Lane line visual detection method and system under expected function safety triggering condition and vehicle

Abstract

The invention discloses a lane line visual detection method and system under an expected function safety triggering condition and a vehicle, and the detection method comprises the steps: obtaining an original image of a lane line, constructing a backbone network based on a residual structure, processing the original image through the backbone network, obtaining a global feature map, and obtaining a lane line visual detection result; using the feature of the starting point position in the global feature map as the representative feature of the lane line instance, and associating the correlation between the representative feature of the current lane line instance and the representative features of other lane line instances into the representative feature of the current lane line instance by using an attention mechanism; according to the method, the lane line instance mask is established through the lane line instance representative features and the lane line starting point coordinates, the single lane line instance features are separated from the global feature map by using the mask, and the final lane line is obtained after the single lane line instance features are separated from the global feature map, so that the stability and the detection effect of a detector are improved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving, and in particular to a lane line visual detection method, system and vehicle under expected functional safety triggering conditions. Background Art

[0002] In autonomous driving, lane detection is a key link in autonomous driving technology, an important part of expected functional safety, and an important factor in determining whether the L3 autonomous driving system can operate. At present, lane detection methods mainly include segmentation-based, anchor-based, and key point-based methods. Among them, the segmentation-based method requires classification of each pixel in the image, ignoring the topological information of the lane line itself, and has poor real-time performance. The anchor-based method requires a large number of anchors to be set in advance, and the setting of anchors has a great impact on the overall detection effect. The key point-based method is relatively flexible and can fit various topological structures of lane lines. The anchor-based method and the key point-based method have certain effects on detection effect and detection speed, but in dark light conditions, the images collected by the sensor can only identify nearby lane lines. Due to the loss of global semantic information, the detection effect of the detector is reduced. Therefore, it is still a challenging task to study lane detection solutions with high accuracy and real-time performance to ensure the safety of autonomous vehicles.

[0003] The disclosure of the above background technology content is only used to assist in understanding the inventive concept and technical solution of the present invention. It does not necessarily belong to the prior art of this patent application, nor does it necessarily provide technical guidance. In the absence of clear evidence that the above content has been disclosed before the filing date of this patent application, the above background technology should not be used to evaluate the novelty and creativity of the present application. Summary of the invention

[0004] The purpose of the present invention is to provide a method, system and vehicle for visual detection of lane lines under expected functional safety triggering conditions, which can accurately detect lane lines under safety triggering conditions such as insufficient light to meet the expected functional safety requirements required for autonomous driving.

[0005] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] A lane line detection method under an expected functional safety triggering condition, comprising:

[0007] Get the original lane line image;

[0008] Constructing a backbone network based on a residual structure, processing the original image through the backbone network to obtain a global feature map, and passing the global feature map through two convolution blocks to obtain the instance acquisition input feature and the lane line separation input feature;

[0009] Regressing the global feature map into a starting point heat map, selecting the first position corresponding to the heat map as the starting point, and using the feature of the starting point position in the global feature map as a representative feature of the lane line instance;

[0010] Use the attention mechanism to associate the correlation between the representative features of the current lane line instance and the representative features of other lane line instances with the representative features of the current lane line instance;

[0011] A lane line instance mask is established using the lane line instance representative features and the lane line starting point coordinates, and the single lane line instance feature is separated from the global feature map using the mask;

[0012] After separating the single lane line instance features from the global feature map, the key points of the entire lane line are located through heat map regression, and the final lane line is obtained according to the key point information of the lane line.

[0013] Further, based on any one of the above technical solutions or a combination of multiple technical solutions, regressing the global feature map into a starting point heat map, selecting the corresponding highest position in the heat map as the starting point, and using the feature of the starting point position in the global feature map as a representative feature of the lane line instance, includes:

[0014] Regressing the global feature map into a starting point heat map, wherein the starting point heat map indicates the probability of the lane line starting point corresponding to each pixel point, and obtaining the starting point coordinates of each lane line according to the lane line starting point heat map;

[0015] The position with the highest response in the starting point heat map is selected as the starting point of the lane line, and the feature of the starting point position in the global feature map is used as the representative feature of the lane line instance.

[0016] Further, based on any one of the above-mentioned technical solutions or a combination of multiple technical solutions, the use of the attention mechanism to associate the correlation between the representative feature of the current lane line instance and the representative features of other lane line instances to the representative feature of the current lane line instance includes:

[0017] Transform the representative features of the lane line instance into the embedding space to obtain the embedding space representation of the representative features of the lane line instance;

[0018] The correlation between the two lane lines in the embedding space is established by a dot product operation;

[0019] After the embedding space representation and the dot product operation, the correlation between the current lane line and all other lane lines is established to obtain a correlation feature set, and the features in the set are fused along the channel dimension to obtain the correlation feature of the current lane line.

[0020] Further, based on any one of the above-mentioned technical solutions or a combination of multiple technical solutions, the use of the attention mechanism to associate the correlation between the representative feature of the current lane line instance and the representative features of other lane line instances to the representative feature of the current lane line instance also includes:

[0021] The correlation feature of the current lane line and the representative feature of the current lane line instance are fused along the channel dimension to obtain a relationship perception feature; the relationship perception feature is passed through two convolutional layers to obtain an attention value;

[0022] The attention value and the lane line instance representative feature are weighted to obtain the lane line instance representative feature including the correlation between the current lane line and other lane lines.

[0023] Further, based on any one of the above technical solutions or a combination of multiple technical solutions, the lane line instance mask is established by using the lane line instance representative feature and the lane line starting point coordinates, and the single lane line instance feature is separated from the global feature map using the mask, including:

[0024] Calculate the inner product of the representative feature of the lane line instance and the feature at each spatial position on the global feature map to obtain a mapping relationship of the lane line instance on the global feature map;

[0025] Calculating the mapping relationship, the starting point coordinates and the relative coordinates of each position on the global feature map to obtain a relative coordinate map;

[0026] The mapping relationship is fused with the relative coordinate map along the channel direction and convolved to obtain a spatial mask of the lane line instance;

[0027] The spatial mask is used to separate a single lane line instance feature from the global feature map.

[0028] Further, based on any one of the above-mentioned technical solutions or a combination of multiple technical solutions, after separating the single lane line instance feature from the global feature map, locating the key points of the entire lane line by heat map regression, and obtaining the final lane line according to the key point information of the lane line, including:

[0029] Regressing a heat map of all key points on a single lane line using the single lane line instance feature;

[0030] The position with the highest response in the heat map is selected as the coordinate of the key point of the lane line, and the coordinates of all the key points of the single lane line are connected to obtain the final lane line.

[0031] Furthermore, based on any one of the above-mentioned technical solutions or a combination of multiple technical solutions, the method further includes: the method further includes a loss function calculation step:

[0032] Among them, the starting point heat map loss and the heat map loss of a single lane line both use focal loss.

[0033] According to another aspect of the present invention, the present invention provides a lane line detection system under an expected functional safety triggering condition, comprising the following modules: an image acquisition module, for acquiring an original image of a lane line;

[0034] A feature map generation module is configured to construct a backbone network based on a residual structure, process the original image through the backbone network to obtain a global feature map, and pass the global feature map through two convolution blocks to obtain the instance acquisition input feature and the lane line separation input feature;

[0035] A lane line instance representative feature acquisition module is used to regress the global feature map into a starting point heat map, select the corresponding first position in the heat map as the starting point, and use the feature of the starting point position in the global feature map as the lane line instance representative feature;

[0036] A module for establishing correlation between different lane lines, which is used to associate the correlation between the representative features of the current lane line instance and the representative features of other lane line instances with the representative features of the current lane line instance using an attention mechanism;

[0037] A lane line instance feature separation module is used to establish a lane line instance mask through the lane line instance representative feature and the lane line starting point coordinates, and use the mask to separate a single lane line instance feature from the global feature map;

[0038] The lane line determination module is used to separate the single lane line instance features from the global feature map, locate the key points of the entire lane line through heat map regression, and obtain the final lane line based on the key point information of the lane line.

[0039] According to another aspect of the present invention, the present invention provides a lane line detection system under expected functional safety triggering conditions, comprising:

[0040] A memory storing executable program code;

[0041] a central processing unit coupled to the memory;

[0042] The central processing unit calls the executable program code stored in the memory to execute the lane line detection method under the above-mentioned expected functional safety triggering condition.

[0043] According to another aspect of the present invention, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein the computer program enables a computer to execute the above-mentioned lane line detection method under the expected functional safety triggering condition.

[0044] According to another aspect of the present invention, the present invention provides a vehicle, the vehicle comprising the lane line detection system under the aforementioned expected functional safety triggering condition

[0045] The beneficial effects brought by the technical solution provided by the present invention are as follows:

[0046] a. The present invention provides a lane line detection method under expected functional safety triggering conditions: by establishing the correlation between lane line instances, the problem of inaccurate detection in lane line detection based on key points in insufficient light is solved, the stability of the model is improved, and the effect of dark light lane line detection is improved.

[0047] b. The present invention provides a lane line detection system under the expected functional safety triggering condition: the image collected by the sensor is passed through the backbone network to obtain a feature map, the attention mechanism is used to associate the correlation between the representative feature of the current lane line instance and the representative features of other lane line instances to the representative feature of the current lane line instance, a lane line instance mask is established through the lane line instance representative feature and the coordinates of the lane line starting point, and the mask is used to separate the single lane line instance feature from the global feature map, the key points of the entire lane line are located through heat map regression, and the final lane line is obtained according to the key point information of the lane line, so that the lane line detection effect under dark light conditions is further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0049] Figure 1 A schematic flow chart of a lane line detection method under expected functional safety triggering conditions provided for an exemplary embodiment of the present invention;

[0050] Figure 2 A model framework diagram of a lane line detection method under expected functional safety triggering conditions provided by an exemplary embodiment of the present invention;

[0051] Figure 3 A flowchart for establishing correlation between different lane lines provided for an exemplary embodiment of the present invention;

[0052] Figure 4 A schematic diagram of a lane detection system under expected functional safety triggering conditions provided by an exemplary embodiment of the present invention;

[0053] Figure 5A schematic diagram of a vehicle is provided for an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0055] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or equipment that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0056] In the related art, the lane detection method based on key points locates the lane line by regressing the heat map of the key points on the lane line, and associates the points on the lane line with the entire lane line. Although good detection results are obtained in simple scenes, under dark conditions, the lane line features are not obvious, which causes great difficulties for the heat map regression of the key points. In order to solve this problem, it has been proposed to express lane detection as a key point estimation problem to more flexibly describe the shape of the lane line and gradually group adjacent key points belonging to the same lane line in a point-by-point manner. However, this method does not work well when the lighting is greatly affected. Therefore, the present invention proposes an innovative technology that first extracts lane line instances, then establishes the correlation between lane line instances, and uses the attention mechanism to associate the correlation between lane lines to the lane line instances, thereby improving the stability of the detector and the detection effect.

[0057] In one embodiment of the present invention, Figure 1The present invention provides a lane line detection method under the condition of expected functional safety triggering, including: obtaining the original image of the lane line; constructing a backbone network based on the residual structure, processing the original image through the backbone network to obtain a global feature map; regressing the global feature map to a starting point heat map, selecting the first position in the heat map as the starting point, and using the feature of the starting point position in the global feature map as the lane line instance representative feature; using the attention mechanism to associate the correlation between the current lane line instance representative feature and other lane line instance representative features to the current lane line instance representative feature; establishing a lane line instance mask through the lane line instance representative feature and the lane line starting point coordinates, and using the mask to separate a single lane line instance feature from the global feature map; after separating the single lane line instance feature from the global feature map, locating the key points of the entire lane line through heat map regression, and obtaining the final lane line according to the key point information of the lane line.

[0058] It should be noted that the road image can be collected by a camera located at any position in the vehicle. For example, when the vehicle is traveling in a straight line, the road image is collected by a camera located in front of the vehicle. When the vehicle turns, the road image is collected by a camera located on the left or right side of the vehicle. The present invention does not specifically limit this, as long as the collected road image contains lane lines. It is also possible to shoot a road video by a camera device on the vehicle while the vehicle is traveling, and then collect the road image corresponding to the current frame of the captured video. The collected road image reflects the road conditions of the vehicle's environment, including at least one lane line. It should be noted that the lane lines in the embodiments of the present invention include the vehicle's driving lane lines, such as various indicator lines on the road surface, and also include lane boundary lines, such as the curb of the road.

[0059] In one embodiment of the present invention, the present invention provides a model framework, such as Figure 2 As shown in the figure, the backbone network extracts the global feature map from the original image. By regressing the starting point heat map, the coordinates of the lane line starting point can be obtained, and the features at this position are extracted as the representative features of the lane line instance. Then, the correlation between the representative features of different lane line instances is established, and the attention mechanism is used to associate the correlation between the representative features of the current lane line instance and the representative features of other lane line instances to the representative features of the current lane line instance. Next, a lane line instance mask is established through the lane line instance representative features and the lane line starting point coordinates, and the lane line instance features are separated from the global feature map using the mask.

[0060] In one embodiment of the present invention, a backbone network based on a residual structure is constructed, and the original image is processed by the backbone network to obtain a global feature map represented as follows:

[0061] X=backbone(X orig )

[0062] Among them, the original image H orig and W orig is the height and width of the original image, and the obtained feature map H and W are the height and width of the downsampled feature map, and C is the number of channels.

[0063] In one embodiment of the present invention, after obtaining the global feature map, the global feature map is passed through two convolution blocks to obtain the instance acquisition input feature and the lane line separation input feature, specifically:

[0064] The feature map X is passed through two convolution blocks to obtain the instance input features and the lane line separation input features:

[0065]

[0066] Get the input features of the instance The graph is regressed into a starting point heat map, which indicates the probability of the lane line starting point corresponding to each pixel point. According to the lane line starting point heat map, the starting point heat map of each lane line is obtained:

[0067]

[0068] Among them, H i is the starting point heat map of the i-th lane line, i = 1, ..., N, N is the total number of lane lines.

[0069] In one embodiment of the present invention, the starting point heat map H is selected i The position with the highest response is taken as the starting point of the lane line, and the starting point coordinate is represented as l i , the features of the starting position in the feature map are considered to be the representative features of the current lane line instance, and the features of each starting position are extracted as the representative features of the lane line instance:

[0070]

[0071] In one embodiment of the present invention, the attention mechanism is used to associate the correlation between the representative features of the current lane line instance and the representative features of other lane line instances with the representative features of the current lane line instance. For details, please refer to Figure 3 ,include:

[0072] Step 1: First, the lane line instance represents the feature f i and f j Transform into embedding space:

[0073] θ=ReLU(conv(f i ))

[0074] φ=ReLU(conv(f j ))

[0075] Where θ and φ are the representations of the representative features of the i-th lane line and the j-th lane line in the embedding space respectively.

[0076] Step 2: Establish the correlation between the two lane lines in the embedding space through the dot product operation:

[0077] r i，j =θ·φ, i≠j; where, r i，j Represents the correlation feature between the i-th lane line and the j-th lane line.

[0078] Repeat the calculations in steps 1 and 2 to establish the correlation between the current lane line and all other lane lines, and obtain the correlation feature set The features in the set are fused together along the channel dimension to obtain the correlation feature R of the current lane line. i .

[0079] The correlation feature R of the current lane line i The relationship perception feature y is obtained by fusing the representative feature of the current lane line instance along the channel dimension i

[0080] ψ=conv(f i )

[0081] y i =concat(ψ, R i )

[0082] Where ψ represents the representation of the representative features of the current lane line in the embedding space, and concat(·) represents the concatenation of features along the channel direction.

[0083] The relation-aware features are passed through two convolutional layers to obtain the attention value a i .

[0084] The attention value a i The obtained lane line instance representative feature is weighted, and the obtained lane line instance representative feature includes the correlation between the lane line and other lane lines

[0085]

[0086] In one embodiment of the present invention, the step of establishing a lane line instance mask using lane line instance representative features and lane line starting point coordinates, and using the mask to separate a single lane line instance feature from a global feature map includes:

[0087] The obtained lane line instance represents the feature Separate the input features from the obtained lane lines The inner product of the features at each spatial position on the lane line is calculated to obtain the representative features of the lane line instance Separating input features in lane lines The mapping relation M′ on i .

[0088] Calculate the starting point coordinates H i Separate input features from lane lines The relative coordinates of each position on the graph O are obtained. i .

[0089] The mapping relationship M′ i With the relative coordinate diagram O i Fusion and convolution operations are performed along the channel direction to obtain the spatial mask M of the lane line instance. i .

[0090] M i =Sigmoid(Conv(concat(O i , M′ i )))

[0091] Use the spatial mask M i Separate input features from lane lines Separate the feature Y of the i-th lane line i :

[0092]

[0093] In one embodiment of the present invention, after separating the single lane line instance feature from the global feature map, locating the key points of the entire lane line through heat map regression, and obtaining the final lane line according to the key point information of the lane line, including:

[0094] Use the feature Y of a single lane line i Regress a heatmap of all key points on a single lane line:

[0095]

[0096] in, Represents the heat map of the kth key point of the i-th lane line, k=1,...,K.

[0097] Select Heatmap The position with the highest response is taken as the coordinate of the lane line key point, and the coordinates of all the key points of a single lane are connected to form the final lane line.

[0098] In one embodiment of the present invention, the above method further includes a loss function calculation step:

[0099] The starting point heat map loss and the heat map loss of a single lane line both use focal loss:

[0100]

[0101] Among them, the response of the true value of the key point on the heat map is expressed as x, y are the horizontal and vertical coordinates of each real key point, H x，y is the lane line starting point heat map or key point heat map output by the model, and β and α are the hyperparameters of the focal loss.

[0102] In summary, an embodiment of the present invention provides a lane line detection method under expected functional safety trigger conditions, which first extracts lane line instances, then establishes the correlation between lane line instances, and uses an attention mechanism to associate the correlation between lane lines with the lane line instances, thereby improving the stability of the detector and the detection effect.

[0103] In one embodiment of the present invention, Figure 4 The present invention provides a lane line detection system under the expected functional safety triggering condition, including:

[0104] An image acquisition module is used to acquire the original image of the lane line;

[0105] A feature map generation module is configured to construct a backbone network based on a residual structure, process the original image through the backbone network to obtain a global feature map, and pass the global feature map through two convolution blocks to obtain the instance acquisition input feature and the lane line separation input feature;

[0106] A lane line instance representative feature acquisition module is used to regress the global feature map into a starting point heat map, select the corresponding first position in the heat map as the starting point, and use the feature of the starting point position in the global feature map as the lane line instance representative feature;

[0107] A module for establishing correlation between different lane lines, which is used to associate the correlation between the representative features of the current lane line instance and the representative features of other lane line instances with the representative features of the current lane line instance using an attention mechanism;

[0108] A lane line instance feature separation module is used to establish a lane line instance mask through the lane line instance representative feature and the lane line starting point coordinates, and use the mask to separate a single lane line instance feature from the global feature map;

[0109] The lane line determination module is used to separate the single lane line instance features from the global feature map, locate the key points of the entire lane line through heat map regression, and obtain the final lane line based on the key point information of the lane line.

[0110] The lane line detection system under expected functional safety triggering conditions of this embodiment and the lane line detection system method under expected functional safety triggering conditions provided in the above embodiment belong to the same concept. Here, by quoting the full text, the entire content of the lane line detection method embodiment is introduced into the present lane line detection system embodiment.

[0111] In one embodiment of the present invention, the present invention also provides a lane line detection system under expected functional safety triggering conditions, including: a memory storing executable program code; a central processing unit coupled to the memory; the central processing unit calls the executable program code stored in the memory to execute the lane line detection method under the expected functional safety triggering conditions described above. The concept of this detection system embodiment and the working process of the lane line detection method under the expected functional safety triggering conditions of the above embodiment are the same concept. The entire contents of the above embodiment of the lane line detection method under the expected functional safety triggering conditions are incorporated into the embodiment of the lane line detection system under the expected functional safety triggering conditions by full reference, and no further description is given.

[0112] In one embodiment of the present invention, the present invention also provides a vehicle, such as Figure 5 As shown, it includes a processor 501, a memory 502, and a computer program 5021 stored in the memory and executable on the processor. When the processor 501 executes the computer program, the lane line detection method under the aforementioned expected functional safety triggering condition is implemented.

[0113] In one embodiment of the present invention, the embodiment of the present invention further provides a computer-readable storage medium, in which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned detection method is implemented.

[0114] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0115] The above is only a specific implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A lane line detection method under expected functional safety triggering conditions, characterized in that: include: Get the original lane line image; Constructing a backbone network based on a residual structure, processing the original image through the backbone network to obtain a global feature map, and passing the global feature map through two convolution blocks to obtain the instance acquisition input feature and the lane line separation input feature; Regressing the input features obtained from the instance into a starting point heat map, selecting the first position in the heat map as the starting point, and using the features of the starting point position in the global feature map as the representative features of the lane line instance; Use the attention mechanism to associate the correlation between the representative features of the current lane line instance and the representative features of other lane line instances with the representative features of the current lane line instance; A lane line instance mask is established using the lane line instance representative features and the lane line starting point coordinates, and the single lane line instance feature is separated from the global feature map using the mask; After separating the single lane line instance features from the global feature map, the key points of the entire lane line are located through heat map regression, and the final lane line is obtained according to the key point information of the lane line.

2. The lane line detection method under the expected functional safety triggering condition according to claim 1 is characterized in that: The step of regressing the input features obtained from the instance into a starting point heat map, selecting the corresponding first position in the heat map as the starting point, and using the features of the starting point position of the input features obtained from the full instance as the representative features of the lane line instance includes: Regressing the global feature map into a starting point heat map, wherein the starting point heat map indicates the probability of the lane line starting point corresponding to each pixel point, and obtaining the starting point coordinates of each lane line according to the lane line starting point heat map; The position with the highest response in the starting point heat map is selected as the starting point of the lane line, and the feature of the starting point position in the global feature map is used as the representative feature of the lane line instance.

3. The lane line detection method under the expected functional safety triggering condition according to claim 1 is characterized in that: The use of the attention mechanism to associate the correlation between the representative feature of the current lane line instance and the representative features of other lane line instances with the representative feature of the current lane line instance includes: Transform the representative features of the lane line instance into the embedding space to obtain the embedding space representation of the representative features of the lane line instance; The correlation between the two lane lines in the embedding space is established by a dot product operation; After the embedding space representation and the dot product operation, the correlation between the current lane line and all other lane lines is established to obtain a correlation feature set, and the features in the set are fused along the channel dimension to obtain the correlation feature of the current lane line.

4. The lane line detection method under the expected functional safety triggering condition according to claim 3 is characterized in that: The method of using the attention mechanism to associate the correlation between the representative feature of the current lane line instance and the representative features of other lane line instances with the representative feature of the current lane line instance further includes: The correlation feature of the current lane line and the representative feature of the current lane line instance are fused along the channel dimension to obtain a relationship perception feature; the relationship perception feature is passed through two convolutional layers to obtain an attention value; The attention value and the lane line instance representative feature are weighted to obtain the lane line instance representative feature including the correlation between the current lane line and other lane lines.

5. The lane line detection method under the expected functional safety triggering condition according to claim 2 is characterized in that: The method of establishing a lane line instance mask by using the lane line instance representative feature and the lane line starting point coordinates, and using the mask to separate a single lane line instance feature from the global feature map, includes: Calculate the inner product of the representative feature of the lane line instance and the feature at each spatial position on the global feature map to obtain a mapping relationship of the lane line instance on the global feature map; Calculating the mapping relationship, the starting point coordinates and the relative coordinates of each position on the global feature map to obtain a relative coordinate map; The mapping relationship is fused with the relative coordinate map along the channel direction and convolved to obtain a spatial mask of the lane line instance; The spatial mask is used to separate a single lane line instance feature from the global feature map.

6. The lane line detection method under the expected functional safety triggering condition according to claim 1 is characterized in that: After separating the single lane line instance feature from the global feature map, locating the key points of the entire lane line through heat map regression, and obtaining the final lane line according to the key point information of the lane line, including: Regressing a heat map of all key points on a single lane line using the single lane line instance feature; The position with the highest response in the heat map is selected as the coordinate of the key point of the lane line, and the coordinates of all the key points of the single lane line are connected to obtain the final lane line.

7. The lane line detection method under the expected functional safety triggering condition according to claim 1 is characterized in that: The method also includes a loss function calculation step: Among them, the starting point heat map loss and the heat map loss of a single lane line both use focal loss.

8. A lane detection system under expected functional safety triggering conditions, characterized in that: include: An image acquisition module is used to acquire the original image of the lane line; A feature map generation module is configured to construct a backbone network based on a residual structure, process the original image through the backbone network to obtain a global feature map, and pass the global feature map through two convolution blocks to obtain the instance acquisition input feature and the lane line separation input feature; A lane line instance representative feature acquisition module is used to regress the global feature map into a starting point heat map, select the corresponding first position in the heat map as the starting point, and use the feature of the starting point position in the global feature map as the lane line instance representative feature; A module for establishing correlation between different lane lines, which is used to associate the correlation between the representative features of the current lane line instance and the representative features of other lane line instances with the representative features of the current lane line instance using an attention mechanism; A lane line instance feature separation module is used to establish a lane line instance mask through the lane line instance representative feature and the lane line starting point coordinates, and use the mask to separate a single lane line instance feature from the global feature map; The lane line determination module is used to separate the single lane line instance features from the global feature map, locate the key points of the entire lane line through heat map regression, and obtain the final lane line based on the key point information of the lane line.

9. A lane detection system under expected functional safety triggering conditions, characterized in that: include: A memory storing executable program code; a central processing unit coupled to the memory; The central processing unit calls the executable program code stored in the memory to execute the lane line detection method under the expected functional safety triggering condition described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program enables a computer to execute the lane line detection method under the expected functional safety triggering condition as described in any one of claims 1-7.

11. A vehicle, characterized in that: comprising a processor, a memory and a computer program stored in the memory and executable on the processor, Wherein, when the processor executes the computer program, the lane line detection method under the expected functional safety triggering condition as described in any one of claims 1-7 is implemented.

Images