Data processing method, device, electronic device and storage medium

Abstract

The present disclosure provides a data processing method, device, electronic device and storage medium, which relate to the computer field, and in particular to the intelligent transportation field. The specific implementation scheme is: obtaining the first dimensional features of the image of the parking area; extracting the second dimensional features from the first dimensional features, wherein the second dimensional features have a causal relationship with the guide points of the parking area, and the guide points are used to guide vehicles in the parking area; and determining the guide points of the parking area based on the second dimensional features.

Description

Technical Field

[0001] The present disclosure relates to the field of computers, and in particular to a data processing method, device, electronic device, and storage medium in the field of intelligent transportation. Background Art

[0002] At present, deep learning methods are usually used to predict parking lot guide points. However, deep learning models are association-driven and most of them are black box models. They do not distinguish the causal relationships in the data, which leads to poor scenario generalization ability of the causal inference model of parking lot guide points and cannot be applied to all parking lots. Summary of the invention

[0003] The present disclosure provides a method, an apparatus, an electronic device, and a storage medium for data processing.

[0004] According to one aspect of the present disclosure, a data processing method is provided. The method may include: acquiring a first dimensional feature of an image of a parking area; extracting a second dimensional feature from the first dimensional feature, wherein the second dimensional feature has a causal relationship with a guide point of the parking area, and the guide point is used to guide a vehicle in the parking area; and determining the guide point of the parking area based on the second dimensional feature.

[0005] According to another aspect of the present disclosure, another data processing method is also provided. The method may include: acquiring an image of a parking area; inputting the image into a guide point prediction model, wherein the guide point prediction model is obtained based on deep stable learning training and is used to extract second dimensional features from first dimensional features of the image, and determine the guide point of the parking area based on the second dimensional features, wherein the second dimensional features and the guide point of the parking area have a causal relationship, and the guide point is used to guide the vehicle in the parking area; and acquiring the guide point output by the guide point prediction model.

[0006] According to another aspect of the present disclosure, a data processing device is also provided. The device may include: a first acquisition unit, configured to acquire a first dimensional feature of an image of a parking area; an extraction unit, configured to extract a second dimensional feature from the first dimensional feature, wherein the second dimensional feature has a causal relationship with a guide point of the parking area, and the guide point is used to guide a vehicle in the parking area; and a determination unit, configured to determine the guide point of the parking area based on the second dimensional feature.

[0007] According to another aspect of the present disclosure, another data processing device is also provided. The device may include: a second acquisition unit, used to acquire an image of a parking area; an input unit, used to input the image into a guide point prediction model, wherein the guide point prediction model is obtained based on deep stable learning training, and is used to extract second dimensional features from first dimensional features of the image, and determine the guide point of the parking area based on the second dimensional features, wherein the second dimensional features and the guide point of the parking area have a causal relationship, and the guide point is used to guide the vehicle in the parking area; a third acquisition unit, used to acquire the guide point output by the guide point prediction model.

[0008] According to another aspect of the present disclosure, an electronic device is also provided. The electronic device may include: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the data processing method of the embodiment of the present disclosure.

[0009] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is further provided, wherein the computer instructions are used to enable a computer to execute the data processing method of an embodiment of the present disclosure.

[0010] According to another aspect of the present disclosure, a computer program product is further provided, which may include a computer program. When the computer program is executed by a processor, the data processing method of the embodiment of the present disclosure is implemented.

[0011] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0013] Figure 1 is a flow chart of a data processing method according to an embodiment of the present disclosure;

[0014] Figure 2 is a flow chart of another data processing method according to an embodiment of the present disclosure;

[0015] Figure 3 It is a schematic diagram of predicting parking lot guide points based on a model according to the related technology of the present disclosure;

[0016] Figure 4 is a flow chart of predicting parking lot guidance points based on a model according to an embodiment of the present disclosure;

[0017] Figure 5 is a schematic diagram of a data processing device according to an embodiment of the present disclosure;

[0018] Figure 6 is a schematic diagram of another data processing device according to an embodiment of the present disclosure;

[0019] Figure 7 is a schematic diagram of an electronic device used to implement an embodiment of the present disclosure. DETAILED DESCRIPTION

[0020] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0021] The data processing method of the embodiment of the present disclosure is introduced below from the server side.

[0022] Figure 1 FIG. 1 is a flow chart of a data processing method according to an embodiment of the present disclosure. Figure 1 As shown, the method may include the following steps:

[0023] S102, obtaining a first dimension feature of an image of a parking area.

[0024] In the technical solution provided in the above step S102 of the present disclosure, an image of the parking area is acquired, and based on the image of the parking area, a first dimensional feature is extracted in the parking lot scene, wherein the image of the parking area may be an image of the parking lot, and the first dimensional feature may be a main feature in the parking lot scene.

[0025] Optionally, the first dimensional features may include, but are not limited to, road data such as road grade, length, width, image data such as parking lot signs, parking poles, parking spaces, trajectory data and positioning data describing user behavior, and geographic location data (point of information, POI for short) of parking lot entrances and exits, etc., which are not specifically limited here.

[0026] S104: extracting a second dimensional feature from the first dimensional feature, wherein the second dimensional feature has a causal relationship with a guide point in the parking area, and the guide point is used to guide the vehicle in the parking area.

[0027] In the technical solution provided in the above step S104 of the present disclosure, features having a causal relationship with the guide points of the parking area are extracted from the first dimensional features and used as the second dimensional features, wherein the guide points may be signs for guiding vehicles in the parking area, and the second dimensional data features may be features other than irrelevant features and false features in the first dimensional data features.

[0028] For example, to determine whether there is a causal relationship between the position point corresponding to a certain feature in the first dimension feature and the guide point of the parking area, the position corresponding to a certain feature in the first dimension feature can be regarded as variable A, and the guide point of the parking area can be regarded as variable B. The causal relationship between variable A and variable B (there is an interference variable C) is determined. If variable A is a discrete binary variable (with a value of 0 or 1), that is, whether a point around the parking lot is a parking lot guide point, then the overall sample can be divided into two groups according to the value of A: A = 0 or A = 1, and each sample is assigned a different weight, so that when A = 0 and A = 1, the interference variable C The distribution of D(B|A=0)=D(C|A=1), where D represents the variable distribution), then judging whether D(B|A=0) and D(B|A=1) are the same can determine whether A has a causal relationship with B. If D(B|A=0) and D(B|A=1) are the same, it means that A has a causal relationship with B, that is, a feature in the first dimension feature has a causal relationship with the guide point of the parking area. If D(B|A=0) and D(B|A=1) are not the same, it means that A has no causal relationship with B, that is, a feature in the first dimension feature has no causal relationship with the guide point of the parking area.

[0029] In this embodiment, studying the causal relationship refers to finding the true causal relationship between things and removing the mixed pseudo causal relationship to avoid introducing additional features, which is of great significance for feature selection of deep learning.

[0030] Optionally, the second dimensional features are extracted from the first dimensional features, that is, the false associations between irrelevant features in the first dimensional features and the guide points of the parking area are eliminated. Prior to this, all features in the first dimensional features need to be independent of each other, which can include: first using a random Fourier feature method to map the first dimensional features to a low-dimensional Euclidean space, and then mapping the features to a high-dimensional space based on a kernel function method, and expanding the sample dimension to ensure that the features in the first dimensional features are strictly independent.

[0031] S106: Determine a guide point of the parking area based on the second dimensional feature.

[0032] In the technical solution provided in the above step S106 of the present disclosure, a second dimensional feature having a causal relationship with the guide point of the parking area is extracted from the first dimensional feature, and the guide point of the parking area can be determined based on the second dimensional feature.

[0033] Through the above steps S102 to S106, the first dimensional features of the image of the parking area are obtained; the second dimensional features are extracted from the first dimensional features, wherein the second dimensional features have a causal relationship with the guide points of the parking area, and the guide points are used to guide the vehicle in the parking area; the guide points of the parking area are determined based on the second dimensional features. That is, the present disclosure determines the essential features that have a causal relationship with the guide points of the parking area from the image of the parking area, and predicts the guide points of the parking area based only on the essential features, thereby achieving a stable prediction effect and solving the technical problem of low efficiency in predicting the guide points of the parking lot.

[0034] The above method of this embodiment is further described in detail below.

[0035] As an optional implementation, the method further includes: multiple features in the second dimensional feature are independent of each other.

[0036] In this embodiment, in order to avoid mutual influence, all features in the first dimensional features need to be independent of each other. Therefore, multiple features in the second dimensional features extracted from the first dimensional features are also independent of each other, thereby avoiding the association of false features to a certain extent.

[0037] Optionally, using parking lot images, on the basis of deep learning, features can be regarded as variables, and a set of sample weights can be found in the deep learning network by the method of interference variable balance, so that all variables can be independent of each other, that is, any variable is selected as the target variable, and the distribution of the target variable does not change with the change of the values ​​of other variables.

[0038] As an optional implementation, the method also includes: extracting the second dimensional feature from the first dimensional feature includes: mapping the first dimensional feature to the first target dimensional space, wherein the dimension corresponding to the first target dimensional space is higher than the first dimensional threshold; in the first dimensional feature in the first target dimensional space, multiple features that have a causal relationship with the guide points of the parking area and are independent of each other are determined as second dimensional features.

[0039] In this embodiment, the first dimensional feature can be mapped to the first target dimensional space through a kernel function, and then in the first target dimensional space, multiple first dimensional features that have a causal relationship with the guide points of the parking area and are independent of each other are determined as second dimensional features, wherein the first target dimensional space can be a high-dimensional space.

[0040] Optionally, the first dimensional feature is mapped to a high-dimensional space through a kernel function. After the kernel function mapping, the feature map dimension of the first dimensional feature is expanded to infinite dimensions, which makes it impossible to calculate the correlation between variables of each dimension.

[0041] Optionally, the kernel function can map the data to a high-dimensional space, and the learning algorithm can be implemented by only calculating k(x, y) without the implicit feature mapping function F. However, this convenience may cause some problems. For example, if a nonlinear classification algorithm is to be implemented and the data set is large, calculating the kernel function will cost more time and space. Therefore, in the case of nonlinear kernel functions and large data sets, the computational cost needs to be controlled.

[0042] As an optional implementation, the method also includes: mapping the first dimensional feature to the first target dimensional space includes: mapping the first dimensional feature to the second target dimensional space, wherein the dimension corresponding to the second target dimensional space is lower than the second dimensional threshold; based on the inner product of two features in the first dimensional feature in the second target dimensional space, mapping the first dimensional feature in the second target dimensional space to the first target dimensional space.

[0043] In this embodiment, the first dimensional features can be first mapped to the second target dimensional space by using the random Fourier feature method, and then the first dimensional features in the second target dimensional space are mapped to the first target dimensional space based on the inner product of two features in the first dimensional features in the second target dimensional space, wherein the second target dimensional space can be a low-dimensional space.

[0044] Optionally, the second target dimensional space may be a low-dimensional random feature space or a low-dimensional Euclidean space, which is not specifically limited here.

[0045] Optionally, the inner product of two random features in the first target dimensional space is approximately equal to a kernel function, and the kernel function can map the first dimensional feature into the first target dimensional space.

[0046] Optionally, the random Fourier feature method mainly considers the estimation of shift-invariant kernels, such as Gaussian kernel functions, Laplace kernel functions, etc., and uses linear learning methods to approximate nonlinear learning algorithms. The random Fourier feature method maps two original data x and y to a low-dimensional Euclidean space (equivalent to the mapping function) through an explicit mapping function Z.

[0047] As an optional implementation, the method further includes: mapping the first dimensional feature to a second target dimensional space, including: performing a random Fourier feature transform on the first dimensional feature to map the first dimensional feature to the second target dimensional space.

[0048] In this embodiment, a random Fourier feature transform is performed on the first dimensional feature based on a random Fourier feature method, the first dimensional feature is mapped to a second target dimensional space, and is expanded in the sample dimension to ensure that the first dimensional feature is strictly independent.

[0049] Optionally, the random Fourier feature method has excellent properties in approximating kernel functions and measuring feature independence, and a linear learning method can be applied to approximate a nonlinear learning algorithm, thereby avoiding the need for more time and space costs when implementing a nonlinear classification algorithm and the data set is relatively large.

[0050] As an optional implementation, the method also includes: in the first dimensional features in the first target dimensional space, multiple features that have a causal relationship with the guide points of the parking area and are independent of each other are determined as second dimensional features, including: obtaining multiple sample data corresponding to the features in the first dimensional features in the first target dimensional space, wherein each sample data includes the value of the corresponding feature, and the value is used to indicate whether the position point corresponding to the feature in the parking area is a guide point or not a guide point; determining the second dimensional feature based on multiple sample data.

[0051] In this embodiment, the second dimensional feature is determined based on multiple sample data corresponding to the feature in the first dimensional feature in the first target dimensional space, wherein the position point corresponding to the feature in the first dimensional feature can be a discrete binary variable, and the sample data can be the value corresponding to the position point of the feature. When the sample data value is 1, it indicates that the position point corresponding to the feature in the parking area is a guide point, and when the sample data value is 0, it indicates that the position point corresponding to the feature in the parking area is not a guide point.

[0052] For example, the position point corresponding to a certain feature in the parking area can be regarded as variable A, and the sample data corresponding to the feature can be the value of variable A (0 or 1), indicating whether the position point corresponding to the feature in the parking area is a guide point, that is, when A=1, it means that the position point corresponding to the feature in the parking area is a guide point, and when A=0, it means that the position point corresponding to the feature in the parking area is not a guide point.

[0053] As an optional implementation, the method also includes: determining a second dimensional feature based on multiple sample data, including: in response to the feature distribution of the first feature being the same under the value of each sample data, determining the feature corresponding to each sample data and the first feature as features in the second dimensional feature, wherein the first feature is a feature in the first dimensional feature in the first target dimensional space except the feature corresponding to each sample data.

[0054] In this embodiment, based on the different values ​​of sample data, when the feature distribution of the first feature is the same, the feature corresponding to the sample data and the first feature can be used as features in the second dimensional feature, wherein the first feature can be a feature other than irrelevant features and false features in the first dimensional data features, that is, a non-interference feature.

[0055] For example, the first feature can be regarded as a non-interfering variable B. When A=0 and A=1, the distribution of variable B is the same, that is, D(B|A=0)=D(B|A=1), where D represents the variable distribution.

[0056] As an optional embodiment, the method also includes: in response to the feature distribution of the second feature being the same under each sample data value, determining whether the feature distribution of the first feature is the same under each sample data value, wherein the second feature is a feature associated with the feature corresponding to each sample data in the first dimensional feature in the first target dimensional space.

[0057] In this embodiment, when the feature distribution of the second feature has the same value in the sample data, it is determined whether the feature distribution of the first feature is the same under each value of the sample data, wherein the second feature may be an interference feature.

[0058] For example, the second feature is regarded as the interference variable C. When A=0 and A=1, the distribution of the interference variable C is the same (that is, D(C|A=0)=D(C|A=1), where D represents the variable distribution). Therefore, by further judging whether D(B|A=0) and D(B|A=1) are the same, it can be concluded whether A has a causal relationship with B.

[0059] As an optional implementation, the method further includes: based on the weight of each sample data, determining that the feature distribution of the second feature is the same under the value of each sample data.

[0060] In this embodiment, different weights are assigned to each sample data, so that when the features in the first dimension features in the first target dimensional space correspond to different sample data, the feature distributions of the second features are the same.

[0061] For example, based on the weight assigned to each sample data, when A=0 and A=1, the distribution of the interference variable C is the same, that is, D(C|A=0)=D(C|A=1), where D represents the variable distribution.

[0062] The data processing method of the embodiment of the present disclosure is introduced below from the client side.

[0063] Figure 2 FIG. 1 is a flow chart of another data processing method according to an embodiment of the present disclosure. Figure 2 As shown, the method may include the following steps:

[0064] Step S202, acquiring an image of the parking area.

[0065] In the technical solution provided in the above step S202 of the present disclosure, an image of the parking area is acquired and the image is input into the guidance point prediction model, wherein the image of the parking area may be an image of the parking lot.

[0066] Optionally, the image of the parking area can be collected on-site, or obtained from images uploaded by netizens using navigation software, which is not specifically limited here.

[0067] Step S204, inputting the image into a guide point prediction model, wherein the guide point prediction model is obtained based on deep stable learning training, and is used to extract second dimensional features from first dimensional features of the image, and determine the guide points of the parking area based on the second dimensional features, wherein there is a causal relationship between the second dimensional features and the guide points of the parking area, and the guide points are used to guide vehicles in the parking area.

[0068] In the technical solution provided in the above step S204 of the present disclosure, the image is input into the guide point prediction model, and based on the image of the parking area, first dimensional features are extracted in the parking lot scene, and then features with a causal relationship with the guide points of the parking area are extracted from the first dimensional features and used as second dimensional features. Finally, the guide points of the parking area are determined based on the second dimensional features.

[0069] Step S206, obtaining the guidance point output by the guidance point prediction model.

[0070] In the technical solution provided in the above step S206 of the present disclosure, the client inputs the image of the parking area into the guidance point prediction model, thereby obtaining the guidance point of the parking area predicted by the guidance point prediction model.

[0071] Through the above steps S202 to S206, an image of the parking area is obtained; the image is input into the guide point prediction model, wherein the guide point prediction model is obtained based on deep stable learning training, and is used to extract second dimensional features from first dimensional features of the image, and determine the guide points of the parking area based on the second dimensional features, wherein the second dimensional features and the guide points of the parking area have a causal relationship, and the guide points are used to guide vehicles in the parking area; the guide points output by the guide point prediction model are obtained to achieve the purpose of improving the scene generalization ability of the causal inference model of the parking lot guide points, thereby solving the technical problem of low efficiency in predicting the guide points of the parking lot.

[0072] The above technical solution of the embodiment of the present disclosure is further introduced below with examples in combination with preferred embodiments.

[0073] Driving accounts for an increasingly larger proportion of national travel modes, so the "difficult parking" scene occurs frequently. The displayed coordinates of the parking lot can only represent the geometric position of the parking lot on the map, and the guide point determines the end point of the planned driving route. Therefore, the construction of the parking lot's guide points is crucial to the user's driving travel experience.

[0074] Parking is a common need that most drivers have when traveling, and parking lots are very important for drivers. However, in some scenarios, the parking lot and the destination are not close together, such as shopping malls, scenic spots, office buildings, etc., especially underground parking lots, where the entrance is often difficult to find. In this case, parking lot guidance points can guide users to the entrance or exit of the parking lot, solving the problem of users having to search twice because they cannot find the entrance to the parking lot near the destination. Parking lot guidance points can effectively improve the parking experience of drivers at the destination and provide users with convenient travel services.

[0075] In related technologies, developers can use models to predict parking guide points based on deep learning methods. Figure 3 is a schematic diagram of predicting parking lot guide points based on a model in the related technology of the present disclosure, such as Figure 3 As shown in , it can include the following two stages: In the first stage, the deep learning method is used to model the road network, trajectory, real-life image and other data into features, which are input into the model to let the model predict the parking lot guide point; In the second stage, the scheme currently in use is causal inference, which is mainly divided into two parts: generating candidate points and data verification. First, the guide point candidate points are generated according to the features of the parking lot entrance and exit, trajectory clustering points, and externally acquired data to obtain the prior probability P, that is, assuming that the candidate point is the guide point. After data processing and feature processing, the probability of the current point being taken as the end point of the trajectory is calculated. If the probability value is greater than a certain threshold, it is considered that the guide point is OK. Then, among several candidate points, the probability of reaching the non-current guide point from the same starting point is calculated, that is, the observation probability. The lower the observation probability, the more likely the low-probability event occurs, and the more likely it is to overturn the proposition that the current candidate point is the guide point (the original hypothesis). Finally, the multi-source features are fused. Data verification is cross-validation based on other data, and the parking lot guide point is inferred based on the probability of the candidate point.

[0076] At present, although deep learning methods have been applied to the construction of parking guide points and have achieved great success in many scenarios, and deep learning models are widely used, they lack interpretability and stability. Once the model predicts an error in the guide point, it may bring extremely bad travel experience to users. The main reason is that the current deep learning method is association-driven. Deep learning models are mostly black box models, and do not distinguish between causal associations and false associations in the data. As far as current artificial intelligence technology is concerned, due to the lack of interpretability, users of the model cannot understand the reasons why the model makes predictions and decision results; on the other hand, if the model cannot maintain sufficiently stable high performance in various scenarios, we cannot fully trust these models. Although causal inference can explain the causal relationship in the data, due to its limited computational complexity, the various types of parking lots, and the complex environment, the causal inference model of parking guide points has poor scene generalization ability and cannot be applied to all parking lots.

[0077] In view of the above situation, this embodiment proposes a method for constructing parking lot guide points based on a deep stable learning method. The method of this embodiment is further introduced below.

[0078] Stable learning refers to learning a model using data whose training data distribution is distribution 1, so that the prediction accuracy of the model on any possible test data distribution 1, distribution 2, ..., distribution n is high and similar. If we know in advance that the distribution of the test data is distribution 1, then the problem is an ideal independent and identically distributed (IID) sample learning problem. If we know in advance that the distribution of the test data is distribution n, we may use transfer learning to solve it. However, in practical problems where it is difficult to predict the distribution of future test data, the desired effect is that the model has a stable prediction effect on test data with unknown distribution.

[0079] The basic idea of ​​deep stable learning is to extract the essential features of different categories, remove irrelevant features and false associations, and make predictions based only on essential features (features that are causally related to labels).

[0080] Therefore, using parking lot images, on the basis of deep learning, the method of interfering with variable balance is used to enable the neural network model to infer causal relationships, and a set of sample weights is found in the deep learning network so that all variables can be independent of each other, that is, any variable is selected as the target variable, and the distribution of the target variable does not change with the change of the values ​​of other variables.

[0081] Causality is of great significance for feature selection in deep learning. One of the biggest goals of studying causality is to find the true causal relationship between things, remove those mixed pseudo-causal relationships, and avoid introducing additional features. Causality learning refers to learning variables from data, which is equivalent to directly extracting a more detailed coarse-grained model of the real world.

[0082] Determine the causal relationship between variable A and variable B (there is an interfering variable C). If variable A is a discrete binary variable (taking the value as 0 or 1), divide the total sample into two groups according to the value of A: A=0 or A=1 (yes or no), and assign different weights to each sample so that the distribution of the interfering variable C is the same when A=0 and A=1 (that is, D(C|A=0)=D(C|A=1), where D represents the variable distribution). At this time, judging whether D(B|A=0) and D(B|A=1) are the same can determine whether A has a causal relationship with B.

[0083] Figure 4 The present invention is a flowchart of using a deep stable learning method to predict parking lot guide points based on a model according to an embodiment of the present disclosure. The method may include:

[0084] Step S401: input the parking lot image into the model.

[0085] Get an image of a parking lot and feed it into the model.

[0086] Step S402, extracting features.

[0087] In the parking lot scene, features are extracted, where the features may include: road data such as road grade, length, width, etc., image data such as parking lot signs, parking poles, parking spaces, etc., trajectory data and positioning data describing user behavior, and geographic location data (point of information, referred to as POI) of parking lot entrances and exits.

[0088] Step S403, processing features.

[0089] Based on the extracted features, the features are processed, for example, irrelevant features and false features such as parking lot exits are removed.

[0090] Step S404: Fourier transform.

[0091] The random Fourier feature method is used to map the original features into a high-dimensional space, expand them in the sample dimension, and eliminate the linear correlation between the new features, thereby ensuring that the original features are strictly independent.

[0092] Since there are complex dependencies between the various dimensional features of deep networks, simply removing the linear correlation between variables is not enough to completely eliminate the false associations between irrelevant features and labels. In related technologies, the kernel method is used to map the features to a high-dimensional space for processing, but after the kernel function mapping, the feature map dimension of the original features is expanded to infinite dimensions, making it impossible to calculate the correlation between the various dimensional variables. The random Fourier feature method has excellent properties in approximating the kernel function and measuring feature independence.

[0093] Optionally, the random Fourier feature (RFF) method mainly considers the estimation of shift-invariant kernels, such as Gaussian kernel functions, Laplace kernel functions, etc., and applies linear learning methods to approximate nonlinear learning algorithms. The random Fourier feature maps two original data x and y to a low-dimensional Euclidean space (relative to the mapping function F) through an explicit mapping function Z.

[0094] Optionally, the features are independent in all dimensions after sample reweighting, so that the classifier is more likely to focus on features related to the parking lot guide point when predicting whether the parking lot guide point exists.

[0095] Step S405: output the prediction result of the parking lot guidance point.

[0096] Based on the parking lot guidance points predicted by the classifier, the prediction results are output.

[0097] This solution can bring the following beneficial effects: the parking lot guidance point can help users quickly find the parking lot entrance, avoid detours, save parking time, and effectively improve the user's driving destination experience.

[0098] The present invention obtains parking lot images, and then uses the parking lot images, based on deep learning, to enable the neural network model to infer causal relationships through the method of interfering variable balance, and finds a set of sample weights in the deep learning network so that all variables can be independent of each other. In order to completely eliminate the false association between irrelevant features and labels, the original features are mapped to a high-dimensional space using random Fourier features, and the sample dimensions are expanded to eliminate the linear correlation between new features, thereby ensuring that the original features are strictly independent, thereby achieving the purpose of improving the scene generalization ability of the causal inference model of parking lot guide points, and solving the technical problem of low efficiency in predicting parking lot guide points.

[0099] The present disclosure also provides a method for executing Figure 1 The data processing apparatus of the illustrated embodiment.

[0100] Figure 5 Schematic diagram of a data processing device according to an embodiment of the present disclosure. Figure 5As shown, the data processing device 500 may include: a first acquiring unit 501 , an extracting unit 502 , and a determining unit 503 .

[0101] The first acquisition unit 501 is used to acquire a first dimensional feature of an image of a parking area.

[0102] The extraction unit 502 is used to extract a second dimensional feature from the first dimensional feature, wherein the second dimensional feature has a causal relationship with a guide point in the parking area, and the guide point is used to guide the vehicle in the parking area.

[0103] The determining unit 503 is configured to determine a guide point of the parking area based on the second dimensional feature.

[0104] Optionally, multiple features in the second dimensional features are independent of each other.

[0105] Optionally, the extraction unit 502 includes: a mapping module, used to map the first dimensional feature to the first target dimensional space, wherein the dimension corresponding to the first target dimensional space is higher than the first dimensional threshold; a determination module, used to determine, among the first dimensional features in the first target dimensional space, multiple features that have a causal relationship with the guide points of the parking area and are independent of each other as second dimensional features.

[0106] Optionally, the mapping module includes: a first mapping submodule, used to map the first dimensional feature to the second target dimensional space, wherein the dimension corresponding to the second target dimensional space is lower than the second dimensional threshold; a second mapping submodule, used to map the first dimensional feature in the second target dimensional space to the first target dimensional space based on the inner product of two features in the first dimensional feature in the second target dimensional space.

[0107] Optionally, the first mapping submodule is used to map the first dimensional feature into the second target dimensional space by the following steps: performing a random Fourier feature transform on the first dimensional feature to map the first dimensional feature into the second target dimensional space.

[0108] Optionally, the determination module includes: an acquisition submodule, used to acquire multiple sample data corresponding to features in the first dimensional features in the first target dimensional space, wherein each sample data includes a value of the corresponding feature, and the value is used to indicate whether a position point corresponding to the feature in the parking area is a guide point or not; a determination submodule, used to determine the second dimensional features based on multiple sample data.

[0109] Optionally, the determination submodule is used to determine the second dimensional feature based on multiple sample data through the following steps: in response to the feature distribution of the first feature being the same under the value of each sample data, the feature corresponding to each sample data and the first feature are determined as features in the second dimensional feature, wherein the first feature is a feature in the first dimensional feature in the first target dimensional space except the feature corresponding to each sample data.

[0110] Optionally, in response to the fact that the feature distribution of the second feature is the same under each value of the sample data, it is determined whether the feature distribution of the first feature is the same under each value of the sample data, wherein the second feature is a feature associated with the feature corresponding to each sample data in the first dimensional feature in the first target dimensional space.

[0111] Optionally, based on the weight of each sample data, it is determined that the feature distribution of the second feature is the same under the value of each sample data.

[0112] In the data processing device of this embodiment, a first acquisition unit is used to acquire a first dimensional feature of an image of a parking area; an extraction unit is used to extract a second dimensional feature from the first dimensional feature, wherein the second dimensional feature has a causal relationship with a guide point of the parking area, and the guide point is used to guide the vehicle in the parking area; a determination unit is used to determine the guide point of the parking area based on the second dimensional feature, thereby solving the technical problem of low efficiency in predicting the guide point of the parking lot.

[0113] The present disclosure also provides a method for executing Figure 2 The data processing apparatus of the illustrated embodiment.

[0114] Figure 6 is a schematic diagram of another data processing device according to an embodiment of the present disclosure. Figure 6 As shown, the data processing device 600 may include: a second acquisition unit 601 , an input unit 602 , and a third acquisition unit 603 .

[0115] The second acquisition unit 601 is used to acquire an image of the parking area.

[0116] The input unit 602 is used to input the image into the guide point prediction model, wherein the guide point prediction model is obtained based on deep stable learning training, and is used to extract second dimensional features from the first dimensional features of the image, and determine the guide points of the parking area based on the second dimensional features, wherein the second dimensional features and the guide points of the parking area have a causal relationship, and the guide points are used to guide vehicles in the parking area.

[0117] The third acquisition unit 603 is used to acquire the guidance point output by the guidance point prediction model.

[0118] In the data processing device of this embodiment, an image of the parking area is acquired by a second acquisition unit; the input unit inputs the image into a guide point prediction model, wherein the guide point prediction model is obtained based on deep stable learning training, and is used to extract second dimensional features from first dimensional features of the image, and determine the guide points of the parking area based on the second dimensional features, wherein there is a causal relationship between the second dimensional features and the guide points of the parking area, and the guide points are used to guide vehicles in the parking area; the third acquisition unit acquires the guide points output by the guide point prediction model, thereby solving the technical problem of low efficiency in predicting the guide points of the parking lot.

[0119] In the technical solution disclosed herein, the acquisition, storage and application of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0120] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0121] An embodiment of the present disclosure provides an electronic device, which may include: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the data processing method of the embodiment of the present disclosure.

[0122] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0123] According to an embodiment of the present disclosure, the present disclosure further provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute the data processing method of the embodiment of the present disclosure.

[0124] Optionally, in this embodiment, the non-volatile storage medium may be configured to store a computer program for performing the following steps:

[0125] S1, obtaining the first dimension feature of the image of the parking area;

[0126] S2, extracting a second dimensional feature from the first dimensional feature, wherein the second dimensional feature has a causal relationship with a guide point of the parking area, and the guide point is used to guide the vehicle in the parking area;

[0127] S3, determining a guide point of the parking area based on the second dimensional feature.

[0128] Optionally, in this embodiment, the non-volatile storage medium may also be configured to store a computer program for performing the following steps:

[0129] S1, acquiring an image of the parking area;

[0130] S2, inputting the image into a guide point prediction model, wherein the guide point prediction model is obtained based on deep stable learning training and is used to extract second dimensional features from first dimensional features of the image, and determine the guide point of the parking area based on the second dimensional features, wherein the second dimensional features and the guide point of the parking area have a causal relationship, and the guide point is used to guide the vehicle in the parking area;

[0131] S3, obtaining the guiding point output by the guiding point prediction model.

[0132] Optionally, in this embodiment, the non-transitory computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the above. More specific examples of readable storage media may include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0133] According to an embodiment of the present disclosure, the present disclosure further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0134] S1, obtaining the first dimension feature of the image of the parking area;

[0135] S2, extracting a second dimensional feature from the first dimensional feature, wherein the second dimensional feature has a causal relationship with a guide point of the parking area, and the guide point is used to guide the vehicle in the parking area;

[0136] S3, determining a guide point of the parking area based on the second dimensional feature.

[0137] When the computer program is executed by the processor, the following steps are also implemented:

[0138] S1, acquiring an image of the parking area;

[0139] S2, inputting the image into a guide point prediction model, wherein the guide point prediction model is obtained based on deep stable learning training and is used to extract second dimensional features from first dimensional features of the image, and determine the guide point of the parking area based on the second dimensional features, wherein the second dimensional features and the guide point of the parking area have a causal relationship, and the guide point is used to guide the vehicle in the parking area;

[0140] S3, obtaining the guiding point output by the guiding point prediction model.

[0141] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0142] Figure 7 A schematic block diagram of an example electronic device 700 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0143] like Figure 7 As shown, the device 700 includes a computing unit 701, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 into a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the device 700 can also be stored. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0144] A number of components in the device 700 are connected to the I / O interface 705, including: an input unit 706, such as a keyboard, a mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a storage unit 708, such as a disk, an optical disk, etc.; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 709 allows the device 700 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0145] The computing unit 701 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 701 performs the various methods and processes described above, such as data processing methods. For example, in some embodiments, the data processing method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 708. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 700 via ROM 702 and / or communication unit 709. When the computer program is loaded into RAM 703 and executed by the computing unit 701, one or more steps of the data processing method described above may be performed. Alternatively, in other embodiments, the computing unit 701 may be configured to perform the data processing method in any other appropriate manner (e.g., by means of firmware).

[0146] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0147] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0148] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0149] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0150] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0151] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0152] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0153] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A data processing method, comprising: Obtaining first dimension features of the image of the parking area; Extracting a second dimensional feature from the first dimensional feature, wherein the second dimensional feature has a causal relationship with a guide point of the parking area, and the guide point is used to guide a vehicle in the parking area; determining a guide point of the parking area based on the second dimensional feature; Among them, extracting the second dimensional feature from the first dimensional feature includes: obtaining a plurality of sample data corresponding to the features in the first dimensional feature in the first target dimensional space, wherein each of the sample data includes a value of the corresponding feature, and the value is used to indicate whether the position point corresponding to the feature in the parking area is the guide point or not the guide point; based on the plurality of sample data, determining the second dimensional feature.

2. The method according to claim 1, wherein: Multiple features in the second dimensional features are independent of each other.

3. The method according to claim 1, wherein: Extracting the second dimensional feature from the first dimensional feature includes: Mapping the first dimensional feature to the first target dimensional space, wherein the dimension corresponding to the first target dimensional space is higher than a first dimensional threshold; Among the first dimensional features in the first target dimensional space, a plurality of features that have the causal relationship with the guidance point of the parking area and are independent of each other are determined as the second dimensional features.

4. The method according to claim 3, wherein: Mapping the first dimensional feature to a first target dimensional space includes: Mapping the first dimensional feature to a second target dimensional space, wherein the dimension corresponding to the second target dimensional space is lower than a second dimensional threshold; Based on the inner product of two features in the first dimensional feature in the second target dimensional space, the first dimensional feature in the second target dimensional space is mapped to the first target dimensional space.

5. The method according to claim 4, wherein: Mapping the first dimensional feature to a second target dimensional space includes: Performing a random Fourier feature transform on the first dimensional feature to map the first dimensional feature into the second target dimensional space.

6. The method according to claim 1, determining the second dimensional feature based on the plurality of sample data, comprising: In response to the fact that the feature distribution of the first feature is the same under each value of the sample data, the feature corresponding to each sample data and the first feature are determined as features in the second dimensional features, wherein the first feature is a feature of the first dimensional features in the first target dimensional space except the feature corresponding to each sample data.

7. The method according to claim 6, further comprising: In response to the fact that the feature distribution of the second feature is the same under each value of the sample data, it is determined whether the feature distribution of the first feature is the same under each value of the sample data, wherein the second feature is a feature associated with the feature corresponding to each sample data in the first dimensional feature in the first target dimensional space.

8. The method according to claim 7, further comprising: Based on the weight of each of the sample data, it is determined that the feature distribution of the second feature is the same under the value of each of the sample data.

9. A data processing method, comprising: Get an image of the parking area; The image is input into a guide point prediction model, wherein the guide point prediction model is obtained based on deep stable learning training and is used to extract second dimensional features from first dimensional features of the image, and determine the guide point of the parking area based on the second dimensional features, wherein there is a causal relationship between the second dimensional features and the guide points of the parking area, the guide points are used to guide vehicles in the parking area, and the second dimensional features are determined based on a plurality of sample data corresponding to features in the first dimensional features in a first target dimensional space, each of the sample data includes a value of the corresponding feature, and the value is used to indicate whether a position point corresponding to the feature in the parking area is the guide point or not; The guide point output by the guide point prediction model is obtained.

10. A data processing device, comprising: A first acquisition unit, used to acquire a first dimensional feature of the image of the parking area; an extraction unit, configured to extract a second dimensional feature from the first dimensional feature, wherein the second dimensional feature has a causal relationship with a guide point in the parking area, and the guide point is used to guide a vehicle in the parking area; a determining unit, configured to determine a guide point of the parking area based on the second dimensional feature; Wherein, the extraction unit is used to extract the second dimensional feature through the following steps: obtaining multiple sample data corresponding to the features in the first dimensional feature in the first target dimensional space, wherein each of the sample data includes a corresponding value of the feature, and the value is used to indicate whether the position point corresponding to the feature in the parking area is the guide point or not the guide point; based on the multiple sample data, determining the second dimensional feature.

11. A data processing device, comprising: A second acquisition unit, used to acquire an image of the parking area; An input unit, used for inputting the image into a guide point prediction model, wherein the guide point prediction model is obtained based on deep stable learning training, and is used for extracting second dimensional features from first dimensional features of the image, and determining the guide point of the parking area based on the second dimensional features, wherein there is a causal relationship between the second dimensional features and the guide points of the parking area, the guide points are used to guide vehicles in the parking area, and the second dimensional features are determined based on a plurality of sample data corresponding to features in the first dimensional features in a first target dimensional space, each of the sample data includes a value of the corresponding feature, and the value is used to indicate whether a position point in the parking area corresponding to the feature is the guide point or not; The third acquisition unit is used to acquire the guidance point output by the guidance point prediction model.

12. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 9.

13. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-9.

14. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 9.

Images