Visual parking space correction method and system in complex environment

Abstract

The invention relates to the technical field of visual parking space correction, and discloses a visual parking space correction method and system in a complex environment, and the method comprises the steps: obtaining a bird's-eye view data set of different environment index parameters and a data set with a weather label; constructing a visual parking space recognition model, and training the visual parking space recognition model to obtain a model training result; the method comprises the following steps: marking external parameters of a vehicle around-view camera under different driving states of a vehicle, and constructing an external parameter calibration parameter library; and through the trained visual parking space identification model, correcting all-round perception data, and identifying an accurate parking space image. According to the method, the robustness and accuracy of parking space recognition in a complex environment are improved, the method can adapt to different brightness, exposure, contrast, weather, illumination intensity and other conditions, and the influence of the vehicle on parking space recognition in different states can be better dealt with through construction of an external parameter calibration parameter library and targeted training.

Description

Technical Field

[0001] The present invention relates to the technical field of visual parking space correction, and in particular to a method and system for visual parking space correction in a complex environment. Background Art

[0002] In autonomous driving and automatic parking systems, accurate parking space identification is one of the key technologies to achieve automated parking. However, different environmental factors such as slope, weather, and light intensity will affect the camera's visual perception ability, thereby affecting the accuracy of parking space corner detection. Currently, there is an urgent need for a visual parking space corner correction solution in complex environments that can effectively correct the bird's-eye view of parking spaces in different environments. Summary of the invention

[0003] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0004] In view of the above existing problems, the present invention is proposed. Therefore, the present invention provides a visual parking space correction method in a complex environment to solve the above problems.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] In a first aspect, the present invention provides a method for visual parking space correction in a complex environment, comprising:

[0007] Obtain a bird's-eye view dataset of different environmental indicator parameters and a dataset with weather labels;

[0008] Constructing a visual parking space recognition model, and performing a first training and a second training on the visual parking space recognition model based on the bird's-eye view data set and the weather label data set, respectively, to obtain a model training result;

[0009] Under different driving conditions of the vehicle, the external parameters of the vehicle surround view camera are annotated to build an external parameter calibration parameter library;

[0010] Based on the external parameter calibration parameter library and the model training results, the surround perception data is corrected through the trained visual parking space recognition model to identify an accurate parking space image.

[0011] As a preferred solution of the visual parking space correction method under complex environment described in the present invention, the bird's-eye view data set for obtaining different environmental index parameters includes:

[0012] According to the preset indicator ratio, different environmental indicator parameters are collected during the parking process to obtain an initial bird's-eye view data set;

[0013] Image processing is performed based on the bird's-eye view data set to produce a correction test set for different environmental index parameters under different weather conditions.

[0014] As a preferred solution of the visual parking space correction method under complex environment described in the present invention, obtaining the model training result includes:

[0015] The model training result includes a first training result obtained by training the visual parking space recognition model once and a second training result obtained by training the visual parking space recognition model twice;

[0016] The step of training the visual parking space recognition model includes:

[0017] Constructing the visual parking space recognition model through a deep learning algorithm, inputting the bird's-eye view data set into the visual parking space recognition model for training, and testing the model through the modified test set;

[0018] According to the recall rate of the visual parking space recognition model for different correction test sets, a correction scheme with the highest recall rate is obtained as a correction array of surround view parameters for environmental indicator parameters.

[0019] As a preferred solution of the visual parking space correction method under a complex environment described in the present invention, the secondary training of the visual parking space recognition model includes:

[0020] According to the obtained data set with weather labels, the data set is divided;

[0021] The visual parking space recognition model is trained and verified, and model parameters are adjusted to obtain a visual parking space recognition model that recognizes surrounding weather.

[0022] As a preferred solution of the visual parking space correction method under complex environment described in the present invention, the construction of the external parameter calibration parameter library includes:

[0023] Simulate different driving conditions of the vehicle, set reference points at different locations around the vehicle and collect surround images, and use the collected image data to calibrate the intrinsic parameters of each camera;

[0024] Based on the calibrated internal parameters, feature matching is performed on the reference point surround images captured by different cameras to determine the positions of the reference points in different images;

[0025] According to the results of feature matching, the extrinsic parameters of each camera under different driving conditions of the vehicle are calculated and updated, the extrinsic parameter values ​​of each calibration and the vehicle status are recorded, the extrinsic parameter matrix is ​​generated, and the extrinsic parameter calibration parameter library is constructed.

[0026] As a preferred solution of the visual parking space correction method in a complex environment described in the present invention, the correcting of surround perception data and identifying an accurate parking space image includes:

[0027] Get current vehicle operation data;

[0028] According to the real-time surround view image of the vehicle, the current weather and light intensity are identified through the visual parking space recognition model, and the corresponding surround view parameter correction data is selected to adjust the parameters of the surround view image;

[0029] According to the vehicle operation data, a corresponding extrinsic parameter matrix is ​​selected from the extrinsic parameter calibration database, and the surround perception data is corrected to generate a bird's-eye view of an accurate parking space.

[0030] As a preferred solution of the visual parking space correction method under complex environment described in the present invention, the acquisition of current vehicle operation data includes:

[0031] The vehicle's brake pressure, speed and steering wheel torque are obtained through the vehicle body CAN to calculate the current road slope;

[0032] The light intensity of the vehicle's surrounding environment is collected in real time through the light sensor in the adaptive headlights;

[0033] The turning radius of the vehicle is obtained according to the steering wheel torque.

[0034] In a second aspect, the present invention provides a visual parking space correction system in a complex environment, comprising:

[0035] The acquisition module is used to obtain the bird's-eye view data set of different environmental indicator parameters and the weather label data set;

[0036] A model building module, used for building a visual parking space recognition model, and performing a primary training and a secondary training on the visual parking space recognition model based on the bird's-eye view data set and the weather label data set, respectively, to obtain a model training result;

[0037] The parameter library construction module is used to annotate the external parameters of the vehicle surround view camera under different driving conditions of the vehicle and build an external parameter calibration parameter library;

[0038] The recognition module is used to correct the surround perception data and recognize the accurate parking space image through the trained visual parking space recognition model based on the external parameter calibration parameter library and the model training results.

[0039] In a third aspect, the present invention provides an electronic device, comprising:

[0040] Memory and processor;

[0041] The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the visual parking space correction method under a complex environment are implemented.

[0042] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the visual parking space correction method in a complex environment.

[0043] Compared with the prior art, the present invention has the following beneficial effects: the present invention improves the robustness and accuracy of parking space recognition in complex environments, can adapt to different conditions such as brightness, exposure, contrast, weather and light intensity, and can better cope with the impact of different vehicle states on parking space recognition by constructing an external parameter calibration parameter library and targeted training. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0045] Figure 1 The figure is a schematic diagram of the overall process of a method for correcting a visual parking space in a complex environment according to an embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0047] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0048] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0049] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0050] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0051] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0052] Reference Figure 1 , is an embodiment of the present invention, and provides a visual parking space correction method in a complex environment, comprising:

[0053] S100, obtains a bird’s-eye view dataset of different environmental indicator parameters and a dataset with weather labels;

[0054] S200, constructing a visual parking space recognition model, and performing a primary training and a secondary training on the visual parking space recognition model based on a bird's-eye view data set and a weather label data set, respectively, to obtain a model training result;

[0055] S300, annotating the external parameters of the vehicle surround view camera under different driving conditions of the vehicle, and building an external parameter calibration parameter library;

[0056] S400, based on the external calibration parameter library and model training results, corrects the surround perception data through the trained visual parking space recognition model and recognizes the accurate parking space image.

[0057] It should be noted that the present invention produces a correction array of surround view parameters for different environmental indicators. Before the surround view is transmitted to the model for inference, the correction array is used to correct the surround view, and overexposure and overdarkness will be corrected to a state that can be recognized by the model;

[0058] At the same time, the vehicle surround view image is calibrated for different environments (slope, torque). When the vehicle travels on roads with different slopes, the surround view image stitching will use the corresponding external parameters of the environment for stitching, avoiding the surround view stitching misalignment caused by the slope and torque, causing the parking space line to be broken and the actual position to be inconsistent with the image position, thereby avoiding the parking space line recognition model from missing or misdetecting broken parking spaces.

[0059] In a preferred embodiment, in step S100, obtaining a bird's-eye view data set of different environmental indicator parameters includes:

[0060] Different environmental index parameters are set according to preset index ratios, and surround view image data during parking is collected to obtain an initial bird's-eye view data set;

[0061] Image processing is performed based on the bird's-eye view dataset to produce a corrected test set for different environmental indicator parameters under different weather conditions.

[0062] In an optional embodiment, the environmental index parameters include parameters such as brightness, exposure and contrast in different weather and light intensities; the preset index ratio can be natural light: overexposure: too dark: snowy day: rainy day = 4:2:2:1:1.

[0063] Specifically, according to the actual needs of the solution application, for example, the brightness ranges from 0 to 255. It is only about 30 on cloudy days or at night, and can reach 255 under direct sunlight. Multiple correction schemes can be set. For example, the brightness increases by 5 degrees on cloudy days and decreases by 5 degrees under direct sunlight, with a total of 10 gradients, and a 10-gradient test set is constructed.

[0064] In an optional implementation, the opencv linear transformation method and function parameter adjustment can be used to produce a bird's-eye view data set with different parameters such as brightness, exposure and contrast, and a test set after correcting the brightness, exposure and contrast for different weather and light intensities;

[0065] Surround view image data during parking is collected according to the ratio of natural light: overexposure: too dark: snowy day: rainy day = 4:2:2:1:1 to obtain an initial bird's-eye view data set;

[0066] The brightness contrast adjustment of an image belongs to the grayscale linear transformation of the image, and the formula is as follows:

[0067] y=[x-127.5*(1-B)]*k+127.5*(1+B)

[0068] Where x represents the pixel value before adjustment, y represents the pixel value after adjustment, B represents the brightness adjustment, and its value is [-1, 1]. k represents the contrast adjustment, and arctan(k) has a value of [1, 89]. Here, k = tan((45 + 44*c) / 180*pi), and c has a value of [-1, 1]. This value is usually used to set the contrast.

[0069] Specifically: when B=0, y=(x-127.5)*k+127.5, only the contrast is adjusted;

[0070] When c=0, k=1: y=x+255*B, only the brightness is adjusted.

[0071] The comvertTo function of opencv is similar to this principle. For a cv::Mat src, scr.comvertTo(dst, -1, alpha, beta) is used; by adjusting alpha and beta, its exposure and contrast can be adjusted, and a new cv::Mat src is output. If B and c are arbitrarily set parameters between [-1, 1], then:

[0072] k = tan((45+44*c) / 180*M_PI)

[0073] alpha=k

[0074] beta=127.5*(1+B)-k*127.5*(1-B)

[0075] According to the minMaxLoc function of opnecv, the maximum and minimum values ​​of the overall brightness of the image (min, max) and the maximum and minimum values ​​of the red, green and blue levels (maxVal, minVal) are calculated. According to the natural light, no processing, overexposure (maximum, small color level / overall minimum brightness-2a), too dark (maximum, small color level / overall minimum brightness+2a), snowy day (maximum, small color level / overall minimum brightness-a), rainy day (maximum, small color level / overall minimum brightness+a); a is a gradient of 5, and 25 gradients are taken from [0,125] to produce 25 test sets.

[0076] In another optional implementation, a bird's-eye view data set with different parameters such as brightness, exposure and contrast and a test set after correcting the brightness, exposure and contrast for different weather and light intensities can also be prepared by the PIL method;

[0077] Specifically, we first preprocess the images under different weather and light intensities, convert the images into RGB mode, and use the PIL function to adjust the brightness, contrast, and exposure of the images. We randomly select the adjustment parameters and generate a large number of images under different conditions.

[0078] In a preferred embodiment, in step S200, obtaining the model training result includes:

[0079] The model training result includes a first training result obtained by training the visual parking space recognition model once and a second training result obtained by training the visual parking space recognition model twice;

[0080] Among them, training the visual parking space recognition model once includes:

[0081] A visual parking space recognition model was built using a deep learning algorithm. The bird's-eye view data set was input into the visual parking space recognition model for training, and the model was tested using a modified test set.

[0082] According to the recall rate of the visual parking space recognition model for different correction test sets, the correction scheme with the highest recall rate is obtained as the surround view parameter correction array for the environmental indicator parameters.

[0083] The secondary training of the visual parking space recognition model includes:

[0084] According to the obtained data set with weather labels, the data set is divided;

[0085] The visual parking space recognition model is trained and verified, and the model parameters are adjusted to obtain a visual parking space recognition model that can recognize the surrounding weather.

[0086] In an optional implementation, a visual parking space recognition model can be constructed through yolov5 deep learning. Specifically, in order to obtain correction parameters, a parking data set is collected according to the ratio of initial natural light: overexposure: too dark: snowy day: rainy day = 4:2:2:1:1, and yolov5 model training is performed. The trained model is tested on a processed 25 gradient test set, and based on the test results, the best performing test set correction parameters are selected as image correction parameters. Different models have different reflection results. The model with the highest recall rate in a certain parameter gradient test set is the best correction scheme for the model. Specifically, the ratio of the actual number of parking spaces to the number of recognized parking spaces can be calculated. The closer to 1, the higher the recall rate.

[0087] In another optional implementation, the visual parking space recognition model can also be constructed using methods such as DMPR and CNN.

[0088] In an optional implementation, a data set with a weather label means that weather selection is completed during data collection, weather types are classified, and lighting types are distinguished for subsequent selection of a lighting parameter table.

[0089] Exemplarily, the weather is divided into cloudy, sunny, rainy, snowy, and nighttime, and the lighting types are divided into dim light, natural light, overexposure, street lights, and indoor lights.

[0090] Specifically, the surround view camera data under different light intensities and weather environments are collected and made into standard training sets; for surround view images in different environments, the proportions are adjusted according to parameters such as image brightness, exposure and contrast, and the collected surround view data are made into different test sets; the training set is used to train visual parking space recognition, and the recall rates of the models on different test sets are compared; the test set adjustment standard with the highest recall rate is selected as the surround view parameter correction array.

[0091] In a preferred embodiment, in step S300, building an external parameter calibration parameter library includes:

[0092] Simulate different driving conditions of the vehicle, set reference points at different locations around the vehicle and collect surround images to ensure that the reference points can be clearly captured by each surround camera, the collected angles cover the entire field of view, and use the collected image data to calibrate the internal parameters of each camera;

[0093] Based on the calibrated internal parameters, feature matching is performed on the reference point surround images captured by different cameras to determine the position of the reference point in different images;

[0094] According to the results of feature matching, the extrinsic parameters of each camera under different driving conditions of the vehicle are calculated and updated to reflect the changes under different conditions. The extrinsic parameter values ​​and vehicle status of each calibration are recorded to build an extrinsic parameter calibration parameter library.

[0095] In an optional embodiment, different driving states of the vehicle may include different driving conditions such as different slopes, tire pressures, and turning radii, that is, the vehicle surround view camera external parameter calibration is performed under different slope road environments and vehicle turning radius postures to obtain an external parameter calibration parameter library; specifically, the tire pressure range may be {50, 75, 100}, the road slope range may be [-30, 30] degrees, with each 10 degrees being a gradient, and the torque is divided into {-180, 0, 180}, and the vehicle external parameter data under 48 scenarios are calibrated.

[0096] In a preferred embodiment, in step S400, correcting the surround view perception data and identifying an accurate parking space image includes:

[0097] Get current vehicle operation data;

[0098] According to the real-time surround view image of the vehicle, the visual parking space recognition model is used to identify the current weather and light intensity, and the corresponding surround view parameter correction data is selected to adjust the parameters of the surround view image;

[0099] According to the vehicle operation data, the corresponding extrinsic parameter matrix is ​​selected from the extrinsic parameter calibration database, and the surround perception data is corrected to generate an accurate bird's-eye view of the parking space.

[0100] Exemplarily, first, based on the identified weather and detected light intensity, the corresponding surround view correction data is selected to adjust the parameters such as the brightness, exposure and contrast of the surround view image; then, based on the calculated road slope and vehicle turning radius, the corresponding extrinsic parameter matrix is ​​accurately selected from the extrinsic parameter calibration parameter library.

[0101] It should be noted that by correcting the surround perception data in this way, the parking spaces in the generated bird's-eye view can always maintain a relatively correct state even in complex and diverse environments, and present features suitable for accurate identification by the parking space recognition model.

[0102] In a preferred embodiment, obtaining current vehicle operation data includes:

[0103] The vehicle's brake pressure, speed and steering wheel torque are obtained through the vehicle body CAN to calculate the current road slope;

[0104] The light intensity of the vehicle's surrounding environment is collected in real time through the light sensor in the adaptive headlights;

[0105] According to the steering wheel torque, the turning radius of the vehicle is obtained.

[0106] In an optional implementation, the road slope can be directly acquired by a slope sensor installed on the vehicle;

[0107] In another optional implementation, first, the vehicle's acceleration is calculated using the vehicle wheel speed data. Combined with the throttle pressure, the theoretical acceleration of the vehicle on a flat road can be inferred. Then, the difference between the actual acceleration and the theoretical acceleration can reflect the effect of the slope.

[0108] Specifically, the weather category is identified through the previous frame of visual parking space recognition model and the vehicle optical sensor is used to obtain weather parameters and light intensity. The brake pressure, wheel speed and steering wheel torque are obtained according to the vehicle body CAN, and the current road slope and vehicle turning radius are calculated. The corresponding surround view correction array is selected according to the weather parameters and light intensity, and the surround view image is corrected. The corresponding vehicle external parameters are selected according to the road slope and the vehicle turning radius, and the corrected surround view is corrected to BEV. BEV is passed into the visual parking space recognition model to identify the parking space information and weather parameters.

[0109] It should be noted that the present invention improves the robustness and accuracy of parking space recognition in complex environments, and can adapt to different conditions such as brightness, exposure, contrast, weather and light intensity. By constructing an external parameter calibration parameter library and targeted training, it can better cope with the impact of different vehicle states on parking space recognition.

[0110] The above is a schematic scheme of a visual parking space correction method under a complex environment of this embodiment. It should be noted that the technical scheme of the visual parking space correction system under a complex environment and the technical scheme of the visual parking space correction method under a complex environment described above belong to the same concept. For details not described in detail in the technical scheme of the visual parking space correction system under a complex environment of this embodiment, please refer to the description of the technical scheme of the visual parking space correction method under a complex environment described above.

[0111] In this embodiment, the visual parking space correction system in a complex environment includes:

[0112] The acquisition module is used to obtain the bird's-eye view data set of different environmental indicator parameters and the weather label data set;

[0113] The model building module is used to build a visual parking space recognition model, and based on the bird's-eye view data set and the weather label data set, the visual parking space recognition model is trained once and twice to obtain the model training results;

[0114] The parameter library construction module is used to annotate the external parameters of the vehicle surround view camera under different driving conditions of the vehicle and build an external parameter calibration parameter library;

[0115] The recognition module is used to correct the surround perception data and identify the accurate parking space image based on the external parameter calibration parameter library and model training results through the trained visual parking space recognition model.

[0116] This embodiment further provides an electronic device, which is applicable to the case of visual parking space correction in a complex environment, and includes:

[0117] Memory and processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the visual parking space correction method in a complex environment as proposed in the above embodiment.

[0118] This embodiment further provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for correcting a visual parking space in a complex environment as proposed in the above embodiment is implemented.

[0119] The storage medium proposed in this embodiment and the method for realizing visual parking space correction in a complex environment proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0120] Through the above description of the implementation methods, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ReadOnly, Memory, ROM), random access memory (RandomAccess Memory, RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform the methods of various embodiments of the present invention.

[0121] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A visual parking space correction method in a complex environment, characterized in that: include: Obtain a bird's-eye view dataset of different environmental indicator parameters and a dataset with weather labels; Constructing a visual parking space recognition model, and performing a first training and a second training on the visual parking space recognition model based on the bird's-eye view data set and the weather label data set, respectively, to obtain a model training result; Under different driving conditions of the vehicle, the external parameters of the vehicle surround view camera are annotated to build an external parameter calibration parameter library; Based on the external parameter calibration parameter library and the model training results, the surround perception data is corrected through the trained visual parking space recognition model to identify an accurate parking space image.

2. The visual parking space correction method in a complex environment as claimed in claim 1, characterized in that: The bird's-eye view datasets for obtaining different environmental indicator parameters include: According to the preset indicator ratio, different environmental indicator parameters are collected during the parking process to obtain an initial bird's-eye view data set; Image processing is performed based on the bird's-eye view data set to produce a correction test set for different environmental index parameters under different weather conditions.

3. The visual parking space correction method in a complex environment as claimed in claim 2, characterized in that: The model training results include: The model training result includes a first training result obtained by training the visual parking space recognition model once and a second training result obtained by training the visual parking space recognition model twice; The step of training the visual parking space recognition model includes: Constructing the visual parking space recognition model through a deep learning algorithm, inputting the bird's-eye view data set into the visual parking space recognition model for training, and testing the model through the modified test set; According to the recall rate of the visual parking space recognition model for different correction test sets, a correction scheme with the highest recall rate is obtained as a correction array of surround view parameters for environmental indicator parameters.

4. The visual parking space correction method in a complex environment as claimed in claim 3, characterized in that: The secondary training of the visual parking space recognition model includes: According to the obtained data set with weather labels, the data set is divided; The visual parking space recognition model is trained and verified, and model parameters are adjusted to obtain a visual parking space recognition model that recognizes surrounding weather.

5. The visual parking space correction method in a complex environment as claimed in claim 1, characterized in that: The construction of the external parameter calibration parameter library comprises: Simulate different driving conditions of the vehicle, set reference points at different locations around the vehicle and collect surround images, and use the collected image data to calibrate the intrinsic parameters of each camera; Based on the calibrated internal parameters, feature matching is performed on the reference point surround images captured by different cameras to determine the positions of the reference points in different images; According to the results of feature matching, the extrinsic parameters of each camera under different driving conditions of the vehicle are calculated and updated, the extrinsic parameter values ​​of each calibration and the vehicle status are recorded, the extrinsic parameter matrix is ​​generated, and the extrinsic parameter calibration parameter library is constructed.

6. The visual parking space correction method in a complex environment as claimed in claim 4 or 5, characterized in that: The correcting of surround view perception data to identify an accurate parking space image includes: Get current vehicle operation data; According to the real-time surround view image of the vehicle, the current weather and light intensity are identified through the visual parking space recognition model, and the corresponding surround view parameter correction data is selected to adjust the parameters of the surround view image; According to the vehicle operation data, a corresponding extrinsic parameter matrix is ​​selected from the extrinsic parameter calibration database, and the surround perception data is corrected to generate a bird's-eye view of an accurate parking space.

7. The visual parking space correction method in a complex environment as claimed in claim 6, characterized in that: The obtaining of current vehicle operation data comprises: The vehicle's brake pressure, speed and steering wheel torque are obtained through the vehicle body CAN to calculate the current road slope; The light intensity of the vehicle's surrounding environment is collected in real time through the light sensor in the adaptive headlights; The turning radius of the vehicle is obtained according to the steering wheel torque.

8. A visual parking space correction system in a complex environment, characterized in that: include, The acquisition module is used to obtain the bird's-eye view data set of different environmental indicator parameters and the weather label data set; A model building module, used for building a visual parking space recognition model, and performing a primary training and a secondary training on the visual parking space recognition model based on the bird's-eye view data set and the weather label data set, respectively, to obtain a model training result; The parameter library construction module is used to annotate the external parameters of the vehicle surround view camera under different driving conditions of the vehicle and build an external parameter calibration parameter library; The recognition module is used to correct the surround perception data and recognize the accurate parking space image through the trained visual parking space recognition model based on the external parameter calibration parameter library and the model training results.

9. An electronic device, characterized in that: include: Memory and processor; The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the visual parking space correction method in a complex environment according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: It stores computer executable instructions, which, when executed by a processor, implement the steps of the visual parking space correction method in a complex environment as described in any one of claims 1 to 7.

Images