WO2020253010A1

WO2020253010A1 - Method and apparatus for positioning parking entrance in parking positioning, and vehicle-mounted terminal

Info

Publication number: WO2020253010A1
Application number: PCT/CN2019/113487
Authority: WO
Inventors: 姜秀宝; 施泽南; 谢国富
Original assignee: 魔门塔(苏州)科技有限公司; 北京初速度科技有限公司
Priority date: 2019-06-17
Filing date: 2019-10-26
Publication date: 2020-12-24
Also published as: CN112102646A; CN112102646B

Abstract

A method and apparatus for positioning a parking entrance in parking positioning, and a vehicle-mounted terminal. The method comprises: when it is detected that a vehicle enters an entrance of a parking lot, and it is determined that the location indicated by an initial vehicle posture of a positioning module is within a preset initialized identification region, obtaining a parking lot image collected by a camera module (S120); detecting semantic information of the parking lot image (S130); determining a first vehicle posture of the vehicle by means of a posture regression model based on the semantic information of the parking lot image and the initial vehicle posture (S140); and matching the semantic information of the parking lot image with semantic information of each location point in a preset map according to the first vehicle posture, and determining a second vehicle posture of the vehicle according to the matching result (S150), wherein the posture regression model is obtained by training according to a plurality of sample parking lot images collected in the initialized identification region in advance. The method can improve the positioning accuracy at the entrance of the parking lot.

Description

Parking lot entrance positioning method, device and vehicle-mounted terminal in parking positioning

Technical field

The present invention relates to the technical field of intelligent driving, in particular to a parking lot entrance positioning method, device and vehicle-mounted terminal in parking positioning.

Background technique

Intelligent parking technology can intelligently control vehicles to enter the parking spaces of the parking lot. Among them, when the vehicle enters the parking lot, it is necessary to accurately locate the position of the vehicle in the parking lot, so as to better control the vehicle to find a suitable parking space in the parking lot and park the car. In related technologies, the vehicle can usually be positioned based on GPS signals. However, in indoor parking lots or underground garages, GPS signals are often blocked.

In this case, precise positioning can be performed based on the match between the semantic information in the visual image and the semantic information in the map. However, this positioning method requires a more accurate initial pose as input to start. In the prior art, GPS signals at the entrance of the parking lot are usually used to determine the initial pose of the vehicle in the parking lot. However, the positioning accuracy of the GPS signal is not high enough, which results in the determined initial pose accuracy at the entrance of the parking lot is not high enough, which in turn leads to poor positioning accuracy inside the parking lot.

Summary of the invention

The invention provides a parking lot entrance positioning method, device and vehicle-mounted terminal in parking positioning, so as to improve the positioning accuracy at the entrance of the parking lot. The specific technical solution is as follows.

In the first aspect, an embodiment of the present invention discloses a parking lot entrance positioning method in parking positioning, including:

When it is detected that the vehicle has entered the entrance of the parking lot, acquiring the initial vehicle pose determined by the positioning module, and determining whether the position indicated by the initial vehicle pose is in the preset initialization recognition area;

If it is, acquire a parking lot image collected by the camera module; wherein the parking lot image is an image collected in the initial recognition area;

Detecting semantic information of the parking lot image; wherein the semantic information is information used to identify landmarks around the vehicle;

Based on the semantic information of the parking lot image and the initial vehicle pose, the first vehicle pose of the vehicle is determined by a pose regression model; wherein, the pose regression model is based on the initial recognition area A number of sample parking lot images collected inside and the corresponding sample initial vehicle pose and labeled vehicle pose are trained;

According to the first vehicle pose, the semantic information of the parking lot image is matched with the semantic information of each location point in the preset map, and the second vehicle pose of the vehicle is determined according to the matching result.

Optionally, the pose regression model is obtained by training in the following manner:

Acquiring a plurality of sample parking lot images collected in the initial recognition area, and a sample initial vehicle pose and a marked vehicle pose corresponding to each sample parking lot image;

Detect the sample semantic information of each sample parking lot image;

Based on the semantic information of each sample and the corresponding initial vehicle pose, the reference vehicle pose is determined through the model parameters in the pose regression model;

Determining the amount of difference between the reference vehicle pose and the marked vehicle pose;

When the difference amount is greater than the preset difference amount threshold, modify the model parameters, return to execute the initial vehicle pose based on the semantic information of each sample and the corresponding sample, and determine the reference vehicle based on the model parameters in the pose regression model Posture steps

When the difference amount is not greater than the preset difference amount threshold, it is determined that the training of the pose regression model is completed.

Optionally, the semantic information of the parking lot image is matched with the semantic information of each location point in the preset map according to the pose of the first vehicle, and the second vehicle position of the vehicle is determined according to the matching result The steps of posture include:

Matching the semantic information of the parking lot image with the semantic information of each location point in the preset map to obtain the target location corresponding to the successfully matched semantic information in the preset map;

Use one of the following iterative methods to determine the second vehicle pose of the vehicle;

The first vehicle pose is used as the initial value of the estimated pose, and the parking lot is calculated according to the value of the estimated pose and the position of the semantic information of the parking lot image in the parking lot image The semantic information of the image is mapped to the first mapping position in the preset map;

Determining a first error between the first mapping position and the target position;

When the first error is greater than a preset error threshold, the value of the estimated pose is adjusted, and the execution of the value of the estimated pose and the semantic information of the parking lot image is performed in the parking lot. The position in the field image, the step of calculating the semantic information of the parking lot image to be mapped to the first mapping position in the preset map;

When the first error is not greater than the preset error threshold, determine the second vehicle pose of the vehicle according to the current value of the estimated pose;

or,

Taking the first vehicle pose as the initial value of the estimated pose, and according to the value of the estimated pose and the target position, calculate the semantic information that is successfully matched in the preset map in the parking lot image The second mapping position in;

Determining a second error between the second mapping position and the position of the semantic information of the parking lot image in the parking lot image;

When the second error is greater than the preset error threshold, adjust the value of the estimated pose, and return to execute the calculation of the preset map based on the value of the estimated pose and the target position The step of matching the second mapping position of the successfully matched semantic information in the parking lot image;

When the second error is not greater than the preset error threshold, the second vehicle pose of the vehicle is determined according to the current value of the estimated pose.

Optionally, after determining the second vehicle pose of the vehicle, the following methods are used to verify whether the parking lot entrance positioning is successful:

Acquiring the second vehicle pose of the vehicle corresponding to the multiple parking lot image frames when the positions indicated by the multiple initial vehicle poses determined by the positioning module are in the initialization recognition area;

Acquire multiple third vehicle poses determined according to the odometer information collected by the odometer;

Determine residuals between multiple second vehicle poses and multiple third vehicle poses;

When the residual error is less than the preset residual error threshold, it is determined that the vehicle is successfully positioned at the entrance of the parking lot, and the plurality of second vehicle poses are used as the successful positioning information of the vehicle at the entrance of the parking lot.

Optionally, the step of determining residuals between multiple second vehicle poses and multiple third vehicle poses includes:

Solve the following function by the least square method to obtain the rigid transformation matrix T between multiple second vehicle poses and multiple third vehicle poses:

Substitute the T obtained by the solution into

Obtain residuals between multiple second vehicle poses and multiple third vehicle poses;

Among them, the

Is the i-th second vehicle pose, the

Is the i-th third vehicle pose, the N is the total number of the second vehicle pose or the third vehicle pose, the min is the minimum value function, and the ‖·‖ is the norm symbol.

Optionally, the step of detecting the semantic information of the parking lot image includes:

Convert the parking lot image to the top view coordinate system to obtain a ground image;

Binarize the ground image to obtain a processed image;

Determine the semantic information of the parking lot image according to the information in the processed image.

In a second aspect, an embodiment of the present invention discloses a parking lot entrance positioning device in parking positioning, including:

The judgment module is configured to obtain the initial vehicle pose determined by the positioning module when it is detected that the vehicle enters the entrance of the parking lot, and judge whether the position indicated by the initial vehicle pose is in the preset initialization recognition area;

The acquiring module is configured to acquire the parking lot image collected by the camera module when the position indicated by the initial vehicle pose is in the preset initialization recognition area; wherein, the parking lot image is in the initialization recognition area Captured images;

The detection module is configured to detect semantic information of the parking lot image; wherein the semantic information is information used to identify landmarks around the vehicle;

The first determining module is configured to determine the first vehicle pose of the vehicle through a pose regression model based on the semantic information of the parking lot image and the initial vehicle pose; wherein, the pose regression model is It is obtained by training in advance based on a plurality of sample parking lot images collected in the initial recognition area and the corresponding sample initial vehicle pose and the marked vehicle pose;

The second determining module is configured to match the semantic information of the parking lot image with the semantic information of each position point in the preset map according to the pose of the first vehicle, and determine the second vehicle position of the vehicle according to the matching result posture.

Optionally, the device further includes a training module; the training module is configured to train to obtain the pose regression model using the following operations:

Detect the sample semantic information of each sample parking lot image;

Based on the semantic information of each sample and the corresponding sample initial vehicle pose, the reference vehicle pose is determined according to the model parameters through the pose regression model;

When the difference amount is greater than the preset difference amount threshold, modify the model parameters, return to execute the initial vehicle pose based on the semantic information of each sample and the corresponding sample, and determine the reference vehicle position based on the model parameters through the pose regression model Posture operation

Optionally, the second determining module is specifically configured as:

Use one of the following iterative operations to determine the second vehicle pose of the vehicle;

When the first error is greater than the preset error threshold, the value of the estimated pose is adjusted, and the execution returns to the parking lot image according to the value of the estimated pose and the semantic information of the parking lot image. Calculate the semantic information of the parking lot image to be mapped to the first mapping position in the preset map;

or,

When the second error is greater than the preset error threshold, adjust the value of the estimated pose, and return to execute the calculation based on the value of the estimated pose and the target position to calculate the matching success in the preset map The second mapping position of the semantic information of in the parking lot image;

Optionally, the device further includes a verification module; the verification module is configured to use the following operations to verify whether the parking lot entrance location is successful:

After determining the second vehicle pose of the vehicle, obtain the position of the vehicle corresponding to the multiple parking lot image frames when the position indicated by the multiple initial vehicle poses determined by the positioning module is in the initialization recognition area The second vehicle pose;

Optionally, when the verification module determines the residuals between multiple second vehicle poses and multiple third vehicle poses, it includes:

Substitute the T obtained by the solution into

Among them, the

Is the i-th second vehicle pose, the

Optionally, the detection module is specifically configured as:

Binarize the ground image to obtain a processed image;

In the third aspect, an embodiment of the present invention discloses a vehicle-mounted terminal, including: a processor, an image acquisition device, and a positioning device; wherein the processor includes: a judgment module, an acquisition module, a detection module, a first determination module, and a second 2. Determine the module;

The judging module is configured to obtain the initial vehicle pose determined by the positioning device when it is detected that the vehicle enters the entrance of the parking lot, and determine whether the position indicated by the initial vehicle pose is in a preset initialization recognition area;

The acquiring module is configured to acquire the parking lot image collected by the image acquisition device when the position indicated by the initial vehicle pose is in the preset initial recognition area; wherein, the parking lot image is in the initial recognition area Images collected in

Optionally, the processor further includes a training module; the training module is configured to train to obtain the pose regression model using the following operations:

Detect the sample semantic information of each sample parking lot image;

When the difference amount is greater than the preset difference amount threshold, modify the model parameters, return to execute the initial vehicle pose based on the semantic information of each sample and the corresponding sample, and determine the reference vehicle based on the model parameters in the pose regression model Posture operation;

Optionally, the second determining module is specifically configured as:

or,

Optionally, the processor further includes a verification module; the verification module is configured to use the following operations to verify whether the parking lot entrance location is successful:

After determining the second vehicle pose of the vehicle, obtain the position of the vehicle corresponding to the multiple parking lot image frames when the position indicated by the multiple initial vehicle poses determined by the positioning device is in the initialization recognition area The second vehicle pose;

Substitute the T obtained by the solution into

Among them, the

Is the i-th second vehicle pose, the

Optionally, the detection module is specifically configured as:

Binarize the ground image to obtain a processed image;

It can be seen from the above content that the parking lot entrance locating method, device and vehicle-mounted terminal in parking locating provided by the embodiments of the present invention can combine the semantic information of the parking lot image with the positioning module when the vehicle is in the preset initialization recognition area. The determined initial vehicle pose is used as the input of the pose regression model, and the first vehicle pose of the vehicle is determined by the pose regression model. The pose regression model is based on the sample parking lot image in the initial recognition area and the corresponding sample initial vehicle position The pose and the labeled vehicle pose are trained, and the first vehicle pose determined according to the pose regression model has a higher accuracy than the initial vehicle pose; then by matching with the semantic information in the preset map, Based on the pose of the first vehicle, the positioning range is further reduced. Therefore, the embodiment of the present invention can improve the positioning accuracy at the entrance of the parking lot. Of course, implementing any product or method of the present invention does not necessarily need to achieve all the advantages described above at the same time.

The innovative points of the embodiments of the present invention include:

1. At the entrance of the parking lot, when the GPS positioning result is used to determine that the vehicle enters the initial recognition area, the semantic information in the image can be used to narrow the positioning range of the vehicle, and then the semantic information and semantic map matching can be used to further reduce the vehicle In order to determine the more accurate vehicle pose at the entrance of the parking lot, as the initial pose of the vehicle.

2. Cross-validate the initialization pose determined for the multi-frame images in the initialization recognition area with the pose determined according to the odometer to determine whether the pose initialization is successful, and it can more accurately determine whether the vehicle's pose initialization is successful, that is Determine whether the accuracy of the determined vehicle pose is sufficient.

Description of the drawings

In order to explain the embodiments of the present invention or the technical solutions in the prior art more clearly, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained from these drawings without creative work.

FIG. 1 is a schematic flowchart of a parking lot entrance positioning method in parking positioning according to an embodiment of the present invention;

2 is a schematic diagram of a parking lot ground marking line and an initial recognition area provided by an embodiment of the present invention;

Figure 3 is a schematic diagram of a ground image determined according to a parking lot image;

Fig. 4 is a reference diagram of the driving track of the vehicle in the initial recognition area at the entrance of the parking lot in Fig. 2;

5 is a schematic structural diagram of a parking lot entrance positioning device in parking positioning according to an embodiment of the present invention;

Fig. 6 is a schematic structural diagram of a vehicle-mounted terminal provided by an embodiment of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

It should be noted that the terms "including" and "having" in the embodiments of the present invention and the drawings and any variations thereof are intended to cover non-exclusive inclusions. For example, the process, method, system, product or device that contains a series of steps or units is not limited to the listed steps or units, but optionally also includes unlisted steps or units, or optionally also includes Other steps or units inherent to these processes, methods, products or equipment.

The embodiment of the present invention discloses a parking lot entrance positioning method, device and vehicle-mounted terminal in parking positioning, which can improve the positioning accuracy at the entrance of the parking lot. The embodiments of the present invention will be described in detail below.

FIG. 1 is a schematic flowchart of a parking lot entrance positioning method in parking positioning according to an embodiment of the present invention. This method is applied to electronic equipment. The electronic device may be an ordinary computer, a server, or an intelligent terminal device, etc., or may be a vehicle-mounted terminal installed in the vehicle. Among them, the parking lot can be an indoor parking lot or an underground garage. The method specifically includes the following steps.

Step S110: When it is detected that the vehicle has entered the entrance of the parking lot, the initial vehicle pose determined by the positioning module is acquired, and it is determined whether the position indicated by the initial vehicle pose is in the preset initialization recognition area.

If it is, proceed to step S120; if it is not, continue to obtain the new initial vehicle pose determined by the positioning module while the vehicle is running, and execute to determine whether the position indicated by the initial vehicle pose is in the preset initialization recognition area A step of.

In this embodiment, the position of the vehicle on the preset map can be determined in real time according to the data collected by the positioning module in the vehicle. When the vehicle is detected to enter the location of the entrance of the parking lot, the location when entering the entrance of the parking lot is obtained. The initial vehicle pose determined by the module.

Among them, the vehicle pose may include the coordinate position of the vehicle and the vehicle orientation information. The preset map may be a high-precision map established in advance. The positioning module may be a global positioning system (Global Positioning System, GPS) module or a BeiDou Navigation Satellite System (BeiDou Navigation Satellite System, BDS) module. The vehicle in this application can be understood as an intelligent vehicle.

The initial recognition area is a preset coordinate area in the preset map. In the initial recognition area, there are significant differences in the observation of any two positions or the observation of the same position from different angles. In the initial recognition area, the position of the vehicle can be accurately determined as the initial positioning position when the vehicle enters the parking lot. When entering the parking lot, real-time positioning will be performed according to the initial positioning position. The initial recognition area may be a circular area with the preset location point at the entrance of the parking lot as the center and the preset distance as the radius. For example, the preset distance can be 15m or other values.

Refer to FIG. 2, which is a schematic diagram of a parking lot ground marking line and an initial recognition area provided by an embodiment of the present invention. Among them, the marking line on the ground of the parking lot and the wall of the entrance passage of the parking lot are shown (indicated by thick lines), and the initial recognition area at the entrance of the parking lot is represented by a larger circular area. When the vehicle is at point A, the positioning module can be positioned in a larger circular area. The smaller circle range in Figure 2 represents the initial pose range that can normally start the positioning system.

In this step, the function of GPS and other signals is to determine that the vehicle has entered the initial recognition area with a radius of 15m, so as to avoid misdetection in areas with similar terrain. There may also be multiple initialization recognition areas, and the target initialization recognition area that the vehicle has entered can be determined from the multiple initialization recognition areas according to GPS data.

Step S120: Acquire parking lot images collected by the camera module.

Among them, the parking lot image is the image collected in the initial recognition area. The camera module and positioning module in the vehicle can both collect data according to a certain period. In this step, the acquired parking lot image may be: the initial vehicle pose and the indicated position in the initial recognition area are collected at the appointed time. The appointed time can be understood as the same time or two moments with a short time difference.

When the vehicle enters the parking lot from the entrance of the parking lot, the parking lot image collected by the camera module may be an image containing the internal environment of the parking lot.

Step S130: Detect the semantic information of the parking lot image.

Among them, the semantic information is information used to identify landmarks around the vehicle. The semantic information may include, but is not limited to, information corresponding to landmarks such as lane lines, garage lines, indicating arrows, road signs, buildings, and sidewalks on the road surface. The semantic information can be the relative position information between various markers in the image.

In an embodiment, the step of detecting the semantic information of the parking lot image may specifically include:

The parking lot image is converted to the top view coordinate system to obtain the ground image; the ground image is binarized to obtain the processed image; the semantic information of the parking lot image is determined according to the information in the processed image.

Among them, the ground image can be a grayscale image. When the ground image is binarized, the Otsu method can be used to determine the pixel threshold used to distinguish the foreground and background part of the ground image, and the ground image is binarized according to the determined pixel threshold to obtain the processed foreground part image.

When determining the semantic information of the parking lot image based on the information in the processed image, the processed image can be directly used as the semantic information, or the relative position information between the various landmarks in the processed image can be used as the semantic information.

Refer to Fig. 3, which is a schematic diagram of a ground image determined according to a parking lot image. The lines are wall lines and lane lines on the ground. After binarizing the ground image, an image containing semantic information can be obtained. The semantic information can be the relative position between various lines. The image after binarization can be called a semantic observation image.

Step S140: Based on the semantic information of the parking lot image and the initial vehicle pose, the first vehicle pose of the vehicle is determined through the pose regression model.

Among them, the pose regression model is obtained by pre-training based on multiple sample parking lot images collected in the initial recognition area and the corresponding sample initial vehicle pose and the marked vehicle pose. The pose regression model can associate the semantic information of the parking lot image and the initial vehicle pose with the first vehicle pose according to the trained model parameters.

This step may specifically include: inputting the semantic information of the parking lot image and the initial vehicle pose as input information into the pose regression model, and obtaining the first vehicle pose of the vehicle output by the pose regression model. Among them, the first vehicle pose is a vehicle pose that is more accurate than the initial vehicle pose. According to the trained model parameters, the pose regression model can perform regression on the basis of the initial vehicle pose and the feature vector extracted from the semantic information of the parking lot image to obtain the first vehicle pose.

The pose regression module can use a multi-stage pose regression (Cascaded Pose Regression, CPR). The multi-stage pose regression adopts the following principle formula to determine the pose of the first vehicle:

P _reg ＝CPR(P _GPS ,I _seg )

Among them, P _GPS is the initial vehicle pose, and _Iseg is the semantic observation image, that is, the semantic information of the parking lot image. P _GPS and I _seg are the input information of CPR, and P _reg is the first vehicle pose output by CPR.

In this step, based on the semantic information and the initial vehicle pose, the more accurate pose of the vehicle can be determined by the multi-level pose regression. In step S110, it is determined that the vehicle enters the initial recognition area with a radius of 15m, so that the positioning pose more accurate. This step can also be understood as identifying the position of FIG. 3 in FIG. 2.

Step S150: According to the first vehicle pose, match the semantic information of the parking lot image with the semantic information of each location point in the preset map, and determine the second vehicle pose of the vehicle according to the matching result.

Since the semantic information in the parking lot image of the vehicle will be affected by external factors such as occlusion, there may be a certain deviation or misdetection between the first vehicle pose and the real vehicle pose. Therefore, this step can further improve the accuracy of the vehicle pose.

In this step, after the first vehicle pose is obtained, the semantic information of the parking lot image can be matched with the semantic information of each location point in the preset map, and a more accurate second vehicle pose can be determined according to the location point that is successfully matched. .

The second vehicle pose can be understood as the initial pose of the vehicle that is located in the initial recognition area and meets a certain accuracy requirement. When determining the initial pose, it can start real-time positioning of the vehicle based on the visual and semantic map in the parking lot.

It can be seen from the above content that this embodiment can use the semantic information of the parking lot image and the initial vehicle pose determined by the positioning module as the input of the pose regression model when the vehicle is in the preset initialization recognition area. Determine the first vehicle pose of the vehicle. The pose regression model is trained based on the sample parking lot image in the initial recognition area and the corresponding sample initial vehicle pose and the marked vehicle pose. The pose regression model is determined Compared with the initial vehicle pose, the first vehicle pose has a higher accuracy; and by matching with the semantic information in the preset map, the positioning range can be further reduced on the basis of the first vehicle pose. Therefore, this embodiment can Improve the positioning accuracy at the entrance of the parking lot. That is, this embodiment can use positioning signals such as GPS to accurately provide an initial value for the positioning system in the entrance area of the parking area, so that the positioning system can be started normally.

This embodiment is less dependent on the position accuracy of positioning modules such as GPS. At the same time, it also has strong robustness against the situation that the signs in the parking lot are blocked. In this embodiment, the efficiency of determining the pose of the second vehicle is high, and real-time operation can be achieved on an embedded computing device with limited computing power.

In another embodiment of the present invention, based on the embodiment shown in FIG. 1, the pose regression model can be obtained by training in the following steps 1a to 5a.

Step 1a: Acquire multiple sample parking lot images collected in the initial recognition area, and the sample initial vehicle pose and the marked vehicle pose corresponding to each sample parking lot image.

Among them, the marked vehicle pose can be understood as the true value and standard value of the vehicle pose corresponding to the sample parking lot image. The sample initial vehicle pose may be the vehicle pose determined by the positioning module when collecting each sample parking lot image, or it may be the vehicle pose obtained by adding preset disturbances to the marked vehicle pose. The preset disturbance can be understood as a preset modification. The sample initial vehicle pose can be understood as the initial value of the vehicle pose used to input the pose regression model, and the pose regression model regresses the sample parking lot image on the basis of the initial value of the vehicle pose.

In one embodiment, a large number of sample parking lot images can be collected in advance through the camera module in the initial recognition area, and the sample initial vehicle pose determined by the positioning module. When collecting each sample parking lot image, the marked vehicle pose corresponding to the sample parking lot image can be determined by offline positioning.

In another embodiment, the semantic information in the preset map and multiple virtual driving trajectories can be directly used to simulate the collection process of the camera module in the vehicle to obtain a large number of simulated images as sample parking lot images. In addition, the marked vehicle pose corresponding to the simulated image can be determined directly according to the preset map.

Refer to Figure 4, which is a reference diagram of the vehicle's trajectory when collecting sample data in the initial recognition area at the entrance of the parking lot in Figure 2, where the irregular gray lines at the entrance of the parking lot represent the trajectory of the vehicle.

Step 2a: Detect the sample semantic information of each sample parking lot image.

For a specific description of this step, refer to the description part of step S130.

Step 3a: Based on the semantic information of each sample and the corresponding sample initial vehicle pose, the reference vehicle pose is determined through the model parameters in the pose regression model.

Among them, when the pose regression model adopts a multi-level pose regressor, the model parameters that have been trained in other aspects in the multi-level pose regressor can be directly used as the initial values of the model parameters in this step. Through a large number of training processes, the model parameters are constantly revised to gradually approach the true value.

Step 4a: Determine the amount of difference between the reference vehicle pose and the marked vehicle pose.

Specifically, the residual function may be used to determine the amount of difference between the reference vehicle pose and the marked vehicle pose.

Step 5a: When the above difference amount is greater than the preset difference amount threshold, correct the model parameters and return to step 3a. When the difference amount is not greater than the preset difference amount threshold, it is determined that the pose regression model training is completed.

Among them, the preset difference threshold is a value set in advance based on experience. When the difference amount is greater than the preset difference amount threshold, it is considered that the model needs to be continuously trained. When modifying the model parameters, the model parameters can be modified according to the difference. For example, the model parameters can be corrected based on the difference amount and the change trend obtained from the difference amount in the previous training process.

In summary, this embodiment provides a specific implementation manner for training the pose regression model, which can improve the accuracy of the pose regression model, thereby improving the accuracy of positioning.

In another embodiment of the present invention, based on the embodiment shown in FIG. 1, in step S150, according to the first vehicle pose, the semantic information of the parking lot image is matched with the semantic information of each location point in the preset map, and according to the matching result The step of determining the second vehicle pose of the vehicle may include the following embodiments.

The first embodiment includes the following steps 1b to 5b.

Step 1b: Match the semantic information of the parking lot image with the semantic information of each location point in the preset map, and obtain the target location of the successfully matched semantic information in the preset map.

Step 2b: Use the first vehicle pose as the initial value of the estimated pose, and calculate the semantic information mapping of the parking lot image to the prediction based on the estimated pose value and the position of the semantic information of the parking lot image in the parking lot image. Set the first mapping position in the map.

In this step, mapping the semantic information of the parking lot image to the first mapping position in the preset map can be understood as mapping the semantic information of the parking lot image to the coordinate system where the preset map is located, and the mapped position is the first A mapping location.

Step 3b: Determine the first error between the first mapping position and the target position.

Step 4b: When the first error is greater than the preset error threshold, adjust the value of the estimated pose, and return to perform step 2b according to the value of the estimated pose and the position of the semantic information of the parking lot image in the parking lot image, The step of calculating the semantic information of the parking lot image to be mapped to the first mapping position in the preset map.

Step 5b: When the first error is not greater than the preset error threshold, the second vehicle pose of the vehicle is determined according to the current value of the estimated pose.

In this step, when determining the second vehicle pose of the vehicle according to the current value of the estimated pose, it may specifically include directly determining the current value of the estimated pose as the second vehicle pose of the vehicle; it may also include, judging Whether the difference between the current value of the estimated pose and the first vehicle pose is less than a preset threshold, if it is smaller, the current value of the estimated pose is directly determined as the second vehicle pose of the vehicle.

In this embodiment, when the first error is greater than the preset error threshold, it is considered that the positioning accuracy has not yet reached the requirement, and iterating needs to be continued. When the first error is not greater than the preset error threshold, it is considered that the positioning accuracy has reached the requirement, and the iteration can be stopped to obtain the precise positioning pose of the vehicle.

The second embodiment includes the following steps 1c to 5c.

Step 1c: match the semantic information of the parking lot image with the semantic information of each location point in the preset map, and obtain the target location of the successfully matched semantic information in the preset map.

Step 2c: Use the first vehicle pose as the initial value of the estimated pose, and calculate the second mapping position in the parking lot image of the successfully matched semantic information in the preset map according to the estimated pose value and the target position.

Among them, calculating the second mapping position of the successfully matched semantic information in the preset map in the parking lot image can be understood as mapping the successfully matched semantic information in the preset map to the coordinate system where the parking lot image is located, and the mapping obtained The location is the second mapping location.

Step 3c: Determine the second error between the second mapping position and the position of the semantic information of the parking lot image in the parking lot image.

Step 4c: When the second error is greater than the preset error threshold, adjust the value of the estimated pose, and return to perform step 2c according to the value of the estimated pose and the target position to calculate the semantic information of the preset map matching successfully. Step of second mapping position in parking lot image.

Step 5c: When the second error is not greater than the preset error threshold, the second vehicle pose of the vehicle is determined according to the current value of the estimated pose.

The iterative process in this embodiment can be represented by the following data model:

P _i =argmin(||X _ij -f(P _i ,A _j )||);

Wherein, P _i is the posture of the vehicle information at time i, A _j is the preset target position map the j-th semantic information, X _ij is a j-th semantic information and semantic information matches the location in the parking lot image , F(.) is the projection equation, used to project the j-th semantic information to the image imaging plane according to A _j and P _i , and the projection result is in the same coordinate system as X _ij . In this way, the error of the current value mapping of the estimated pose and the actual observation can be obtained, and the vehicle pose can be optimized by means of nonlinear optimization to obtain the most likely pose.

In this embodiment, according to the value of the estimated pose, the semantic information in the preset map is mapped to the parking lot image, or the semantic information in the parking lot image is mapped to the preset map through two mapping methods. The difference between the positions of the mapping information is calculated, and the value of the estimated pose is continuously adjusted according to the difference. When the condition is met, the second vehicle pose of the vehicle is determined according to the estimated pose. In this way, the vehicle pose can be iterated relatively quickly, and a certain positioning accuracy can be guaranteed.

In another embodiment of the present invention, based on the foregoing embodiment, when the vehicle enters the initial recognition area to leave, the electronic device can perform multiple initial positioning based on multiple image frames. For example, the electronic device can perform multiple initial positioning based on a preset initial positioning frequency. Perform the initial positioning shown in Figure 1 in the initial recognition area. Among them, the 5 steps in Figure 1 constitute an initial positioning. In this embodiment, after determining the second vehicle pose of the vehicle, the following steps 1d to 4d can be used to verify whether the parking lot entrance location is successful.

Step 1d: Acquire the second vehicle poses of the vehicles corresponding to the multiple parking lot image frames when the positions indicated by the multiple initial vehicle poses determined by the positioning module are in the initialization recognition area.

Specifically, according to a preset initial positioning frequency, a plurality of second vehicle poses may be determined using the five steps in FIG. 1, and each second vehicle pose may be stored in a preset storage space. When acquiring the second vehicle pose of the vehicle corresponding to the multiple parking lot image frames, it may be acquired from a preset storage space.

Step 2d: Obtain multiple third vehicle poses determined according to the odometer information collected by the odometer.

In this embodiment, the odometer in the vehicle can periodically collect odometer information, and based on the previous odometer information and the odometer information, the vehicle pose can be estimated as the third vehicle pose.

Among them, the multiple third vehicle poses determined according to the odometer information may be the pose information in the preset odometer map.

In one embodiment, the frequency of determining the second vehicle pose and determining the third vehicle pose may be the same, and when the vehicle travels to a certain position, the second vehicle pose and the third vehicle pose are determined simultaneously Operation. That is, the second vehicle pose and the third vehicle pose may have a one-to-one correspondence.

Step 3d: Determine residuals between multiple second vehicle poses and multiple third vehicle poses.

This step may specifically include: determining a residual error between the second vehicle pose and the third vehicle pose in a one-to-one correspondence. The determined residual may be the sum of the residuals between each second vehicle pose and the corresponding third vehicle pose, or it may be the sum of the residuals between each second vehicle pose and the corresponding third vehicle pose. The residual vector composed of the residuals between.

Step 4d: When the residual error is less than the preset residual error threshold, it is determined that the vehicle is successfully positioned at the entrance of the parking lot, and multiple second vehicle poses are used as the successful positioning information of the vehicle at the entrance of the parking lot.

Wherein, the preset residual threshold may be a value determined in advance based on experience. When the residual is not less than the preset residual threshold, it is considered that the vehicle has failed to locate at the entrance of the parking lot.

In summary, this embodiment adopts in the initialization recognition area to cross-validate the vehicle pose of multiple parking lot image frames and the vehicle pose of the odometer, which can effectively reduce the false detection rate of positioning initialization. When the cross-validation is successful, it is determined that the initial positioning process of the entrance of the parking lot is successful, which can more accurately verify whether the accuracy of the initial positioning meets the requirements and improve the positioning accuracy.

In another embodiment of the present invention, based on the above-mentioned embodiment, step 3d, the step of determining the residuals between the multiple second vehicle poses and the multiple third vehicle poses may specifically include:

Solve the following function by the least square method to obtain the rigid transformation matrix T between multiple second vehicle poses and multiple third vehicle poses, and substitute the obtained T into

Obtain the residuals between multiple second vehicle poses and multiple third vehicle poses:

among them,

Is the i-th second vehicle pose,

Is the i-th third vehicle pose, N is the total number of the second vehicle pose or the third vehicle pose, min is the minimum function, and ‖·‖ is the norm symbol.

In the initial recognition area, the second vehicle pose determined from multiple parking lot image frames can adopt the first trajectory

Indicates that multiple third vehicle poses determined according to the odometer information can adopt the second trajectory

Said. Solve

The formula can be understood as determining the minimum amount of change when changing the first trajectory to the second trajectory. Calculate the residual size of each item in trace _init and trace _odom , that is, calculate

The matching degree of the two trajectories can be obtained, and when the matching degree is greater than the preset matching degree threshold, it is determined that the initial positioning is successful.

FIG. 5 is a schematic structural diagram of a parking lot entrance positioning device in parking positioning according to an embodiment of the present invention. The device is applied to electronic equipment. The electronic device may be an ordinary computer, a server, or an intelligent terminal device, etc., or may be a vehicle-mounted terminal installed in the vehicle. This device embodiment corresponds to the method embodiment shown in FIG. 1. The device includes the following modules.

The judging module 510 is configured to obtain the initial vehicle pose determined by the positioning module when it is detected that the vehicle has entered the entrance of the parking lot, and judge whether the position indicated by the initial vehicle pose is in the preset initialization recognition area;

The acquiring module 520 is configured to acquire the parking lot image collected by the camera module when the position indicated by the initial vehicle pose is in the preset initial recognition area; wherein the parking lot image is an image acquired in the initial recognition area;

The detection module 530 is configured to detect semantic information of the parking lot image; wherein the semantic information is information used to identify landmarks around the vehicle;

The first determining module 540 is configured to determine the first vehicle pose of the vehicle through the pose regression model based on the semantic information of the parking lot image and the initial vehicle pose; wherein, the pose regression model is based on the initial recognition area The collected multiple sample parking lot images and the corresponding sample initial vehicle pose and the marked vehicle pose are trained;

The second determining module 550 is configured to match the semantic information of the parking lot image with the semantic information of each location point in the preset map according to the first vehicle pose, and determine the second vehicle pose of the vehicle according to the matching result.

In another embodiment of the present invention, based on the embodiment shown in FIG. 5, the device may further include a training module (not shown in the figure); the training module is configured to train the pose regression model by using the following operations:

Acquire multiple sample parking lot images collected in the initial recognition area, as well as the sample initial vehicle pose and the marked vehicle pose corresponding to each sample parking lot image;

Detect the sample semantic information of each sample parking lot image;

Based on the semantic information of each sample and the corresponding sample initial vehicle pose, the reference vehicle pose is determined through the model parameters in the pose regression model;

Determine the amount of difference between the reference vehicle pose and the marked vehicle pose;

When the difference amount is greater than the preset difference amount threshold, modify the model parameters, return to execute the operation of determining the reference vehicle pose based on the semantic information of each sample and the corresponding sample initial vehicle pose, and determine the reference vehicle pose based on the model parameters in the pose regression model;

When the difference amount is not greater than the preset difference amount threshold, it is determined that the pose regression model training is completed.

In another embodiment of the present invention, based on the embodiment shown in FIG. 5, the second determining module 550 is specifically configured to:

Taking the first vehicle pose as the initial value of the estimated pose, and according to the value of the estimated pose and the position of the semantic information of the parking lot image in the parking lot image, the semantic information of the parking lot image is mapped to The first mapping position in the preset map;

Determine the first error between the first mapping position and the target position;

When the first error is greater than the preset error threshold, adjust the value of the estimated pose, and return to execute the calculation of parking based on the value of the estimated pose and the position of the semantic information of the parking lot image in the parking lot image The semantic information of the field image is mapped to the first mapping position in the preset map;

or,

Use the first vehicle pose as the initial value of the estimated pose, and calculate the second mapping position in the parking lot image of the semantic information successfully matched in the preset map according to the estimated pose value and the target position;

Determine the second error between the second mapping position and the position of the semantic information of the parking lot image in the parking lot image;

When the second error is greater than the preset error threshold, adjust the value of the estimated pose, and return to execute the calculation based on the value of the estimated pose and the target position to calculate the semantic information that is successfully matched in the preset map in the parking lot. The second mapping position in the image;

In another embodiment of the present invention, based on the embodiment shown in FIG. 5, the device may further include a verification module (not shown in the figure); the verification module is configured to use the following operations to verify whether the parking lot entrance location is successful :

After determining the second vehicle pose of the vehicle, obtain the second vehicle pose of the vehicle corresponding to the multiple parking lot image frames when the position indicated by the multiple initial vehicle poses determined by the positioning module is in the initialization recognition area;

When the residual is less than the preset residual threshold, it is determined that the vehicle is successfully positioned at the entrance of the parking lot, and multiple second vehicle poses are used as the successful positioning information of the vehicle at the entrance of the parking lot.

In another embodiment of the present invention, based on the embodiment shown above, the verification module, when determining the residuals between the multiple second vehicle poses and the multiple third vehicle poses, includes:

Substitute the obtained T into

among them,

Is the i-th second vehicle pose,

In another embodiment of the present invention, based on the embodiment shown in FIG. 5, the detection module 530 is specifically configured to:

Convert the parking lot image to the top view coordinate system to obtain the ground image;

Binarize the ground image to obtain the processed image;

According to the information in the processed image, the semantic information of the parking lot image is determined.

The foregoing device embodiment corresponds to the method embodiment, and has the same technical effect as the method embodiment. For specific description, refer to the method embodiment. The device embodiment is obtained based on the method embodiment, and the specific description can be found in the method embodiment part, which will not be repeated here.

Fig. 6 is a schematic structural diagram of a vehicle-mounted terminal provided by an embodiment of the present invention. The vehicle-mounted terminal includes a processor 610, an image acquisition device 620, and a positioning device 630. Wherein, the processor includes: a judgment module, an acquisition module, a detection module, a first determination module, and a second determination module (not shown in the figure).

The judging module is configured to obtain the initial vehicle pose determined by the positioning device 630 when it is detected that the vehicle has entered the entrance of the parking lot, and determine whether the position indicated by the initial vehicle pose is in the preset initialization recognition area;

The acquiring module is configured to acquire the parking lot image collected by the image acquisition device 620 when the position indicated by the initial vehicle pose is in the preset initial recognition area; wherein, the parking lot image is an image acquired in the initial recognition area;

The first determination module is configured to determine the first vehicle pose of the vehicle through the pose regression model based on the semantic information of the parking lot image and the initial vehicle pose; wherein, the pose regression model is based on pre-collecting in the initialization recognition area Multiple sample parking lot images and corresponding sample initial vehicle poses and labeled vehicle poses are trained;

The second determining module is configured to match the semantic information of the parking lot image with the semantic information of each location point in the preset map according to the first vehicle pose, and determine the second vehicle pose of the vehicle according to the matching result.

In another embodiment of the present invention, based on the embodiment shown in FIG. 6, the processor 610 further includes a training module (not shown in the figure); the training module is configured to use the following operations to train to obtain the pose regression model:

Detect the sample semantic information of each sample parking lot image;

In another embodiment of the present invention, based on the embodiment shown in FIG. 6, the second determining module is specifically configured as:

or,

In another embodiment of the present invention, based on the embodiment shown in FIG. 6, the processor 610 may also include a verification module (not shown in the figure); the verification module is configured to use the following operations to verify whether the parking lot entrance location is successful :

After determining the second vehicle pose of the vehicle, obtain the second vehicle pose of the vehicle corresponding to the multiple parking lot image frames when the position indicated by the multiple initial vehicle poses determined by the positioning device 630 is in the initialization recognition area;

In another embodiment of the present invention, based on the embodiment shown in FIG. 6, when the verification module determines the residuals between the multiple second vehicle poses and the multiple third vehicle poses, it includes:

Substitute the obtained T into

among them,

Is the i-th second vehicle pose,

In another embodiment of the present invention, based on the embodiment shown in FIG. 6, the detection module is specifically configured as:

Binarize the ground image to obtain the processed image;

This embodiment of the terminal and the embodiment of the method shown in FIG. 1 are embodiments based on the same inventive concept, and relevant points may be referred to each other. The foregoing terminal embodiment corresponds to the method embodiment, and has the same technical effect as the method embodiment. For specific description, refer to the method embodiment.

A person of ordinary skill in the art can understand that the drawings are only schematic diagrams of an embodiment, and the modules or processes in the drawings are not necessarily necessary for implementing the present invention.

A person of ordinary skill in the art can understand that the modules in the device in the embodiment may be distributed in the device in the embodiment according to the description of the embodiment, or may be located in one or more devices different from this embodiment with corresponding changes. The modules of the above-mentioned embodiments can be combined into one module or further divided into multiple sub-modules.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: The technical solutions recorded in the foregoing embodiments are modified, or some of the technical features are equivalently replaced; these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

A parking lot entrance location method in parking location, which is characterized in that it comprises:

When it is detected that the vehicle has entered the entrance of the parking lot, acquiring the initial vehicle pose determined by the positioning module, and determining whether the position indicated by the initial vehicle pose is in the preset initialization recognition area;

If it is, acquire a parking lot image collected by the camera module; wherein the parking lot image is an image collected in the initial recognition area;

Detecting semantic information of the parking lot image; wherein the semantic information is information used to identify landmarks around the vehicle;

Based on the semantic information of the parking lot image and the initial vehicle pose, the first vehicle pose of the vehicle is determined by a pose regression model; wherein, the pose regression model is based on the initial recognition area A number of sample parking lot images collected inside and the corresponding sample initial vehicle pose and labeled vehicle pose are trained;

According to the first vehicle pose, the semantic information of the parking lot image is matched with the semantic information of each location point in the preset map, and the second vehicle pose of the vehicle is determined according to the matching result.
The method of claim 1, wherein the pose regression model is trained in the following manner:

Acquiring a plurality of sample parking lot images collected in the initial recognition area, and a sample initial vehicle pose and a marked vehicle pose corresponding to each sample parking lot image;

Detect the sample semantic information of each sample parking lot image;

Based on the semantic information of each sample and the corresponding initial vehicle pose, the reference vehicle pose is determined through the model parameters in the pose regression model;

Determining the amount of difference between the reference vehicle pose and the marked vehicle pose;

When the difference amount is greater than the preset difference amount threshold, modify the model parameters, return to execute the initial vehicle pose based on the semantic information of each sample and the corresponding sample, and determine the reference vehicle based on the model parameters in the pose regression model Posture steps

When the difference amount is not greater than the preset difference amount threshold, it is determined that the training of the pose regression model is completed.
The method according to claim 1, wherein the matching semantic information of the parking lot image with the semantic information of each location point in the preset map according to the first vehicle pose, according to the matching result The step of determining the second vehicle pose of the vehicle includes:

Matching the semantic information of the parking lot image with the semantic information of each location point in the preset map to obtain the target location corresponding to the successfully matched semantic information in the preset map;

Use one of the following iterative methods to determine the second vehicle pose of the vehicle;

The first vehicle pose is used as the initial value of the estimated pose, and the parking lot is calculated according to the value of the estimated pose and the position of the semantic information of the parking lot image in the parking lot image The semantic information of the image is mapped to the first mapping position in the preset map;

Determining a first error between the first mapping position and the target position;

When the first error is greater than a preset error threshold, the value of the estimated pose is adjusted, and the execution of the value of the estimated pose and the semantic information of the parking lot image is performed in the parking lot. The position in the field image, the step of calculating the semantic information of the parking lot image to be mapped to the first mapping position in the preset map;

When the first error is not greater than the preset error threshold, determine the second vehicle pose of the vehicle according to the current value of the estimated pose;

or,

Taking the first vehicle pose as the initial value of the estimated pose, and according to the value of the estimated pose and the target position, calculate the semantic information that is successfully matched in the preset map in the parking lot image The second mapping position in;

Determining a second error between the second mapping position and the position of the semantic information of the parking lot image in the parking lot image;

When the second error is greater than the preset error threshold, adjust the value of the estimated pose, and return to execute the calculation of the preset map based on the value of the estimated pose and the target position The step of matching the second mapping position of the successfully matched semantic information in the parking lot image;

When the second error is not greater than the preset error threshold, the second vehicle pose of the vehicle is determined according to the current value of the estimated pose.
The method of claim 1, wherein after determining the second vehicle pose of the vehicle, the following methods are used to verify whether the parking lot entrance location is successful:

Acquiring the second vehicle pose of the vehicle corresponding to the multiple parking lot image frames when the positions indicated by the multiple initial vehicle poses determined by the positioning module are in the initialization recognition area;

Acquiring multiple third vehicle poses determined according to the odometer information collected by the odometer;

Determine residuals between multiple second vehicle poses and multiple third vehicle poses;

When the residual error is less than the preset residual error threshold, it is determined that the vehicle is successfully positioned at the entrance of the parking lot, and the plurality of second vehicle poses are used as the successful positioning information of the vehicle at the entrance of the parking lot.
The method of claim 4, wherein the step of determining the residuals between the plurality of second vehicle poses and the plurality of third vehicle poses comprises:

Solve the following function by the least square method to obtain the rigid transformation matrix T between multiple second vehicle poses and multiple third vehicle poses:

Substitute the T obtained by the solution into
Obtain residuals between multiple second vehicle poses and multiple third vehicle poses;

Among them, the
Is the i-th second vehicle pose, the
Is the i-th third vehicle pose, the N is the total number of the second vehicle pose or the third vehicle pose, the min is the minimum value function, and the ‖·‖ is the norm symbol.
The method of claim 1, wherein the step of detecting the semantic information of the parking lot image comprises:

Convert the parking lot image to the top view coordinate system to obtain a ground image;

Binarize the ground image to obtain a processed image;

Determine the semantic information of the parking lot image according to the information in the processed image.
A parking lot entrance positioning device in parking positioning, characterized in that it comprises:

The judgment module is configured to obtain the initial vehicle pose determined by the positioning module when it is detected that the vehicle enters the entrance of the parking lot, and judge whether the position indicated by the initial vehicle pose is in the preset initialization recognition area;

The acquiring module is configured to acquire the parking lot image collected by the camera module when the position indicated by the initial vehicle pose is in the preset initialization recognition area; wherein, the parking lot image is in the initialization recognition area Captured images;

The detection module is configured to detect semantic information of the parking lot image; wherein the semantic information is information used to identify landmarks around the vehicle;

The first determining module is configured to determine the first vehicle pose of the vehicle through a pose regression model based on the semantic information of the parking lot image and the initial vehicle pose; wherein, the pose regression model is It is obtained by training in advance based on a plurality of sample parking lot images collected in the initial recognition area and the corresponding sample initial vehicle pose and the marked vehicle pose;

The second determining module is configured to match the semantic information of the parking lot image with the semantic information of each position point in the preset map according to the pose of the first vehicle, and determine the second vehicle position of the vehicle according to the matching result posture.
7. The device according to claim 7, wherein the device further comprises a verification module; the verification module is configured to verify whether the parking lot entrance location is successful or not by adopting the following operations:

After determining the second vehicle pose of the vehicle, obtain the position of the vehicle corresponding to the multiple parking lot image frames when the position indicated by the multiple initial vehicle poses determined by the positioning module is in the initialization recognition area The second vehicle pose;

Acquire multiple third vehicle poses determined according to the odometer information collected by the odometer;

Determine residuals between multiple second vehicle poses and multiple third vehicle poses;

When the residual error is less than the preset residual error threshold, it is determined that the vehicle is successfully positioned at the entrance of the parking lot, and the plurality of second vehicle poses are used as the successful positioning information of the vehicle at the entrance of the parking lot.
A vehicle-mounted terminal, characterized by comprising: a processor, an image acquisition device, and a positioning device; wherein the processor includes: a judgment module, an acquisition module, a detection module, a first determination module, and a second determination module;

The judging module is configured to obtain the initial vehicle pose determined by the positioning device when it is detected that the vehicle enters the entrance of the parking lot, and determine whether the position indicated by the initial vehicle pose is in a preset initialization recognition area;

The acquiring module is configured to acquire the parking lot image collected by the image acquisition device when the position indicated by the initial vehicle pose is in the preset initial recognition area; wherein, the parking lot image is in the initial recognition area Images collected in

The detection module is configured to detect semantic information of the parking lot image; wherein the semantic information is information used to identify landmarks around the vehicle;

The first determining module is configured to determine the first vehicle pose of the vehicle through a pose regression model based on the semantic information of the parking lot image and the initial vehicle pose; wherein, the pose regression model is It is obtained by training in advance based on a plurality of sample parking lot images collected in the initial recognition area and the corresponding sample initial vehicle pose and the marked vehicle pose;

The second determining module is configured to match the semantic information of the parking lot image with the semantic information of each position point in the preset map according to the pose of the first vehicle, and determine the second vehicle position of the vehicle according to the matching result posture.
The terminal according to claim 9, wherein the processor further comprises a verification module; the verification module is configured to use the following operations to verify whether the parking lot entrance location is successful:

After determining the second vehicle pose of the vehicle, obtain the position of the vehicle corresponding to the multiple parking lot image frames when the position indicated by the multiple initial vehicle poses determined by the positioning device is in the initialization recognition area The second vehicle pose;

Acquire multiple third vehicle poses determined according to the odometer information collected by the odometer;

Determine residuals between multiple second vehicle poses and multiple third vehicle poses;

When the residual error is less than the preset residual error threshold, it is determined that the vehicle is successfully positioned at the entrance of the parking lot, and the plurality of second vehicle poses are used as the successful positioning information of the vehicle at the entrance of the parking lot.