CN-121973791-A - Lane departure intention recognition method, device, electronic equipment and storage medium

Abstract

The application discloses a lane departure intention recognition method, a lane departure intention recognition device, electronic equipment and a storage medium, which relate to the technical field of driving safety recognition, wherein the lane departure intention recognition method comprises the steps of acquiring a sight line region transfer sequence of a target driving object in a vehicle under a first preset time window before the vehicle is transversely deviated, wherein the first preset time window is a time window with preset duration before the current moment; and determining the intention recognition probability of the active lane departure intention of the target driving object based on the intention recognition probability and a preset probability threshold value, and determining whether the active lane departure intention exists in the target driving object at the current moment or not. The method and the device can improve the accuracy of identifying the lane departure intention, and further improve the experience of the user and the acceptance of the user to the driving assistance system.

Inventors

• LUO JI
• Ding xuan
• SHEN KAI
• LI JIXUAN
• PENG PAI

Assignees

• 岚图汽车科技股份有限公司

Dates

Publication Date

20260505

Application Date

20260309

Claims (20)

1. 1. A lane departure intention recognition method, characterized in that the lane departure intention recognition method comprises: before the vehicle is laterally deviated, acquiring a sight line region transfer sequence of a target driving object in the vehicle under a first preset time window, wherein the first preset time window is a time window with preset duration before the current moment; Determining intention recognition probability of the target driving object with the driving lane departure intention based on the sight line region transfer sequence and the trained intention recognition model under the condition that the sight line region transfer sequence does not accord with a preset sight line transfer mode; and determining whether the target driving object has an active lane departure intention at the current moment based on the intention recognition probability and a preset probability threshold.
2. 2. The lane departure intention recognition method according to claim 1, wherein the determining whether the target driving object has an active lane departure intention at the current time based on the intention recognition probability and a preset probability threshold value includes: if the intention recognition probability is greater than or equal to the preset probability threshold, determining that the driving lane departure intention exists in the target driving object at the current moment; And if the intention recognition probability is smaller than the preset probability threshold, determining that the driving lane departure intention does not exist in the target driving object at the current moment.
3. 3. The lane departure intention recognition method according to claim 1, wherein the method further comprises: And under the condition that the sight line area transfer sequence accords with the preset sight line transfer mode, determining that the target driving object has the initiative lane departure intention at the current moment.
4. 4. The lane departure intention recognition method of claim 3, wherein the preset gaze transfer mode comprises a scanning observation mode and a continuous gaze mode, the method further comprising: when the sight line area transfer sequence does not accord with the scanning observation mode and does not accord with the continuous fixation mode, determining that the sight line area transfer sequence does not accord with the preset sight line transfer mode; When the sight-line area transfer sequence accords with any one of the scanning observation mode and the continuous fixation mode, determining that the sight-line area transfer sequence accords with the preset sight-line transfer mode; The continuous fixation mode is used for representing a fixation mode that the continuous fixation time length of the sight line in at least one target monitoring area exceeds a preset time length.
5. 5. The lane departure intention recognition method according to claim 1, wherein the acquiring the line-of-sight region transfer sequence of the target driving object in the vehicle under the first preset time window includes: acquiring real-time sight-line area information of a target driving object under the condition that the target driving object is determined not to actively control the vehicle based on operation information of the target driving object on the vehicle in the vehicle; the real-time sight line area information is obtained by performing sight line tracking processing on the face image of the target driving object and then mapping the face image to a plurality of preset discretized gazing areas, wherein the preset discretized gazing areas comprise a front road area and a target monitoring area; and determining a sight line region transfer sequence of the target driving object under a first preset time window based on the real-time sight line region information.
6. 6. The lane departure intention recognition method according to claim 5, wherein the operation information includes turn signal light information, brake pedal opening information, accelerator pedal opening information, and hand torque information of the target driving object, and the determining that the target driving object actively steers the vehicle based on the operation information of the target driving object in the vehicle includes: If the information of the steering lamp controlled by the target driving object in the vehicle is that the steering lamp is turned on, determining that the target driving object actively controls the vehicle, or If the brake pedal opening information of the target driving object in the vehicle is that the brake pedal opening is larger than a first preset opening value, determining that the target driving object actively controls the vehicle, or If the accelerator pedal opening information of the target driving object in the vehicle is that the accelerator pedal opening is larger than a second preset opening value, determining that the target driving object actively controls the vehicle, or And if the hand torque information is that the hand torque is larger than a preset hand torque threshold value, determining that the target driving object actively controls the vehicle.
7. 7. A lane departure intention recognition method according to claim 2 or 3, wherein after the determination that the target driving object has the active lane departure intention at the present time, the method further comprises: Judging whether the vehicle generates the transverse offset under a second preset time window, wherein the second preset time window is an effective time window corresponding to the deviation intention of the driving lane, and the starting moment of the second preset time window is the current moment; If yes, acquiring the state information of the vehicle and the hand moment information of the target driving object, and verifying the driving lane departure intention based on the state information and the hand moment information.
8. 8. The lane departure intention recognition method according to claim 7, wherein the state information includes a vehicle lateral offset direction, the hand moment information includes a hand moment direction and a hand moment magnitude, and the verifying the driving lane departure intention based on the state information and the hand moment information includes: And if the transverse deviation direction of the vehicle is consistent with the hand torque direction and the hand torque is greater than or equal to a preset hand torque deviation threshold, determining that the driving lane deviation intention of the target driving object at the current moment is a real driving lane deviation intention.
9. 9. The lane departure intention recognition method according to claim 1, wherein before determining that the target driving object has an intention recognition probability of an active lane departure intention based on the line-of-sight region transfer sequence and a trained intention recognition model, the method further comprises: Constructing graph nodes corresponding to each preset discretized gazing area to obtain a node set; Determining edges between the graph nodes based on the spatial association relation among the preset discrete sight areas to obtain an edge set; a target graph structure is generated based on the set of nodes and the set of edges.
10. 10. The method for identifying the intention of a lane departure according to claim 9, wherein said constructing the graph node corresponding to each of the preset discretized gazing regions, to obtain a node set, includes: Determining an area category of the preset discretized gazing area, wherein the area category comprises a front road area, a left rearview mirror area, a right rearview mirror area, a left window area, a right window area, an in-vehicle rearview mirror area, an instrument panel area, a center console area and an in-vehicle sight discrete area; Constructing a graph node corresponding to each preset discretized gazing area based on the area category of each preset discretized gazing area; The set of graph nodes is determined as the set of nodes.
11. 11. The method for recognizing the intention of the lane departure according to claim 9, wherein the intention recognition model comprises an embedded layer, a space-time convolution module, an averaging pooling layer, a full connection layer and an output layer which are sequentially cascaded, wherein the determining the intention recognition probability that the target driving object has the intention of the lane departure based on the sight area transfer sequence and the trained intention recognition model comprises: Inputting the sight area transfer sequence and the target graph structure into the trained intention recognition model so as to obtain an initial node characteristic tensor through extraction of the embedding layer; Performing space-time enhancement convolution processing on the initial node characteristic tensor through the space-time convolution module to obtain a space-time enhancement characteristic tensor; carrying out global average pooling treatment on the space-time enhancement feature tensor through the average pooling layer to obtain a global feature vector; performing deviation intention analysis processing on the global feature vector through the full connection layer to obtain intention scores; and converting the intention score into probability distribution through the output layer to obtain intention recognition probability that the target driving object has the intention of deviating from the driving lane.
12. 12. The method of claim 11, wherein the spatio-temporal convolution module includes at least two sequentially cascaded spatio-temporal convolution blocks, each of the spatio-temporal convolution blocks including sequentially cascaded temporal convolution layers and spatial map convolution layers, and wherein the performing, by the spatio-temporal convolution module, a spatio-temporal enhancement convolution process on the initial node feature tensor to obtain a spatio-temporal enhancement feature tensor includes: for each space-time convolution block, carrying out feature sequence convolution processing on the initial node feature tensor in the time dimension through the time convolution layer to obtain a time sequence enhanced feature tensor; And carrying out graph convolution processing on the time sequence enhancement characteristic tensor in a space dimension through the space graph convolution layer to obtain the time-space enhancement characteristic tensor.
13. 13. The lane departure intention recognition method according to claim 1, wherein the trained intention recognition model is determined by: Determining at least one historical sight-line region transfer sequence of the historical driving object under a historical preset time window based on at least one historical sight-line region information of the historical driving object output by the driver monitoring system; Constructing a historical training data set based on all the historical sight line area transfer sequences and the historical intention labels corresponding to each historical sight line area transfer sequence, wherein the historical intention labels are used for representing whether the historical driving object has an active lane departure intention or not; inputting the historical training data set into an initial space-time diagram convolution model, and determining a predicted historical intent recognition probability that the historical driving object has the driving lane departure intent; and carrying out iterative training on the initial space-time diagram convolution model based on the historical intention labels and the predicted historical intention recognition probabilities, and determining the trained intention recognition model.
14. 14. The lane departure intention recognition method of claim 13, wherein the historical intention labels comprise positive sample labels, the historical intention labels corresponding to the historical line-of-sight region transfer sequence being determined by: if the historical driving state information of the historical driving object meets a first preset historical driving condition under the historical preset time window, determining that a historical intention label corresponding to the historical sight line area transfer sequence is a positive sample label; wherein the first preset historical driving condition includes: The historical driving object turns on a turn signal of the vehicle; and the history driving object controls the vehicle to finish lane change operation; and the hand torque direction of the history driving object is consistent with the lane change direction of the vehicle; And the residence time of the historical driving object in the sight line area on the same side of the lane change direction is longer than the preset residence time.
15. 15. The lane departure intention recognition method of claim 13, wherein the historical intention labels comprise negative sample labels, the historical intention labels corresponding to the historical gaze region transfer sequence being determined by: If the historical driving state information of the historical driving object meets a second preset historical driving condition under the historical preset time window, determining that a historical intent tag corresponding to the historical sight line region transfer sequence is a negative sample tag; wherein the second preset historical driving condition includes: The history driving object does not turn on a turn signal of the vehicle; And the hand torque direction of the historic driving object is inconsistent with the lane change direction of the vehicle; And the residence time of the historical driving object in the sight line area on the same side of the lane change direction is smaller than or equal to the preset residence time.
16. 16. An intention recognition device of a lane departure, characterized in that the intention recognition device of a lane departure includes: The acquisition module is used for acquiring a sight line region transfer sequence of a target driving object in the vehicle under a first preset time window before the vehicle is transversely deviated, wherein the first preset time window is a time window with preset duration before the current moment; The first determining module is used for determining intention recognition probability of the active lane departure intention of the target driving object based on the sight area transfer sequence and the trained intention recognition model under the condition that the sight area transfer sequence does not accord with a preset sight area transfer mode; And the second determining module is used for determining whether the target driving object has an active lane departure intention at the current moment based on the intention recognition probability and a preset probability threshold.
17. 17. An electronic device comprising a processor, a memory and a bus, the memory storing machine-readable instructions executable by the processor, the processor and the memory in communication over the bus when the electronic device is in operation, the machine-readable instructions being executable by the processor to perform the steps of the lane departure intention recognition method as claimed in any one of claims 1 to 15.
18. 18. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a computer program which, when executed by a processor, performs the steps of the method for identifying the intention of a lane departure as claimed in any one of the preceding claims 1-15.
19. 19. A vehicle mounted with the lane departure intention recognition device according to claim 16.
20. 20. A computer program product comprising a computer program or computer-executable instructions which, when executed by a processor, implement the steps of the method for identifying the intention of a lane departure as claimed in any one of claims 1 to 15.

Description

Lane departure intention recognition method, device, electronic equipment and storage medium Technical Field The present application relates to the field of driving safety recognition technologies, and in particular, to a method and apparatus for recognizing intention of lane departure, an electronic device, and a storage medium. Background At present, with the development of society and the advancement of technology, advanced driving assistance systems have become the standard of intelligent vehicles, and conventional lane departure warning systems can effectively avoid a large number of traffic accidents by monitoring the behavior of vehicles which unintentionally deviate from lanes. However, the existing system generally has the problem of poor accuracy of identifying the intention of the lane departure intention, namely, unnecessary alarm interference can be generated when a driver consciously changes lanes, and the user experience and the acceptance of the driving assistance system by the user are affected. Disclosure of Invention The embodiment of the application provides a lane departure intention recognition method, a lane departure intention recognition device, electronic equipment and a storage medium, solves the problem that in the prior art, the intention accuracy of recognizing the lane departure intention is poor, and influences the user experience and the technical problem of the user acceptance of a driving assistance system. In a first aspect of the present embodiment, the present embodiment provides a method for identifying an intention of a lane departure, the method for identifying an intention of a lane departure including: Before the vehicle is laterally deviated, acquiring a sight line region transfer sequence of a target driving object in the vehicle under a first preset time window, wherein the first preset time window is a time window with preset duration before the current moment; under the condition that the sight line area transfer sequence does not accord with a preset sight line transfer mode, determining intention recognition probability of the active lane departure intention of the target driving object based on the sight line area transfer sequence and the trained intention recognition model; and determining whether the target driving object has the driving lane departure intention at the current moment based on the intention recognition probability and the preset probability threshold. In a possible embodiment, determining whether the target driving object has an intention to deviate from the driving lane at the current moment based on the intention recognition probability and a preset probability threshold includes: if the intention recognition probability is greater than or equal to a preset probability threshold, determining that the target driving object has an active lane departure intention at the current moment; if the intention recognition probability is smaller than the preset probability threshold value, determining that the target driving object does not have the driving lane departure intention at the current moment. In one possible embodiment, the method further comprises: and under the condition that the sight line area transfer sequence accords with a preset sight line transfer mode, determining that the target driving object has an active lane departure intention at the current moment. In a possible embodiment, the preset gaze transfer mode includes a scanning observation mode and a continuous gaze mode, the method further comprising: when the sight line area transfer sequence does not accord with the scanning observation mode and does not accord with the continuous fixation mode, determining that the sight line area transfer sequence does not accord with the preset sight line transfer mode; When the sight line area transfer sequence accords with any one of a scanning observation mode and a continuous fixation mode, determining that the sight line area transfer sequence accords with a preset sight line transfer mode; The continuous fixation mode is used for representing a fixation mode that the continuous fixation time length of the sight line in the at least one target monitoring area exceeds a preset time length. In one possible embodiment, acquiring a line-of-sight region transfer sequence of a target driving object in a vehicle under a first preset time window includes: acquiring real-time sight area information of a target driving object under the condition that the target driving object is determined to not actively control the vehicle based on the operation information of the target driving object in the vehicle; The real-time sight line area information is obtained by mapping a face image of a target driving object to a plurality of preset discretized gazing areas after sight line tracking processing, wherein the preset discretized gazing areas comprise a front road area and a target monitoring area; and determining a sight-line region transfer sequ