CN111213154A

CN111213154A - Lane line detection method, lane line detection equipment, mobile platform and storage medium

Info

Publication number: CN111213154A
Application number: CN201980004927.3A
Authority: CN
Inventors: 唐蔚博; 许睿
Original assignee: SZ DJI Technology Co Ltd
Current assignee: SZ DJI Technology Co Ltd
Priority date: 2019-03-08
Filing date: 2019-03-08
Publication date: 2020-05-29
Also published as: WO2020181426A1

Abstract

The embodiment of the invention provides a lane line detection method, a device, a mobile platform and a storage medium, wherein the method comprises the following steps: acquiring an image of a lane line, and determining a local lane line set; constructing an assumed lane line according to the prior information of the lane line and the local lane line set to obtain an assumed lane line set, wherein the prior information comprises: characteristic attributes of the lane lines; and determining a target lane line set from the assumed lane line set, so that the accuracy of lane line detection can be improved.

Description

Lane line detection method, lane line detection equipment, mobile platform and storage medium

The disclosure of this patent document contains material which is subject to copyright protection. The copyright is owned by the copyright owner. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document or the patent disclosure, as it appears in the patent and trademark office official records and records.

Technical Field

The embodiment of the invention relates to the technical field of control, in particular to a lane line detection method, lane line detection equipment, a mobile platform and a storage medium.

Background

The lane line local map is mainly applied to the field of automatic driving, a driving plan of a current driving vehicle can be planned based on the lane line local map, the establishment of the local map mainly depends on the detection of lane lines, the currently adopted lane line detection method is mainly based on images collected by image sensors such as cameras, and after the images are collected by the image sensors on the automatic driving vehicle, the lane lines in the field angle of the image sensors are determined through image recognition.

Research shows that the lane line constructed based on the image recognition mode is easily interfered by factors such as image noise, certain errors exist in the determination of the lane line, and an optimization space also exists.

Disclosure of Invention

In view of this, embodiments of the present invention provide a lane line detection method, a lane line detection device, a mobile platform, and a storage medium, which can improve the accuracy of lane line detection.

In one aspect, an embodiment of the present invention provides a lane line detection method, including:

acquiring an image of a lane line, and determining a local lane line set;

constructing an assumed lane line according to the prior information of the lane line and the local lane line set to obtain an assumed lane line set, wherein the prior information comprises: characteristic attributes of the lane lines;

determining a target lane line set from the assumed lane line set;

the target lane line set comprises at least one target lane line, and each target lane line is formed by one assumed lane line in the assumed lane line set or by a combination of at least two assumed lane lines.

In another aspect, an embodiment of the present invention provides a lane line detection apparatus, including:

the determining unit is used for acquiring images of lane lines and determining a local lane line set;

a constructing unit, configured to construct an assumed lane line according to prior information of the lane line and the local lane line set, to obtain an assumed lane line set, where the prior information includes: characteristic attributes of the lane lines;

the determining unit is further configured to determine a target lane line set from the assumed lane line set;

In another aspect, an embodiment of the present invention provides a lane line detection apparatus, where the lane line detection apparatus is built in a mobile platform, and the lane line detection apparatus includes: a memory and a processor;

the memory is used for storing program codes;

the processor, invoking the program code, when executed, is configured to:

acquiring an image of a lane line, and determining a local lane line set;

determining a target lane line set from the assumed lane line set;

In another aspect, an embodiment of the present invention provides a mobile platform, including:

the power system is used for providing power for the mobile platform;

and a lane line detection apparatus as described in the third aspect.

In the embodiment of the invention, the mobile platform takes the local lane line obtained by fitting the pixel points in the lane line overhead view as the minimum unit to construct the assumed lane line, and completes the detection process of the lane line based on the constructed assumed lane line, since the partial lane lines used in constructing the assumed lane lines are small, the construction speed of the mobile platform can be increased, and the accuracy of the assumed lane line constructed based on the local lane lines is higher, so when the target lane line set is determined from the constructed assumed lane lines based on the preset multi-hypothesis solution rule, the accuracy of the determined set of target lane lines is also improved, thereby reducing errors caused by lane line detection, and, in addition, when the target lane line is determined from the constructed assumed lane lines, the method can be carried out based on a multi-hypothesis model set by a preset multi-hypothesis solution rule, and the problem that the lane line in the overhead view of the lane line is difficult to obtain global optimization is solved.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.

Fig. 1 is a schematic flow chart of a lane line detection method according to an embodiment of the present invention;

fig. 2 is a view of a lane line detection scenario provided in an embodiment of the present invention;

fig. 3 is a schematic flow chart of a lane line detection method according to another embodiment of the present invention;

fig. 4 is a schematic flow chart of a lane line detection method according to another embodiment of the present invention;

FIG. 5 is a schematic diagram of a connected domain tag provided by an embodiment of the present invention;

fig. 6 is a schematic flow chart of a lane line detection method according to another embodiment of the present invention;

FIG. 7a is a schematic diagram of a hypothetical lane line provided in an embodiment of the present invention;

fig. 7b is a schematic diagram of an embodiment of the present invention, which is used to split the assumed lane line shown in fig. 7 a;

fig. 7c is a schematic diagram illustrating a corresponding connectivity graph established for the assumed lane line split as shown in fig. 7b according to an embodiment of the present invention;

FIG. 8 is a schematic illustration of a connectivity graph provided by an embodiment of the present invention;

fig. 9 is a schematic block diagram of a lane line detection apparatus according to an embodiment of the present invention;

fig. 10 is a schematic block diagram of a lane line detection apparatus according to an embodiment of the present invention.

Detailed Description

In order to reduce errors in lane line detection and improve lane line detection accuracy, the embodiment of the invention provides a lane line detection method based on multiple hypotheses. Specifically, the mobile platform can perform lane line fitting according to pixel points in a small range in the image to obtain a local lane line in each small range image area, so as to form a local lane line set. Further, based on the characteristic information for the lane lines, the obtained local lane lines may be configured as one or more assumed lane lines, and then a target lane line is screened from the assumed lane lines, and the screened target lane line may be considered as a lane line that best matches the actual lane line. Compared with the currently adopted method for obtaining the complete lane line by directly fitting all pixel points in the image, the method for constructing the whole lane line starting from the local lane line can improve the fitting precision of the lane line to a certain extent, improve the accuracy of lane line identification and save the calculation speed of the mobile platform during lane line fitting to a certain extent.

In one embodiment, referring to the schematic flowchart of the lane line detection method shown in fig. 1, the mobile platform may determine a lane line overhead view, so as to determine a local lane line set for the lane line overhead view on the lane line overhead view, and use each local lane line in the local lane line set as a minimum unit for constructing a hypothetical lane line to increase the calculation speed of the mobile platform processor. Further, a set of assumed lane lines may be constructed based on the geometric relationship between the local lane lines of the minimum unit, and a set of target lane lines may be determined from the set of assumed lane lines to complete the detection of the lane lines.

When determining the target lane line set from the assumed lane line set, the mobile platform may first determine an assumed score corresponding to each assumed lane line in the assumed lane line set, where the assumed score is used to indicate a probability that the corresponding assumed lane line is an actual lane line, the assumed score is determined according to a weight value of a local lane line used to construct the assumed lane line, and after determining the assumed score corresponding to each assumed lane line, the mobile platform may determine the target lane line set from the assumed lane line set based on the assumed score. Wherein, the assumed lane line satisfying the preset score threshold value in the assumed scores may be determined as the target lane line, thereby forming the target lane line set.

In one embodiment, the lane line detection method may be applied to the lane line detection scenario shown in fig. 2, in which an overhead view of a lane line acquired by a mobile platform (e.g., an unmanned vehicle shown in fig. 2) is generated based on images of the lane line acquired at a current time and a plurality of historical times by one or more visual sensors (e.g., a camera sensor disposed in front of the mobile platform in fig. 2) disposed on the mobile platform. The visual sensors may be disposed in front of, behind, and/or on a roof of the mobile platform, and one or more visual sensors disposed in the mobile platform may be disposed in the same position or in different positions, which is not limited in the embodiments of the present invention.

Typically, a vision sensor of the mobile platform is mounted in front of the mobile platform for acquiring images of the front. After the image in front, or called front view, is obtained, projection, segmentation and fusion processing can be performed on pixel points in the front view based on the front view, so that detection of the lane line is completed. The projection is an aerial view of a local running vehicle transformed by geometric calculation according to the position relation between the vision sensor and the mobile platform and the position of the assumed ground based on the internal reference and the external reference of the vision sensor; the segmentation means clustering the pixel points of the lane lines on the front view or the overhead view and endowing each lane line pixel point with a label corresponding to the lane line; the fusion is to perform data association on a plurality of frames of overhead views with labels, merge lane lines of the plurality of frames in a time sequence, and optimize curve parameters of the lane lines so as to complete detection of the lane lines around the mobile platform.

The lane line overhead view generated based on the images of the lane lines acquired at multiple current and historical moments increases the visible range around the mobile platform due to the reference of the images of the lane lines at multiple moments, and the lane line overhead view generated by the images of the lane lines including a large curvature curve and converging lane lines such as lanes and the like may exist, so that if a traditional lane line detection strategy is still adopted, some irregular lane lines cannot be correctly detected to determine a target lane line set, and lane line detection is more prone to errors. Such as the case where the lane lines at the branch road junction in fig. 2 overlap and are not in the same direction, such a lane line scene cannot be correctly detected if a conventional lane line detection strategy is used, for example, to detect with the parallelism of the lane lines.

Therefore, in order to better detect some irregular lane line problems, a more general lane line detection method is provided, which can generate a local lane line set from a lane line overhead view, and construct a plurality of assumed lane lines according to the local lane line set as pending selections of a target lane line; and then, a target lane line set is determined from the assumed lane line set, so that the method can be better suitable for various lane line scenes, and the universality is improved.

Fig. 3 is a schematic flow chart of a lane line detection method according to an embodiment of the present invention, where the lane line detection method may be applied to various mobile platforms such as the automobiles shown in fig. 2, and in some embodiments, may be specifically executed by a mobile controller in the mobile platform, where the method includes the following steps:

s301, acquiring an image of the lane line, and determining a local lane line set. In one embodiment, the mobile platform may determine the set of local lane lines from a lane line overhead view generated from images of lane lines taken at current and/or historical times by one or more visual sensors on the mobile platform. Specifically, the mobile platform may first obtain a front view of an environment where the mobile platform is located based on a vision sensor disposed in front of the mobile platform, and further, may first perform a preliminary detection on a lane line with respect to the front view, so that after the front view is converted into a local overhead view of the mobile platform, a lane line region in the converted overhead view may be preliminarily determined based on a result of the preliminary detection on the lane line in the front view, and in one embodiment, the front view may be subjected to the preliminary lane line detection using a Convolutional Neural Network (CNN).

After the lane line area included in the lane line overhead view is preliminarily determined, a local lane line set corresponding to the lane line overhead view can be determined based on the flow diagram shown in fig. 4. Specifically, the mobile platform can extract the connected domain according to the lane line overhead view, so that the connected domain labels of the pixels in the small range can be determined according to the connected domain detection result, lane line fitting can be performed based on the connected domain labels corresponding to the pixels, one or more local lane lines corresponding to the pixels in the small range are determined, and by analogy, all local lane lines in the lane line overhead view can be determined, post-processing is performed on all the obtained local lane lines, and a local lane line set corresponding to the lane line overhead view can be obtained.

In one embodiment, the connected domain labels of the pixels in the lane line overhead view are added to the pixels by the mobile platform in advance, the mobile platform may add the same connected domain labels to the pixels belonging to the same lane line in the lane line overhead view in advance, as shown in fig. 5, the mobile platform is a schematic diagram of the connected domain labels added to some pixels in the lane line overhead view, as shown in the figure, each square represents a pixel, and the number in the square represents the connected domain label corresponding to the pixel, wherein the pixel with the connected domain label of 0 may be a pixel not belonging to the lane line in the lane line overhead view, or may be a pixel corresponding to the lane line a in the lane line overhead view; the pixel points with the connected domain labels of 1 are the pixel points belonging to the lane line B in the lane line overhead view; and the pixel point with the communicated domain label of 2 is the pixel point corresponding to the lane line which is different from the lane line A and the lane line B in the lane line overhead view, wherein the lane line A and the lane line B are two different lane lines in the lane line overhead view.

After the connected domain labels are added to the pixel points in the lane line overhead view, connected domain detection (namely connected domain extraction) can be carried out to determine the connected domain labels corresponding to the pixel points. When the mobile platform extracts the connected domain in the lane line overhead view, small-range pixel points can be selected from the lane line overhead view according to an image detection window with a preset size, and then the connected domain label of the pixel point in the image detection window can be determined in advance. Since the pixel points in the same connected domain correspond to the same connected domain label, based on the characteristic that whether the connected domain labels are the same or not, each connected domain of the image detection window can be determined, that is, the pixel points included in each connected domain are determined, wherein the preset size of the image detection window may be, for example, 3 pixels by 3 pixels or 5 pixels by 5 pixels.

After determining the pixel points included in each connected domain, fitting the pixel points corresponding to each connected domain according to an optimal solution algorithm to obtain the local lane lines corresponding to each connected domain, and so on, determining the local lane lines corresponding to the pixel points included when the image detection window is located at different positions in the lane line overhead view, thereby determining all the local lane lines of the lane line overhead view.

For example, if the connected domain label corresponding to the pixel point in the image detection window selected from the lane line overhead view is shown in fig. 4, a local lane line can be fitted to all the pixel points with the connected domain label of 1 according to the optimal solution, a local lane line can be fitted to all the pixel points with the connected domain label of 2 according to the optimal solution, and by analogy, all the pixel points in the image detection window can be fitted into a plurality of local lane lines, so that all the local lane lines for the lane line overhead view can be determined.

After all the local lane lines of the lane line overhead view are determined, post-processing can be performed based on preset prior information of the local lane lines, so that all the local lane lines of the lane line overhead view are filtered, and wrong lane lines in all the local lane lines are filtered. The preset prior information for the local lane line is a preset range or a preset value set based on the national standard of the lane line, and the prior information specifically includes length information, width information and the like, and assuming that the length of the lane line of the national standard is 1.5 meters, the length information included in the prior information for the local lane line can be set to be a partial length smaller than the length of the standard lane line, for example, a range of 10 centimeters to 15 centimeters and the like; assuming that the lane line width of the national standard is 15 cm, the width information included in the prior information may be set to a range of 13 cm to 17 cm, or the like. Correspondingly, the wrong lane line is a lane line which does not meet the preset prior information.

After the mobile platform filters the wrong lane line in all the local lane lines corresponding to the lane line overhead view, the local lane line set aiming at the lane line overhead view can be determined, the processing of the wrong lane line in the subsequent steps is avoided, the processing resources can be liberated, and the calculation speed is improved.

S302, constructing a hypothetical lane line according to the prior information of the lane line and the local lane line set, and obtaining a hypothetical lane line set. In one embodiment, the a priori information for the lane lines includes: characteristic attributes of the lane lines, the characteristic attributes comprising: the lane lines have geometric features and color features, and the geometric features include any one or more of length features, width features, and parallel features between lane lines.

In one embodiment, whether each local lane line in the local lane line set belongs to the same lane line may be determined based on prior information for the lane lines, so that the local lane lines belonging to the same lane line may be divided into a local lane line subset, and the local lane lines in the local lane line subset are connected, so that one or more assumed lane lines constructed based on the local lane line set may be determined, where any local lane line subset may correspond to one or more assumed lane lines.

When determining whether the local lane lines in the local lane line set belong to the same lane line, the determination may be made based on a direction angle and/or an euler distance between any two of the local lane lines, and when the direction angle between any two of the local lane lines is less than or equal to a preset direction angle threshold, and/or when the euler distance between any two of the local lane lines is less than or equal to a preset distance threshold, it may be determined that any two of the local lane lines belong to the same lane line, that is, belong to the same hypothetical lane line.

In one embodiment, the mobile platform may determine a local lane line C from the local lane line set, and use the local lane line C as an element in the local lane line subset H, where the other local lane lines except C in the local lane line set may be elements in the remaining local lane line set R, and after determining the local lane line subset H and the remaining local lane line set R, because the local lane line subset H includes only one local lane line C, a local lane line with a minimum euclidean distance from the local lane line C may be determined from the remaining local lane line set R, and assuming that the local lane line C and the local lane line D form a local lane line group.

Further, the local lane line subset H and the remaining local lane line set R are updated, the local lane line subset H at this time includes local lane lines C and D, and the remaining local lane lines in the remaining local lane line set R are the remaining local lane lines except for the local lane lines C and D, at this time, the local lane lines in the updated remaining local lane line set R may be compared with the local lane lines D, and the local lane line E with the shortest euclidean distance to the local lane line D is determined, so that the E may be added to the local lane line subset H, where the local lane line D and the local lane line E also form a local lane line group. By analogy, all local lane line groups in the local lane line set can be determined, and the local lane line subsets belonging to the same hypothetical lane line can be determined.

In another embodiment, after determining the local lane line subset H and the remaining local lane line set R, the mobile platform may further determine a local lane line with the smallest direction difference with the local lane line C from the remaining local lane line set R, and if the local lane line is the local lane line a, the local lane line a and the local lane line C form a local lane line group, and so on, may determine all the local lane line groups from the local lane line set, and determine the local lane line subset belonging to the same assumed lane line.

In another embodiment, after determining the local lane line subset H and the remaining local lane line set R, the mobile platform may further determine a local lane line with the smallest direction difference with the local lane line C from the remaining local lane line set R, and if the local lane line is the local lane line B, the local lane line B and the local lane line C form a local lane line group.

In one embodiment, the determined local lane lines in the local lane line group are connected according to the sequence of adding to the local lane line subset, and a corresponding assumed lane line can be constructed.

S303, determining a target lane line set from the assumed lane line set. In one embodiment, the set of target lane lines includes at least one target lane line, and each target lane line is formed by one hypothetical lane line in the set of hypothetical lane lines, or by a combination of at least two hypothetical lane lines.

When the target lane line set is determined from the assumed lane line set, a connected graph may be established based on the assumed lane lines constructed in step S302, specifically, each assumed lane line in the assumed lane line set may be used as a vertex of the connected graph to be established, and if the assumed lane lines corresponding to any two vertices do not include the same local lane line, it is described that the two assumed lane lines are compatible, and then the compatible vertices may be connected by one edge, so as to establish the connected graph corresponding to the assumed lane lines.

In one embodiment, if the assumed lane lines corresponding to any two vertices include the same local lane line, it is necessary to split the one group of assumed lane lines including the same local lane line to obtain two groups of assumed lane lines, where the two obtained groups of assumed lane lines do not include the same local lane line, and the two obtained groups of assumed lane lines after splitting are respectively used as vertices of the connected graph to be established, so as to ensure that the assumed lane lines (or the assumed lane line groups) corresponding to any two vertices in the established connected graph do not include the same local lane line.

After the corresponding connected graph is established based on the assumed lane lines, the assumed score corresponding to each assumed lane line can be obtained and is used as the weight value of the top point of the corresponding connected graph, wherein the higher the assumed score is, the higher the possibility that the assumed lane line corresponding to the assumed score is the actual lane line is. Further, a weighted maximum group is solved based on the connected graph, the assumed lane lines corresponding to the vertexes in the solved maximum group are mutually compatible and have the highest score, and the target lane line formed by the assumed lane lines corresponding to the vertexes of the solved maximum group is the lane line which is most consistent with the actual lane line.

In the embodiment of the invention, the mobile platform performs the assumed lane line structure by taking the local lane line obtained by fitting the pixel points in the overhead view of the lane line as the minimum unit, and completes the detection process of the lane line based on the assumed lane line of the structure. In addition, when the target lane line is determined from the constructed assumed lane lines, the determination can be carried out based on a multi-hypothesis model set by a preset multi-hypothesis solving rule, and the problem that the lane lines in the lane line overhead view are difficult to obtain global optimization is solved.

Referring to fig. 6, a schematic flowchart of a lane detection method according to another embodiment of the present invention is shown, where the method is also applicable to various mobile platforms such as automobiles shown in fig. 2, and in some embodiments, the method may be specifically executed by a mobile controller in the mobile platform, as shown in fig. 6, and the method includes the following steps.

S601, acquiring an image of the lane line, and determining a local lane line set. In one embodiment, when the mobile platform determines the local lane line set, images of the lane lines may be captured at current and historical times according to one or more visual sensors arranged on the mobile platform, so that an overhead view of the lane lines may be generated based on the images of the lane lines.

After confirming lane line overlook view, can be according to lane line overlook view confirms local lane line set, specifically, mobile platform can be right earlier lane line overlook view carries out connected domain analysis and processing, confirms the connected domain label of each pixel in lane line overlook view, wherein, the connected domain label of each pixel is that mobile platform adds in advance and saves in the lane line overlook view, and the pixel that belongs to same connected domain is recorded with the same connected domain label.

Further, a target pixel point may be selected from a connected domain of the lane line overhead view, and a related pixel point having the same connected domain label as the target pixel point is determined, in an embodiment, the mobile platform may detect a connected domain on the lane line overhead view according to an image detection window of a preset size, and select a target pixel point from a connected domain located in the image detection window on the lane line overhead view, and further, may take other pixel points located in the connected domain in the image detection window and having the same connected domain label as the target pixel point as related pixel points. The preset size includes the above-mentioned sizes of 3 pixels by 3 pixels or 5 pixels by 5 pixels.

And selecting target pixel points from the connected domain of the lane line overhead view on the mobile platform, and after determining associated pixel points with the same connected domain labels as the target pixel points, performing curve fitting processing on the target pixel points and the associated pixel points according to a preset curve fitting algorithm to obtain a local lane line set. In one embodiment, the mobile platform may fit the target pixel points and the associated pixel points based on an optimal solution algorithm to obtain an initial local lane line, so that the local lane line set may be obtained according to the initial local lane lines obtained by the plurality of image detection windows, where the optimal solution algorithm includes a least square algorithm.

When the mobile platform obtains a local lane line set according to initial local lane lines obtained by a plurality of image detection windows, the initial local lane lines obtained by the plurality of image detection windows can be filtered according to a preset filtering algorithm, wherein the preset filtering algorithm is generated according to prior information of the lane lines, and the prior information comprises: the characteristic attribute of the lane line, the attribute characteristic of the lane line includes: the lane lines have geometric features and color features, and the geometric features include any one or more of length features, width features, and parallel features between lane lines. Based on the filtering of the initial local lane lines, wrong local lane lines in the initial local lane lines can be filtered out, and a local lane line set corresponding to the lane line overlooking image is obtained.

When the target pixel points and the associated pixel points are fitted based on an optimal solution algorithm to obtain initial local lane lines, fitting information when the target pixel points and the associated pixel points are fitted can be determined, the fitting information is used for describing aggregation degree or dispersion degree between the target pixel points and the associated pixel points, so that fitting weight corresponding to the initial local lane lines can be determined based on the fitting information, the fitting weight is used for representing the possibility that the fitted initial local lane lines are local to the actual lane lines, and the possibility that the initial local lane lines are local to the actual lane lines is represented by a value with the larger fitting weight. Further, the fitting weight corresponding to each local lane line in the local lane line set after the filtering operation can be determined.

S602, according to a preset judgment condition for judging whether the local lane lines belong to the same lane line, determining a local lane line subset from the local lane line set. In one embodiment, the preset determination condition for determining whether the local lane lines belong to the same lane line includes: a condition set based on a direction difference and/or a euclidean distance between the local lane lines; the local lane line subset comprises a local lane line group; the direction difference between at least two local lane lines forming the local lane line group is smaller than or equal to a preset direction difference threshold value; and/or the Euclidean distance between at least two local lane lines forming the local lane line group is smaller than or equal to a preset distance threshold value.

In one embodiment, at least two local lane lines of the local lane line group are adjacent local lane lines that constitute the same lane line, e.g., the local lane line subset includes local lane line groups A, B and C, where local lane line group a includes local lane lines a, B, and C, local lane line group B includes local lane lines d, e, and local lane line group C includes local lane lines f, g, h, and i. If the local lane lines a, b, c, d, e, f, g, h and i belong to the same lane line, the direction difference between a and b in the local lane line group A is necessarily less than or equal to a preset direction difference threshold value; and/or the Euclidean distance between a and b is smaller than or equal to a preset distance threshold; the direction difference between b and c is less than or equal to a preset direction difference threshold value; and/or the Euclidean distance between b and c is smaller than or equal to a preset distance threshold value. The direction difference between d and e is necessarily smaller than or equal to a preset direction difference threshold value in the local lane line group B; and/or the Euclidean distance between d and e is smaller than or equal to a preset distance threshold value. And the direction difference between c and d is necessarily smaller than or equal to a preset direction difference threshold value in the local lane line groups A and B; and/or the Euclidean distance between c and d is smaller than or equal to a preset distance threshold value.

After determining the partial lane line subset from the partial lane lines, step S603 may be executed to construct a corresponding assumed lane line according to each partial lane line in the partial lane line subset.

S603, constructing a hypothetical lane line corresponding to each local lane line subset according to the local lane lines in each local lane line subset to obtain a hypothetical lane line set.

In one embodiment, the mobile platform may first construct assumed lane lines corresponding to the respective subsets according to local lane lines in the local lane line subsets, and split the assumed lane line groups satisfying the intersection condition if the constructed assumed lane lines include assumed lane line groups satisfying the intersection condition, so as to obtain two groups of assumed lane lines, where each group of the split assumed lane lines does not include the same local lane line.

After a plurality of assumed lane lines are obtained through construction, when a target lane line set is determined from the assumed lane line set based on a preset multi-assumption solving rule, a connected graph with mutually compatible vertexes needs to be constructed, and the vertexes in the connected graph correspond to the assumed lane lines in the assumed lane line set, so that the assumed lane lines corresponding to the vertexes in the established connected graph need to be ensured to be mutually compatible to ensure that the established connected graph has an optimal solution, and therefore, the assumed lane lines corresponding to the vertexes in the established connected graph only need to be ensured not to include the same local lane lines. In one embodiment, whether the constructed hypothetical lane lines include the same local lane line is determined by whether the intersection condition is satisfied, and the same local lane line exists in the two hypothetical lane lines satisfying the intersection condition.

In one embodiment, if the constructed hypothetical lane line includes lane lines corresponding to hypothesis 1 and hypothesis 2 in fig. 7a according to the local lane lines in the local lane line subset, it may be determined that there is a hypothetical lane line group satisfying the intersection condition in the constructed hypothetical lane line, i.e., a hypothetical lane line group consisting of the lane line corresponding to hypothesis 1 and the lane line corresponding to hypothesis 2. In order to make the constructed assumed lane lines not include the same local lane line, therefore, the assumed lane line group composed of assumption 1 and assumption 2 may be split, and the splitting process may be as shown in fig. 7b, an overlapping portion of the assumed lane line corresponding to assumption 1 and the assumed lane line corresponding to assumption 2 in the assumed lane line group, that is, a portion indicated by a dashed line in the figure, may be determined; further, the assumed lane line group may be split based on the overlapping portion and the non-overlapping portion, so as to obtain two assumed lane lines, where the obtained two assumed lane lines may be, as shown in the figure, respectively, one assumed lane line group composed of assumed lane lines corresponding to assumption 1 and assumption 3, and one assumed lane line group composed of assumed lane lines corresponding to assumption 2 and assumption 3.

After splitting the set of hypothetical lane lines that satisfy the intersection condition, a connectivity graph whose vertices are mutually exclusive may be established, and based on the 4 hypothetical lane lines obtained after the splitting in fig. 7b, the established connectivity graph may be as shown in fig. 7c, where a vertex labeled 1 in the connectivity graph corresponds to the hypothetical lane line of hypothesis 1 after the splitting in fig. 7b, a vertex labeled 2 corresponds to the hypothetical lane line of hypothesis 2 after the splitting in fig. 7b, a vertex labeled 3 corresponds to the hypothetical lane line of hypothesis 3 after the splitting in fig. 7b, a vertex labeled 4 corresponds to the hypothetical lane line of hypothesis 4 after the splitting in fig. 7b, and since the same local lane line is not included between the lane lines of hypothesis 1 and hypothesis 3, vertex 1 and vertex 3 are connected by one line in fig. 7c for indicating that hypothesis 1 and hypothesis 3 are compatible, similarly, since hypothesis 2 and hypothesis 4 are compatible, vertex 2 and vertex 4 are connected by a line.

After the connected graph corresponding to the assumed lane line set is established, a target lane line set may be determined from the assumed lane line set based on a preset multi-hypothesis solution rule and the connected graph, that is, step S604 is performed instead.

S604, determining a target lane line set from the assumed lane line set based on a preset multi-hypothesis solution rule. In one embodiment, the process of solving based on multiple hypotheses is a process of obtaining mutually compatible optimal combinations of hypotheses. Of course, in other embodiments, the final lane line may be determined by performing image recognition (or performing connected component detection on the image) and multi-hypothesis solution directly on the basis of one or more captured images based on the captured image of the driving environment.

In one embodiment, the multi-hypothesis-based solution rule is to process the data source (i.e., the set of local lane lines) by using a multi-hypothesis model, that is, if there is no reasonable flow to obtain an optimal solution, list all solutions that may be generated by the current data source, determine the set of local lane lines corresponding to the overhead view of the lane lines, and use all listed assumed lane lines as hypotheses, and further score all assumed lane lines through prior information until finding out the better-scored assumed lane lines, that is, the better-scored lane lines are finally used to make the local map of lane lines or lane lines for other purposes. The found lane line with better score may be formed by combining one or more hypotheses, and when the lane line is formed by combining multiple hypotheses, mutual exclusivity may exist between different hypotheses, that is, when a is true, B is false. Therefore, when solving a plurality of hypotheses, not only the scores of the hypotheses need to be considered, but also the hypotheses in the optimal solution need to be ensured not to be mutually exclusive, that is, each lane line in the target lane line set obtained by the solution is not mutually exclusive.

In order to ensure that all target lane lines in the target lane line set determined from the assumed lane line set based on the multi-hypothesis solution rule are not mutually exclusive, a connected graph can be established in pairs according to the assumed lane lines in the assumed lane line set. Each vertex in the connected graph corresponds to one hypothetical lane line in the hypothetical lane line set, the vertex further takes the score corresponding to each hypothetical lane line as the weight corresponding to the vertex, if two hypotheses do not repel each other (i.e., are compatible and can exist at the same time), one edge is used for connecting the corresponding vertices, so that a solution with better score and hypothesis compatibility can be obtained by calculating the weighted maximum group to form the target lane line set. Wherein, the maximum weighted group means: in the connected graph, edges exist on all vertexes, and are connected pairwise (the fact that the connected edges exist between the vertexes means that assumed lane lines corresponding to the connected vertexes are not mutually exclusive), and the group with the largest weight sum of all vertexes is formed. The two hypothetical lane lines may be simultaneously present, for example, hypothetical lane lines corresponding to two lane lines at a turnout in an actual environment.

When establishing a connected graph corresponding to the assumed lane line set, if one or more assumed lane lines constructed by the same local lane lines exist, the one or more assumed lane lines need to be split, the maximum clique can be obtained by solving the connected graph established based on the assumed lane line set, and the problem that the global optimal solution is difficult to obtain when the lane lines are divided to overlook the aerial view can be solved.

If the assumed lane line set includes 5 assumed lane lines a, b, c, d, e, when establishing a corresponding connected graph based on the assumed lane line set, vertices in the connected graph may be determined first, where each vertex in the connected graph corresponds to each assumed lane line in the assumed lane line set, for example, vertex 1 corresponds to assumed lane line a, vertex 2 corresponds to assumed lane line b, vertex 3 corresponds to assumed lane line c, vertex 4 corresponds to assumed lane line d, and vertex 5 corresponds to assumed lane line e, and after determining the vertices in the connected graph, the connection relationship between the vertices may be determined according to whether there is an overlapping portion (i.e., whether the same local lane line is included) in each assumed lane line, so that the connected graph shown in fig. 8 may be established based on the assumed lane line set.

In one embodiment, after determining a connected graph corresponding to the set of assumed lane lines, solving a maximum clique based on the connected graph, where in the connected graph as shown in the figure, the vertex set V is {1, 2, 3, 4, 5}, the edge set E is { (1, 2) (1, 4) (1, 5) (2, 3) (2, 5) (3, 5) (4, 5) }, and the maximum clique obtained by the solution may be {1, 2, 5}, also may be {1, 4, 5} or {2, 3, 5 }. Based on the respective assumed weight scores corresponding to the vertexes, the weight value corresponding to each maximum group can be calculated, the weighted maximum group corresponding to the connected graph can be determined by comparing the weight values of the maximum groups, and the weight value corresponding to the assumed maximum group {1, 2 and 5} is 25; the weight value corresponding to the maximum cluster {1, 4, 5} is 18; and if the weight value corresponding to the maximum clique {2, 3 and 5} is 7, the determined weighted maximum clique is {1, 2 and 5 }.

According to the determined weighted maximum clique, a target lane line set consisting of assumed lane lines can be determined, in one embodiment, the assumed lane lines corresponding to the vertexes in the weighted maximum clique can be taken as elements forming the target lane line set, that is, if the determined weighted maximum clique is {1, 2, 5}, the target lane line set comprises a lane line a, a lane line b and a lane line e.

In an embodiment, the step S604 specifically includes steps S11 and S12, where, first in S11, the mobile platform divides the set of assumed lane lines into at least two assumed lane line groups based on an indication of a preset multi-hypothesis solution rule, and determines an assumed weight score corresponding to each assumed lane line group. In one embodiment, in order to avoid the problem that the connection graph established based on the assumed lane lines does not have solution due to the existence of the same local lane lines between the assumed lane lines included in the assumed lane line set, the assumed lane line set may be divided according to the above steps to obtain at least two assumed lane line groups, where the assumed lane lines included in each lane line group do not include the same local lane line; further, based on the split obtained assumed lane line groups, an assumed weight score corresponding to an assumed lane line in each assumed lane line group may be determined.

In an embodiment, the assumed weight score corresponding to the assumed lane line is determined by the fitting weight corresponding to one or more local lane lines constituting the assumed lane line, and the determination manner of the fitting weight corresponding to the local lane line may specifically refer to the related description in the step S601, which is not described herein again.

After the hypothetical weight scores corresponding to the hypothetical lane line groups are determined, step S605 may be executed to determine a target lane line set from the hypothetical lane line sets.

In s12, the mobile platform composes a target lane line set from the assumed lane lines in the assumed lane line group corresponding to the maximum assumed weight score. In one embodiment, the hypothesized weight score is determined from weight values of hypothesized lane lines in the hypothesized set of lane lines, the weight values of the hypothesized lane lines being determined from the a priori information for the lane lines.

After the hypothesis weight scores corresponding to the hypothesis lane line groups are determined, a weighted maximum group problem can be solved based on the established connected graph, wherein the weight corresponding to each vertex of the connected graph is the hypothesis weight score corresponding to the hypothesis lane line corresponding to the vertex, so that the weighted maximum group of the connected graph can be solved, that is, the hypothesis lane lines in the hypothesis lane line group corresponding to the maximum hypothesis weight score are solved to form a target lane line set.

For example, for two hypothetical lane line groups as shown in fig. 7b, hypothetical weight scores for hypothetical lane lines corresponding to hypothetical 1 can be, for example, 15, hypothetical weight scores corresponding to hypothetical 2 can be, for example, 14, hypothetical weight scores corresponding to hypothetical 3 can be, for example, 10, hypothetical weight scores corresponding to hypothetical 4 can be, for example, 8, the weight corresponding to vertex 1 in the connected graph of fig. 7c is 15, the weight corresponding to vertex 2 is 14, the weight corresponding to vertex 3 is 10, the weight corresponding to vertex 4 is 8, the maximum clique solved based on the connected graph includes (1, 3) and (2, 4), the weights are 25 and 22, respectively, therefore, the weighted maximum cluster of the connected graph is (1, 3), and thus the solved target lane line set includes the assumed lane line corresponding to the assumption 1 and the assumed lane line corresponding to the assumption 3.

In the embodiment of the invention, after determining the local lane line set aiming at the lane line overhead view, the mobile platform determines the local lane line subset from the local lane line set according to a preset judgment condition for judging whether the local lane line belongs to the same lane line, and constructs the assumed lane line corresponding to each local lane line subset based on the local lane line group included in the local lane line subset, so as to obtain the assumed lane line set. Further, the assumed lane line set can be divided based on the preset indication of the multi-hypothesis solution rule to obtain two assumed lane line sets, so as to ensure that each assumed lane line in the assumed lane line set does not include the same local lane line.

An embodiment of the present invention provides a lane line detection apparatus, where the lane line detection apparatus is used to execute a unit of any one of the foregoing methods, and specifically, refer to fig. 9, which is a schematic block diagram of a lane line detection apparatus provided in an embodiment of the present invention, and the lane line detection apparatus of the present embodiment may be disposed in a mobile platform of an automatic driving automobile or the like, for example, and the lane line detection apparatus includes: a determination unit 901 and a construction unit 902.

The determining unit 901 is configured to acquire an image of a lane line and determine a local lane line set; a constructing unit 902, configured to construct an assumed lane line according to prior information of the lane line and the local lane line set, to obtain an assumed lane line set, where the prior information includes: characteristic attributes of the lane lines; the determining unit 901 is further configured to determine a target lane line set from the assumed lane line set based on a preset multi-hypothesis solution rule. The target lane line set comprises at least one target lane line, and each target lane line is formed by one assumed lane line in the assumed lane line set or by a combination of at least two assumed lane lines.

In an embodiment, the determining unit 901 is specifically configured to: shooting at current and historical moments according to one or more visual sensors arranged on a mobile platform to obtain images of the lane line; generating a lane line aerial view according to the image of the lane line; and determining a local lane line set according to the lane line aerial view.

In an embodiment, the determining unit 901 is further specifically configured to, according to images of the lane lines at least two capturing angles captured by one or more visual sensors disposed on the mobile platform at current and historical times: and generating a lane line aerial view based on the images of the lane line under the at least two shooting angles.

In an embodiment, the determining unit 901 is further specifically configured to: analyzing and processing a connected domain of the lane line aerial view, and determining a connected domain label of each pixel point in the lane line aerial view, wherein the pixel points belonging to the same connected domain are recorded with the same connected domain label; selecting target pixel points from the connected domain of the lane line aerial view, and determining associated pixel points with the same connected domain labels as the target pixel points; and performing curve fitting processing on the target pixel points and the associated pixel points according to a preset curve fitting algorithm to obtain a local lane line set.

In an embodiment, the determining unit 901 is further specifically configured to: detecting a connected domain on the lane line aerial view according to an image detection window with a preset size; selecting target pixel points from a connected domain positioned in the image detection window on the lane line aerial view; and taking other pixel points which are positioned in the connected domain in the image detection window and have the same connected domain label with the target pixel point as associated pixel points.

In an embodiment, the determining unit 901 is further specifically configured to: fitting the target pixel points and the associated pixel points based on an optimal solution algorithm to obtain initial local lane lines; and obtaining a local lane line set according to the initial local lane lines obtained by the plurality of image detection windows.

In one embodiment, the optimal solution comprises a least squares algorithm.

In an embodiment, the determining unit 901 is further specifically configured to: filtering each initial local lane line in the initial local lane lines obtained by the plurality of image detection windows according to a preset filtering algorithm; filtering error local lane lines in the initial local lane lines to obtain a local lane line set corresponding to the lane line overlook image; the preset filtering algorithm is generated according to prior information for the lane lines.

In one embodiment, the constructing unit 902 is specifically configured to: determining a local lane line subset from the local lane line set according to a preset judgment condition for judging whether the local lane lines belong to the same lane line; and constructing a hypothetical lane line corresponding to each local lane line subset according to the local lane lines in each local lane line subset to obtain a hypothetical lane line set.

In one embodiment, the preset determination condition for determining whether the local lane lines belong to the same lane line includes: conditions set based on a difference in direction between the local lane lines and/or a euclidean distance.

In one embodiment, the subset of local lane lines includes a set of local lane lines; the direction difference between at least two local lane lines forming the local lane line group is smaller than or equal to a preset direction difference threshold value; and/or the Euclidean distance between at least two local lane lines forming the local lane line group is smaller than or equal to a preset distance threshold value.

In an embodiment, the constructing unit 902 is further specifically configured to: constructing a hypothetical lane line corresponding to each local lane line subset according to the local lane lines in each local lane line subset; if the constructed assumed lane line group meeting the intersection condition exists in the constructed assumed lane lines, splitting the assumed lane line group meeting the intersection condition to obtain two groups of assumed lane lines; wherein, each split group of assumed lane lines does not include the same local lane line.

In an embodiment, the constructing unit 902 is further specifically configured to: dividing the assumed lane line set to obtain at least two assumed lane line groups based on the indication of a preset multi-hypothesis solution rule, and determining hypothesis weight scores corresponding to the assumed lane line groups; forming a target lane line set by the assumed lane lines in the assumed lane line group corresponding to the maximum assumed weight score; wherein the hypothetical weight score is determined from weight values of hypothetical lane lines in the set of hypothetical lane lines, the weight values of the hypothetical lane lines being determined from the a priori information for the lane lines.

In one embodiment, the attribute features of the lane lines include: the lane lines have geometric features and color features, and the geometric features include any one or more of length features, width features, and parallel features between lane lines.

In an embodiment, the lane line detection apparatus provided in this embodiment can execute the lane line detection method shown in fig. 3 and 6 provided in the foregoing embodiment, and the execution manner and the beneficial effects are similar, and are not described herein again.

An embodiment of the present invention provides a lane line detection apparatus, which may be applied to the mobile platform mentioned in the above embodiment, where fig. 10 is a structural diagram of the lane line detection apparatus provided in the embodiment of the present invention, and as shown in fig. 10, the lane line detection apparatus 100 includes a memory 101, a processor 102, and a vision sensor 103.

The processor 102 may be a Central Processing Unit (CPU). The processor 102 may be a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a Programmable Logic Device (PLD), or a combination thereof. The PLD may be a Complex Programmable Logic Device (CPLD), a field-programmable gate array (FPGA), a General Array Logic (GAL), or any combination thereof.

The memory 102 stores program code, the processor 102 calls the program code in the memory, and when the program code is executed, the processor 102 performs the following operations: acquiring an image of a lane line, and determining a local lane line set; constructing an assumed lane line according to the prior information of the lane line and the local lane line set to obtain an assumed lane line set, wherein the prior information comprises: characteristic attributes of the lane lines; determining a target lane line set from the assumed lane line set; the target lane line set comprises at least one target lane line, and each target lane line is formed by one assumed lane line in the assumed lane line set or by a combination of at least two assumed lane lines.

In one embodiment, the processor 102, when acquiring images of lane lines and determining the set of local lane lines, is configured to: shooting at the current and historical moments according to one or more visual sensors 103 arranged on the mobile platform to obtain images of the lane line; generating a lane line aerial view according to the image of the lane line; and determining a local lane line set according to the lane line aerial view.

In one embodiment, the images of the lane line at least two capturing angles are captured at the current and historical time according to one or more visual sensors disposed on the mobile platform, and the processor 102, when generating the overhead view of the lane line according to the images of the lane line, is configured to: and generating a lane line aerial view based on the images of the lane line under the at least two shooting angles.

In one embodiment, the processor 102, when determining the set of local lane lines from the lane line overhead view, is configured to: analyzing and processing a connected domain of the lane line aerial view, and determining a connected domain label of each pixel point in the lane line aerial view, wherein the pixel points belonging to the same connected domain are recorded with the same connected domain label; selecting target pixel points from the connected domain of the lane line aerial view, and determining associated pixel points with the same connected domain labels as the target pixel points; and performing curve fitting processing on the target pixel points and the associated pixel points according to a preset curve fitting algorithm to obtain a local lane line set.

In one embodiment, the processor 102, when selecting a target pixel point from the connected domain of the lane line overhead view and determining an associated pixel point having the same connected domain label as the target pixel point, is configured to: detecting a connected domain on the lane line aerial view according to an image detection window with a preset size; selecting target pixel points from a connected domain positioned in the image detection window on the lane line aerial view; and taking other pixel points which are positioned in the connected domain in the image detection window and have the same connected domain label with the target pixel point as associated pixel points.

In an embodiment, the processor 102 is configured to, when performing curve fitting processing on the target pixel and the associated pixel according to a preset curve fitting algorithm to obtain a local lane line set,: fitting the target pixel points and the associated pixel points based on an optimal solution algorithm to obtain initial local lane lines; and obtaining a local lane line set according to the initial local lane lines obtained by the plurality of image detection windows.

In one embodiment, the optimal solution comprises a least squares algorithm.

In one embodiment, the processor 102, when obtaining the set of local lane lines from the initial local lane lines obtained from the plurality of image detection windows, is configured to: filtering each initial local lane line in the initial local lane lines obtained by the plurality of image detection windows according to a preset filtering algorithm; filtering error local lane lines in the initial local lane lines to obtain a local lane line set corresponding to the lane line overlook image; the preset filtering algorithm is generated according to prior information for the lane lines.

In one embodiment, the processor 102, when constructing the assumed lane lines according to the priori information of the lane lines and the local lane line sets, and obtaining an assumed lane line set, is configured to: determining a local lane line subset from the local lane line set according to a preset judgment condition for judging whether the local lane lines belong to the same lane line; and constructing a hypothetical lane line corresponding to each local lane line subset according to the local lane lines in each local lane line subset to obtain a hypothetical lane line set.

In one embodiment, the processor 102, when constructing the assumed lane line corresponding to each of the local lane line subsets according to the local lane lines in each of the local lane line subsets to obtain the assumed lane line set, is configured to: constructing a hypothetical lane line corresponding to each local lane line subset according to the local lane lines in each local lane line subset; if the constructed assumed lane line group meeting the intersection condition exists in the constructed assumed lane lines, splitting the assumed lane line group meeting the intersection condition to obtain two groups of assumed lane lines; wherein, each split group of assumed lane lines does not include the same local lane line.

In one embodiment, the processor 102, when determining the target lane line set from the assumed lane line set based on a preset multi-hypothesis solution rule, is configured to: dividing the assumed lane line set to obtain at least two assumed lane line groups based on the indication of a preset multi-hypothesis solution rule, and determining hypothesis weight scores corresponding to the assumed lane line groups; forming a target lane line set by the assumed lane lines in the assumed lane line group corresponding to the maximum assumed weight score; wherein the hypothetical weight score is determined from weight values of hypothetical lane lines in the set of hypothetical lane lines, the weight values of the hypothetical lane lines being determined from the a priori information for the lane lines.

The mobile platform provided in this embodiment can execute the lane line detection method provided in the foregoing embodiment and shown in fig. 3 and 6, and the execution manner and the beneficial effects are similar and will not be described again here.

Embodiments of the present invention further provide a computer program product including instructions, which when run on a computer, enable the computer to perform the relevant steps of the lane line detection method described in the above method embodiments.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.

The above disclosure is intended to be illustrative of only some embodiments of the invention, and is not intended to limit the scope of the invention.

Claims

1. A lane line detection method is characterized by comprising the following steps:

acquiring an image of a lane line, and determining a local lane line set;

determining a target lane line set from the assumed lane line set;

2. The method of claim 1, wherein obtaining the image of the lane lines and determining the set of local lane lines comprises:

shooting at current and historical moments according to one or more visual sensors arranged on a mobile platform to obtain images of the lane line;

generating a lane line aerial view according to the image of the lane line;

and determining a local lane line set according to the lane line aerial view.

3. The method of claim 2, wherein the capturing of the images of the lane lines at least two capturing angles from the one or more vision sensors disposed on the mobile platform at the current and historical times comprises generating a line overhead view from the images of the lane lines, comprising:

and generating a lane line aerial view based on the images of the lane line under the at least two shooting angles.

4. The method of claim 2, wherein determining the set of local lane lines from the lane line overhead view comprises:

analyzing and processing a connected domain of the lane line aerial view, and determining a connected domain label of each pixel point in the lane line aerial view, wherein the pixel points belonging to the same connected domain are recorded with the same connected domain label;

selecting target pixel points from the connected domain of the lane line aerial view, and determining associated pixel points with the same connected domain labels as the target pixel points;

and performing curve fitting processing on the target pixel points and the associated pixel points according to a preset curve fitting algorithm to obtain a local lane line set.

5. The method of claim 4, wherein the selecting a target pixel point from the connected domain of the lane line overhead view and determining an associated pixel point having the same connected domain label as the target pixel point comprises:

detecting a connected domain on the lane line aerial view according to an image detection window with a preset size;

selecting target pixel points from a connected domain positioned in the image detection window on the lane line aerial view;

and taking other pixel points which are positioned in the connected domain in the image detection window and have the same connected domain label with the target pixel point as associated pixel points.

6. The method according to claim 5, wherein the curve-fitting the target pixel point and the associated pixel point according to a preset curve-fitting algorithm to obtain a local lane line set comprises:

fitting the target pixel points and the associated pixel points based on an optimal solution algorithm to obtain initial local lane lines;

and obtaining a local lane line set according to the initial local lane lines obtained by the plurality of image detection windows.

7. The method of claim 6, wherein the optimal solution comprises a least squares algorithm.

8. The method of claim 6, wherein obtaining a set of local lane lines from the initial local lane lines obtained from the plurality of image detection windows comprises:

filtering each initial local lane line in the initial local lane lines obtained by the plurality of image detection windows according to a preset filtering algorithm;

filtering error local lane lines in the initial local lane lines to obtain a local lane line set corresponding to the lane line overlook image;

the preset filtering algorithm is generated according to prior information for the lane lines.

9. The method of claim 8, wherein constructing the assumed lane lines according to the priori information of the lane lines and the local lane line sets to obtain an assumed lane line set comprises:

determining a local lane line subset from the local lane line set according to a preset judgment condition for judging whether the local lane lines belong to the same lane line;

and constructing a hypothetical lane line corresponding to each local lane line subset according to the local lane lines in each local lane line subset to obtain a hypothetical lane line set.

10. The method of claim 9, wherein the preset determination condition for determining whether the local lane lines belong to the same lane line comprises: conditions set based on a difference in direction between the local lane lines and/or a euclidean distance.

11. The method of claim 9, wherein the subset of local lane lines includes a set of local lane lines;

the direction difference between at least two local lane lines forming the local lane line group is smaller than or equal to a preset direction difference threshold value;

and/or the Euclidean distance between at least two local lane lines forming the local lane line group is smaller than or equal to a preset distance threshold value.

12. The method of claim 9, wherein constructing the hypothetical lane lines corresponding to each of the subsets of local lane lines according to the local lane lines in each of the subsets of local lane lines to obtain a set of hypothetical lane lines comprises:

constructing a hypothetical lane line corresponding to each local lane line subset according to the local lane lines in each local lane line subset;

if the constructed assumed lane line group meeting the intersection condition exists in the constructed assumed lane lines, splitting the assumed lane line group meeting the intersection condition to obtain two groups of assumed lane lines;

wherein, each split group of assumed lane lines does not include the same local lane line.

13. The method of claim 12, wherein determining a set of target lane lines from the set of hypothetical lane lines comprises:

and determining a target lane line set from the assumed lane line set based on a preset multi-hypothesis solution rule.

14. The method of claim 13, wherein determining a set of target lane lines from the set of assumed lane lines based on a predetermined multi-hypothesis solution rule comprises:

dividing the assumed lane line set to obtain at least two assumed lane line groups based on the indication of a preset multi-hypothesis solution rule, and determining hypothesis weight scores corresponding to the assumed lane line groups;

forming a target lane line set by the assumed lane lines in the assumed lane line group corresponding to the maximum assumed weight score;

wherein the hypothetical weight score is determined from weight values of hypothetical lane lines in the set of hypothetical lane lines, the weight values of the hypothetical lane lines being determined from the a priori information for the lane lines.

15. The method according to any one of claims 1-14, wherein the attribute characteristics of the lane lines include: the lane lines have geometric features and color features, and the geometric features include any one or more of length features, width features, and parallel features between lane lines.

16. The utility model provides a lane line check out test set, the lane line check out test set is arranged in the moving platform in, its characterized in that, lane line check out test set includes: a memory and a processor;

the memory is used for storing program codes;

the processor, invoking the program code, when executed, is configured to:

acquiring an image of a lane line, and determining a local lane line set;

determining a target lane line set from the assumed lane line set;

17. The lane line detection apparatus according to claim 16, further comprising a vision sensor, wherein the processor performs the following operations when acquiring an image of a lane line and determining a local set of lane lines:

generating a lane line aerial view according to the image of the lane line;

and determining a local lane line set according to the lane line aerial view.

18. The lane line detection apparatus according to claim 17, wherein the images of the lane line at least two capturing angles are captured at current and historical times based on one or more visual sensors provided on the mobile platform, and the processor performs the following operations when generating a line overhead view from the images of the lane line:

19. The lane line detection apparatus of claim 17, wherein the processor, when determining the set of local lane lines from the lane line overhead view, performs the following:

20. The lane line detection apparatus according to claim 19, wherein the processor, when selecting a target pixel point from the connected components of the lane line overhead view and determining an associated pixel point having the same connected component label as the target pixel point, performs:

21. The lane line detection device according to claim 20, wherein the processor performs the following operations when performing curve fitting processing on the target pixel point and the associated pixel point according to a preset curve fitting algorithm to obtain a local lane line set:

22. The lane line detection apparatus of claim 21, wherein the optimal solution comprises a least squares algorithm.

23. The lane line detecting apparatus according to claim 21, wherein the processor performs the following operation when obtaining a set of local lane lines from initial local lane lines obtained from a plurality of image detection windows:

24. The lane line detection apparatus according to claim 23, wherein the processor performs the following operations when constructing the assumed lane lines from the priori information of the lane lines and the local set of lane lines to obtain a set of assumed lane lines:

25. The lane line detection apparatus according to claim 24, wherein the preset determination condition for determining whether or not the local lane lines belong to the same lane line includes: conditions set based on a difference in direction between the local lane lines and/or a euclidean distance.

26. The lane line detection apparatus of claim 24, wherein the subset of local lane lines includes a set of local lane lines;

27. The lane line detection apparatus according to claim 24, wherein the processor, when constructing the assumed lane line corresponding to each of the partial lane line subsets from the partial lane lines in each of the partial lane line subsets to obtain the assumed lane line set, performs the following operations:

28. The lane line detection apparatus of claim 27, wherein the processor, when determining the set of target lane lines from the set of assumed lane lines, performs the following:

29. The lane line detection apparatus according to claim 28, wherein the processor, when determining a target lane line set from the assumed lane line set based on a preset multi-hypothesis solution rule, performs:

30. The lane line detection apparatus according to any one of claims 16 to 29, wherein the attribute characteristics of the lane line include: the lane lines have geometric features and color features, and the geometric features include any one or more of length features, width features, and parallel features between lane lines.

31. A mobile platform, comprising:

the power system is used for providing power for the mobile platform;

and a lane line detection apparatus according to any one of claims 16-30.

32. The mobile platform of claim 30, further comprising: a vision sensor;

the vision sensor is arranged on the mobile platform and used for shooting the images of the lane lines so as to obtain the images of the lane lines at least two shooting angles.

33. The mobile platform of claim 32, wherein the mobile platform is a vehicle.

34. A computer-readable storage medium, in which program instructions are stored, which, when run on a processor, implement the lane line detection method of any one of claims 1-15.