CN115810184A

CN115810184A - Method and device for determining road associated with traffic signal lamp

Info

Publication number: CN115810184A
Application number: CN202310138416.9A
Authority: CN
Inventors: 蒋丹宁; 刘志成; 孙赞; 廖易天; 关焕康; 宁宇光
Original assignee: Autonavi Software Co Ltd
Current assignee: Autonavi Software Co Ltd
Priority date: 2023-02-14
Filing date: 2023-02-14
Publication date: 2023-03-17
Anticipated expiration: 2043-02-14
Also published as: CN115810184B

Abstract

The specification discloses a method and a device for determining a road associated with a traffic signal lamp. The method comprises the following steps: determining a plurality of candidate roads corresponding to the target traffic signal lamp; determining a characteristic vector of each candidate road and a characteristic vector of the target traffic signal lamp; for each candidate road, inputting the feature vector of the candidate road and the feature vector of the target traffic signal lamp into a classification model, and outputting the probability of the candidate road being associated with the target traffic signal lamp through the classification model; and determining a target road associated with the target traffic signal lamp in the candidate roads according to the probability. By adopting the scheme provided by the specification, the automatic determination of the traffic signal lamp-associated road can be realized, the whole process does not need manual operation, the efficiency can be greatly improved, and the labor cost is saved.

Description

Method and device for determining road associated with traffic signal lamp

Technical Field

The specification relates to the technical field of high-precision maps, in particular to a method and a device for determining a road associated with a traffic signal lamp.

Background

The traffic signal lamp related road is important data in a high-precision map, and is important prior data of the traffic signal lamp corresponding to the road judged to be driven by the automatic driving automobile. At present, the associated road of the traffic signal lamp is determined by a manual operation mode, and the problems of high cost, low efficiency and the like exist.

Disclosure of Invention

In view of the above, the present specification provides a method and an apparatus for determining a road associated with a traffic signal.

Specifically, the description is realized by the following technical scheme:

a method for determining a road associated with a traffic signal lamp comprises the following steps:

determining a plurality of candidate roads corresponding to the target traffic signal lamp;

determining a feature vector of each candidate road and a feature vector of the target traffic signal lamp;

for each candidate road, inputting the feature vector of the candidate road and the feature vector of the target traffic signal lamp into a classification model, and outputting the probability of the candidate road being associated with the target traffic signal lamp through the classification model;

and determining the target road associated with the target traffic signal lamp in the candidate roads according to the probability.

Optionally, the determining a plurality of candidate roads corresponding to the target traffic signal lamp includes:

acquiring each road in the area to which the target traffic signal lamp belongs;

determining a road associated with a stop line as a lamp control road in each road in the area of the target traffic signal lamp;

determining a number of candidate roads for the target traffic signal light in the light-controlled road.

Optionally, the determining, as a light-controlled road, a road associated with a stop line in each road in the area to which the target traffic signal lamp belongs includes:

determining a left end point and a right end point of the tail end of the road according to the driving direction of the road aiming at each road in the area where the target traffic signal lamp belongs;

extending a preset first distance to the driving direction of the road by taking the connecting line of the left end point and the right end point as a reference to obtain a rectangular area taking the left end point and the right end point as sides, and taking the rectangular area as an extension area corresponding to the road;

determining the intersection proportion of the extension area corresponding to the road and each stop line in the area where the target traffic signal lamp belongs;

and if the stop line with the intersection proportion larger than or equal to the proportion threshold value exists, determining that the road is the lamp control road.

Optionally, the determining a plurality of candidate roads for the target traffic signal lamp in the lamp-controlled road includes:

determining the relative position and relative orientation of the light control road and the target traffic signal lamp;

inputting the relative position and the relative orientation into a Gaussian model, and outputting probability density that the position orientations of the lamp control road and the target traffic signal lamp accord with Gaussian distribution through the Gaussian model;

and selecting N light control roads with the maximum probability density as the candidate roads, wherein N is a natural number greater than 1.

Optionally, the method for dividing the area to which the target traffic signal lamp belongs includes:

clustering the traffic signal lamps according to the position information of the traffic signal lamps to obtain a plurality of traffic signal lamp clusters;

and expanding the envelope surface formed by the traffic lights in the traffic light cluster outwards by a preset second distance to obtain the area of each traffic light in the traffic light cluster.

Optionally, the determining the feature vector of each candidate road and the feature vector of the target traffic signal lamp includes:

acquiring various traffic elements corresponding to the target traffic signal lamp, wherein the traffic elements comprise roads and traffic signal lamps;

coding the traffic elements to obtain initial vectors of the traffic elements;

performing attention processing on the initial vectors of the traffic elements by adopting an attention mechanism to obtain the attention vectors of the traffic elements as feature vectors of the corresponding traffic elements;

acquiring the feature vector of the candidate road and the feature vector of the target traffic signal lamp from the feature vector of each traffic element;

the feature vector of the candidate road comprises incidence relation information between the corresponding candidate road and other traffic elements in the area to which the target traffic signal lamp belongs, and the feature vector of the target traffic signal lamp comprises incidence relation information between the target traffic signal lamp and other traffic elements in the area to which the target traffic signal lamp belongs.

Optionally, after determining the target road associated with the target traffic signal lamp, the method further includes:

judging whether the target road is a road in any road group or not aiming at each target road, wherein each road in the road group is associated with the same traffic signal lamp;

and if the target road is a road in a certain road group, determining other roads in the road group to which the target road belongs as the target road associated with the target traffic signal lamp.

Optionally, the method for dividing the road group includes:

aiming at any two lamp control roads in the area where the target traffic signal lamp belongs, judging whether an included angle between the two lamp control roads is smaller than a threshold value;

if the included angle between the two lamp control roads is smaller than the threshold value, judging whether the distance between the two lamp control roads meets the distance constraint;

and if the distance between the two lamp control roads accords with the distance constraint and no other road exists between the two lamp control roads, dividing the two lamp control roads into the same road group.

judging whether the transverse distance between the target road and the target traffic signal lamp is larger than a preset third distance or not according to each target road;

and if the transverse distance is greater than the third distance, determining that the confidence degree of the association between the target road and the target traffic signal lamp is low, and outputting a prompt of manual checking.

for each target road, when the target road is simultaneously associated with a plurality of target traffic signal lamps, judging whether the orientations of the plurality of target traffic signal lamps associated with the target road are orthogonal or not;

and if the target road is associated with two target traffic signal lamps with orthogonal directions, determining that the confidence degrees of the association between the target road and the two target traffic signal lamps with the orthogonal directions are low, and outputting a prompt of manual checking.

A traffic signal light associated road determination apparatus, comprising:

the candidate road determining unit is used for determining a plurality of candidate roads corresponding to the target traffic signal lamp;

the characteristic vector determining unit is used for determining the characteristic vector of each candidate road and the characteristic vector of the target traffic signal lamp;

the association probability prediction unit inputs the feature vector of the candidate road and the feature vector of the target traffic signal lamp into a classification model aiming at each candidate road, and outputs the association probability of the candidate road and the target traffic signal lamp through the classification model;

and the associated road determining unit is used for determining the target road associated with the target traffic signal lamp in the candidate roads according to the probability.

An electronic device, comprising:

a processor;

a memory for storing processor-executable instructions;

wherein the processor executes the executable instructions to realize the method

A computer readable storage medium having stored thereon computer instructions which, when executed by a processor, carry out the steps of the aforementioned method.

By adopting the embodiment, a plurality of candidate roads can be determined for the target traffic signal lamp, then the feature vector of the candidate road and the feature vector of the target traffic signal lamp are respectively input into the classification model aiming at each candidate road, the probability of the candidate road being associated with the target traffic signal lamp is output through the classification model, and then each item road mark associated with the target traffic signal lamp is determined in the candidate road, so that the automatic determination of the road associated with the traffic signal lamp is realized, the whole process does not need manual operation, the efficiency can be greatly improved, and the labor cost is saved. Meanwhile, compared with manual operation, the method for determining the road associated with the traffic signal lamp by adopting the machine learning mode has more stable and controllable quality.

Drawings

Fig. 1 is a flowchart illustrating a method for determining a road associated with a traffic signal according to an exemplary embodiment of the present disclosure.

Fig. 2 is a flowchart illustrating a method for determining a candidate road according to an exemplary embodiment of the present disclosure.

Fig. 3 is a schematic view of a road shown in an exemplary embodiment of the present description.

Fig. 4 is a schematic view of a road corresponding to an extended area according to an exemplary embodiment of the present disclosure.

FIG. 5 is a schematic diagram showing one relative position and relative orientation for an exemplary embodiment of the present description.

Fig. 6 is a flow chart illustrating a method for determining feature vectors using an attention mechanism in an exemplary embodiment of the present description.

Fig. 7 is a flowchart illustrating another method for determining feature vectors using an attention mechanism in accordance with an exemplary embodiment of the present disclosure.

Fig. 8 is a vector difference diagram shown in an exemplary embodiment of the present description.

Fig. 9 is a hardware configuration diagram of an electronic device in which a traffic signal lamp-associated road determination device is shown in an exemplary embodiment of the present specification.

Fig. 10 is a block diagram of a traffic signal lamp-associated road determination device according to an exemplary embodiment of the present disclosure.

Detailed Description

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary examples do not represent all implementations consistent with this description. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the specification, as detailed in the appended claims.

The terminology used in the description herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the description. As used in this specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

It should be understood that although the terms first, second, third, etc. may be used herein to describe various information, these information should not be limited to these terms. These terms are only used to distinguish one type of information from another. For example, first information may also be referred to as second information, and similarly, second information may also be referred to as first information, without departing from the scope of the present description. The word "if" as used herein may be interpreted as "at" \8230; "or" when 8230; \8230; "or" in response to a determination ", depending on the context.

The traffic signal lamp associated road is important data in a high-precision map and is important prior data for judging the corresponding traffic signal lamp of the road driven by the automatic driving automobile.

For example, road 1 is controlled by traffic light a, road 2 is controlled by traffic light B, and an autonomous automobile is controlled by traffic light a when traveling on road 1 and by traffic light B when traveling on road 2.

At present, the related roads of the traffic signal lamp are determined by a manual operation mode, and the problems of high cost, low efficiency, unstable quality and the like exist.

The specification provides a scheme for determining a road associated with a traffic signal lamp, which can automatically determine the road associated with the traffic signal lamp, can greatly improve the production efficiency, reduce the cost and ensure the stability and controllability of the quality compared with a manual operation mode.

Fig. 1 is a flowchart illustrating a method for determining a road associated with a traffic signal lamp according to an exemplary embodiment of the present disclosure.

The method for determining a traffic signal-associated road shown in fig. 1 may be applied to a server, such as a server of an internet map company, and the like, and includes the steps of:

and 102, determining a plurality of candidate roads corresponding to the target traffic signal lamp.

In this specification, a road associated with a traffic signal refers to a road controlled by the traffic signal. In the real world, a real road is divided into a plurality of links due to a fork, a t-junction, an entrance, and the like, and the road described in this specification refers to a link obtained by dividing the real road.

The target traffic signal lamp is a traffic signal lamp needing to determine the associated road, and when the candidate road corresponding to the target traffic signal lamp is determined, all roads in the area where the target traffic signal lamp belongs can be obtained firstly, and then the candidate road is determined from the roads. Wherein the area to which the target traffic signal light belongs is generally a range in the vicinity of the target traffic signal light.

In one example, the area to which the target traffic signal lamp belongs may be a circular area with a radius of 200 meters and the target traffic signal lamp as a center point.

In another example, the target traffic lights may be divided into regions in advance, and all the traffic lights that need to be associated with the road may be divided into multiple regions. By adopting the pre-division mode, when the associated road is determined, the area to which each traffic signal lamp belongs does not need to be divided separately, so that the calculation amount can be greatly reduced.

For example, the traffic lights may be clustered according to the position information of each traffic light to obtain a plurality of traffic light clusters, and then the envelope surface formed by the traffic lights in each traffic light cluster is extended outward by a preset second distance (such as 200 meters), so as to obtain the area to which each traffic light in the traffic light cluster belongs.

The Clustering may use a Density-Based Clustering algorithm, such as a DBSCAN algorithm (Density-Based Spatial Clustering of Applications with Noise, density-Based Clustering method with Noise), and the like.

In this specification, each road in the area to which the target traffic light belongs may be obtained from the high-precision map database, that is, the roads near the target traffic light are obtained first, then several candidate roads are determined for the target traffic light in the obtained roads, and then the road associated with the target traffic light may be further determined in the candidate roads.

When candidate roads are determined, roads associated with stop lines can be screened out from all acquired roads, roads controlled by traffic lights are all associated with the stop lines, so that the roads associated with the stop lines can be called light control roads, and then the candidate roads are further screened out from the light control roads. The specific screening of candidate roads will be described in detail in the following embodiments.

Of course, in other examples, all roads in the area where the target traffic signal light belongs may also be taken as candidate roads, which is not particularly limited in this specification.

And 104, determining the characteristic vector of each candidate road and the characteristic vector of the target traffic signal lamp.

Based on the foregoing step 102, after the candidate roads of the target traffic light are determined, the feature vectors of the target traffic light and each candidate road may be respectively determined, so as to facilitate subsequent model processing.

In one example, the road data and the traffic signal lamp data stored in the database may be directly encoded to determine the corresponding feature vector.

In another example, when determining the feature vector, the traffic elements corresponding to the target traffic light may be all encoded to obtain an initial vector corresponding to each traffic element, for example, the traffic elements in the area of the target traffic light may be encoded to obtain a corresponding initial vector, and then the initial vector may be subjected to attention processing by using an attention mechanism, so as to obtain an attention vector of each traffic element as the feature vector.

The traffic elements in the area to which the target traffic signal lamp belongs include the target traffic signal lamp itself, and may also include other elements such as each road, stop line, pedestrian safety island and the like in the area. The specific implementation of the feature vector determination using the attention mechanism in this example will be described in detail in the following embodiments.

And 106, inputting the feature vector of the candidate road and the feature vector of the target traffic signal lamp into a classification model for each candidate road, and outputting the probability of the candidate road and the target traffic signal lamp association through the classification model.

In this specification, each candidate road and the target traffic signal lamp may be combined respectively to obtain a plurality of sets of "traffic signal lamp-candidate road" combinations (hereinafter, simply referred to as "lamp combination"), then the feature vectors of each set of lamp combination are input into the classification model, and the association probability between the traffic signal lamp and the candidate road in the set of lamp combination is output through the classification model.

For example, if there are 6 candidate roads of the target traffic signal a, which are roads 1-6, please refer to table 1, a combination relationship of six groups of roads, such as target traffic signal a-road 1, target traffic signal a-road 2, etc., can be obtained.

In this specification, for each group of light road combination, the feature vector of the target traffic signal lamp in the group of light road combination may be spliced with the feature vector of the candidate road to serve as the feature vector of the group of light road combination; the feature vector of the target traffic signal in the set of light combination and the feature vector of the candidate road may also be calculated, and the calculation result is used as the feature vector thereof, for example, addition calculation, averaging calculation, and the like, which is not limited in this specification.

In this specification, the classification model may be a classification model such as an MLP (multi layer Perceptron) model, and may be trained in advance based on a combination of sample light paths, and a specific training process may refer to related technologies, which are not described herein again.

And 108, determining a target road associated with the target traffic signal lamp in the candidate roads according to the probability.

Based on the foregoing step 106, after the probability associated with each group of light circuit combinations is predicted by the classification model, it can be respectively determined whether the probability associated with the group of light circuit combinations is greater than a probability threshold, for example: 0.5, 0.6, etc.

If the probability associated with a certain group of light path combination is greater than or equal to the probability threshold value, determining that the candidate road in the group of light path combination is the associated road of the corresponding target traffic signal lamp; if the probability associated with a certain group of light path combination is smaller than the probability threshold, determining that the candidate road in the group of light path combination is not the associated road of the corresponding target traffic signal lamp, and further determining each item road associated with the target traffic signal lamp from the candidate road.

Still taking the light and road combination shown in table 1 as an example, please refer to the light and road combination association probability shown in table 2, and assuming that the probability threshold is 0.6, it can be determined that the target roads associated with the target traffic signal lamp a are road 1 and road 2.

Of course, after the target road associated with the target traffic light is determined, the result may be corrected, so as to ensure the accuracy of the target road, and the specific correction scheme will be described in detail in the following embodiments.

As can be seen from the above description, according to the technical scheme provided by the present specification, a plurality of candidate roads can be determined for the target traffic signal lamp, then, for each candidate road, the feature vector of the candidate road and the feature vector of the target traffic signal lamp are input into the classification model, the probability that the candidate road is associated with the target traffic signal lamp is output through the classification model, and then, each entry road associated with the target traffic signal lamp is determined in the candidate road, so that the automatic determination of the road associated with the traffic signal lamp is realized, the whole process does not need manual operation, the efficiency can be greatly improved, and the labor cost can be saved. Meanwhile, compared with manual operation, the method for determining the road associated with the traffic signal lamp by adopting the machine learning mode has more stable and controllable quality.

The technical solutions provided in the present specification are described in detail below in terms of determination of candidate roads, determination of feature vectors, and correction of target roads.

1. Determination of candidate roads

Referring to fig. 2, the following method may be used to determine candidate roads for the target traffic signal:

step 202, obtaining all roads in the area where the target traffic signal lamp belongs, and determining the road associated with the stop line as a lamp control road in all the roads in the area where the target traffic signal lamp belongs.

In this specification, when determining a light-controlled road, all roads in an area to which a target traffic light belongs may be acquired first, and then, for each road in the area to which the target traffic light belongs, a left end point and a right end point of a road end may be determined first according to a driving direction of the road. Because the roads are represented by lane lines, and one road comprises a left lane line and a right lane line, the end points of the tail ends of the left lane line and the right lane line can be determined according to the driving direction of the road.

Referring to fig. 3, fig. 3 shows two left and right lane lines of the road 1, the driving direction of the road 1 is the direction indicated by the arrow, i.e. the driving direction is from bottom to top, and the end points of the two left and right lane lines in the driving direction are point a and point B, respectively.

After the left end point and the right end point of the tail end of the road are determined, a preset first distance is extended towards the driving direction of the road by taking the connecting line of the left end point and the right end point as a reference, and a rectangular area with the left end point and the right end point as sides is obtained and is used as an extension area corresponding to the road. Wherein the first distance may be preset, for example, 1 meter, 1.1 meter, etc.

Still taking the lane 1 shown in fig. 3 as an example, please refer to fig. 4 again, after extending upward (the driving direction) for a first distance by taking a connecting line AB between the left and right endpoints as a reference, a rectangular area ABCD with AB as a side, that is, an area enclosed by the dotted line shown in fig. 4, can be obtained, and the rectangular ABCD is an extending area corresponding to the lane 1. Wherein the length of sides AD and BC of the rectangle ABCD is the first distance.

After generating a corresponding extended area for the road, the intersection proportion of each stop line and the extended area in the area to which the target traffic signal lamp belongs may be calculated, where the intersection proportion may be the proportion of the intersection of the stop line and the extended area occupying the stop line area, and may also be the proportion of the intersection of the stop line and the extended area occupying the extended area, which is not limited in this specification.

Generally, the stop line is located at the end of the road controlled by the stop line, for example, at the connecting line of the left end point and the right end point of the road controlled by the stop line, or the distance between the stop line and the connecting line is short, so that a proportion threshold value can be set according to the actual situation, and if the intersection proportion obtained by the calculation is greater than or equal to the proportion threshold value, the corresponding road can be determined to be associated with the stop line and is the light-controlled road.

For example, assuming a threshold value of 40%, this represents the proportion of the stop-line area that the stop-line intersects with the extended area. Still taking the extended area ABCD of the road 1 shown in fig. 4 as an example, if the intersection of a certain stop line and the extended area ABCD is calculated to account for 80% of the stop line area, that is, the intersection proportion is 80%, it may be determined that the intersection proportion is greater than the proportion threshold, and then it is determined that the road 1 is a light-controlled road.

It should be noted that, in the database, the stop line is generally stored in a polygonal form, so the above lamp control road determination scheme based on the intersection ratio is adopted. If the stop line is stored in the form of a line, the position relationship between the stop line and the connecting line of the left end point and the right end point of the tail end of the road can be calculated, and then whether the road is the lamp control road or not can be determined according to the position relationship, for example, if the stop line is parallel to the connecting line of the left end point and the right end point and the distance is smaller than a threshold value, the road is determined as the lamp control road and the like.

In the specification, the light control road is determined in the area to which the target traffic signal lamp belongs, and then the candidate roads are screened in the light control road, so that the calculation amount of screening the candidate roads can be reduced, and the calculation efficiency is improved.

Step 204, determining a plurality of candidate roads for the target traffic signal lamp in the lamp-controlled road.

Based on the foregoing step 202, after the light-controlled roads are determined in the area to which the target traffic light belongs, for each light-controlled road, the light-controlled road controlled by the target traffic light with a relatively high probability may be screened out according to the position relationship and the orientation relationship between the light-controlled road and the target traffic light as a candidate road, where the value of the number N of candidate roads may be preset, for example, 6 or 8.

In this specification, it is found through analysis that the position relationship and the orientation relationship between the destination road associated with the traffic signal and the traffic signal approximately conform to gaussian distribution, and therefore, a gaussian model may be used to determine the candidate road, where the input of the gaussian model is the relative position and the relative orientation between the light-controlled road and the destination traffic signal, and then the probability density that the position and the orientation relationship between the light-controlled road and the destination traffic signal conform to gaussian distribution is output through the gaussian model, and then N light-controlled roads with the largest probability density value may be selected as the candidate road.

When the relative position and the relative orientation are calculated, a coordinate system can be established by taking the central point of the target traffic signal lamp as the original point and taking the orientation direction of the target traffic signal lamp as a y axis, then the position information of the lamp control road is converted into the newly established coordinate system according to the position relation between the target traffic signal lamp and the lamp control road, the position coordinate of the lamp control road in the newly established coordinate system is further obtained and used as the relative position, and the included angle between the driving direction of the lamp control road and the y axis is used as the relative orientation.

When the light control road is calculated, the central point of the connection line of the left end point and the right end point of the tail end of the light control road can represent the light control road to carry out position information conversion, and the relative position is obtained.

Referring to fig. 5, taking the target traffic signal shown in fig. 5 as an example, if the center point of the target traffic signal is point O, a coordinate system can be established with point O as the origin and the orientation of the target traffic signal (downward in fig. 5) as the y-axis. Taking the light control road shown in fig. 5 as an example, if the central point of the connection line AB between the left and right end points at the tail end of the light control road is the H point, the coordinates of the H point can be transformed into the coordinates with O as the originIn the system xOy, the position coordinates of the H point in the coordinate system xOy are obtained

As the relative position between the light-controlled road and the target traffic signal light.

Referring to fig. 5, if the driving direction of the light-controlled road in fig. 5 is upward and is parallel to the y-axis of the orientation of the target traffic signal light, i.e. the included angle is 0 degree, the relative orientation between the light-controlled road and the target traffic signal light is 0 degree.

In the example shown in FIG. 5, the relative positions may be compared

And the relative orientation of 0 degree is used as the input of a Gaussian model, and the probability density that the position orientation between the lamp control road and the target traffic signal lamp is in accordance with Gaussian distribution shown in the figure 5 is obtained through the Gaussian model.

In this step, assuming that 20 light-controlled roads are counted in the area to which the target traffic signal lamp belongs, and the value of N is preset to 6, the 6 light-controlled roads with the highest probability of meeting the gaussian distribution can be screened out from the 20 light-controlled roads as candidate roads.

Of course, in other examples, the candidate roads may be determined in other manners, for example, determining the corresponding control range for the target traffic signal according to the position and the orientation of the target traffic signal, then determining whether each light-controlled road is within the control range, and if so, determining the light-controlled road as the candidate road. The control range may be a sector area with the target traffic signal lamp as a vertex, and the like, which is not particularly limited in this specification.

2. Determination of feature vectors

The embodiment introduces an implementation method for determining feature vectors of candidate roads and target traffic signal lamps by adopting an attention mechanism.

Referring to fig. 6 and 7, the method for determining feature vectors using the attention mechanism may include the following steps:

step 602, obtaining traffic elements corresponding to the target traffic signal lamp, where the traffic elements include roads and traffic signal lamps.

In the present specification, when determining the feature vector, a traffic element corresponding to the target traffic signal, for example, a traffic element in an area to which the target traffic signal belongs, or the like, may be acquired.

In one example, in determining the feature vector, only two traffic elements, namely, a road and a traffic light in the area to which the target traffic light belongs, may be acquired.

In another example, when determining the feature vector, all types of traffic elements in the area to which the target traffic light belongs, such as roads, traffic lights, stop lines, zebra crossings, pedestrian safety islands, no-parking zones, and the like, may also be acquired. Generally, the more traffic element types, the more comprehensive the information carried by the subsequently determined feature vectors.

In this step, the acquired traffic elements include the target traffic signal lamp itself and also include each candidate road.

And step 604, coding the traffic elements to obtain initial vectors of the traffic elements.

Based on the foregoing step 602, after the traffic element is obtained, the traffic element may be encoded to generate a corresponding vector token, and for convenience of distinguishing, the vector token obtained by directly encoding the traffic element is referred to as an initial vector.

Generally, the traffic element data stored in the database includes geometric data of traffic elements such as position coordinates and the like, and also includes type data of traffic elements such as a turn type of a road (straight, left, and the like), a type of traffic signal (indicating a straight turn signal, indicating a left turn signal), and the like.

The geometric data of the traffic elements are often stored in the form of vectors, and for convenience of subsequent encoding and calculation, the vector data can be subjected to differential operation. For example, lane lines used for representing roads are often stored in a high-precision map database in the form of line strings (LineString), which is an ordered spatial point set, and the ordered spatial point set can be converted into an unordered spatial point set through vector difference processing.

Referring to fig. 8, the left side of fig. 8 shows an ordered spatial point set including 3 points with coordinates (x 1, y 1), (x 2, y 2) and (x 3, y 3), respectively, and the arrows show the sequential relationship. Through vector difference processing, the 3 points can be converted into 3 unordered points on the right side of fig. 8, and the coordinates of each point are (x 1, y1, dx1, dy 1), (x 2, y2, dx2, dy 2) and (x 3, y3, dx3, dy 3), respectively, where dx and dy can represent the direction information of each point pointing to the next point, so that the sequential relationship of the 3 points is embodied in the coordinates.

When the vector difference calculation is performed, the coordinates of two adjacent points can be subtracted, and then normalization processing is performed to make the modulus of the coordinate be 1, so as to obtain the values of dx and dy.

Through vector difference processing, the ordered point set used for representing the geometric data of the traffic elements can be converted into an unordered point set, the subsequent calculation difficulty is reduced, and vector encoding is facilitated.

In this step, after the vector difference is performed on the geometric data of the traffic element, coordinate transformation can be performed to further reduce the subsequent calculation amount.

The method is simple, and when the target road associated with the target traffic signal lamp is determined, the geometric data after vector difference of all traffic elements in the area to which the target traffic signal lamp belongs can be converted into a coordinate system with the target traffic signal lamp as an origin. The positive y-axis direction of the coordinate system with the target traffic signal lamp as the origin may be a direction toward which the target traffic signal lamp faces, and then a direction in which the positive y-axis direction is rotated clockwise by 90 degrees may be the positive x-axis direction. Of course, the above-mentioned establishment of the coordinate system is only an exemplary one, and in other examples, a direction in which the positive y-axis direction is rotated 90 degrees counterclockwise may be taken as the positive x-axis direction, and the like, and the present specification does not particularly limit this.

After the geometric data of the traffic elements are subjected to coordinate transformation, an encoder can be adopted for encoding, and then geometric encoding vectors are obtained. For example, a point encoder can be adopted to encode the geometric data of the traffic signal lamp to obtain a geometric encoding vector of the traffic signal lamp; encoding the geometric data of the road by adopting a line encoder to obtain a geometric encoding vector of the road; and (4) encoding the surface-shaped elements such as the pedestrian safety island by adopting a surface encoder to obtain geometric encoding vectors of the surface-shaped elements such as the pedestrian safety island. Wherein, the encoder can adopt aggregation functions such as PointNet and the like.

For type data, embedding processing can be performed on the type data to obtain a corresponding type data vector. For example, a coding vector matrix of a corresponding type may be preset or learned, and when a type coding vector is determined, a row or a column of vectors of the corresponding type may be extracted from the coding vector matrix.

In this specification, after determining the geometric code vector and the type code vector of each traffic element, the geometric code vector and the type code vector may be added together or spliced together bitwise to obtain an initial vector corresponding to the traffic element.

And 606, performing attention processing on the initial vector of each traffic element by using an attention mechanism to obtain the attention vector of each traffic element as the feature vector of the corresponding traffic element.

Based on step 604, after the initial vector of each traffic element is determined, attention processing may be performed on the initial vector, for example, the initial vector of each traffic element is input to a module having an attention mechanism such as a Transformer, and the attention vector of each traffic element is output as a feature vector thereof by the module. After the attention mechanism processing, the output feature vector includes information of the traffic element itself, and may also include association relationship information between the corresponding traffic element and other traffic elements.

Step 608, obtaining the feature vector of the candidate road and the feature vector of the target traffic signal lamp from the feature vectors of the traffic elements.

Based on the foregoing step 606, after the feature vector of each traffic element is determined, the feature vector of each candidate road and the feature vector of the target traffic signal lamp can be obtained therefrom.

In this specification, the feature vector of each candidate road includes, in addition to the feature of the corresponding candidate road itself, association relationship information between the candidate road and other traffic elements in the area to which the target traffic signal lamp belongs; similarly, the feature vector of the target traffic light includes, in addition to the feature of the target traffic light itself, association relationship information between the target traffic light and other traffic elements in the area to which the target traffic light belongs. The method and the device have the advantages that more information in the region to which the target traffic signal lamp belongs is carried in the feature vector through the attention mechanism, and the accuracy of road prediction related to the subsequent target traffic signal lamp is improved.

3. Correction of target road

In this specification, after the target road associated with the target traffic light is determined through the embodiment shown in fig. 1, the target road may be verified and corrected, so as to further improve the accuracy of determining the target road.

In the real world, there may be a case where a plurality of roads are controlled by the same traffic light, for example, a road in the same driving direction in a main road and a sub road is controlled by the same traffic light at an intersection, or a road and a road waiting to turn are controlled by the same traffic light at an intersection, and the like. In this case, with the target road determining method shown in fig. 1, omission may occur, for example, when the distance between the main road and the auxiliary road is long, the model may be caused to match the main road with the target road, and the auxiliary road may be omitted.

In order to solve the problem, the specification may previously divide a plurality of roads controlled by the same traffic light into the same road group, that is, each road in the road group is associated with the same traffic light, after a target road associated with the target traffic light is determined, it may be determined, for each target road, whether the target road is a road in any road group, and if the target road is a road in any road group, other roads in the road group may also be determined as target roads associated with the target traffic light.

For example, assuming that the target roads associated with the target traffic signal a are road 1 and road 2, where the road 1 belongs to the road group (road 1, road 8, road 9), the road 8 and the road 9 in the road group may also be determined as the target roads associated with the target traffic signal a.

The method for dividing the road group can be preset according to the characteristics of roads belonging to the same road group in the real world.

Taking the main and auxiliary road group as an example, for any two light-controlled roads in the area to which the target traffic signal lamp belongs, whether an included angle between the two light-controlled roads is smaller than a threshold value is judged, for example, whether the included angle is smaller than 30 degrees, and if the included angle between the two light-controlled roads is smaller than the threshold value, whether a distance between the two light-controlled roads meets a distance constraint is judged, for example, the distance between the two light-controlled roads is not more than 3 meters. If the distance between the two light-controlled roads meets the distance constraint and no other road exists between the two light-controlled roads, it is presumed that green belts may exist between the two light-controlled roads, which are respectively a main road and a sub road, and the two light-controlled roads are divided into the same road group.

Of course, the above method for dividing the road group is only an exemplary illustration, and in practical applications, other methods for dividing the road group may be set, or the road group may be manually divided in advance, and the like, and the description does not limit this.

By adopting the road group-based target road correction method provided by the specification, the problem of missing matching of the target road can be effectively solved, and the accuracy of determining the final target road is improved.

In addition to the above-mentioned missing matching problem of the target road, there may be a wrong matching situation, i.e., the determined target road is not actually the road associated with the target traffic signal lamp.

For the situation, some verification rules can be preset, and after the classification model is adopted to determine the target roads associated with the target traffic signal lamp, each target road is verified.

For example, for each target road, it may be determined whether a lateral distance between the target road and a target traffic light is greater than a preset third distance, for example, 10 meters, 12 meters, and the like, and if the lateral distance is greater than the third distance, it indicates that the lateral distance between the target road and the target traffic light is far and is not in accordance with a situation of the real world, and a problem of road missing may exist, and it may be determined that the confidence degree associated with the target traffic light is low, and the checking is performed manually, for example, a manual checking prompt is output. The transverse distance between the target road and the target traffic signal lamp can be the distance between the center line of the target road and the center line of the target traffic signal lamp, and the like.

Alternatively, for each target road, when the target road is simultaneously matched with a plurality of target traffic lights, the orientation relationship between the target traffic lights may be determined, and if there are two target traffic lights with almost orthogonal orientations (for example, the orientation angle is greater than the angle threshold), it may be determined that the confidence of the association between the target road and the two target traffic lights with almost orthogonal orientations is low, and manual checking may be performed.

For example, if the target road 11 is associated with a target traffic signal a, the target road 11 is associated with a target traffic signal B, and the angle between the target traffic signal a and the target traffic signal B is 85 degrees and greater than the angle threshold value of 70 degrees, it may be determined that the confidence level between the target road 11 and the target traffic signal a is low, the confidence level between the target road 11 and the target traffic signal B is low, and the prediction of at least one association relationship is incorrect, i.e., the target road 11 is incorrectly matched with at least one of the target traffic signal a and the target traffic signal B, and manual checking may be performed.

Therefore, the target road related to the target traffic signal lamp can be predicted through the classification model, and the target road can be verified and corrected through a plurality of methods, so that the accuracy of the final target road is ensured.

Corresponding to the embodiment of the method for determining a road associated with a traffic signal lamp, the present specification also provides an embodiment of a device for determining a road associated with a traffic signal lamp.

The embodiment of the traffic signal lamp-associated road determining device can be applied to electronic equipment. The device embodiments may be implemented by software, or by hardware, or by a combination of hardware and software. Taking a software implementation as an example, as a logical device, the device is formed by reading, by a processor of the electronic device where the device is located, a corresponding computer program instruction in the nonvolatile memory into the memory for operation. From a hardware aspect, as shown in fig. 9, the hardware structure diagram of the electronic device where the device for determining a road associated with a traffic signal lamp is located in this specification is shown, except for the processor, the memory, the network interface, and the nonvolatile memory shown in fig. 9, the electronic device where the device is located in the embodiment may also include other hardware according to the actual function of the 8230instruction, which is not described again.

Referring to fig. 10, the apparatus for determining a road associated with a traffic signal lamp may be applied to the electronic device shown in fig. 9, for example, a server of an internet map company, and the apparatus includes:

the association probability prediction unit is used for inputting the feature vector of the candidate road and the feature vector of the target traffic signal lamp into a classification model aiming at each candidate road and outputting the association probability of the candidate road and the target traffic signal lamp through the classification model;

Optionally, the step of determining a plurality of candidate roads corresponding to the target traffic signal lamp includes:

Optionally, the step of determining a road associated with a stop line as a light-controlled road in each road in the area where the target traffic light belongs includes:

determining a left end point and a right end point of the tail end of the road according to the driving direction of the road aiming at each road in the region to which the target traffic signal lamp belongs;

determining the intersection proportion of the extension area corresponding to the road and each stop line in the area to which the target traffic signal lamp belongs;

and if the stop line with the intersection proportion larger than or equal to the proportion threshold exists, determining that the road is the lamp control road.

Optionally, the step of determining a plurality of candidate roads for the target traffic light in the light-controlled road includes:

clustering the traffic signal lamps according to the position information of each traffic signal lamp to obtain a plurality of traffic signal lamp clusters;

Optionally, the step of determining the feature vector of each candidate road and the feature vector of the target traffic signal lamp includes:

acquiring traffic elements corresponding to the target traffic signal lamp, wherein the traffic elements comprise roads and traffic signal lamps;

coding the traffic elements to obtain initial vectors of the traffic elements;

performing attention processing on the initial vector of each traffic element by adopting an attention mechanism to obtain the attention vector of each traffic element as a characteristic vector of the corresponding traffic element;

acquiring the characteristic vector of the candidate road and the characteristic vector of the target traffic signal lamp from the characteristic vector of each traffic element;

Optionally, the method for dividing the road group includes:

aiming at any two lamp control roads in the area to which the target traffic signal lamp belongs, judging whether an included angle between the two lamp control roads is smaller than a threshold value;

for each target road, when the target road is simultaneously associated with a plurality of target traffic lights, judging whether the directions of the plurality of target traffic lights associated with the target road are orthogonal or not;

The implementation process of the functions and actions of each unit in the above device is specifically described in the implementation process of the corresponding step in the above method, and is not described herein again.

For the device embodiments, since they substantially correspond to the method embodiments, reference may be made to the partial description of the method embodiments for relevant points. The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution in the specification. One of ordinary skill in the art can understand and implement without inventive effort.

The systems, apparatuses, modules or units described in the above embodiments may be specifically implemented by a computer chip or an entity, or implemented by a product with certain functions. A typical implementation device is a computer, which may take the form of a personal computer, laptop computer, cellular telephone, camera phone, smart phone, personal digital assistant, media player, navigation device, email messaging device, game console, tablet computer, wearable device, or a combination of any of these devices.

In a typical configuration, a computer includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include forms of volatile memory in a computer readable medium, random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.

Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic disk storage, quantum memory, graphene-based storage media or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.

In correspondence with the aforementioned embodiments of the traffic signal lamp-associated road determination method, the present specification also provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of:

Optionally, the step of determining a road associated with a stop line as a light control road in each road in the area to which the target traffic signal lamp belongs includes:

coding the traffic elements to obtain initial vectors of the traffic elements;

Optionally, the method for dividing the road group includes:

and if the target road is associated with two target traffic signal lamps with orthogonal directions, determining that the confidence degrees of the association between the target road and the two target traffic signal lamps with orthogonal directions are low, and outputting a prompt of manual checking.

It should be noted that, the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) referred to in this specification are information and data authorized by the user or sufficiently authorized by each party, and the collection, use and processing of the relevant data need to comply with relevant laws and regulations and standards of relevant countries and regions, and are provided with corresponding operation entrances for the user to choose to authorize or reject.

The foregoing description has been directed to specific embodiments of this disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.

The above description is only a preferred embodiment of the present disclosure, and should not be taken as limiting the present disclosure, and any modifications, equivalents, improvements, etc. made within the spirit and principle of the present disclosure should be included in the scope of the present disclosure.

Claims

1. A method for determining a road associated with a traffic signal lamp comprises the following steps:

determining a characteristic vector of each candidate road and a characteristic vector of the target traffic signal lamp;

2. The method of claim 1, wherein the determining a number of candidate roads corresponding to the target traffic signal comprises:

determining a road associated with a stop line as a lamp control road in each road in the area where the target traffic signal lamp belongs;

3. The method of claim 2, wherein the determining a road associated with a stop line as a light-controlled road among roads in the area of the target traffic signal comprises:

4. The method of claim 2, the determining a number of candidate roads in the light-controlled road for the target traffic signal light, comprising:

5. The method of claim 1, wherein the method for dividing the area to which the target traffic signal lamp belongs comprises the following steps:

and extending the envelope surface formed by the traffic lights in the traffic light cluster outwards by a preset second distance to obtain the area of each traffic light in the traffic light cluster.

6. The method of claim 1, the determining the feature vector of each candidate road and the feature vector of the target traffic signal comprises:

coding the traffic elements to obtain initial vectors of the traffic elements;

7. The method of claim 1, after determining the target road associated with the target traffic signal, the method further comprising:

judging whether each target road is a road in any road group or not aiming at each target road, wherein each road in the road group is associated with the same traffic signal lamp;

8. The method of any of claims 1-7, after determining the target road associated with the target traffic signal, the method further comprising:

9. A traffic signal light associated road determination apparatus, comprising:

10. An electronic device, comprising:

a processor;

a memory for storing processor-executable instructions;

wherein the processor implements the method of any one of claims 1-8 by executing the executable instructions.

11. A computer readable storage medium having stored thereon computer instructions which, when executed by a processor, carry out the steps of the method according to any one of claims 1 to 8.