CN117912242A

CN117912242A - Steering intention recognition method, device, computer equipment and storage medium

Info

Publication number: CN117912242A
Application number: CN202311867779.2A
Authority: CN
Inventors: 凡俊生; 杨唐涛; 王邓江
Original assignee: Suzhou Wanji Iov Technology Co ltd
Current assignee: Suzhou Wanji Iov Technology Co ltd
Priority date: 2023-12-29
Filing date: 2023-12-29
Publication date: 2024-04-19

Abstract

The application relates to a steering intention recognition method, a device, computer equipment and a storage medium, which are used for preliminarily acquiring a steering intention set of a vehicle according to lane function information and lanes where the vehicle is located by acquiring scene data of a current intersection and track data of the vehicle in the current intersection, processing the track data through a pre-trained neural network model to obtain probability sets of steering intentions of the vehicle, determining the steering intention of the vehicle according to the steering intention sets and the probability sets, accurately recognizing the steering intention of the vehicle and improving the driving safety of the vehicle.

Description

Steering intention recognition method, device, computer equipment and storage medium

Technical Field

The present application relates to the field of autopilot technology, and in particular, to a steering intention recognition method, apparatus, computer device, and storage medium.

Background

In the process of passing through an intersection, the driving intention of the vehicle is accurately identified, so that traffic accidents can be effectively avoided, wherein the steering intention refers to the motor way in which the vehicle passes through the intersection when entering the intersection, such as left turn, straight run and right turn.

In the related art, the steering intention is primarily judged based on the vehicle information acquired by the vehicle-mounted sensing sensor, however, the traffic environment of the existing intersection is complex, more shared lanes exist, and the steering intention of the vehicle is difficult to accurately identify.

Disclosure of Invention

In view of the foregoing, it is desirable to provide a steering intention recognition method, apparatus, computer device, computer-readable storage medium, and computer program product that are capable of accurately recognizing a steering intention of a vehicle.

In a first aspect, the present application provides a steering intention recognition method, including:

Acquiring road side perception data of a current intersection; the road side perception data comprise scene data of the current intersection and track data of vehicles in the current intersection; the scene data comprises lane lines, lane position information and lane function information; the track data comprises position information, speed information, course angle information and lanes of the vehicle at each moment;

acquiring a steering intention set of a vehicle according to lane function information and a lane where the vehicle is located; the steering intent set includes at least one of a left turn, a straight turn, and a right turn;

Processing the track data through a pre-trained neural network model to obtain probability sets of steering intentions of the vehicle;

And determining the steering intention of the vehicle according to the steering intention set and the probability set.

In one embodiment, the training process of the neural network model includes:

acquiring historical scene data and historical track data;

Constructing a steering category boundary according to the historical scene data;

acquiring track sequence data corresponding to each steering intention according to the steering category boundary and the historical track data;

The neural network model is trained based on the trajectory sequence data.

In one embodiment, the step of constructing a turn category boundary from historical scene data includes:

Determining intersection points of the central line, the left road side line and the right road side line of each road section and corresponding extension lines of the parking line according to the historical scene data to obtain a central point, a left side edge point and a right side edge point of the road section;

and under the condition that each road section has a corresponding straight-going outlet, taking a connecting line between a road section central point corresponding to the road section and a road section central point corresponding to the corresponding straight-going outlet as a left-turning dividing line, taking a left-turning dividing line corresponding to the road section intersected with a corresponding extending line of the road section as a straight-going dividing line, and taking a connecting line between a right edge point of the road section and a left edge point corresponding to the corresponding straight-going outlet as a right-turning dividing line.

In one embodiment, the method further comprises:

and under the condition that the corresponding straight-going exit does not exist in the road section, extending the central line of the road section to be intersected with the corresponding opposite-side road edge line, taking the extension line of the central line of the road section as a left-turning boundary, extending the right-side road edge line of the road section to be intersected with the corresponding opposite-side road edge line, and taking the extension line of the right-side road edge line as a right-turning boundary.

In one embodiment, the step of obtaining track sequence data corresponding to each steering intention according to the steering category boundary and the historical track data includes:

Acquiring a first intersection point of a historical track and a parking line corresponding to a road section; the first intersection point comprises at least one of a left turn starting point, a straight-going starting point and a right turn starting point;

For each boundary of the road section, acquiring a second intersection point of the boundary corresponding to the first intersection point in the history track;

Aiming at each historical track, acquiring track sequence data and characteristics corresponding to each sequence data point in the track sequence data from the historical track according to the corresponding first intersection point and second intersection point to obtain track sequence data corresponding to each steering intention; the data points in the trace sequence data corresponding to the ends of the time series are located between the first intersection point and the second intersection point.

In one embodiment, the step of determining the steering intent of the vehicle from the set of steering intents and the set of probabilities comprises:

in the case where there is only one type of intention in the steering intention set, determining the intention as the steering intention of the vehicle;

when at least two types of intentions exist in the steering intention set and probabilities in probability sets corresponding to the intentions are different, determining the intention corresponding to the maximum probability value in the probability sets as the steering intention of the vehicle;

under the condition that at least two types of intents exist in the steering intention set and probabilities in the probability set corresponding to the at least two types of intents are the same, acquiring steering priority corresponding to each type of intents, and determining the intention with the highest steering priority as the steering intention of the vehicle

In a second aspect, the present application also provides a steering intention recognition apparatus, including:

The model training module is used for training the neural network model in advance;

the data acquisition module is used for acquiring road side perception data of the current intersection; the road side perception data comprise scene data of the current intersection and track data of vehicles in the current intersection; the scene data comprises lane lines, lane position information and lane function information; the track data comprises position information, speed information, course angle information and lanes of the vehicle at each moment;

The intention acquisition module is used for acquiring a steering intention set of the vehicle according to the lane function information and the lane in which the vehicle is positioned; the steering intent set includes at least one of a left turn, a straight turn, and a right turn;

The probability acquisition module is used for processing the track data through a pre-trained neural network model to obtain probability sets of steering intentions of the vehicle;

The intention determining module is used for determining the steering intention of the vehicle according to the steering intention set and the probability set.

In one embodiment, the model training module comprises:

the track acquisition unit is used for acquiring historical scene data and historical track data;

the boundary construction unit is used for constructing a steering category boundary according to the historical scene data;

The sequence acquisition unit is used for acquiring track sequence data corresponding to each steering intention according to the steering category boundary and the historical track data;

And the model training unit is used for training the neural network model based on the track sequence data.

In a third aspect, the present application also provides a computer device comprising a memory storing a computer program and a processor implementing the method steps of any one of the first aspects when the processor executes the computer program.

In a fourth aspect, the present application also provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method steps of any of the first aspects.

In a fifth aspect, the application also provides a computer program product comprising a computer program which, when executed by a processor, implements the method steps of any of the first aspects.

According to the steering intention recognition method, the device, the computer equipment, the storage medium and the computer program product, the scene data of the current intersection and the track data of the vehicle in the current intersection are obtained, the steering intention set of the vehicle is preliminarily obtained according to the lane function information and the lane where the vehicle is located, the track data is processed through the pre-trained neural network model to obtain the probability set of each steering intention of the vehicle, the steering intention of the vehicle is determined according to the steering intention set and the probability set, the steering intention of the vehicle can be accurately recognized, and the running safety of the vehicle is improved.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the related art, the drawings that are required to be used in the embodiments or the related technical descriptions will be briefly described, and it is apparent that the drawings in the following description are only some embodiments of the present application, and other drawings may be obtained according to the drawings without inventive effort for those skilled in the art.

FIG. 1 is an application environment diagram of a method of steering intent recognition in one embodiment;

FIG. 2 is a flow diagram of a method of identifying intent to turn in one embodiment;

FIG. 3 is a schematic illustration of one of the parting lines in one embodiment;

FIG. 4 is a schematic illustration of one of the parting lines in one embodiment;

FIG. 5 is a schematic diagram of trace sequence data in one embodiment;

FIG. 6 is a flow diagram of a method of identifying intent to turn in one embodiment;

FIG. 7 is a block diagram of a steering intent recognition device in one embodiment;

fig. 8 is an internal structural diagram of a computer device in one embodiment.

Detailed Description

The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.

The steering intention recognition method provided by the embodiment of the application can be applied to an application environment shown in fig. 1. The roadside traffic device 102 and the roadside sensor 104 may be connected in a wired manner, or may be connected in a wireless communication manner, which is not particularly limited herein. The road side sensor 104 is used for acquiring road side perception data of the current intersection; the road side perception data comprise scene data of the current intersection and track data of vehicles in the current intersection; the scene data comprises lane lines, lane position information and lane function information; the trajectory data includes position information, speed information, heading angle information, and a lane in which the vehicle is located at each moment of time of the vehicle. The road side traffic device 102 is configured to obtain road side perception data from the road side sensor 104; acquiring a steering intention set of a vehicle according to lane function information and a lane where the vehicle is located; the steering intent set includes at least one of a left turn, a straight turn, and a right turn; processing the track data through a pre-trained neural network model to obtain probability sets of steering intentions of the vehicle; and determining the steering intention of the vehicle according to the steering intention set and the probability set. The road side traffic device 102 may be, but not limited to, a road side unit RSU, an edge computing unit MEC, an intelligent base station, etc., where the road side unit RSU or the intelligent base station may be integrated with a computing unit, or may be connected to the edge computing unit MEC by a wire. Roadside sensors 104 include lidar, cameras, and other sensors, among others.

The traffic environment where the vehicle is located in the range of the intersection is quite complex, the accurate recognition of the steering intention of the vehicle is still a challenge, the steering intention refers to the maneuvering mode of the vehicle when the vehicle enters the intersection, such as left turn, right turn and straight run, after the vehicle enters the intersection, the steering intention can be primarily judged through the lanes where the history is located, however, more shared lanes exist in the existing intersection, such as straight run and left turn shared lanes, in this case, the steering intention of the vehicle cannot be accurately recognized, if the steering information is predicted by using only the intention recognition model based on the neural network, the recognition error is large and fluctuation exists. Based on the above, the embodiment of the application provides a steering intention recognition method, which is used for recognizing the driving intention of a vehicle based on road side perception data, does not depend on vehicle end data, comprehensively considers lane information and model real-time prediction results to determine the current steering intention of the vehicle, and ensures the accuracy of the steering intention.

In an exemplary embodiment, as shown in fig. 2, a steering intention recognition method is provided, which is illustrated by taking the application of the method to the road side traffic device 102 in fig. 1 as an example, and includes the following steps 202 to 208. Wherein:

S202: acquiring road side perception data of a current intersection; the road side perception data comprise scene data of the current intersection and track data of vehicles in the current intersection; the scene data comprises lane lines, lane position information and lane function information; the trajectory data includes position information, speed information, heading angle information, and a lane in which the vehicle is located at each moment of time of the vehicle.

The road side traffic equipment comprises one of a road side computing unit RSU, an edge computing unit MEC, an intelligent base station and the like, and the road side sensor comprises but is not limited to a road side laser radar, a millimeter wave radar, a camera and the like. The track data includes position information, speed information (including speed and acceleration), heading angle information and lanes in which the vehicle is located at each moment of time of the vehicle.

S204: acquiring a steering intention set of a vehicle according to lane function information and a lane where the vehicle is located; the steering intent set includes at least one of a left turn, a straight turn, and a right turn.

The lane function information indicates a steering direction allowed by a corresponding lane, for example, the steering direction allowed by a left-turn lane is left-turn, and according to a lane where the vehicle is located when the vehicle is located in front of an entrance stop line of an intersection, a possible steering intention set of the vehicle can be primarily determined, for example, if the lane function information of the lane where the vehicle is located is a straight right lane, the possible steering intention type is C _v = { straight, right-turn }. Here, the lane function information does not consider the presence of a turning function.

S206: and processing the track data through a pre-trained neural network model to obtain a probability set of each steering intention of the vehicle.

The track data is used as input data of a trained neural network model, output data of the neural network model is prediction probability of each steering intention of the vehicle, and a probability set of each steering intention of the vehicle is obtained. The track data are used for extracting a historical track sequence of the vehicle, and the steering intention of the vehicle is predicted by extracting a continuous n-frame track sequence of the current moment of the vehicle and n-1 frames before the current moment of the vehicle, wherein each characteristic point in the sequence characteristics comprises the position, the speed, the acceleration, the course angle and the lane where the vehicle is positioned. Specifically, the sequence characteristics can be expressed as:

I_f＝[x,y,v_x,v_y,a_x,a_y,angle,lane]

V _x、v_y is the speed in the x and y directions, a _x、a_y is the acceleration in the x and y directions, angle is the course angle, lane is the lane where the vehicle is located, wherein the left lane is denoted by 0, the straight lane is denoted by 2, the straight left lane is denoted by 1, the right lane is denoted by 4, the straight right lane is denoted by 3, and the straight left and right lanes are denoted by 5.

S208: and determining the steering intention of the vehicle according to the steering intention set and the probability set.

When only one type of steering intention exists in the steering intention set, the current steering intention is determined as the final steering intention of the vehicle without considering the probability of each steering intention in the probability set. In the case that two or more types of steering intentions exist in the steering intention set, determining the final steering intention of the vehicle according to the probability of the corresponding steering intention in the probability set.

According to the steering intention recognition method, the scene data of the current intersection and the track data of the vehicle in the current intersection are obtained, the steering intention set of the vehicle is preliminarily obtained according to the lane function information and the lane where the vehicle is located, the track data is processed through the pre-trained neural network model to obtain the probability set of each steering intention of the vehicle, the steering intention of the vehicle is determined according to the steering intention set and the probability set, the steering intention of the vehicle can be accurately recognized, and the running safety of the vehicle is improved.

In one exemplary embodiment, the training process of the neural network model includes: acquiring historical scene data and historical track data; constructing a steering category boundary according to the historical scene data; acquiring track sequence data corresponding to each steering intention according to the steering category boundary and the historical track data; the neural network model is trained based on the trajectory sequence data.

The method comprises the steps of respectively extracting complete historical tracks of different graph categories according to historical track data, wherein the complete historical tracks comprise track data of a large number of vehicles which have completed steering, the steering intention comprises left-turning intention, straight-going intention and right-turning intention, the complete historical tracks refer to complete tracks detected by all vehicles in an intersection scene, and the complete tracks of the left-turning vehicles, the complete tracks of the straight-going vehicles and the complete tracks of the right-turning vehicles are respectively extracted based on position information. Further, only part of track sequence data in the historical track data is extracted as input data of the neural network model, wherein the steering category boundary is used for dividing track sequence data extraction areas of different categories of steering intents.

Further, the track sequence data of the disagreeable graph category is divided into a training set and a verification set, wherein the proportion of the track sequences of the three intention categories in the training set and the verification set is 1:1:1, the proportion of the training set and the verification set is 8:2, and the proportion relation can be set according to actual requirements, and is only an example. Training is completed based on the training set, the verification set and the built model, wherein the neural network model can be a long-short-term memory network (LSTM) model, different types of intention sequence classification is carried out by adopting a Softmax multi-classification method, different types of track sequences and corresponding labels contained in the training set and the verification set are input, the probability of different graph types to which each sequence belongs is output, and a loss function is calculated based on the input labels and the output results until the loss function converges.

In the embodiment, the historical scene data and the historical track data are acquired, the steering category boundary is constructed according to the historical scene data, the track sequence data corresponding to each steering intention is acquired according to the steering category boundary and the historical track data, and the neural network model is trained based on the track sequence data, so that the model training efficiency can be improved, the model precision is ensured, and the steering intention of the vehicle is accurately predicted.

In one exemplary embodiment, constructing a turn category boundary from historical scene data includes: determining intersection points of the central line, the left road side line and the right road side line of each road section and corresponding extension lines of the parking line according to the historical scene data to obtain a central point, a left side edge point and a right side edge point of the road section; and under the condition that each road section has a corresponding straight-going outlet, taking a connecting line between a road section central point corresponding to the road section and a road section central point corresponding to the corresponding straight-going outlet as a left-turning dividing line, taking a left-turning dividing line corresponding to the road section intersected with a corresponding extending line of the road section as a straight-going dividing line, and taking a connecting line between a right edge point of the road section and a left edge point corresponding to the corresponding straight-going outlet as a right-turning dividing line.

As shown in fig. 3, fig. 3 is a schematic diagram of a demarcation line in the case that each road section in the current intersection has a corresponding straight-going exit, where, according to historical scene data, the intersection points of the road section center line, the left road side line, and the right road side line of each road section with the corresponding extension lines of the parking line and the parking line are respectively determined, so as to obtain a corresponding road section center point, a left side edge point, and a right side edge point.

Further, taking fig. 3 as an example, the left turn boundary L1 of the road section a in fig. 3 is a line between the road section center point A2 and the road section center point corresponding to the corresponding straight-going exit, and the right turn boundary L3 of the road section a is a line between the right side edge point A3 and the left side edge point corresponding to the corresponding straight-going exit. The straight-going boundary L2 of the road segment a is a left-turn boundary of the road segment B intersecting with the extension line of the road segment a.

In this embodiment, the steering category boundary of each road section is constructed through the historical scene data, so that the extracted track sequence data is the data of the front section of the execution intention corresponding to the historical track, and the neural network model is trained based on the track sequence data, so that the steering intention of the vehicle can be accurately predicted.

In an exemplary embodiment, the method further comprises: and under the condition that the corresponding straight-going exit does not exist in the road section, extending the central line of the road section to be intersected with the corresponding opposite-side road edge line, taking the extension line of the central line of the road section as a left-turning boundary, extending the right-side road edge line of the road section to be intersected with the corresponding opposite-side road edge line, and taking the extension line of the right-side road edge line as a right-turning boundary.

As shown in fig. 4, fig. 4 is a schematic diagram of a demarcation line in the case where a corresponding straight exit does not exist in a road segment in the current intersection, and taking fig. 4 as an example for explanation, a road segment center line of the road segment E is extended to intersect with a corresponding opposite side road edge line, an extension line of the road segment center line is regarded as a left turn demarcation line S1, a right side road edge line of the road segment is extended to intersect with a corresponding opposite side road edge line, and an extension line of the right side road edge line is regarded as a right turn demarcation line S3.

It should be noted that, since the road section E does not have a corresponding straight-going exit, that is, the vehicle located on the road section E does not have a straight-going intention, only track sequence data corresponding to a left-turn intention and a right-turn intention need to be extracted when training the neural network model. Alternatively, in order to avoid the influence of the historical track data corresponding to the straight-going intention on training, a boundary corresponding to the straight-going category may be constructed, as shown in fig. 4, where the straight-going boundary S2 of the road segment E is a left-turn boundary of the road segment F intersecting with the extension line of the road segment E.

In an exemplary embodiment, obtaining track sequence data corresponding to each steering intention according to the steering category boundary and the historical track data includes: acquiring a first intersection point of a historical track and a parking line corresponding to a road section; the first intersection point comprises at least one of a left turn starting point, a straight-going starting point and a right turn starting point; for each boundary of the road section, acquiring a second intersection point of the boundary corresponding to the first intersection point in the history track; aiming at each historical track, acquiring track sequence data and characteristics corresponding to each sequence data point in the track sequence data from the historical track according to the corresponding first intersection point and second intersection point to obtain track sequence data corresponding to each steering intention; the data points in the trace sequence data corresponding to the ends of the time series are located between the first intersection point and the second intersection point.

As shown in fig. 5, fig. 5 is a schematic diagram of track sequence data with different types of intentions, taking the road section a in fig. 5 as an example, a first intersection point of a history track of each steering type and a parking line corresponding to the road section is defined as a steering starting point, that is, a left-turn starting point is t _l1 in fig. 5, a straight-turn starting point is t _s1, and a right-turn starting point is t _r1, where it is to be noted that the history track may only include tracks corresponding to left-turn, right-turn or straight-turn, or tracks corresponding to both left-turn and straight-turn, and accordingly, the first intersection point includes at least one of a left-turn starting point, a straight-turn starting point, and a right-turn starting point.

Further, a second intersection point of the history track corresponding to the left turn and the left turn boundary is defined as a left turn point, a second intersection point of the history track corresponding to the straight run and the straight run boundary is defined as a straight run point, a second intersection point of the history track corresponding to the right turn and the right turn boundary is defined as a right turn point, that is, the left turn point is t _l2, the straight run point is t _s2, and the right turn point is t _r2 in fig. 5, wherein the second intersection point is an intersection point of the boundary line corresponding to the history track and the first intersection point.

Further, for each category of historical track, extracting continuous n frames of sequence data, wherein data points corresponding to the tail ends of the time sequences in the extracted sequence data are located between a first intersection point and a second intersection point, and extracting the characteristics corresponding to each data point to obtain track sequence data corresponding to each steering intention.

In this embodiment, by acquiring a first intersection point of a historical track and a parking line corresponding to a road segment, acquiring a second intersection point of a boundary line corresponding to the first intersection point in the historical track for each boundary line of the road segment, acquiring track sequence data and features corresponding to each sequence data point in the track sequence data from the historical track according to the corresponding first intersection point and the second intersection point for each historical track, acquiring track sequence data corresponding to each steering intention, and training a neural network model based on the track sequence data, thereby ensuring accurate prediction of the steering intention of the vehicle.

In one exemplary embodiment, determining a steering intent of a vehicle from a set of steering intents and a set of probabilities includes: in the case where there is only one type of intention in the steering intention set, determining the intention as the steering intention of the vehicle; when at least two types of intentions exist in the steering intention set and probabilities in probability sets corresponding to the intentions are different, determining the intention corresponding to the maximum probability value in the probability sets as the steering intention of the vehicle; when at least two types of intentions exist in the steering intention set and probabilities in the probability set corresponding to the at least two types of intentions are the same, acquiring steering priority corresponding to each type of intentions, and determining the intention with the highest steering priority as the steering intention of the vehicle.

When the steering intention set has at least two types of intentions, the steering intention of the vehicle is determined according to the probability corresponding to each type of intention in the probability set.

Specifically, in the case where probabilities corresponding to each type of intentions in the set of turning intentions are different, an intention corresponding to the maximum probability is determined as a turning intention of the vehicle, and in the case where probabilities corresponding to each type of intentions in the set of turning intentions are the same, an intention having the highest turning priority is determined as a turning intention of the vehicle, wherein the turning priority is set according to scene data of the current intersection, and for example, it may be that a straight-going turning priority is greater than a turning priority of a right turn, which is greater than a turning priority of a left turn.

In this embodiment, by comprehensively considering the lane information and the prediction information, the steering intention of the vehicle can be accurately recognized in real time to determine the expected behavior of the vehicle.

In one exemplary embodiment, as shown in fig. 6, there is provided a steering intention recognition method including the steps of:

Historical scene data and historical track data are acquired.

Determining intersection points of the central line, the left road side line and the right road side line of each road section and corresponding extension lines of the parking line according to the historical scene data to obtain a central point, a left side edge point and a right side edge point of the road section; and under the condition that each road section has a corresponding straight-going outlet, taking a connecting line between a road section central point corresponding to the road section and a road section central point corresponding to the corresponding straight-going outlet as a left-turning dividing line, taking a left-turning dividing line corresponding to the road section intersected with a corresponding extending line of the road section as a straight-going dividing line, and taking a connecting line between a right edge point of the road section and a left edge point corresponding to the corresponding straight-going outlet as a right-turning dividing line.

Acquiring a first intersection point of a historical track and a parking line corresponding to a road section; the first intersection point comprises at least one of a left turn starting point, a straight-going starting point and a right turn starting point; for each boundary of the road section, acquiring a second intersection point of the boundary corresponding to the first intersection point in the history track; aiming at each historical track, acquiring track sequence data and characteristics corresponding to each sequence data point in the track sequence data from the historical track according to the corresponding first intersection point and second intersection point to obtain track sequence data corresponding to each steering intention; the data points in the trace sequence data corresponding to the ends of the time series are located between the first intersection point and the second intersection point.

The neural network model is trained based on the trajectory sequence data.

Acquiring road side perception data of a current intersection; the road side perception data comprise scene data of the current intersection and track data of vehicles in the current intersection; the scene data comprises lane lines, lane position information and lane function information; the trajectory data includes position information, speed information, heading angle information, and a lane in which the vehicle is located at each moment of time of the vehicle.

Acquiring a steering intention set of a vehicle according to lane function information and a lane where the vehicle is located; the steering intent set includes at least one of a left turn, a straight turn, and a right turn.

And processing the track data through a pre-trained neural network model to obtain a probability set of each steering intention of the vehicle.

In the case where there is only one type of intention in the steering intention set, determining the intention as the steering intention of the vehicle; when at least two types of intentions exist in the steering intention set and probabilities in probability sets corresponding to the intentions are different, determining the intention corresponding to the maximum probability value in the probability sets as the steering intention of the vehicle; when at least two types of intentions exist in the steering intention set and probabilities in the probability set corresponding to the at least two types of intentions are the same, acquiring steering priority corresponding to each type of intentions, and determining the intention with the highest steering priority as the steering intention of the vehicle.

In this embodiment, the scene data of the current intersection and the track data of the vehicle in the current intersection are obtained, the steering intention set of the vehicle is initially obtained according to the lane function information and the lane where the vehicle is located, the track data is processed through the pre-trained neural network model to obtain the probability set of each steering intention to which the vehicle belongs, and the steering intention of the vehicle is determined according to the steering intention set and the probability set, so that the steering intention of the vehicle can be accurately identified, and the running safety of the vehicle is improved.

It should be understood that, although the steps in the flowcharts related to the above embodiments are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in the flowcharts described in the above embodiments may include a plurality of steps or a plurality of stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of the steps or stages is not necessarily performed sequentially, but may be performed alternately or alternately with at least some of the other steps or stages.

Based on the same inventive concept, the embodiment of the application also provides a steering intention recognition device for realizing the steering intention recognition method. The implementation of the solution provided by the device is similar to the implementation described in the above method, so the specific limitation in the embodiments of the steering intention identifying device or devices provided below may be referred to the limitation of the steering intention identifying method hereinabove, and will not be repeated here.

In an exemplary embodiment, as shown in fig. 7, there is provided a steering intention recognition apparatus including: model training module 10, data acquisition module 20, intent acquisition module 30, probability acquisition module 40, and intent determination module 50, wherein:

the model training module 10 is configured to train the neural network model in advance.

The data acquisition module 20 is used for acquiring road side perception data of the current intersection; the road side perception data comprise scene data of the current intersection and track data of vehicles in the current intersection; the scene data comprises lane lines, lane position information and lane function information; the trajectory data includes position information, speed information, heading angle information, and a lane in which the vehicle is located at each moment of time of the vehicle.

An intention obtaining module 30, configured to obtain a steering intention set of the vehicle according to the lane function information and the lane in which the vehicle is located; the steering intent set includes at least one of a left turn, a straight turn, and a right turn.

The probability acquisition module 40 is configured to process the trajectory data through a neural network model trained in advance, so as to obtain a probability set of each steering intention to which the vehicle belongs.

The intention determination module 50 is configured to determine a steering intention of the vehicle based on the steering intention set and the probability set.

In one exemplary embodiment, model training module 10 includes: the system comprises a track acquisition unit, a boundary construction unit, a sequence acquisition unit and a model training unit, wherein:

And the track acquisition unit is used for acquiring the historical scene data and the historical track data.

And the boundary line construction unit is used for constructing a steering category boundary line according to the historical scene data.

And the sequence acquisition unit is used for acquiring track sequence data corresponding to each steering intention according to the steering category boundary and the historical track data.

In an exemplary embodiment, the boundary construction unit is further configured to determine, according to the historical scene data, intersections of the center line of each road segment, the left road edge line, and the right road edge line with the respective extension lines of the parking line and the parking line, and obtain a road segment center point, a left edge point, and a right edge point; and under the condition that each road section has a corresponding straight-going outlet, taking a connecting line between a road section central point corresponding to the road section and a road section central point corresponding to the corresponding straight-going outlet as a left-turning dividing line, taking a left-turning dividing line corresponding to the road section intersected with a corresponding extending line of the road section as a straight-going dividing line, and taking a connecting line between a right edge point of the road section and a left edge point corresponding to the corresponding straight-going outlet as a right-turning dividing line.

In an exemplary embodiment, the boundary construction unit is further configured to extend a road segment center line of the road segment to intersect with the corresponding opposite-side road edge line, to extend a right-side road edge line of the road segment to intersect with the corresponding opposite-side road edge line, and to extend an extension line of the right-side road edge line as the right-turn boundary, in a case where the corresponding straight-going exit is not present in the road segment.

In an exemplary embodiment, the sequence obtaining unit is further configured to obtain a first intersection of the historical track and the parking line corresponding to the road segment; the first intersection point comprises at least one of a left turn starting point, a straight-going starting point and a right turn starting point; for each boundary of the road section, acquiring a second intersection point of the boundary corresponding to the first intersection point in the history track; aiming at each historical track, acquiring track sequence data and characteristics corresponding to each sequence data point in the track sequence data from the historical track according to the corresponding first intersection point and second intersection point to obtain track sequence data corresponding to each steering intention; the data points in the trace sequence data corresponding to the ends of the time series are located between the first intersection point and the second intersection point.

In one exemplary embodiment, the intent determination module 40 is further configured to determine an intent as a steering intent of the vehicle if there is only one type of intent in the set of steering intents; when at least two types of intentions exist in the steering intention set and probabilities in probability sets corresponding to the intentions are different, determining the intention corresponding to the maximum probability value in the probability sets as the steering intention of the vehicle; when at least two types of intentions exist in the steering intention set and probabilities in the probability set corresponding to the at least two types of intentions are the same, acquiring steering priority corresponding to each type of intentions, and determining the intention with the highest steering priority as the steering intention of the vehicle.

The respective modules in the steering intention recognition apparatus described above may be implemented in whole or in part by software, hardware, and a combination thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.

In one exemplary embodiment, a computer device is provided, which may be a road side traffic device, the internal structure of which may be as shown in fig. 8. The computer device includes a processor, a memory, an input/output interface, a communication interface, a display unit, and an input means. The processor, the memory and the input/output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input/output interface. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input/output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used for carrying out wired or wireless communication with external road side traffic equipment, and the wireless mode can be realized through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. The computer program is executed by a processor to implement a steering intent recognition method. The display unit of the computer device is used for forming a visual picture, and can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, can also be a key, a track ball or a touch pad arranged on the shell of the computer equipment, and can also be an external keyboard, a touch pad or a mouse and the like.

It will be appreciated by those skilled in the art that the structure shown in FIG. 8 is merely a block diagram of some of the structures associated with the present inventive arrangements and is not limiting of the computer device to which the present inventive arrangements may be applied, and that a particular computer device may include more or fewer components than shown, or may combine some of the components, or have a different arrangement of components.

In one exemplary embodiment, a computer device is provided comprising a memory and a processor, the memory having stored therein a computer program, the processor when executing the computer program performing the steps of: acquiring road side perception data of a current intersection; the road side perception data comprise scene data of the current intersection and track data of vehicles in the current intersection; the scene data comprises lane lines, lane position information and lane function information; the track data comprises position information, speed information, course angle information and lanes of the vehicle at each moment; acquiring a steering intention set of a vehicle according to lane function information and a lane where the vehicle is located; the steering intent set includes at least one of a left turn, a straight turn, and a right turn; processing the track data through a pre-trained neural network model to obtain probability sets of steering intentions of the vehicle; and determining the steering intention of the vehicle according to the steering intention set and the probability set.

In one embodiment, a training process for a neural network model involved in executing a computer program comprises: acquiring historical scene data and historical track data; constructing a steering category boundary according to the historical scene data; acquiring track sequence data corresponding to each steering intention according to the steering category boundary and the historical track data; the neural network model is trained based on the trajectory sequence data.

In one embodiment, constructing the turn category boundaries from historical scene data involved in executing the computer program by the processor includes: determining intersection points of the central line, the left road side line and the right road side line of each road section and corresponding extension lines of the parking line according to the historical scene data to obtain a central point, a left side edge point and a right side edge point of the road section; and under the condition that each road section has a corresponding straight-going outlet, taking a connecting line between a road section central point corresponding to the road section and a road section central point corresponding to the corresponding straight-going outlet as a left-turning dividing line, taking a left-turning dividing line corresponding to the road section intersected with a corresponding extending line of the road section as a straight-going dividing line, and taking a connecting line between a right edge point of the road section and a left edge point corresponding to the corresponding straight-going outlet as a right-turning dividing line.

In one embodiment, the processor when executing the computer program further performs the steps of: and under the condition that the corresponding straight-going exit does not exist in the road section, extending the central line of the road section to be intersected with the corresponding opposite-side road edge line, taking the extension line of the central line of the road section as a left-turning boundary, extending the right-side road edge line of the road section to be intersected with the corresponding opposite-side road edge line, and taking the extension line of the right-side road edge line as a right-turning boundary.

In one embodiment, the processor, when executing the computer program, obtains track sequence data corresponding to each steering intention according to the steering category boundary and the historical track data, including: acquiring a first intersection point of a historical track and a parking line corresponding to a road section; the first intersection point comprises at least one of a left turn starting point, a straight-going starting point and a right turn starting point; for each boundary of the road section, acquiring a second intersection point of the boundary corresponding to the first intersection point in the history track; aiming at each historical track, acquiring track sequence data and characteristics corresponding to each sequence data point in the track sequence data from the historical track according to the corresponding first intersection point and second intersection point to obtain track sequence data corresponding to each steering intention; the data points in the trace sequence data corresponding to the ends of the time series are located between the first intersection point and the second intersection point.

In one embodiment, determining a steering intent of a vehicle from a set of steering intents and a set of probabilities involved in executing a computer program comprises: in the case where there is only one type of intention in the steering intention set, determining the intention as the steering intention of the vehicle; when at least two types of intentions exist in the steering intention set and probabilities in probability sets corresponding to the intentions are different, determining the intention corresponding to the maximum probability value in the probability sets as the steering intention of the vehicle; when at least two types of intentions exist in the steering intention set and probabilities in the probability set corresponding to the at least two types of intentions are the same, acquiring steering priority corresponding to each type of intentions, and determining the intention with the highest steering priority as the steering intention of the vehicle.

In one embodiment, a computer readable storage medium is provided having a computer program stored thereon, which when executed by a processor, performs the steps of: acquiring road side perception data of a current intersection; the road side perception data comprise scene data of the current intersection and track data of vehicles in the current intersection; the scene data comprises lane lines, lane position information and lane function information; the track data comprises position information, speed information, course angle information and lanes of the vehicle at each moment; acquiring a steering intention set of a vehicle according to lane function information and a lane where the vehicle is located; the steering intent set includes at least one of a left turn, a straight turn, and a right turn; processing the track data through a pre-trained neural network model to obtain probability sets of steering intentions of the vehicle; and determining the steering intention of the vehicle according to the steering intention set and the probability set.

In one embodiment, a training process for a neural network model involved when a computer program is executed by a processor comprises: acquiring historical scene data and historical track data; constructing a steering category boundary according to the historical scene data; acquiring track sequence data corresponding to each steering intention according to the steering category boundary and the historical track data; the neural network model is trained based on the trajectory sequence data.

In one embodiment, constructing the turn category boundaries from historical scene data, which is involved when the computer program is executed by the processor, includes: determining intersection points of the central line, the left road side line and the right road side line of each road section and corresponding extension lines of the parking line according to the historical scene data to obtain a central point, a left side edge point and a right side edge point of the road section; and under the condition that each road section has a corresponding straight-going outlet, taking a connecting line between a road section central point corresponding to the road section and a road section central point corresponding to the corresponding straight-going outlet as a left-turning dividing line, taking a left-turning dividing line corresponding to the road section intersected with a corresponding extending line of the road section as a straight-going dividing line, and taking a connecting line between a right edge point of the road section and a left edge point corresponding to the corresponding straight-going outlet as a right-turning dividing line.

In one embodiment, the computer program when executed by the processor further performs the steps of: and under the condition that the corresponding straight-going exit does not exist in the road section, extending the central line of the road section to be intersected with the corresponding opposite-side road edge line, taking the extension line of the central line of the road section as a left-turning boundary, extending the right-side road edge line of the road section to be intersected with the corresponding opposite-side road edge line, and taking the extension line of the right-side road edge line as a right-turning boundary.

In one embodiment, the method for acquiring track sequence data corresponding to each steering intention according to steering category boundaries and historical track data involved in the execution of the computer program by a processor comprises the following steps: acquiring a first intersection point of a historical track and a parking line corresponding to a road section; the first intersection point comprises at least one of a left turn starting point, a straight-going starting point and a right turn starting point; for each boundary of the road section, acquiring a second intersection point of the boundary corresponding to the first intersection point in the history track; aiming at each historical track, acquiring track sequence data and characteristics corresponding to each sequence data point in the track sequence data from the historical track according to the corresponding first intersection point and second intersection point to obtain track sequence data corresponding to each steering intention; the data points in the trace sequence data corresponding to the ends of the time series are located between the first intersection point and the second intersection point.

In one embodiment, determining a steering intent of a vehicle from a set of steering intents and a set of probabilities involved when the computer program is executed by a processor comprises: in the case where there is only one type of intention in the steering intention set, determining the intention as the steering intention of the vehicle; when at least two types of intentions exist in the steering intention set and probabilities in probability sets corresponding to the intentions are different, determining the intention corresponding to the maximum probability value in the probability sets as the steering intention of the vehicle; when at least two types of intentions exist in the steering intention set and probabilities in the probability set corresponding to the at least two types of intentions are the same, acquiring steering priority corresponding to each type of intentions, and determining the intention with the highest steering priority as the steering intention of the vehicle.

In one embodiment, a computer program product is provided comprising a computer program which, when executed by a processor, performs the steps of: acquiring road side perception data of a current intersection; the road side perception data comprise scene data of the current intersection and track data of vehicles in the current intersection; the scene data comprises lane lines, lane position information and lane function information; the track data comprises position information, speed information, course angle information and lanes of the vehicle at each moment; acquiring a steering intention set of a vehicle according to lane function information and a lane where the vehicle is located; the steering intent set includes at least one of a left turn, a straight turn, and a right turn; processing the track data through a pre-trained neural network model to obtain probability sets of steering intentions of the vehicle; and determining the steering intention of the vehicle according to the steering intention set and the probability set.

Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high density embedded nonvolatile memory, resistive random access memory (ReRAM), magneto-resistive random access memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric memory (Ferroelectric Random Access Memory, FRAM), phase change memory (PHASE CHANGE memory, PCM), graphene memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in various forms such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), etc. The databases referred to in the embodiments provided herein may include at least one of a relational database and a non-relational database. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processor referred to in the embodiments provided in the present application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, or the like, but is not limited thereto.

The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.

The foregoing examples illustrate only a few embodiments of the application and are described in detail herein without thereby limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of the application should be assessed as that of the appended claims.

Claims

1. A steering intent recognition method, the method comprising:

Acquiring road side perception data of a current intersection; the road side perception data comprise scene data of a current intersection and track data of vehicles in the current intersection; the scene data comprises lane lines, lane position information and lane function information; the track data comprises position information, speed information, course angle information and lanes of the vehicle at each moment;

Acquiring a steering intention set of the vehicle according to the lane function information and the lane in which the vehicle is positioned;

Processing the track data through a pre-trained neural network model to obtain a probability set of each steering intention of the vehicle;

2. The method of claim 1, wherein the training process of the neural network model comprises:

acquiring historical scene data and historical track data;

and training a neural network model based on the track sequence data.

3. The method of claim 2, wherein said constructing a turn category boundary from said historical scene data comprises:

determining intersection points of the central line of each road section, the left road side line and the right road side line with corresponding extension lines of the parking line according to the historical scene data to obtain a central point of the road section, a left side edge point and a right side edge point;

And under the condition that each road section has a corresponding straight-going outlet, taking a connecting line between a road section central point corresponding to the road section and a road section central point corresponding to the corresponding straight-going outlet as a left-turning dividing line, taking a left-turning dividing line corresponding to the road section intersected with a corresponding extension line of the road section as a straight-going dividing line, and taking a connecting line between a right edge point of the road section and a left edge point corresponding to the corresponding straight-going outlet as a right-turning dividing line.

4. A method according to claim 3, characterized in that the method further comprises:

And under the condition that the corresponding straight-going exit is not present in the road section, extending the road section central line of the road section to be intersected with the corresponding opposite side road edge line, taking the extension line of the road section central line as a left turning boundary, extending the right side road edge line of the road section to be intersected with the corresponding opposite side road edge line, and taking the extension line of the right side road edge line as a right turning boundary.

5. The method according to claim 4, wherein the obtaining track sequence data corresponding to each steering intention according to the steering category boundary and the historical track data includes:

Acquiring a first intersection point of a historical track and a parking line corresponding to the road section; the first intersection point comprises at least one of a left turn starting point, a straight-going starting point and a right turn starting point;

For each historical track, acquiring track sequence data and characteristics corresponding to each sequence data point in the track sequence data from the historical track according to the corresponding first intersection point and second intersection point to obtain track sequence data corresponding to each steering intention; the data points corresponding to the tail ends of the time series in the track sequence data are located between the first intersection point and the second intersection point.

6. The method of claim 1, wherein the determining the steering intent of the vehicle from the set of steering intents and the set of probabilities comprises:

Determining the intent as a steering intent of the vehicle if there is only one type of intent in the steering intent set;

determining an intention corresponding to a probability maximum value in the probability set as a steering intention of the vehicle when at least two types of intentions exist in the steering intention set and probabilities in the probability set corresponding to the intentions are different;

And under the condition that at least two types of intentions exist in the steering intention set and the probabilities in the probability set corresponding to the at least two types of intentions are the same, acquiring the steering priority corresponding to each type of intentions, and determining the intention with the highest steering priority as the steering intention of the vehicle.

7. A steering intent recognition device, characterized in that the device comprises:

The data acquisition module is used for acquiring road side perception data of the current intersection; the road side perception data comprise scene data of a current intersection and track data of vehicles in the current intersection; the scene data comprises lane lines, lane position information and lane function information; the track data comprises position information, speed information, course angle information and lanes of the vehicle at each moment;

The intention acquisition module is used for acquiring a steering intention set of the vehicle according to the lane function information and the lane in which the vehicle is positioned;

The probability acquisition module is used for processing the track data through a pre-trained neural network model to obtain a probability set of each steering intention of the vehicle;

and the intention determining module is used for determining the steering intention of the vehicle according to the steering intention set and the probability set.

8. The apparatus of claim 7, wherein the model training module comprises:

9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 6 when the computer program is executed.

10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 6.