Mapping method and equipment in automatic parking scene and vehicle

Abstract

The embodiment of the invention provides a mapping method and device in an automatic parking scene and a vehicle, and relates to the technical field of automatic driving. The method comprises the steps that image data collected by multiple fisheye cameras in a vehicle are acquired in real time, and a first image sequence is obtained; performing grouping processing on the first image sequence according to preset image feature dimensions to obtain a second image sequence corresponding to each preset image feature dimension; performing feature extraction processing on the image data in the second image sequence according to preset image feature dimensions to obtain image feature information under each preset image feature dimension; and performing mapping processing according to the image feature information under each preset image feature dimension to obtain a parking map. The method is used for establishing a high-precision parking map, so that the accuracy of automatic parking is improved according to the established high-precision parking map, and the driving experience of a user is improved.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, and particularly to a mapping method, device, and vehicle in an automatic parking scenario. Background Art

[0002] In an automatic parking scenario, a high-precision map needs to be constructed to accurately implement the automatic parking function.

[0003] Exemplarily, for the memory parking function (also known as the parking lot memory parking function) in an automatic parking scenario, it can assist the user to automatically cruise and park in a parking space according to the memorized route, or automatically park out from the parking space and cruise to the parking out point, thereby improving the driving convenience of the user and further enhancing the user's driving experience.

[0004] For the vehicle to complete the above memory parking function, it depends on a pre-generated parking map. Therefore, how to obtain a high-precision parking map has become a technical problem that needs to be solved urgently. Summary of the Invention

[0005] Embodiments of this application provide a mapping method, device, and vehicle in an automatic parking scenario, so as to achieve the effect of establishing a high-precision parking map, thereby improving the accuracy of automatic parking according to the established high-precision parking map, and further enhancing the user's driving experience.

[0006] In a first aspect, an embodiment of this application provides a mapping method in an automatic parking scenario, including:

[0007] Real-time obtain image data respectively collected by multiple fisheye cameras in a vehicle to obtain a first image sequence;

[0008] Perform grouping processing on the first image sequence according to a preset image feature dimension to obtain a second image sequence corresponding to each preset image feature dimension;

[0009] Perform feature extraction processing on the image data in the second image sequence according to the preset image feature dimension to obtain image feature information under each preset image feature dimension;

[0010] Perform mapping processing according to the image feature information under each preset image feature dimension to obtain a parking map.

[0011] Optionally, the first image sequence includes image data collected by each fisheye camera at multiple acquisition times; the preset image feature dimensions at least include a semantic feature dimension and a corner feature dimension; performing grouping processing on the first image sequence according to a preset image feature dimension to obtain a second image sequence corresponding to each preset image feature dimension includes:

[0012] If the preset image feature dimension is the semantic feature dimension, then for the image data collected by multiple fisheye cameras at each acquisition moment included in the first image sequence, splicing processing is performed to obtain first image data; and according to the first image data, a second image sequence in the semantic feature dimension is determined.

[0013] If the preset image feature dimension is the corner feature dimension, then polling processing is performed on the image data included in the first image sequence according to the number of fisheye cameras to obtain second image data, and according to the second image data, a second image sequence in the corner feature dimension is determined.

[0014] Optionally, the preset image feature dimension is the semantic feature dimension; according to the preset image feature dimension, feature extraction processing is performed on the image data in the second image sequence to obtain image feature information in each preset image feature dimension, including:

[0015] Performing image detection processing on each first image data to obtain semantic feature information corresponding to each first image data; wherein, the semantic feature information at least indicates at least one semantic object included in the first image data.

[0016] According to the pose difference of the semantic object at different acquisition moments, determining the semantic feature information of the same semantic object at different acquisition moments, and according to the semantic feature information of each semantic object at different acquisition moments, determining the image feature information in the semantic feature dimension.

[0017] Optionally, the preset image feature dimension is the corner feature dimension; according to the preset image feature dimension, feature extraction processing is performed on the image data in the second image sequence to obtain image feature information in each preset image feature dimension, including:

[0018] Performing corner feature extraction processing on each second image data to obtain corner feature information corresponding to each second image data; wherein, the corner feature information at least indicates at least one feature point included in the second image data.

[0019] According to the similarity between feature points at different acquisition moments, determining the corner feature information of the same feature point at different acquisition moments;

[0020] Obtaining the vehicle pose at different acquisition moments;

[0021] Convert the corner feature information of the same feature point at different acquisition times into three-dimensional corner feature information according to the vehicle poses at different acquisition times and the corner feature information of the same feature point at different acquisition times.

[0022] Determine the image feature information in the corner feature dimension according to the three-dimensional corner feature information of the same feature point at different acquisition times.

[0023] Optionally, the image feature information in each preset image feature dimension includes a feature object and the object features corresponding to the feature object; perform mapping processing according to the image feature information in each preset image feature dimension to obtain a parking map, including:

[0024] Determine the target relative poses of each feature object at different acquisition times; wherein, the different acquisition times include a previous acquisition time and a subsequent acquisition time;

[0025] Determine the first pose of the feature object at the previous acquisition time, and determine the second pose of the feature object at the subsequent acquisition time according to the target relative pose and the first pose;

[0026] Perform mapping processing according to the first pose, the second pose, the feature object and the object features corresponding to the feature object to obtain the parking map.

[0027] Optionally, determining the target relative poses of each feature object at different acquisition times includes:

[0028] Obtain the third pose and the fourth pose corresponding to the feature object; wherein, the third pose is the pose obtained by performing dead reckoning on the feature object at different acquisition times; the fourth pose is the pose of the feature object in the vehicle body coordinate system at different acquisition times;

[0029] Determine the initial relative pose of the feature object at different acquisition times according to the third pose at the previous acquisition time and the third pose at the subsequent acquisition time;

[0030] Project the fourth pose of the feature object at the subsequent acquisition time to the previous acquisition time according to the initial relative pose to obtain the fifth pose corresponding to the feature object;

[0031] After determining the first error between the fourth pose and the fifth pose of the feature object at all previous acquisition times, optimize the initial relative pose according to the first error to obtain the target relative pose.

[0032] Optionally, the number of the preset image feature dimensions is multiple; determining a first error between a fourth pose and the fifth pose of the feature object at all previous acquisition moments includes:

[0033] Determining a reprojection error between the fourth pose and the fifth pose of the feature object under each of the preset image feature dimensions at all previous acquisition moments to obtain a second error;

[0034] After summing up the second errors under each of the preset image feature dimensions, the first error is obtained.

[0035] Optionally, the number of the preset image feature dimensions is multiple; performing mapping processing according to the first pose, the second pose, the feature object, and the object features corresponding to the feature object to obtain the parking map, including:

[0036] Performing mapping processing according to the first pose and the second pose to obtain a vehicle driving sub-map;

[0037] Performing mapping processing according to the feature object and the corresponding object features under each of the preset image feature dimensions to obtain a feature sub-map corresponding to each of the preset image feature dimensions;

[0038] After superimposing the feature sub-maps corresponding to each of the preset image feature dimensions and the vehicle driving sub-map, the parking map is obtained.

[0039] In a second aspect, an embodiment of the present application provides a mapping device in an automatic parking scenario, including:

[0040] An acquisition unit, configured to acquire image data respectively collected by multiple fisheye cameras in a vehicle in real time to obtain a first image sequence;

[0041] A grouping unit, configured to perform grouping processing on the first image sequence according to preset image feature dimensions to obtain a second image sequence corresponding to each of the preset image feature dimensions;

[0042] A processing unit, configured to perform feature extraction processing on the image data in the second image sequence according to the preset image feature dimensions to obtain image feature information under each of the preset image feature dimensions;

[0043] A mapping unit, configured to perform mapping processing according to the image feature information under each of the preset image feature dimensions to obtain a parking map.

[0044] Optionally, the first image sequence includes image data collected by each fisheye camera at multiple acquisition moments; at this time, the grouping unit is configured to:

[0045] If the preset image feature dimension is the semantic feature dimension, perform stitching processing on the image data collected by the multi-channel fisheye cameras at each acquisition moment included in the first image sequence to obtain first image data; and determine a second image sequence in the semantic feature dimension according to the first image data.

[0046] If the preset image feature dimension is the corner feature dimension, perform polling processing on the image data included in the first image sequence according to the number of fisheye cameras to obtain second image data, and determine a second image sequence in the corner feature dimension according to the second image data.

[0047] Optionally, the preset image feature dimension is the semantic feature dimension; in this case, the processing unit is configured to:

[0048] Perform image detection processing on each first image data to obtain semantic feature information corresponding to each first image data; wherein, the semantic feature information at least indicates at least one semantic object included in the first image data.

[0049] Determine the semantic feature information of the same semantic object at different acquisition moments according to the pose difference of the semantic object at different acquisition moments, and determine the image feature information in the semantic feature dimension according to the semantic feature information of each semantic object at different acquisition moments.

[0050] Optionally, the preset image feature dimension is the corner feature dimension; in this case, the processing unit is configured to:

[0051] Perform corner feature extraction processing on each second image data to obtain corner feature information corresponding to each second image data; wherein, the corner feature information at least indicates at least one feature point included in the second image data.

[0052] Determine the corner feature information of the same feature point at different acquisition moments according to the similarity between the feature points at different acquisition moments.

[0053] Obtain the vehicle poses at different acquisition moments.

[0054] Convert the corner feature information of the same feature point at different acquisition moments into three-dimensional corner feature information according to the vehicle poses at different acquisition moments and the corner feature information of the same feature point at different acquisition moments.

[0055] Determine the image feature information in the corner feature dimension based on the three-dimensional corner feature information of the same feature point at different acquisition times.

[0056] Optionally, the image feature information in each preset image feature dimension includes a feature object and the object features corresponding to the feature object; at this time, the mapping unit includes a mapping module for:

[0057] Determine the target relative pose of each feature object at different acquisition times; wherein, the different acquisition times include the previous acquisition time and the subsequent acquisition time;

[0058] Determine the first pose of the feature object at the previous acquisition time, and determine the second pose of the feature object at the subsequent acquisition time according to the target relative pose and the first pose;

[0059] Perform mapping processing according to the first pose, the second pose, the feature object, and the object features corresponding to the feature object to obtain the parking map.

[0060] Optionally, the mapping unit includes an optimization module for:

[0061] Obtain the third pose and the fourth pose corresponding to the feature object; wherein, the third pose is the pose obtained by performing dead reckoning on the feature object at different acquisition times; the fourth pose is the pose of the feature object in the vehicle body coordinate system at different acquisition times;

[0062] Determine the initial relative pose of the feature object at different acquisition times according to the third pose at the previous acquisition time and the third pose at the subsequent acquisition time;

[0063] Project the fourth pose of the feature object at the subsequent acquisition time to the previous acquisition time according to the initial relative pose to obtain the fifth pose corresponding to the feature object;

[0064] After determining the first error between the fourth pose and the fifth pose of the feature object at all previous acquisition times, perform optimization processing on the initial relative pose according to the first error to obtain the target relative pose.

[0065] Optionally, the number of preset image feature dimensions is multiple; at this time, the optimization module is used for:

[0066] Determine the reprojection error between the fourth pose and the fifth pose of the feature object in each preset image feature dimension at all previous acquisition times to obtain a second error;

[0067] After summing the second errors in each of the preset image feature dimensions, the first error is obtained.

[0068] Optionally, the number of the preset image feature dimensions is multiple; in this case, the mapping module is configured to:

[0069] Perform mapping processing according to the first pose and the second pose to obtain a vehicle driving sub-map;

[0070] Perform mapping processing according to the feature object and the corresponding object feature in each of the preset image feature dimensions to obtain a feature sub-map corresponding to each of the preset image feature dimensions;

[0071] After superimposing the feature sub-map corresponding to each of the preset image feature dimensions and the vehicle driving sub-map, the parking map is obtained.

[0072] In a third aspect, an embodiment of the present application provides an electronic device, including: a memory, a processor;

[0073] The memory stores computer execution instructions;

[0074] The processor executes the computer execution instructions stored in the memory, so that the processor executes the above first aspect and / or various possible implementation manners of the first aspect.

[0075] In a fourth aspect, an embodiment of the present application provides a vehicle, and the vehicle includes the electronic device in the third aspect.

[0076] In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer execution instructions are stored, and when the computer execution instructions are executed by a processor, they are used to implement the above first aspect and / or various possible implementation manners of the first aspect.

[0077] In a sixth aspect, an embodiment of the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the above first aspect and / or various possible implementation manners of the first aspect.

[0078] The mapping method, device, and vehicle in the automatic parking scenario provided by the embodiments of the present application can obtain the image data collected by multiple fisheye cameras in the vehicle in real time, obtain a first image sequence including the omnidirectional information of the environment where the vehicle is located, and then, according to the preset image feature dimensions, perform grouping processing on the first image sequence to obtain a second image sequence corresponding to each preset image feature dimension. After that, according to the preset image feature dimensions, perform feature extraction processing on the image data in the second image sequence to obtain the image feature information under each preset image feature dimension. At this time, it is possible to group the image data in the first image sequence according to the preset image feature dimensions, realize on-demand image processing, not only improve the efficiency and accuracy of image processing, but also reduce the computing power overhead. After that, mapping processing can be performed according to the image feature information under each preset image feature dimension to obtain a parking map. At this time, by combining various types of image feature information for mapping processing, the obtained parking map is more accurate, which helps to improve the accuracy of parking and further improve the user's driving experience. Description of the Drawings

[0079] The drawings here are incorporated into the specification and form a part of this specification, showing the embodiments in line with the present application, and are used together with the specification to explain the principles of the present application.

[0080] Figure 1 Flow schematic of a mapping method in the automatic parking scenario provided by the embodiments of the present application Figure 1 ;

[0081] Figure 2 Flow schematic of a mapping method in the automatic parking scenario provided by the embodiments of the present application Figure 2 ;

[0082] Figure 3 Schematic diagram of the implementation process of a mapping method in the automatic parking scenario provided by the embodiments of the present application;

[0083] Figure 4 Schematic diagram of the structure of a mapping device in the automatic parking scenario provided by the embodiments of the present application;

[0084] Figure 5 Schematic diagram of the structure of another mapping device in the automatic parking scenario provided by the embodiments of the present application;

[0085] Figure 6 Schematic diagram of the structure of an electronic device provided by the embodiments of the present application;

[0086] Figure 7 Schematic diagram of the structure of a vehicle provided by the embodiments of the present application.

[0087] With the above-mentioned drawings, specific embodiments of the present application have been shown, and will be described in more detail hereinafter. These drawings and the written description are not intended to limit the scope of the concept of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. Detailed Embodiments

[0088] Exemplary embodiments will be described in detail herein, and examples thereof are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0089] In the automatic parking scenario, it is necessary to construct a high-precision map to accurately implement the automatic parking function. Exemplarily, for the memory parking function in the automatic parking scenario, it can assist the user to automatically cruise and park in the parking space according to the memorized route, or automatically park out from the parking space and cruise to the parking out point, thereby solving the parking problem in a fixed parking lot.

[0090] When the user uses the memory parking function for the first time, route learning needs to be performed first. That is, the user needs to manually drive the vehicle to drive through the parking route once, and after parking is completed, obtain the "memory route", that is, the parking map. After that, the vehicle can perform path planning and obstacle avoidance according to the generated parking map and the pose of the located vehicle in the parking map, so as to automatically complete the parking operation.

[0091] However, in some parking lot scenarios, there are certain challenges in constructing the parking map. First, in the parking lot scenario, the vehicles parked in the parking spaces often change, resulting in easy changes in the parking scenario, which in turn affects the positioning effect of the vehicle in the parking map; second, the lighting conditions in the parking lot (for example, the lighting changes in the ground parking lot, or the lighting changes in the underground parking lot, or the light occlusion in the oncoming vehicle scenario, the headlight irradiation of other vehicles, etc.) may be unstable, which in turn easily affects the positioning effect of the vehicle; third, in the indoor parking lot, there is usually no satellite signal, resulting in the inability of the vehicle to be positioned in combination with navigation, further affecting the positioning effect.

[0092] Therefore, the accuracy of the parking map and the characteristics of easy positioning of the parking map are crucial for the parking effect.

[0093] The mapping method in the automatic parking scenario provided by this application extracts image feature information in multiple preset image feature dimensions by combining the image data collected by multiple fisheye cameras, so as to be able to establish a parking map by combining image features in multiple dimensions, making the established parking map not only have high accuracy but also be easy to locate, thus solving the above technical problems.

[0094] The following uses specific embodiments to elaborate in detail on the technical solution of this application and how the technical solution of this application solves the above technical problems. These several specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0095] Figure 1 Flow schematic of a mapping method in the automatic parking scenario provided by the embodiments of this application Figure 1 , as Figure 1 shown, the method includes:

[0096] S101. Obtain the image data collected by multiple fisheye cameras in the vehicle in real time to obtain a first image sequence.

[0097] In one example, the multiple fisheye cameras in the vehicle can be used to obtain 360-degree environmental information around the vehicle, thereby ensuring the comprehensiveness of the image data included in the first image sequence.

[0098] At this time, the multiple fisheye cameras can be 4 fisheye cameras equipped on the front, rear, left, and right of the vehicle. At this time, the first image sequence can be obtained by obtaining the image data collected by the 4 fisheye cameras in real time. For example, assuming the time series is t1, t2,..., tn, then the obtained first image sequence can be P1, P2,..., Pn, where each image data in the image sequence can correspond to the image data collected by the 4 fisheye cameras, that is, Pi = (C1, C2, C3, C4).

[0099] S102. Perform grouping processing on the first image sequence according to the preset image feature dimensions to obtain a second image sequence corresponding to each preset image feature dimension.

[0100] In one example, the preset image feature dimensions are used to indicate the types of image features. At this time, the preset image feature dimensions can be determined according to the image features required in the process of constructing the parking map. Exemplarily, assume that the embodiments of this application can combine highly robust image features and high-precision image features for mapping processing to ensure the high precision and high robustness of the obtained parking map.

[0101] For different preset image feature dimensions, different image data can be determined to perform feature extraction, thereby ensuring the accuracy of the extracted features.

[0102] Therefore, in the embodiments of the present application, the first image sequence can be grouped according to the preset image feature dimensions to obtain multiple groups of second image sequences.

[0103] S103. Perform feature extraction processing on the image data in the second image sequence according to the preset image feature dimensions to obtain image feature information under each preset image feature dimension.

[0104] At this time, the corresponding image feature extraction method can be selected according to the preset image feature dimensions to perform feature extraction on the image data in the second image sequence corresponding to the preset image feature dimensions, and image feature information under each preset image feature dimension can be obtained.

[0105] S104. Perform mapping processing according to the image feature information under each preset image feature dimension to obtain a parking map.

[0106] In one example, the embodiments of the present application can perform mapping processing according to the image feature information under multiple preset image feature dimensions, so that the obtained parking map is more accurate.

[0107] The mapping method in the automatic parking scenario provided by the embodiments of the present application can obtain the image data collected by multiple fisheye cameras in the vehicle in real time to obtain a first image sequence including all-round information of the environment where the vehicle is located. Then, according to the preset image feature dimensions, the first image sequence is grouped to obtain the second image sequence corresponding to each preset image feature dimension. After that, the image data in the second image sequence can be subjected to feature extraction processing according to the preset image feature dimensions to obtain image feature information under each preset image feature dimension. At this time, the image data in the first image sequence can be grouped according to the preset image feature dimensions to realize on-demand image processing, which not only improves the efficiency and accuracy of image processing, but also reduces the computing power overhead. After that, mapping processing can be performed according to the image feature information under each preset image feature dimension to obtain a parking map. At this time, by combining various types of image feature information for mapping processing, the obtained parking map is more accurate, which helps to improve the accuracy of parking and further improve the driving experience of users.

[0108] Figure 2 It is a flow diagram of a mapping method in an automatic parking scenario provided by an embodiment of the present application Figure 2 As Figure 2 shown, this embodiment is in Figure 1Based on the embodiments, the mapping method in the automatic parking scenario will be described in detail. This method includes:

[0109] S201. Obtain the image data collected by multiple fisheye cameras in the vehicle in real time to obtain a first image sequence.

[0110] In one example, this step can refer to the content described in S101 above and will not be elaborated here.

[0111] S202. Group the first image sequence according to a preset image feature dimension to obtain a second image sequence corresponding to each preset image feature dimension.

[0112] In a possible implementation manner, the first image sequence includes the image data collected by each fisheye camera at multiple acquisition times; the preset image feature dimensions at least include a semantic feature dimension and a corner feature dimension. At this time, the second image sequence corresponding to each preset image feature dimension can be determined according to the process described below.

[0113] For example, if the preset image feature dimension is the semantic feature dimension, first splice the image data collected by multiple fisheye cameras at each acquisition time included in the first image sequence to obtain first image data, and then determine the second image sequence in the semantic feature dimension according to the first image data.

[0114] In one example, when splicing the image data collected by multiple fisheye cameras, IPM (Inverse Perspective Mapping) technology can be used to convert the image data collected by multiple fisheye cameras into a top view to obtain first image data.

[0115] In one example, the first image data at each acquisition time can be aggregated to obtain the second image sequence in the semantic feature dimension.

[0116] This implementation manner can avoid image processing on the original image by splicing the image data collected by multiple fisheye cameras, thereby reducing the computing power overhead, and also making the semantic information included in the first image data richer, which helps to improve the quality of the obtained image features.

[0117] Again, for example, if the preset image feature dimension is the corner feature dimension, the image data included in the first image sequence is polled according to the number of fisheye cameras to obtain second image data, and the second image sequence in the corner feature dimension is determined according to the second image data.

[0118] In one example, if the image data collected by the multi-channel fisheye cameras at acquisition time i is Pi = (C1, C2, C3, C4), then, according to the number of channels 4, the image data included in the first image sequence can be polled. That is, at time t1, the image data C1 collected by the first fisheye camera is processed to obtain second image data; at time t2, the image data C2 collected by the second fisheye camera is processed to obtain second image data; at time t3, the image data C3 collected by the third fisheye camera is processed; at time t4, the image data C4 collected by the fourth fisheye camera is processed to obtain second image data; then, at time t5, the image data C1 collected by the first fisheye camera is processed, and so on, polling in turn to obtain a second image sequence.

[0119] In this implementation, through the polling mechanism, 1-channel image data collected by the multi-channel fisheye cameras is processed each time, so as to reduce the computing power overhead and improve the processing efficiency.

[0120] After obtaining the second image sequence, the image data in the second image sequence can be subjected to feature extraction processing. For example, if the preset image feature dimension is the semantic feature dimension, then the feature extraction processing process can refer to the process described in S203 to S204 below; if the preset image feature dimension is the corner feature dimension, then the feature extraction processing process can refer to the process described in S205 to S209 below.

[0121] S203. Perform image detection processing on each first image data to obtain the semantic feature information corresponding to each first image data.

[0122] Among them, the semantic feature information at least indicates at least one semantic object included in the first image data.

[0123] In one example, based on the deep learning-based image detection technology, the first image data can be subjected to image detection processing to obtain the semantic feature information corresponding to the first image data. At this time, the semantic objects indicated by the semantic feature information may include, but are not limited to: the type of semantic object, the semantic information of the semantic object (for example, the geometric information of the semantic object), and the pose of the semantic object in the vehicle body coordinate system. The semantic objects included in the semantic feature information may include, but are not limited to, parking spaces, columns, etc.

[0124] S204. Determine the semantic feature information of the same semantic object at different acquisition times according to the pose difference of the semantic object at different acquisition times, and determine the image feature information in the semantic feature dimension according to the semantic feature information of each semantic object at different acquisition times.

[0125] In one example, the pose difference of a semantic object at different acquisition times can be determined according to the DR (Dead Reckoning) pose corresponding to the semantic object, so as to track the same semantic object according to the pose difference at different acquisition times, and then determine the semantic feature information of the same semantic object at different acquisition times.

[0126] During specific implementation, for any semantic object k, when tracking the semantic object, the DR poses of the semantic object at different acquisition times ti and tj can be determined as Ti and Tj respectively. Moreover, the poses of the semantic object in the vehicle body coordinate system at different acquisition times ti and tj are determined as Ski and Skj respectively. Then, the relative pose T of the semantic object at two acquisition times can be calculated according to the following formula (1). ji 。

[0127] T ji =T i -1 *T j (1)

[0128] Wherein, T i -1 represents the inverse of T i 。

[0129] After that, the poses in the vehicle body coordinate system at different acquisition times can be transformed to the same acquisition time according to the relative pose. At this time, assuming that the pose in the vehicle body coordinate system at the previous acquisition time is Ski, then Ski can be projected to the acquisition time tj according to the following formula (2).

[0130] jSki=T ji *Ski (2)

[0131] At this time, the reprojection error between jSki and Skj can be used to determine whether the semantic objects k at two acquisition times are the same object, so as to realize the tracking of the semantic object. The calculation method of the reprojection error can be seen in the following formula (3).

[0132] E semantic =norm(Skj - jSki)=norm(Skj - T ji *Ski ) (3)

[0133] Wherein, norm() represents taking the modulus of the reprojection error. At this time, if the value of E_semantic is less than the first preset threshold, it is considered that the semantic objects in the two vehicle body coordinate systems are the same semantic object.

[0134] In this implementation manner, image detection processing can be performed on the first image data to obtain the semantic feature information corresponding to the first image data. Then, according to the pose difference and reprojection error of the semantic object at different acquisition times, the semantic object at different acquisition times is tracked, so that the image feature information in the semantic feature dimension can be determined simply and quickly.

[0135] S205. Perform corner feature extraction processing on each second image data to obtain the corner feature information corresponding to each second image data.

[0136] Among them, the corner feature information at least indicates at least one feature point included in the second image data.

[0137] In one example, according to the corner feature extraction method, for example, SIFT (Scale-Invariant Feature Transform) technology, ORB (Oriented FAST and Rotated BRIEF) technology, etc., corner feature extraction is performed on the second image data to obtain the corner feature information corresponding to each second image data.

[0138] In one example, the corner feature information at least includes the two-dimensional coordinate information and feature descriptor of the indicated feature point.

[0139] S206. Determine the corner feature information of the same feature point at different acquisition times according to the similarity between the feature points at different acquisition times.

[0140] In one example, the similarity between two feature points can be determined according to the similarity between the feature descriptors of the feature points. Suppose that at acquisition time ti and acquisition time tj, the determined corner feature information is fki and fmj respectively. Then, for a certain corner feature information at acquisition time ti, the corner feature information corresponding to each feature point at acquisition time tj can be traversed, so as to determine the similar feature points at different acquisition times according to the similarity, and the similar feature points are determined as the same feature point.

[0141] S207. Obtain the vehicle poses at different acquisition times.

[0142] In one example, according to the DR technology, the vehicle poses at different acquisition times can be determined. Suppose that the poses of the vehicle at different acquisition times are recorded as Ti, Ti+1, ….

[0143] S208. According to the vehicle poses at different acquisition times and the corner feature information of the same feature point at different acquisition times, convert the corner feature information of the same feature point at different acquisition times into three-dimensional corner feature information.

[0144] In one example, an embodiment of the present application can determine the three-dimensional coordinates corresponding to each feature point through triangulation technology according to the corner feature information of the same feature point at different acquisition times and the corresponding vehicle poses at different acquisition times. At this time, the three-dimensional coordinates in the camera coordinate system can also be converted into the three-dimensional coordinates in the vehicle coordinate system through the external camera parameters, so as to realize the conversion of the two-dimensional corner feature information corresponding to each feature point into three-dimensional corner feature information. At this time, the three-dimensional corner feature information includes the three-dimensional coordinates of the feature point and the feature descriptor.

[0145] S209. Determine the image feature information in the corner feature dimension according to the three-dimensional corner feature information of the same feature point at different acquisition times.

[0146] In the above embodiment, after performing corner feature extraction processing on the second image data and determining the corner feature information of each feature point at different acquisition times, the two-dimensional corner feature information can also be converted into three-dimensional corner feature information according to the vehicle pose, so that the parking map established according to the corner feature information can be more accurate.

[0147] S210. Perform mapping processing according to the image feature information in each preset image feature dimension to obtain a parking map.

[0148] In a possible implementation manner, the image feature information in each preset image feature dimension includes a feature object and the object feature corresponding to the feature object. At this time, in the semantic feature dimension, the feature object can be understood as the above semantic object, and the object feature corresponding to the feature object can be understood as the above semantic information, the pose of the semantic object in the vehicle body coordinate system, etc. In the corner feature dimension, the feature object can be understood as the above feature point, and the object feature corresponding to the feature object can be understood as the above feature descriptor.

[0149] At this time, when performing mapping processing according to the image feature information in each preset image feature dimension to obtain a parking map, the target relative pose of each feature object at different acquisition times can be determined first, and then the first pose of the feature object at the previous acquisition time can be determined, and the second pose of the feature object at the subsequent acquisition time can be determined according to the target relative pose and the first pose. Finally, mapping processing is performed according to the first pose, the second pose, the feature object, and the object feature corresponding to the feature object to obtain a parking map. Among them, different acquisition times include the previous acquisition time and the subsequent acquisition time.

[0150] At this time, according to the process described above, the second pose at the next acquisition moment can be determined based on the first pose and the target relative pose at the previous acquisition moment, and then the pose of the feature object at any acquisition moment, that is, the vehicle pose, can be obtained. Thus, map building can be performed based on the vehicle poses at each acquisition moment and the object features corresponding to the feature objects to obtain a parking map.

[0151] As can be seen from the above description, to build an accurate parking map, the accuracy of the target relative pose is particularly important. At this time, the target relative pose of each feature object at different acquisition moments can be determined according to the process described below.

[0152] First, obtain the third pose and the fourth pose corresponding to the feature object. Among them, the third pose is the pose obtained by dead reckoning calculation of the feature object at different acquisition moments, that is, the DR pose; the fourth pose is the pose of the feature object in the vehicle body coordinate system at different acquisition moments.

[0153] Secondly, determine the initial relative pose of the feature object at different acquisition moments according to the third pose at the previous acquisition moment and the third pose at the next acquisition moment.

[0154] Next, project the fourth pose of the feature object at the next acquisition moment to the previous acquisition moment according to the initial relative pose to obtain the fifth pose corresponding to the feature object.

[0155] Finally, after determining the first error between the fourth pose and the fifth pose of the feature object at all previous acquisition moments, optimize the initial relative pose according to the first error to obtain the target relative pose.

[0156] In specific implementation, assume that the third poses obtained by dead reckoning calculation of the feature object at different acquisition moments ti and tj are Ti and Tj, and the fourth poses of the feature object in the vehicle body coordinate system at different acquisition moments ti and tj are Iki and Ikj. At this time, the third pose can be used as the initial value to calculate the initial relative pose. At this time, the calculation method of the initial relative pose can refer to the above formula (1), which will not be elaborated here. Then, according to this initial relative pose, project the fourth pose of the feature object at the next acquisition moment to the previous acquisition moment to obtain the fifth pose corresponding to the feature object. The specific projection method can refer to the above formula (2), which will not be elaborated here.

[0157] After obtaining the fifth pose corresponding to the feature object, the reprojection error between the fourth pose and the fifth pose of the feature object can be calculated according to the process described in the above formula (3). After determining the reprojection error between the fourth pose and the fifth pose of the feature object at each previous acquisition moment among multiple acquisition moments of the feature object, the reprojection errors are summed to obtain the first error. At this time, the optimal relative pose, that is, the target relative pose, can be obtained by solving the minimum value of the first error.

[0158] In one example, if the number of preset image feature dimensions is multiple, then when determining the first error between the fourth pose and the fifth pose of the feature object at all previous acquisition moments, the reprojection error between the fourth pose and the fifth pose of the feature object under each preset image feature dimension can be first determined to obtain the second error. Then, after summing the second errors under each preset image feature dimension, the first error is obtained.

[0159] Exemplarily, the preset image feature dimensions are the above semantic feature dimension and corner feature dimension. Then, assuming that the second error under the semantic feature dimension is expressed as the following formula (4) and the second error under the corner feature dimension is expressed as the following formula (5), then the first error can be seen as shown in formula (6).

[0160]

[0161] Cost(T ji )=cost semantic (T ji )+cost corner (T ji ) (6)

[0162] Among them, SUM() represents summing the reprojection errors between the fourth pose and the fifth pose of the feature object under the preset image feature dimensions at all previous acquisition moments.

[0163] In one example, when optimizing the initial relative pose according to the first error, the optimization process can be carried out according to the process described in the following formula (7).

[0164] T ji -opt=argmin(Cost(T ji )) (7)

[0165] Among them, T ji -opt represents the initial relative pose after optimization processing, that is, the target relative pose, and argmin() is used to take the minimum value.

[0166] In the above embodiments, the reprojection errors in multiple preset image feature dimensions can be combined to determine the relative pose of the target, so that the determined relative pose of the target can combine information from multiple levels, thereby improving the accuracy of the relative pose of the target.

[0167] In a possible implementation, if the number of preset image feature dimensions is multiple, then when performing mapping processing based on the first pose, the second pose, the feature object, and the object features corresponding to the feature object, the mapping processing can be first performed according to the first pose and the second pose at each acquisition moment to obtain a vehicle driving sub-map. At this time, the vehicle driving sub-map can be used to indicate the driving trajectory of the vehicle. At the same time, mapping processing can also be performed according to the feature object and the corresponding object features in each preset image feature dimension to obtain a feature sub-map corresponding to each preset image feature dimension. After that, the feature sub-maps corresponding to each preset image feature dimension and the vehicle driving sub-map can be superimposed to obtain a parking map.

[0168] In an example, if the preset image feature dimensions are the above semantic feature dimension and corner feature dimension, then mapping processing can be performed according to the image feature information in the semantic feature dimension to obtain a parking sub-map in the semantic feature dimension, that is, a semantic map; and mapping processing can be performed according to the image feature information in the corner feature dimension to obtain a parking sub-map in the corner feature dimension, that is, a corner map.

[0169] In an example, when performing mapping processing according to the feature object and the corresponding object features in each preset image feature dimension, the feature object and the object features at different acquisition moments can be fused (for example, including but not limited to weighted fusion, average fusion, etc.), so that the obtained feature object and object features are more accurate.

[0170] This implementation can determine a parking map based on the vehicle driving trajectory and image feature information in multiple dimensions, which can make the information included in the determined parking map richer, thereby improving the accuracy of the established parking map. It also enables the vehicle to accurately position itself according to the feature objects and object features included in the parking map, thereby improving the accuracy of automatic parking and enhancing the user's parking experience.

[0171] Figure 3 The following is a schematic flowchart of the implementation process of a mapping method in an automatic parking scenario provided by an embodiment of the present application. As Figure 3 shown, the mapping method in the automatic parking scenario provided by the embodiment of the present application can obtain 4 fisheye cameras in real time (for example, Figure 3The image data respectively collected by the fish-eye cameras 1 to 4 shown are used to obtain the first image sequence. Then, the image data collected by the four fish-eye cameras can be stitched through the IPM technology to obtain the above-mentioned first image data. After that, image detection processing can be performed on the first image data to obtain the image feature information in the semantic feature dimension. At the same time, the image data collected by each fish-eye camera (i.e., the above-mentioned second image data) can be processed in turn by means of image polling, and corner feature extraction processing is performed on the image data to obtain the image feature information in the corner feature dimension.

[0172] At this time, dead reckoning can be performed according to the IMU (Inertial Measurement Unit) and the wheel speedometer to obtain the DR pose.

[0173] After that, according to the DR pose, the image feature information in the semantic feature dimension, and the image feature information in the corner feature dimension, the first pose and the second pose of the vehicle in the visual odometry coordinate system can be determined, and then the accurate vehicle pose can be determined.

[0174] At this time, map building processing can be performed according to the vehicle pose to obtain a vehicle driving sub-map, and map building processing can be performed according to the image feature information in the semantic feature dimension to obtain a semantic map, and map building processing can be performed according to the image feature information in the corner feature dimension to obtain a corner map.

[0175] After that, the vehicle driving sub-map, the semantic map, and the corner map can be superimposed to obtain a parking map.

[0176] Figure 4 The following is a schematic structural diagram of a map building device in an automatic parking scenario provided by an embodiment of the present application. As Figure 4 shown, the map building device 40 in the automatic parking scenario provided by this embodiment includes:

[0177] An acquisition unit 401, configured to acquire the image data respectively collected by multiple fish-eye cameras in a vehicle in real time to obtain a first image sequence.

[0178] A grouping unit 402, configured to group the first image sequence according to a preset image feature dimension to obtain a second image sequence corresponding to each preset image feature dimension.

[0179] A processing unit 403, configured to perform feature extraction processing on the image data in the second image sequence according to a preset image feature dimension to obtain the image feature information in each preset image feature dimension.

[0180] A mapping unit 404, configured to perform mapping processing based on image feature information in each preset image feature dimension to obtain a parking map.

[0181] Figure 5 FIG. 4 is a schematic structural diagram of another mapping device in an automatic parking scenario provided by an embodiment of the present application. As Figure 5 shown, the mapping device 50 in the automatic parking scenario provided in this embodiment includes:

[0182] An acquisition unit 501, configured to acquire image data collected by multiple fisheye cameras in a vehicle in real time to obtain a first image sequence.

[0183] A grouping unit 502, configured to perform grouping processing on the first image sequence according to a preset image feature dimension to obtain a second image sequence corresponding to each preset image feature dimension.

[0184] A processing unit 503, configured to perform feature extraction processing on the image data in the second image sequence according to a preset image feature dimension to obtain image feature information in each preset image feature dimension.

[0185] A mapping unit 504, configured to perform mapping processing based on the image feature information in each preset image feature dimension to obtain a parking map.

[0186] Optionally, the first image sequence includes image data collected by each fisheye camera at multiple acquisition times; at this time, the grouping unit 502 is configured to:

[0187] If the preset image feature dimension is a semantic feature dimension, perform splicing processing on the image data collected by multiple fisheye cameras at each acquisition time included in the first image sequence to obtain first image data; and determine a second image sequence in the semantic feature dimension according to the first image data;

[0188] If the preset image feature dimension is a corner feature dimension, perform polling processing on the image data included in the first image sequence according to the number of fisheye cameras to obtain second image data, and determine a second image sequence in the corner feature dimension according to the second image data.

[0189] Optionally, the preset image feature dimension is a semantic feature dimension; at this time, the processing unit 503 is configured to:

[0190] Perform image detection processing on each first image data to obtain semantic feature information corresponding to each first image data; wherein the semantic feature information at least indicates at least one semantic object included in the first image data;

[0191] Determine the semantic feature information of the same semantic object at different acquisition times according to the pose difference of the semantic object at different acquisition times, and determine the image feature information in the semantic feature dimension according to the semantic feature information of each semantic object at different acquisition times.

[0192] Optionally, the preset image feature dimension is the corner feature dimension; at this time, the processing unit 503 is used for:

[0193] Perform corner feature extraction processing on each second image data to obtain the corner feature information corresponding to each second image data; wherein, the corner feature information at least indicates at least one feature point included in the second image data;

[0194] Determine the corner feature information of the same feature point at different acquisition times according to the similarity between the feature points at different acquisition times;

[0195] Obtain the vehicle poses at different acquisition times;

[0196] According to the vehicle poses at different acquisition times and the corner feature information of the same feature point at different acquisition times, convert the corner feature information of the same feature point at different acquisition times into three-dimensional corner feature information;

[0197] Determine the image feature information in the corner feature dimension according to the three-dimensional corner feature information of the same feature point at different acquisition times.

[0198] Optionally, the image feature information in each preset image feature dimension includes a feature object and the object feature corresponding to the feature object; at this time, the mapping unit 504 includes a mapping module 5041, which is used for:

[0199] Determine the target relative pose of each feature object at different acquisition times; wherein, different acquisition times include the previous acquisition time and the subsequent acquisition time;

[0200] Determine the first pose of the feature object at the previous acquisition time, and determine the second pose of the feature object at the subsequent acquisition time according to the target relative pose and the first pose;

[0201] Perform mapping processing according to the first pose, the second pose, the feature object and the object feature corresponding to the feature object to obtain a parking map.

[0202] Optionally, the mapping unit 504 includes an optimization module 5042, which is used for:

[0203] Obtain the third pose and the fourth pose of the feature object at different acquisition times; wherein, the third pose is obtained by performing dead reckoning on the feature object; the fourth pose is the pose of the feature object in the vehicle body coordinate system;

[0204] Determine the initial relative pose of the feature object at different acquisition times according to the third pose at the previous acquisition time and the third pose at the subsequent acquisition time;

[0205] Project the feature object at the subsequent acquisition time to the previous acquisition time to obtain the fifth pose corresponding to the feature object;

[0206] After determining the first error between the fourth pose and the fifth pose of the feature object at all previous acquisition times, optimize the initial relative pose according to the first error to obtain the target relative pose.

[0207] Optionally, the number of preset image feature dimensions is multiple; at this time, the optimization module 5042 is used for:

[0208] Determine the reprojection error between the fourth pose and the fifth pose of the feature object under each preset image feature dimension at all previous acquisition times to obtain the second error;

[0209] After summing the second errors under each preset image feature dimension, obtain the first error.

[0210] Optionally, the number of preset image feature dimensions is multiple; at this time, the mapping module 5041 is used for:

[0211] Perform mapping processing according to the first pose and the second pose to obtain a vehicle driving sub-map;

[0212] Perform mapping processing according to the feature object and the corresponding object features under each preset image feature dimension to obtain a feature sub-map corresponding to each preset image feature dimension;

[0213] After superimposing the feature sub-map corresponding to each preset image feature dimension and the vehicle driving sub-map, obtain a parking map.

[0214] The mapping device in the automatic parking scenario provided by the embodiments of the present application can execute the method provided by the above method embodiments, and its implementation principle and technical effects are similar, which will not be elaborated here in this embodiment.

[0215] Figure 6 It is a schematic structural diagram of an electronic device provided by an embodiment of the present application. As Figure 6 shown, the electronic device 60 provided in this embodiment includes: at least one processor 601 and a memory 602. Optionally, the electronic device 60 further includes a communication component 603. Among them, the processor 601, the memory 602, and the communication component 603 are connected through a bus 604.

[0216] In a specific implementation process, at least one processor 601 executes computer-executable instructions stored in a memory 602, such that at least one processor 601 executes the above-described method.

[0217] For the specific implementation process of the processor 601, reference may be made to the above method embodiments. Their implementation principles and technical effects are similar, and thus will not be elaborated herein.

[0218] In the above embodiments, it should be understood that the processor may be a central processing unit (CPU for short), or may also be other general-purpose processors, digital signal processors (DSP for short), application specific integrated circuits (ASIC for short), etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the method disclosed in combination with the invention may be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor.

[0219] The memory may include a high-speed random access memory (RAM), and may also include non-volatile memory (NVM), such as at least one disk memory.

[0220] The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the buses in the drawings of this application are not limited to only one bus or one type of bus.

[0221] Figure 7 FIG. [FIGURE NUMBER] is a schematic structural diagram of a vehicle provided by an embodiment of the present application. As Figure 7 shown, the vehicle provided in this embodiment may include Figure 6 the electronic device shown in FIG. [FIGURE NUMBER].

[0222] The present application further provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0223] Note: The figure numbers in and are placeholders in the original text and should be filled with the actual figure numbers in the complete context. Here they are presented as "[FIGURE NUMBER]" for clarity in translation.The present application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0224] The above-readable storage medium may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, a magnetic disk, or an optical disk. The readable storage medium may be any available medium accessible by a general-purpose or special-purpose computer.

[0225] An exemplary readable storage medium is coupled to the processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium may also be a component of the processor. The processor and the readable storage medium may be located in an application specific integrated circuit (ASIC). Of course, the processor and the readable storage medium may also exist as discrete components in a device.

[0226] The division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms.

[0227] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0228] In addition, the functional units in the various embodiments of the present invention may be integrated in a processing unit, may exist separately physically for each unit, or two or more units may be integrated in one unit.

[0229] If a function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing an electronic device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present invention. The foregoing storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs, etc., which are various media that can store program codes.

[0230] Those of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions. The foregoing program can be stored in a computer-readable storage medium. When this program is executed, it executes the steps including the above method embodiments; and the foregoing storage medium includes: ROMs, RAMs, magnetic disks, or optical discs, etc., which are various media that can store program codes.

[0231] Finally, it should be noted that: After considering the specification and practicing the invention disclosed herein, those skilled in the art will easily think of other implementation manners of the present invention. The present invention is intended to cover any variations, uses, or adaptations of the present invention, and these variations, uses, or adaptations follow the general principles of the present invention and include common general knowledge or conventional technical means in the technical field not disclosed in the present invention. It is not limited to the precise structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present invention is only limited by the appended claims.

Claims

1. A mapping method in an automatic parking scenario, characterized in that Including: Obtaining in real time the image data respectively collected by multiple fisheye cameras in a vehicle to obtain a first image sequence; Performing grouping processing on the first image sequence according to a preset image feature dimension to obtain a second image sequence corresponding to each preset image feature dimension; Performing feature extraction processing on the image data in the second image sequence according to the preset image feature dimension to obtain image feature information under each preset image feature dimension; Performing mapping processing according to the image feature information under each preset image feature dimension to obtain a parking map.

2. The method according to claim 1, wherein The first image sequence includes the image data collected by each fisheye camera at multiple acquisition times; the preset image feature dimensions at least include a semantic feature dimension and a corner feature dimension; Performing grouping processing on the first image sequence according to a preset image feature dimension to obtain a second image sequence corresponding to each preset image feature dimension, including: If the preset image feature dimension is the semantic feature dimension, splicing the image data collected by multiple fisheye cameras at each acquisition time included in the first image sequence to obtain first image data; and determining the second image sequence under the semantic feature dimension according to the first image data; If the preset image feature dimension is the corner feature dimension, polling the image data included in the first image sequence according to the number of fisheye cameras to obtain second image data, and determining the second image sequence under the corner feature dimension according to the second image data.

3. The method according to claim 2, characterized in that, The preset image feature dimension is the semantic feature dimension; performing feature extraction processing on the image data in the second image sequence according to the preset image feature dimension to obtain image feature information under each preset image feature dimension, including: Performing image detection processing on each first image data to obtain semantic feature information corresponding to each first image data; wherein the semantic feature information at least indicates at least one semantic object included in the first image data; Determining the semantic feature information of the same semantic object at different acquisition times according to the pose difference of the semantic object at different acquisition times, and determining the image feature information under the semantic feature dimension according to the semantic feature information of each semantic object at different acquisition times.

4. The method according to claim 2, wherein The preset image feature dimension is the corner feature dimension; performing feature extraction processing on the image data in the second image sequence according to the preset image feature dimension to obtain image feature information under each preset image feature dimension, including: Performing corner feature extraction processing on each second image data to obtain corner feature information corresponding to each second image data; wherein the corner feature information at least indicates at least one feature point included in the second image data; Determining the corner feature information of the same feature point at different acquisition times according to the similarity between feature points at different acquisition times; Obtain the vehicle poses at different acquisition times; According to the vehicle poses at different acquisition times and the corner feature information of the same feature point at different acquisition times, convert the corner feature information of the same feature point at different acquisition times into three-dimensional corner feature information; According to the three-dimensional corner feature information of the same feature point at different acquisition times, determine the image feature information in the corner feature dimension.

5. The method according to claim 1, wherein The image feature information in each preset image feature dimension includes a feature object and the object features corresponding to the feature object; according to the image feature information in each preset image feature dimension, perform mapping processing to obtain a parking map, including: Determine the target relative poses of each feature object at different acquisition times; where the different acquisition times include the previous acquisition time and the subsequent acquisition time; Determine the first pose of the feature object at the previous acquisition time, and determine the second pose of the feature object at the subsequent acquisition time according to the target relative pose and the first pose; According to the first pose, the second pose, the feature object, and the object features corresponding to the feature object, perform mapping processing to obtain the parking map.

6. The method according to claim 5, wherein Determining the target relative poses of each feature object at different acquisition times includes: Obtain the third pose and the fourth pose corresponding to the feature object; where the third pose is the pose obtained by performing dead reckoning on the feature object at different acquisition times; the fourth pose is the pose of the feature object in the vehicle body coordinate system at different acquisition times; According to the third pose at the previous acquisition time and the third pose at the subsequent acquisition time, determine the initial relative pose of the feature object at different acquisition times; According to the initial relative pose, project the fourth pose of the feature object at the subsequent acquisition time to the previous acquisition time to obtain the fifth pose corresponding to the feature object; After determining the first error between the fourth pose and the fifth pose of the feature object at all previous acquisition times, optimize the initial relative pose according to the first error to obtain the target relative pose.

7. The method according to claim 6, wherein The number of preset image feature dimensions is multiple; determining the first error between the fourth pose and the fifth pose of the feature object at all previous acquisition times includes: Determine the reprojection error between the fourth pose and the fifth pose of the feature object in each preset image feature dimension at all previous acquisition times to obtain a second error; After summing up the second errors in each preset image feature dimension, obtain the first error.

8. The method according to claim 5, characterized in that The number of preset image feature dimensions is multiple; according to the first pose, the second pose, the feature object, and the object features corresponding to the feature object, perform mapping processing to obtain the parking map, including: According to the first pose and the second pose, perform mapping processing to obtain a vehicle driving sub-map; Map building processing is performed based on the feature objects and corresponding object features in each of the preset image feature dimensions, and a feature sub-map corresponding to each of the preset image feature dimensions is obtained; After superimposing the feature sub-maps corresponding to each of the preset image feature dimensions and the vehicle driving sub-map, the parking map is obtained.

9. An electronic device, characterized in that, It includes: A memory and a processor; The memory stores computer execution instructions; The processor executes the computer execution instructions stored in the memory, so that the processor executes the method according to any one of claims 1-8.

10. A vehicle, characterized in that, The vehicle includes the electronic device according to claim 9.

Images