution of the present invention, in the decision warning module, the specific method for determining whether the running trajectories between the moving entity and the vehicle collide and generating a warning signal or a safety signal based on the collision result further includes:

[0049] Judge whether the moving entity moving trajectory sequence and the vehicle moving trajectory sequence intersect. If there is no intersection, it is regarded that the vehicle and the moving entity will not collide, and a safety signal is generated without further processing;

[0050] If there is an intersection, extract the intersection moment t of the moving trajectory of the moving entity and the vehicle moving trajectory 交 ;

[0051] Then obtain the position coordinate of the vehicle on the moving trajectory at this intersection moment t 交 , denoted as and the position coordinate B( ) of the moving entity on the moving trajectory, and calculate and The straight-line distance K between them;

[0052] If K≥K 阈 , it is regarded that the vehicle and the moving entity will not collide, and a safety signal is generated, where K 阈 Is the safety straight-line distance preset by the operator;

[0053] If K<K 阈 , it is regarded that the vehicle and the moving entity will collide, and a warning signal is generated.

[0054] As a further solution of the present invention, in the decision warning module, if a warning signal is generated, an information warning reminder is sent to the driver through the vehicle dashboard or the central control screen, and the moment T1 when the warning reminder is sent is recorded, and continuous monitoring is carried out until the moment T2. If T2 - T 阈 >T1, it is regarded that the driver has not made corresponding adjustment actions within the specified time, where T 阈The time threshold preset for the operator;

[0055] Furthermore, a sound alarm is sent to the driver through the speaker equipped on the vehicle, the moment T3 when the sound alarm is sent is recorded, and continuous monitoring is carried out until moment T4. If T4 - T 阈 > T3, it is regarded that the driver has not made corresponding adjustment actions within the specified time, triggering the adaptive adjustment strategy, and the intelligent driving system automatically makes adjustment actions;

[0056] If it is monitored within the specified time that the driver makes corresponding adjustment actions, at this time, the adjusted vehicle running trajectory and the running trajectory of the moving entity are recorded, and it is judged again whether the vehicle running trajectory and the running trajectory of the moving entity will collide, and a warning signal and a safety signal are generated, and the above steps are repeated;

[0057] If it is monitored that the corresponding moving entity leaves the driving blind area of the current vehicle, the monitoring of the moving entity is stopped.

[0058] Advantages of the present invention:

[0059] (1) Through a variety of sensors, the present invention can comprehensively cover the blind area and obtain high-precision blind area-related data to monitor the running trajectories of moving entities and vehicles in the driving blind area in real time, can quickly identify potential collision risks, and can avoid accidents through warnings or automatic adjustment actions when necessary, and can reduce the incidence of traffic accidents caused by blind areas to a certain extent;

[0060] (2) Through technical means such as the sliding window method, normalization processing, and area threshold judgment, the present invention realizes the accurate judgment and dynamic adjustment of the visibility state of the vehicle's environment. The noise and extreme values are removed by the sliding window method and triple standard deviation filtering to ensure the stability and reliability of the data; secondly, the long-term trend of visibility is extracted through normalization and mean processing, and the visibility state is dynamically judged whether it is normal based on the area quantization method of the visibility change broken line; on this basis, the system dynamically adjusts the acquisition frequencies of the camera, ultrasonic sensor, and millimeter wave radar according to the visibility state, adopts the conventional frequency under normal visibility to save resources, and enables the unconventional frequency under abnormal visibility to improve the density and accuracy of data acquisition; this method not only improves the intelligence and adaptive ability of the system, but also reduces misjudgment and misoperation, optimizes the data acquisition efficiency, ensures efficient operation under different environmental conditions, provides reliable guarantee for driving safety, and significantly reduces the incidence of traffic accidents caused by blind areas;

[0061] (3) The present invention effectively identifies collision risks and improves the accuracy of early warnings by accurately judging whether the moving entity intersects with the vehicle trajectory and combining the comparison of the straight-line distance with the preset safety distance; adopts a hierarchical early warning mechanism, first through visual reminders, then sound alarms, and finally automatic adjustment, which not only respects the driver's subjective initiative but also ensures safety in emergency situations; continuously monitors the driver's reaction and the vehicle's operating state, dynamically adjusts the early warning strategy, and enhances the intelligence and adaptability of the system; stops monitoring in a timely manner after the moving entity leaves the blind spot, reduces unnecessary interference, and optimizes the driving experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] The present invention will be further described below with reference to the accompanying drawings.

[0063] Figure 1 is a schematic structural diagram of the system of the present invention;

[0064] Figure 2 is a schematic flow diagram of the method described in Embodiment 2 of the present invention;

[0065] Figure 3 is a schematic flow diagram of the method described in Embodiment 3 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0066] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0067] Embodiment 1

[0068] An artificial intelligence-assisted intelligent driving system based on a driving blind spot, as Figure 1 shown, specifically includes the following:

[0069] This system mainly includes three major modules, namely an environment perception module, a data acquisition module, and a decision-making and early warning module. The environment perception module is as described below:

[0070] The environment perception module contains an infrared sensor for measuring visibility. The visibility of the current vehicle's environment is jointly monitored by multiple infrared sensors equipped in different directions of the vehicle body, and the measured visibility is transmitted to the edge computing node equipped in this module for analysis to lock the visibility state of the current vehicle's environment and generate the acquisition frequency of the data acquisition sensor matching the current visibility state to save computing resources; when the visibility is abnormal, the acquisition frequency of the data acquisition sensor switches to an unconventional frequency to improve the density and accuracy of data acquisition and ensure the reliability of blind spot monitoring;

[0071] Then, the visibility status and the acquisition frequency of the data acquisition sensor are jointly transmitted to the data acquisition module.

[0072] The data acquisition module receives in real time the acquisition frequency of the data acquisition sensor transmitted by the environment perception module and adjusts the acquisition frequencies of the cameras, ultrasonic sensors, and millimeter wave radars in this module;

[0073] The blind - spot images of the vehicle are obtained through the camera, and the heat sources in the blind - spot images are identified using infrared technology (including pedestrians and vehicles) to identify the moving entities in the blind - spot images; the ultrasonic sensors and millimeter wave radars determine the real - time positional relationship between the moving entities in the driving blind - spot and the vehicle in real time.

[0074] The decision - making and warning module obtains the running trajectories of the moving entities in the blind - spot and the moving trajectory of the current vehicle, and combines to judge whether there is a possibility of collision of the moving trajectories. If it is monitored that the vehicle and the moving entity may collide within the specified time, a warning message is generated and sent to the driver through the vehicle display screen or the central control screen for information reminder, and during this period, the adjustment actions made by the driver are monitored in real time. If no adjustment action made by the driver is monitored within the specified time, further reminder is given to the driver in the form of sound and light alarm, and the adjustment actions made by the driver within the specified time are continuously monitored. If the driver still does not make any adjustment action within the specified time, the intelligent driving system (the adjustment actions made by the intelligent driving system are prior art and will not be elaborated here) automatically takes over the vehicle and makes corresponding adjustment actions;

[0075] If the driver makes an adjustment action within any specified time, re - enter the judgment of moving - trajectory collision. If there is still a collision risk, continue to repeat the above steps of sending warning reminders. If there is no collision risk, generate a safety message and do not perform any processing;

[0076] Repeat the above steps until the moving entity leaves the driving blind - spot and stop the monitoring operation.

[0077] Embodiment 2

[0078] This embodiment discloses a method for generating the acquisition frequency of the corresponding data acquisition sensor according to the visibility in the current environment where the vehicle is located, as Figure 2 shown, and specifically includes the following steps:

[0079] The environmental perception module includes several infrared sensors equipped in all directions of the current vehicle body. By collecting the visibility of the environment where the current vehicle is located through multiple infrared sensors and performing an averaging process, the average visibility of the environment where the vehicle is currently located is obtained. Then, using the current moment as the starting moment for measuring visibility, all the average visibilities within a time t are continuously collected and denoted as the visibility set Q. The obtained visibility set Q is transmitted through the bus to the marginalized computing node equipped in this module for further analysis, where t is a time value preset by the operator;

[0080] Based on the visibility set Q of the environment where the vehicle is located within a determined time t, first, instantaneous noise is removed by the sliding window method, and then extreme values in the visibility set Q are filtered using the three - standard - deviation principle, and the finally processed visibility set Q is obtained. Then, the visibility set Q is normalized to obtain the normalized visibility set. After normalizing the visibility set, a mean process is performed on the normalized visibility set to obtain the mean visibility within the current time t, denoted as

[0081] Then, the infrared sensors are used to continuously monitor the visibility of the environment where the current vehicle is located within j time intervals of t and repeat the above - mentioned steps to obtain j mean visibilities, which are denoted as the mean visibility set in chronological order, expressed as: where j is a counting index starting from 1; where j is a counting index starting from 1;

[0082] A two - dimensional coordinate system of mean visibility - time line is constructed, where the mean visibility is used as the vertical axis of the two - dimensional coordinate system and the time line is used as the horizontal axis of the two - dimensional coordinate system. Each mean visibility in the obtained mean visibility set is marked in the constructed two - dimensional coordinate system, and using the first mean visibility as the starting point, all the mean visibilities are successively connected using short lines until the last mean visibility and the entire connected line is used as the visibility change broken line, denoted as Q ～ ;

[0083] Then, a straight line passing through the first mean visibility and perpendicular to the horizontal axis is constructed and denoted as the first - moment line L1. Then, the first - moment line L1 is copied and translated one time - t scale in the positive direction of the horizontal axis to obtain a straight line passing through the second mean visibility denoted as the second - moment line L2;

[0084] At this time, the visibility change broken line, the horizontal - axis first - moment line L1, and the second - moment line L2 will form a closed area. The area of this closed area is obtained and denoted as S1, which is used as the visibility between the first time t and the second time t after quantization;

[0085] Again, according to the above method, copy the time line at the third moment until the time line at the j-th moment, and successively obtain the areas of the enclosed regions formed by two adjacent time lines, the horizontal axis, and the visibility change broken line. A total of j - 1 enclosed regions are obtained, and the corresponding j - 1 areas are calculated. They are successively recorded as the enclosed region area sequence according to the order of the time line: S1, S2,..., S j-1 , where S i is the area of any enclosed region in the enclosed region area sequence, and i does not exceed j;

[0086] Obtain the area threshold S 阈 formulated by the operator based on experience, and compare the area S j-1 of any enclosed region in the obtained enclosed region area sequence: S1, S2,..., S i with the area threshold S 阈 formulated by the operator based on experience;

[0087] If the area S i of any enclosed region is greater than or equal to the area threshold S 阈 , then calibrate the visibility of the environment where the current vehicle is located as normal visibility;

[0088] If the area S i of any enclosed region is less than the area threshold S 阈 , then calibrate the visibility of the environment where the current vehicle is located as abnormal visibility;

[0089] Determine the acquisition frequencies of the data acquisition sensors in the normal state, namely the camera, ultrasonic sensor, and millimeter wave radar, respectively, and combine them. Denote the combined result as the normal frequency feature R = R1, R2, R3, where R1 is the acquisition frequency of the camera in the normal state, R2 is the acquisition frequency of the ultrasonic sensor in the normal state, and R3 is the acquisition frequency of the millimeter wave radar in the normal state;

[0090] Then, obtain the acquisition frequencies of each data acquisition sensor in the non - normal state preset by the operator (the acquisition frequencies of each data acquisition sensor in the non - normal state are the highest acquisition frequencies of each sensor to cope with harsh environments), and combine them. Denote the combined result as the non - normal frequency feature R ′ = R ′ 1, R ′ 2, R ′ 3, where R ′ 1 > R1, R ′ 2 > R2, R ′ 3 > R3, R ′ 1 is the acquisition frequency of the camera in the non - normal state, R ′2 is the acquisition frequency of the ultrasonic sensor under abnormal conditions, R ′ 3 is the acquisition frequency of the millimeter-wave radar under abnormal conditions;

[0091] Based on the determined visibility state of the environment where the current vehicle is located, select the acquisition frequency of the corresponding data acquisition sensor. If the visibility state of the environment where the current vehicle is located is normal visibility, enable the conventional frequency feature R as the acquisition frequency of the data acquisition sensor;

[0092] If the visibility state of the environment where the current vehicle is located is abnormal visibility, enable the abnormal frequency feature R ′ as the acquisition frequency of the data acquisition sensor;

[0093] In this embodiment, the infrared sensors in all directions of the vehicle body continuously collect visibility data, and the average visibility set is obtained through average processing; then, the sliding window method and the three-standard-deviation principle are used to denoise and filter extreme values of the average visibility set, and the mean visibility is obtained after normalization and mean processing; then, continuously monitor the mean visibility within multiple time t, construct a two-dimensional coordinate system of the mean visibility-time line, calculate the area of the closed region enclosed by the adjacent time lines and the visibility change broken line, and form an area sequence; compare these areas with a preset threshold to determine whether the visibility state is normal or abnormal; finally, select the conventional or abnormal frequency feature as the acquisition frequency of the data acquisition sensor (camera, ultrasonic sensor, millimeter-wave radar) according to the visibility state to optimize the data acquisition efficiency and accuracy.

[0094] Embodiment 3

[0095] This embodiment discloses a method for predicting and controlling collisions between a vehicle and a moving entity in a blind area, as Figure 3 shown, and specifically includes the following steps:

[0096] Use the camera equipped on the vehicle body to continuously monitor the blind area image in the driving blind area of the environment where the current vehicle is located, and combine infrared technology to identify the moving entities in the blind area image (the moving entities include other vehicles and pedestrians, and both vehicles and pedestrians will generate heat sources, and the moving entities are identified based on this feature);

[0097] If a moving entity is detected in the blind area image by using infrared technology, record the moment t1 when the moving entity is detected, and then enable dual detection by the ultrasonic sensor and the millimeter-wave radar and determine the positional relationship between the moving entity and the current vehicle, and the positional relationship includes direction and distance;

[0098] Starting from the moment t1, continuously detect until the next moment t2, where the time interval between the moment t1 and the moment t2 is a time value preset by the operator in combination with the actual situation;

[0099] Take the location of the vehicle at time t1 as the origin of the coordinate in a two-dimensional plane coordinate system, and record the position coordinate as A(0, 0). Then, take the positive direction of the vehicle's movement as the vertical axis of the two-dimensional plane coordinate system, and construct a horizontal axis perpendicular to the positive direction of the vehicle's movement and in the same plane as the positive direction of the vehicle's movement. The origin of the coordinate, the horizontal axis, and the vertical axis determined above together form a two-dimensional plane coordinate system;

[0100] Based on the dual detection of the ultrasonic sensor and the millimeter-wave radar, determine the positional relationship between the moving entity and the current vehicle, mark the moving entity in the constructed two-dimensional plane coordinate system, and record the position coordinate of the current moving entity as B(o1, p1);

[0101] Then, at time t2, obtain the moving distance and moving direction of the current vehicle relative to the origin of the coordinate (the position coordinate of the vehicle on the two-dimensional plane coordinate system at time t1), mark them on the constructed two-dimensional plane coordinate system, and record the position coordinate where the current vehicle is located as A(x1, y1);

[0102] Then, at time t2, obtain the moving distance and moving direction of the moving entity detected relative to B(o1, p1) (the position coordinate of the moving entity on the two-dimensional plane coordinate system at time t1) on the constructed two-dimensional plane coordinate system, and record the position coordinate of the current moving entity as B(o2, p2);

[0103] Based on the determined time interval from time t1 to time t2, the two position coordinates of the vehicle A(0, 0) and A(x1, y1), and the two position coordinates of the moving entity B(o1, p1) and B(o2, p2); among them, for the vehicle, take the position coordinate A(0, 0) as the starting point of the ray, construct a ray passing through the position coordinate A(x1, y1), and continue to extend in this direction to obtain ray V1, and obtain the straight-line distance from the position coordinate A(0, 0) to the position coordinate A(x1, y1), which is recorded as the moving trajectory G1 of the current vehicle associated with the time interval from time t1 to time t2;

[0104] For the moving entity, similarly, take the position coordinate B(o1, p1) of the moving entity on the coordinate system at time t1 as the starting point, construct a ray passing through the position coordinate B(o2, p2), and continue to extend in this direction to obtain ray V2, and obtain the straight-line distance between the position coordinate B(o1, p1) and the position coordinate B(o2, p2), which is recorded as the moving trajectory D1 of the current moving entity associated with the time interval from time t1 to time t2;

[0105] Obtain the ray V1 associated with the current vehicle, and continuously obtain n distances along the direction of ray V1 that are the same as the moving trajectory G1 of the vehicle within the time period from time t1 to time t2. These n distances are used as the n moving trajectories of the vehicle when it continuously moves to the (n + 1)-th moment in the current moving state. Together with the first moving trajectory G1 of the vehicle, they are denoted as the vehicle moving trajectory sequence: G1, G2,..., G n , where n is a value preset by the operator;

[0106] Then obtain the ray V2 associated with the current moving entity, and continuously obtain n distances along the direction of ray V2 that are the same as the moving trajectory D1 of the moving entity within the time period from time t1 to time t2. These n distances are used as the n moving trajectories of the moving entity when it continuously moves to the (n + 1)-th moment in the current moving state. Together with the first moving trajectory D1 of the moving entity, they are denoted as the moving entity moving trajectory sequence: D1, D2,..., D n ;

[0107] Determine whether the moving entity moving trajectory sequence: G1, G2,..., G n and the vehicle moving trajectory sequence: D1, D2,..., D n will intersect. If there is no intersection, it is considered that the vehicle and the moving entity are at a safe distance and there will be no collision. At this time, a safety signal is generated and no further processing is done;

[0108] If an intersection is detected, extract the intersection time of the vehicle moving trajectory and the moving entity moving trajectory and denote it as t 交 , and the position coordinates of the intersection position, denoted as A - B (the position coordinates A - B of the intersection position are both on ray V1 and on ray V2);

[0109] Because when the moving trajectory of the vehicle intersects with the moving trajectory of the moving entity, at this intersection time t 交 it is not certain that the vehicle and the moving entity will collide. Therefore, it is necessary to obtain the position coordinates of the vehicle on the vehicle's moving trajectory at the intersection time t 交 , denoted as and the position coordinates of the moving entity on the moving entity's moving trajectory at the intersection time t 交 Then obtain the straight-line distance K between the position coordinates and the position coordinates and the position coordinates ;

[0110] Obtain the preset safe straight-line distance K 阈 set by the operator, and compare the straight-line distance K with the preset safe straight-line distance K 阈Perform a comparison operation. If the distance K of the straight line is greater than or equal to the safe straight-line distance K 阈 , it is regarded that there is a safe distance between the vehicle and the moving entity, and the vehicle and the moving entity will not collide, and a safety signal is generated;

[0111] If the distance K of the straight line is less than the safe straight-line distance K preset by the operator 阈 , it is regarded that the vehicle and the moving entity will collide, and a warning signal is generated;

[0112] This embodiment discloses a method for predicting and controlling the collision between a vehicle and a moving entity in a blind area; by combining a vehicle body camera with infrared technology to identify a moving entity in the blind area (such as other vehicles or pedestrians), if a moving entity is detected, use ultrasonic sensors and millimeter-wave radars to determine its positional relationship with the vehicle; within a preset time interval, construct a two-dimensional plane coordinate system, record the position coordinates and moving trajectories of the vehicle and the moving entity; by predicting the moving trajectories of the vehicle and the moving entity at multiple future moments, determine whether the trajectories will intersect; if the trajectories intersect and the distance at the intersection is less than the preset safe distance, generate a warning signal; if the intersection distance is greater than or equal to the safe distance or the trajectories do not intersect, generate a safety signal; through real-time monitoring and prediction, this method effectively warns of potential collision risks and ensures driving safety.

[0113] Some of the data in the above formulas are numerically calculated after removing their dimensions, and the content not described in detail in this specification belongs to the prior art well-known to those skilled in the art.

[0114] The above content is only an example and explanation of the present invention. Those skilled in the art of this technology can make various modifications or supplements to the described specific embodiments or use similar methods for substitution, as long as they do not deviate from the invention or exceed the scope defined by this claim book, they should all belong to the protection scope of the present invention.

[0115] It should be stated that: all user data collected in this application is collected with the consent and authorization of the user. And the uses of user data are all legal and compliant, and the use and processing of user data comply with the relevant laws, regulations and standards of the relevant regions.

Claims

1. An artificial intelligence-assisted intelligent driving system based on a driving blind spot, characterized in that The system includes the following: An environment perception module that real-time obtains the visibility of the vehicle's surrounding environment, analyzes and locks the visibility state of the current vehicle's surrounding environment in real-time through the edge computing nodes equipped in the environment perception module, and adaptively adjusts the acquisition frequency of the data acquisition sensors based on the visibility state; A data acquisition module that receives the visibility state determined by the environment perception module and the acquisition frequency of the data acquisition sensors adapted to this visibility state; Real-time obtains the blind area related data in the driving blind area through the data acquisition sensors; A decision and warning module that analyzes the blind area related data to determine whether there is a collision in the running trajectories between the moving entity and the vehicle, and generates a warning signal or a safety signal based on the collision result; If it is a warning signal, a warning reminder is sent to the driver, and the adjustment actions made by the driver are monitored in real-time; If the driver does not make any adjustment actions within the specified time, the decision and warning module sends an audible alarm to the driver. If the driver still does not make any adjustment actions within the specified time, the intelligent driving system automatically makes adjustment actions; If it is monitored that the driver makes corresponding adjustment actions, it is determined again whether there will be a collision in the running trajectories between the moving entity and the vehicle, and corresponding safety signals and warning signals are generated; If a safety signal is generated, no processing is performed; If the moving entity leaves the driving blind area, the monitoring stops.

2. The artificial intelligence-assisted intelligent driving system based on a driving blind spot according to claim 1, wherein The data acquisition sensors in the data acquisition module include: cameras, ultrasonic sensors, millimeter wave radars; The blind area related data includes the blind area images of the vehicle obtained in real-time by the camera and the moving entities identified in the blind area images, as well as the real-time position relationship between the moving entities and the vehicle in the driving blind area determined by the ultrasonic sensors and millimeter wave radars.

3. The artificial intelligence-assisted intelligent driving system based on a driving blind area according to claim 1, wherein In the environment perception module, the specific method for analyzing and locking the visibility state of the current vehicle's surrounding environment is: Collect the visibility of the current vehicle's surrounding environment, take the current moment as the start moment, continuously collect the visibility within the time t, and summarize it as the visibility set Q, and transmit it to the edge computing node, where t is a preset value set by the operator; Process the visibility set Q using the sliding window method and the three-standard-deviation principle; Extract the processed visibility set Q, normalize it, and take the mean to obtain the mean visibility within time t Repeat the above method to obtain the average visibility within consecutive j time intervals t Denote it as the average visibility set: where j is the counting index, starting from 1; Taking the mean visibility as the vertical axis and the timeline as the horizontal axis, a two-dimensional coordinate system is constructed, and the set of mean visibilities is marked in the two-dimensional coordinate system. Starting from and successively connecting adjacent mean visibilities backward with short lines until the visibility change broken line Q is obtained. ～ ; For Q ～ Further analyze and lock the visibility states associated with different time points.

4. The artificial intelligence-assisted intelligent driving system based on a driving blind area according to claim 3, wherein, In the environment perception module, the specific method for analyzing and locking the visibility state of the current vehicle's surrounding environment also includes: For Q ～ , after construct a straight line perpendicular to the horizontal axis, denote it as the line L1 at the first moment, copy L1 and translate it one time t in the positive direction of the horizontal axis to obtain a straight line passing through , denote it as the line L2 at the second moment; Obtain the area S1 of the enclosed region formed by the horizontal axis, Q ～ , L1, and L2; By analogy, obtain the areas of j - 1 closed regions, denoted as the closed region area sequence: S1, S2,..., S j-1 , where S i is any one in the closed region area sequence, and i does not exceed j; Extract the area threshold S preset by the operator 阈 , and compare the area S of any closed area in the closed area area sequence i with the area threshold S preset by the operator 阈 ; If the area S of any closed area i ≥ S 阈 , then the visibility of the environment where the vehicle is located is normal visibility; If S i <S 阈 , then the visibility of the environment where the vehicle is located is abnormal visibility.

5. The artificial intelligence-assisted intelligent driving system based on a driving blind area according to claim 4, wherein, In the environment perception module, the specific method for adaptively adjusting the acquisition frequency of the data acquisition sensors based on the visibility state is: Lock the acquisition frequencies R1, R2, R3 of the camera, ultrasonic sensor, and millimeter wave radar in their respective normal states, and the combination is denoted as the normal frequency feature R = R1, R2, R3; Obtain the acquisition frequencies of each data acquisition sensor in the non-conventional state preset by the operator, and the combination is denoted as the non-conventional frequency feature R ′ = R ′ 1, R ′ 2, R ′ 3, where R ′ 1 > R1, R ′ 2 > R2, R ′ 3 > R3; Extract the visibility state of the current vehicle's surrounding environment. If the visibility is normal, enable R as the acquisition frequency; If the visibility is abnormal, enable R ′ as the acquisition frequency.

6. The artificial intelligence-assisted intelligent driving system based on a driving blind area according to claim 5, wherein, In the decision and warning module, the specific method for determining whether there is a collision in the running trajectories between the moving entity and the vehicle and generating a warning signal or a safety signal based on the collision result is: Extract the blind area images of the current vehicle monitored and captured by the camera in real-time, and identify the moving entities through infrared technology; If there is a moving entity in the vehicle's blind spot, record the moment t1 when the moving entity is detected, determine the positional relationship between the moving entity and the current vehicle, and continuously monitor until the next moment t2, where the time interval between moment t1 and moment t2 is a preset value set by the operator; Taking the position of the vehicle at moment t1 as the coordinate origin A(0,0), with the positive direction of vehicle movement as the vertical axis, construct a horizontal axis perpendicular to the positive direction of vehicle movement and in the same plane as the vertical axis, and combine the above to form a two-dimensional plane coordinate system; Extract the positional relationship between the moving entity and the vehicle and mark it in the two-dimensional plane coordinate system. The position coordinates of the moving entity at moment t1 are denoted as B(o1,p1); Obtain the position coordinates of the vehicle at moment t2, denoted as A(x1,y1), and the position coordinates of the moving entity at moment t2, denoted as B(o2,p2); For the vehicle, starting from A(0,0), connect A(x1,y1) with a ray and continue to extend in this direction to obtain ray V1. Calculate the straight-line distance from A(0,0) to A(x1,y1), denoted as the moving trajectory G1 of the vehicle from moment t1 to moment t2; For the moving entity, taking B(o1,p1) as the starting point, connect B(o2,p2) with a ray and continue to extend in this direction to obtain ray V2. Among them, the straight-line distance between B(o1,p1) and B(o2,p2) is denoted as the moving trajectory D1 of the moving entity from moment t1 to moment t2; Along the V1 direction, continuously obtain n distances identical to G1 as the n moving trajectories when the vehicle continuously moves for n + 1 moments in the current moving state, and together with G1, denote them as the vehicle moving trajectory sequence: G1, G2,..., G n , where n is a value preset by the operator; Similarly, the moving trajectory sequence of the moving entity is obtained: D1, D2,..., D n ; Further analyze the vehicle moving trajectory sequence and the moving entity moving trajectory sequence to obtain the collision result, and generate a safety signal and a warning signal associated with the collision result.

7. The artificial intelligence-assisted intelligent driving system based on a driving blind area according to claim 6, wherein In the decision warning module, the specific method of determining whether the running trajectories between the moving entity and the vehicle collide and generating a warning signal or a safety signal based on the collision result also includes: Judge whether the moving entity moving trajectory sequence and the vehicle moving trajectory sequence intersect. If there is no intersection, it is considered that the vehicle and the moving entity will not collide, and a safety signal is generated without further processing; If there is an intersection, extract the intersection time t between the movement trajectory of the moving entity and the vehicle movement trajectory 交 ; Obtain the position coordinates of the vehicle on the moving trajectory at this intersection time t 交 and denote it as as well as the position coordinates of the moving entity on the moving trajectory Calculate and the straight-line distance K between them; If K≥K 阈 , it is considered that the vehicle will not collide with the moving entity, and a safety signal is generated, where K 阈 is the safety straight-line distance preset by the operator; If K < K 阈 , it is considered that the vehicle will collide with the moving entity, and a warning signal is generated.

8. The artificial intelligence-assisted intelligent driving system based on a driving blind area according to claim 7, characterized in that, In the decision-making early warning module, if a warning signal is generated, an information warning reminder is sent to the driver through the vehicle instrument panel or the central control screen, and the moment T1 when the warning reminder is sent is recorded, and continuous monitoring is carried out until the moment T2. If T2 - T 阈 > T1, it is regarded that the driver has not made corresponding adjustment actions within the specified time, where T 阈 is the time threshold preset by the operator; Further, a sound alarm is issued to the driver through the vehicle's equipped speaker, the moment T3 when the sound alarm is issued is recorded, and continuous monitoring is carried out until moment T4. If T4 - T 阈 > T3, it is considered that the driver has not made corresponding adjustment actions within the specified time, triggering the adaptive adjustment strategy, and the intelligent driving system automatically makes adjustment actions; If it is detected that the driver makes a corresponding adjustment action within the specified time, record the adjusted vehicle running trajectory and the running trajectory of the moving entity at this time, and judge again whether the vehicle running trajectory and the running trajectory of the moving entity will collide, and generate a warning signal and a safety signal, and repeat the above steps; If it is detected that the corresponding moving entity leaves the driving blind spot of the current vehicle, stop monitoring the moving entity.

Images