Vehicle calibration method, device, electronic equipment and storage medium

Abstract

The present application relates to the field of vehicle control technology, and provides a vehicle calibration method, device, electronic device and storage medium. The method extracts contour feature points from an image frame to obtain a first stereo feature set and a first plane feature set; determines the shooting posture corresponding to the shooting device according to the first stereo feature set and the first plane feature set; obtains the angular velocity and driving speed of the vehicle, and determines the vehicle odometer posture according to the angular velocity and the driving speed; calculates the error posture based on the shooting posture and the vehicle odometer posture, and calibrates the shooting device and the vehicle based on the error posture, and ensures high precision and automation of the calibration process by self-calibrating the shooting device and the vehicle; then determines the error posture through the shooting posture and the vehicle odometer posture, and subsequently accurately calibrates the shooting device and the vehicle coordinate system through the error posture.

Description

Technical Field

[0001] The present application relates to the field of vehicle control technology, and in particular to a vehicle calibration method, device, electronic equipment and storage medium. Background Art

[0002] The extrinsic calibration of the camera and vehicle coordinate systems is an important step in autonomous vehicles. It involves converting the information observed by the camera into the vehicle coordinate system so that the vehicle can accurately understand and respond to the surrounding environment.

[0003] In the existing technology, the calibration of the camera and vehicle coordinate systems is divided into two categories. One is a static calibration method based on a calibration plate. Specifically, a calibration plate and a total station are used in the calibration room to assist in calculating the external parameters between the calibration plate and the vehicle body by measuring the coordinates of the points on the calibration plate and the coordinates of the four wheels. This method is mainly computationally complex and sensitive to the environment, and cannot accurately calibrate the camera and vehicle coordinate systems. The other is a dynamic calibration method based on specific road lane lines. Specifically, the vehicle is driven in the center as much as possible, and then the distance between the vehicle and the parallel line and the perceived lane line vanishing point are used for calibration. This method is limited to specific roads, and the accuracy of artificial straight-line driving cannot be guaranteed. It has low applicability and is also unable to accurately calibrate the camera and vehicle coordinate systems. Summary of the invention

[0004] In view of this, the embodiments of the present application provide a vehicle calibration method, device, electronic device and storage medium to solve the problem in the prior art that the camera and the vehicle coordinate system cannot be accurately calibrated, resulting in limited automatic driving functions.

[0005] According to a first aspect of an embodiment of the present application, a vehicle calibration method is provided, the method comprising: extracting contour feature points of an image frame to obtain a first stereo feature set and a first plane feature set; determining a shooting posture corresponding to a shooting device according to the first stereo feature set and the first plane feature set; obtaining the angular velocity and driving speed of the vehicle, and determining the vehicle odometer posture according to the angular velocity and the driving speed; calculating an error posture based on the shooting posture and the vehicle odometer posture, and calibrating the shooting device and the vehicle based on the error posture.

[0006] According to a second aspect of an embodiment of the present application, a vehicle calibration device is provided, which includes: a shooting posture acquisition module, configured to extract contour feature points of an image frame to obtain a first stereo feature set and a first plane feature set; determining a shooting posture corresponding to a shooting device according to the first stereo feature set and the first plane feature set; an odometer posture acquisition module, configured to obtain an angular velocity and a driving speed of the vehicle, and determine the vehicle odometer posture according to the angular velocity and the driving speed; a calibration module, configured to calculate an error posture based on the shooting posture and the vehicle odometer posture, and calibrate the shooting device and the vehicle based on the error posture.

[0007] According to a third aspect of an embodiment of the present application, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0008] According to a fourth aspect of an embodiment of the present application, a computer-readable storage medium is provided, which stores a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0009] Compared with the prior art, the embodiments of the present application have the following beneficial effects: the method in the embodiments of the present application obtains a first stereo feature set and a first plane feature set by extracting contour feature points from an image frame; determines the corresponding shooting posture of the shooting device according to the first stereo feature set and the first plane feature set; obtains the angular velocity and the driving speed of the vehicle, and determines the vehicle odometer posture according to the angular velocity and the driving speed; calculates the error posture based on the shooting posture and the vehicle odometer posture, and calibrates the shooting device and the vehicle based on the error posture, with the following beneficial effects: wherein, the above method first obtains the corresponding shooting posture and the vehicle odometer posture respectively by self-calibrating the shooting device and the vehicle, thereby ensuring high precision and automation of the calibration process; then determines the error posture through the shooting posture and the vehicle odometer posture, and the error posture can characterize the relative posture of the shooting device relative to the vehicle, and subsequently the shooting device and the vehicle coordinate system can be accurately calibrated through the error posture, thereby avoiding the problem in the related art that the camera and the vehicle coordinate system cannot be accurately calibrated, resulting in limited automatic driving function. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0011] Figure 1It is a flow chart of a vehicle calibration method provided in an embodiment of the present application;

[0012] Figure 2 is a network diagram of a deep learning model provided in an embodiment of the present application;

[0013] Figure 3 is a flow chart of another vehicle calibration method provided in an embodiment of the present application;

[0014] Figure 4 is a flow chart of another vehicle calibration method provided in an embodiment of the present application;

[0015] Figure 5 is a flow chart of another vehicle calibration method provided in an embodiment of the present application;

[0016] Figure 6 is a flow chart of another optional vehicle calibration method provided in an embodiment of the present application;

[0017] Figure 7 is a flowchart of another optional vehicle calibration method provided in an embodiment of the present application;

[0018] Figure 8 It is a basic schematic diagram of performing linear interpolation on the vehicle odometer posture provided by an embodiment of the present application;

[0019] Fig. 9 is a flowchart of another optional vehicle calibration method provided in an embodiment of the present application;

[0020] Fig.10 is a flowchart of another optional vehicle calibration method provided in an embodiment of the present application;

[0021] Fig.11 is a structural schematic diagram of a vehicle calibration device provided in an embodiment of the present application;

[0022] Fig.12 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0024] A vehicle calibration method and device according to an embodiment of the present application will be described in detail below with reference to the accompanying drawings.

[0025] Figure 1 A vehicle calibration method provided in an embodiment of the present application is as follows: Figure 1 As shown, the method includes:

[0026] S101, extracting contour feature points from an image frame to obtain a first three-dimensional feature set and a first plane feature set;

[0027] S102, determining a shooting posture corresponding to the shooting device according to the first three-dimensional feature set and the first planar feature set;

[0028] S103, obtaining the angular velocity and driving speed of the vehicle, and determining the vehicle odometer posture according to the angular velocity and the driving speed;

[0029] S104: Calculate an error posture based on the shooting posture and the vehicle odometer posture, and calibrate the shooting device and the vehicle based on the error posture.

[0030] It can be understood that the vehicle calibration method provided in this example is applied to vehicles, including vehicles with automatic driving or intelligent driving (including vehicles with passenger functions (such as cars, buses, large buses, minibuses, etc.), vehicles with cargo functions (such as ordinary trucks, vans, trailers, closed trucks, tank trucks, flatbed trucks, container trucks, dump trucks, trucks with special structures), special vehicles (such as logistics distribution vehicles, automatic guided transport vehicles AGV, patrol cars, cranes, cranes, excavators, bulldozers, forklifts, rollers, loaders, off-road engineering vehicles, armored engineering vehicles, sewage treatment vehicles, sanitation vehicles, vacuum trucks, floor washing vehicles, sprinkler trucks, sweeping robots, food delivery robots, shopping guide robots, lawn mowers, golf carts, etc.), vehicles with entertainment functions (such as entertainment vehicles, amusement park automatic driving devices, balance cars, etc.), rescue vehicles (such as fire trucks, ambulances, power repair vehicles, engineering rescue vehicles, etc.)), etc.

[0031] In some examples, the image frame in step S101 is an image that can reflect the current environment captured by a camera installed on the vehicle during the driving process; the camera includes but is not limited to a camera. The camera can obtain semantic information of the image frame based on a deep learning model, and then extract contour feature points after sparseness, and classify the contour feature points (plane elements such as ground arrows, triangle signs, crosswalks, etc.) according to stereo features (stereo elements such as traffic signs, traffic signs, and road poles) and plane features to obtain a first stereo feature set and a first plane feature set.

[0032] The deep learning model includes but is not limited to one of the YoloV7 model, the YoloV8 model, the Faster R-CNN model, and the PP-YOLO model; for example, the deep learning model is the YoloV8 model, such as Figure 2 As shown in the figure, the YoloV8 model includes a backbone network (Backbone) 21, a feature fusion network (Neck) 22, and a detection head (Head) 23. The backbone network 21 extracts features from the input image frame. The backbone network 21 is mainly composed of a CBS module, a C2f module, and an SPPF module. The CBS module represents Conv, BN, and SILU, which are used to convolve the input image or feature map. The result of the convolution is a decrease in resolution and an increase in the number of channels. The C2f module is used to extract features in depth, and the size of the output feature map of the C2f module remains unchanged from the input feature map. SPPF is the spatial pyramid pooling layer of YOLOV8, which is used to aggregate features at multiple scales. The feature fusion network 22 is used to fuse the feature maps output by the backbone network 21, that is, to concat the features extracted from three different scales by upsampling, and output the feature map (80*80) to the Head layer after two upsamplings, and then downsample the feature map through the CBS module to output the remaining feature maps (40*40 and 20*20) to the Head layer. The detection head 23 is used to generate a specific feature map according to the number of categories.

[0033] It can be understood that this example classifies contour feature points according to stereo features and plane features. Compared with the single feature extraction method, the two feature extraction methods of stereo features and plane features can more comprehensively capture the contour information in the image frame. This multi-feature extraction method ensures the stability and accuracy of pose calculation, especially in complex environments, and can effectively cope with the challenges brought by lighting changes or scene occlusion.

[0034] It can be understood that when the vehicle is driving, the camera device will continue to shoot according to its own shooting frequency. Therefore, when the vehicle is driving, the camera device will continue to acquire image frames, and then acquire the corresponding first stereo feature set and first plane feature set based on each image frame.

[0035] In some examples, after determining the first stereo feature set and the first plane feature set, this example will jointly determine the shooting posture of the shooting device in the current environment based on the stereo feature points included in the first stereo feature set and the plane feature points included in the first plane feature set.

[0036] In some examples, this example also obtains the angular velocity and driving speed of the vehicle; in some examples, the vehicle can obtain the driving speed of the vehicle through sensors such as odometer sensors, global positioning systems, and inertial measurement units; in some examples, the vehicle obtains the angular velocity of the vehicle through sensors such as inertial measurement units and visual inertial systems.

[0037] It can be understood that the vehicle can also be simplified into a vehicle single-track model by constructing a vehicle kinematic model based on Ackerman steering. During the driving process of the vehicle, according to the encoder installed on the wheel, the number of revolutions of the left and right wheels in a certain period of time can be obtained, and the distance traveled by the two wheels can be calculated, and then the speed of the two wheels can be obtained. The angular velocity of the vehicle can be determined by the speed and the left and right wheelbases and wheelbases.

[0038] In some examples, after determining the angular velocity and the driving speed, the vehicle can perform a vehicle odometer posture calculation to obtain the vehicle odometer posture.

[0039] Finally, the vehicle calculates the error pose based on the shooting pose and the vehicle odometer pose, and the error pose is used for the relative pose of the shooting device relative to the vehicle; subsequently, based on the error pose and the vehicle odometer pose, the position and posture of the shooting device in the vehicle coordinate system (that is, the pose of the shooting device in the vehicle coordinate system) can be calculated, thereby realizing the calibration between the shooting device and the vehicle.

[0040] According to the technical solution provided in the embodiment of the present application, a first stereo feature set and a first plane feature set are obtained by extracting contour feature points of an image frame; a shooting posture corresponding to a shooting device is determined according to the first stereo feature set and the first plane feature set; the angular velocity and the driving speed of the vehicle are obtained, and the vehicle odometer posture is determined according to the angular velocity and the driving speed; based on the shooting posture and the vehicle odometer posture, an error posture is calculated, and the shooting device and the vehicle are calibrated based on the error posture, with beneficial effects: wherein, the above method first obtains the corresponding shooting posture and the vehicle odometer posture respectively by self-calibrating the shooting device and the vehicle, thereby ensuring high precision and automation of the calibration process; then, the error posture is determined through the shooting posture and the vehicle odometer posture, and the error posture can characterize the relative posture of the shooting device relative to the vehicle, and subsequently the shooting device and the vehicle coordinate system can be accurately calibrated through the error posture, thereby avoiding the problem in the related art that the camera and the vehicle coordinate system cannot be accurately calibrated, resulting in limited automatic driving function.

[0041] In some embodiments, Figure 3 As shown, determining the shooting posture corresponding to the shooting device according to the first three-dimensional feature set and the first plane feature set includes:

[0042] S301, obtaining a map database corresponding to the current position of the vehicle, where the map database contains three-dimensional features and two-dimensional features;

[0043] S302, constructing a second three-dimensional feature set according to the three-dimensional features in the map database, and constructing a second plane feature set according to the plane features in the map database;

[0044] S303, calculating a stereo feature residual according to the first stereo feature set and the second stereo feature set;

[0045] S304, calculating a plane feature residual according to the first plane feature set and the second plane feature set;

[0046] S305 , obtaining the shooting posture by calculating the stereo feature residual and the planar feature residual.

[0047] It can be understood that in order to accurately obtain the shooting posture of the shooting device, the solution provided in this example will compare the feature points of the image frame with the feature points in the high-precision map, and then determine the shooting posture of the shooting device.

[0048] Specifically, this example extracts a section of road with a length of about 3 km and significant road traffic element features from the high-precision map according to the current position of the vehicle, mainly including three-dimensional features (traffic signs, traffic signs, road poles and other elements) and plane features (ground arrows, triangular signs, crosswalks and other elements) to generate a map database, which contains three-dimensional features and plane features.

[0049] Then this example constructs a second three-dimensional feature set based on the three-dimensional features in the map database, and constructs a second plane feature set based on the plane features in the map database; specifically, for the elements in the map database, contour feature points are extracted according to the three-dimensional features and plane features, and unique label encoding is performed to construct the second three-dimensional feature set and the second plane feature set.

[0050] In some examples, this example calculates a stereo feature residual based on the first stereo feature set and the second stereo feature set, and the stereo feature residual is used to characterize the difference between the stereo feature points in the first stereo feature set and the stereo feature points in the second stereo feature set; and calculates a plane feature residual based on the first plane feature set and the second plane feature set, and the plane feature residual is used to characterize the difference between the plane feature points in the first plane feature set and the plane feature points in the second plane feature set.

[0051] It can be understood that since the first stereo feature set and the second stereo feature set contain multiple stereo feature points, the above-mentioned stereo feature residual is expressed by an equation; the first plane feature set and the second plane feature set contain multiple plane feature points, so the plane feature residual is generally expressed by an equation.

[0052] Then the equation of the stereo feature residual is rotated and translated to obtain the first shooting pose. Using this as a constraint, the equations of the plane feature residual are rotated and translated to be derived, and the LM algorithm is used to obtain the second shooting posture The first shooting posture and the second shooting posture together constitute the shooting posture of the shooting device.

[0053] It should be noted that when updating the state quantity, if the number of iterations or the update amount exceeds the threshold, the plane / stereoscopic features of the image frame need to be updated and a new round of iterations needs to be performed. For the optimal transformation matrix T between the shooting device and the map database, the final shooting posture needs to be generated through two-step LM optimization, which achieves higher positioning accuracy than the traditional ICP algorithm on the high-precision map.

[0054] According to the technical solution provided in the embodiment of the present application, a map database corresponding to the current position of the vehicle is obtained, and the map database contains three-dimensional features and plane features; a second three-dimensional feature set is constructed according to the three-dimensional features in the map database, and a second plane feature set is constructed according to the plane features in the map database; a three-dimensional feature residual is calculated according to the first three-dimensional feature set and the second three-dimensional feature set; a plane feature residual is calculated according to the first plane feature set and the second plane feature set; a shooting pose is calculated according to the three-dimensional feature residual and the plane feature residual. By extracting significant road elements from the map database, the system can adapt to a variety of road traffic scenes, and realizes automatic solution of the shooting pose of the shooting device according to the map database, and adopts a two-step LM optimization algorithm to improve the accuracy of shooting pose estimation, and compared with the traditional ICP algorithm, higher positioning accuracy is achieved. The above-mentioned automated process reduces manual intervention, reduces dependence on scene assumptions and manual mechanical parameter measurement, reduces the demand for high-precision calibration equipment, and reduces calibration costs.

[0055] In some examples, such as Figure 4 As shown, according to the first stereo feature set and the second stereo feature set, a stereo feature residual is calculated, including:

[0056] S401, taking each stereo feature point in the first stereo feature set as a target stereo feature point, traversing the second stereo feature set to determine a first stereo comparison point and a second stereo comparison point corresponding to the target stereo feature point;

[0057] S402, obtaining a first stereo difference between a target stereo feature point and a first stereo comparison point, obtaining a second stereo difference between the target stereo feature point and the second stereo comparison point, and obtaining a third stereo difference between the first stereo comparison point and the second stereo comparison point;

[0058] S403 : Calculate a stereo feature residual corresponding to the target stereo feature point according to the first stereo difference, the second stereo difference, and the third stereo difference.

[0059] In some examples, the present application will take each stereo feature point in the first stereo feature set as the target stereo feature point in turn. For example, if the first stereo feature set contains N stereo feature points, the first stereo feature point will be taken as the first target stereo feature point, the second stereo feature point will be taken as the second target stereo feature point... and the Nth stereo feature point will be taken as the Nth target stereo feature point.

[0060] It can be understood that after determining the target stereo feature point, the present application will traverse the second stereo feature set to determine the first stereo comparison point and the second stereo comparison point corresponding to the target feature point; specifically, the target stereo feature point is compared with each stereo feature point in the second stereo feature set, and the two stereo feature points closest to the target stereo feature point are determined from the second stereo feature set, which are respectively recorded as the first stereo comparison point and the second stereo comparison point. It can be understood that the first stereo comparison point and the second stereo comparison point can be connected into a line.

[0061] In some examples, the present application will respectively obtain a first stereo difference between the target stereo feature point and the first stereo comparison point, obtain a second stereo difference between the target stereo feature point and the second stereo comparison point, and obtain a third stereo difference between the first stereo comparison point and the second stereo comparison point; illustratively, the i-th target stereo feature point in the first stereo feature set is recorded as , the first stereo comparison point corresponding to the i-th target stereo feature point is recorded as , the second stereo comparison point corresponding to the i-th target stereo feature point is recorded as , then the first stereo difference between the target stereo feature point and the first stereo comparison point is The second stereo difference between the target stereo feature point and the second stereo comparison point is , the third stereo difference between the first stereo comparison point and the second stereo comparison point is .

[0062] In some examples, the present application calculates the stereo feature residual corresponding to the target stereo feature point according to the first stereo difference, the second stereo difference, and the third stereo difference; specifically, the first stereo difference and the second stereo difference are cross-multiplied, and the absolute value of the paradigm obtained by the cross multiplication is calculated by ratio with the absolute value of the third stereo difference to obtain the stereo feature residual, as follows:

[0063] ;

[0064] in, is the stereo feature residual, is the i-th target stereo feature point, is the first stereo comparison point corresponding to the i target stereo feature points, is the second stereo comparison point corresponding to the i target stereo feature points.

[0065] According to the technical solution provided in the embodiment of the present application, each stereo feature point in the first stereo feature set is taken as the target stereo feature point in turn, and the second stereo feature set is traversed to determine the first stereo comparison point and the second stereo comparison point corresponding to the target feature point; the first stereo difference between the target stereo feature point and the first stereo comparison point is obtained, the second stereo difference between the target stereo feature point and the second stereo comparison point is obtained, and the third stereo difference between the first stereo comparison point and the second stereo comparison point is obtained; according to the first stereo difference, the second stereo difference and the third stereo difference, the stereo feature residual corresponding to the target stereo feature point is calculated. The above method realizes the accurate acquisition of the stereo feature residual corresponding to each target stereo feature point according to the stereo feature points in the map database.

[0066] In some embodiments, Figure 5 As shown, according to the first plane feature set and the second plane feature set, the plane feature residual is calculated, including:

[0067] S501, taking each plane feature point in the first plane feature set as a target plane feature point, traversing the second plane feature set to determine a first plane comparison point, a second plane comparison point, and a third plane comparison point corresponding to the target plane feature point, wherein the first plane comparison point, the second plane comparison point, and the third plane comparison point are three points under one plane;

[0068] S502, obtaining a first plane difference between the target plane feature point and the third plane comparison point, obtaining a second plane difference between the second plane comparison point and the third plane comparison point, and obtaining a third plane difference between the first plane comparison point and the third plane comparison point;

[0069] S503: Calculate the plane feature residual corresponding to the target plane feature point according to the first plane difference, the second plane difference and the third plane difference.

[0070] Specifically, in some examples, the present application will take each plane feature point in the first plane feature set as the target plane feature point in turn. For example, the first plane feature set contains N plane feature points, and the first plane feature point is taken as the first target plane feature point, the second plane feature point is taken as the second target plane feature point... and the Nth plane feature point is taken as the Nth target plane feature point.

[0071] It can be understood that after determining the target plane feature point, the present application will traverse the second plane feature set to determine the first plane comparison point, the second plane comparison point and the third plane comparison point corresponding to the target feature point; specifically, the target plane feature point is compared with each plane feature point in the second plane feature set, and the three plane feature points closest to the target plane feature point are determined from the second plane feature set, which are respectively recorded as the first plane comparison point, the second plane comparison point and the third plane comparison point. It can be understood that the first plane comparison point, the second plane comparison point and the third plane comparison point are three points under a plane, that is, the first plane comparison point, the second plane comparison point and the third plane comparison point can be connected into a plane.

[0072] In some examples, the present application obtains a first plane difference between a target plane feature point and a third plane comparison point, obtains a second plane difference between a second plane comparison point and a third plane comparison point, and obtains a third plane difference between the first plane comparison point and the third plane comparison point; illustratively, the kth target plane feature point in the first plane feature set is recorded as , the first plane comparison point corresponding to the kth target plane feature point is recorded as , the second plane comparison point corresponding to the kth target plane feature point is recorded as , the second plane comparison point corresponding to the kth target plane feature point is recorded as , then the first plane difference between the target plane feature point and the third plane comparison point is , the second plane difference between the second plane comparison point and the third plane comparison point is , get the third plane difference between the first plane comparison point and the third plane comparison point as .

[0073] Finally, the plane feature residual corresponding to the target plane feature point is calculated according to the first plane difference, the second plane difference and the third plane difference; specifically, the first plane difference is multiplied by the cross product of the second plane difference and the third plane difference, and then the absolute value of the paradigm obtained by the product operation is calculated by ratio with the absolute value of the paradigm obtained by the cross product of the second plane difference and the third plane difference to obtain the plane feature residual, as follows:

[0074] ;

[0075] in, is the plane feature residual, is the kth target plane feature point, is the first plane comparison point corresponding to the kth target plane feature point, is the second plane comparison point corresponding to the kth target plane feature point, is the third plane comparison point corresponding to the kth target plane feature point.

[0076] According to the technical solution provided in the embodiment of the present application, each plane feature point in the first plane feature set is taken as the target plane feature point in turn, and the second plane feature set is traversed to determine the first plane comparison point, the second plane comparison point and the third plane comparison point corresponding to the target plane feature point, the first plane comparison point, the second plane comparison point and the third plane comparison point are three points under a plane; the first plane difference between the target plane feature point and the third plane comparison point is obtained, the second plane difference between the second plane comparison point and the third plane comparison point is obtained, and the third plane difference between the first plane comparison point and the third plane comparison point is obtained; according to the first plane difference, the second plane difference and the third plane difference, the plane feature residual corresponding to the target plane feature point is calculated, and the above method realizes the accurate acquisition of the plane feature residual corresponding to each target plane feature point according to the plane feature points in the map database.

[0077] In some embodiments, the vehicle odometer position includes: a predicted horizontal coordinate value, a predicted vertical coordinate value, and a predicted angle; Figure 6 As shown, the vehicle odometer posture at the current moment is determined according to the angular velocity and the driving speed, including:

[0078] S601, constructing a calibration coordinate system and constructing a motion model of the vehicle, wherein the motion model is parallel to the calibration coordinate system;

[0079] S602, obtaining the horizontal coordinate value, the vertical coordinate value and the angle of the motion model of the vehicle at the previous moment in the calibration coordinate system;

[0080] S603, calculating the horizontal axis movement distance of the vehicle in the calibrated coordinate system according to the vehicle's driving speed and the angle at the last moment; calculating the vertical axis movement distance of the vehicle in the calibrated coordinate system according to the vehicle's driving speed and the angle at the last moment; calculating the angular movement value of the vehicle in the calibrated coordinate system according to the angular velocity of the vehicle input into the movement angle determination formula;

[0081] S604, calculate the predicted horizontal coordinate value according to the horizontal coordinate value and the horizontal axis moving distance at the previous moment; calculate the predicted vertical coordinate value according to the vertical coordinate value and the vertical axis moving distance at the previous moment; calculate the predicted angle according to the angle at the previous moment and the angle moving value.

[0082] Specifically, when the vehicle starts, a calibration coordinate system is constructed. This example does not limit the origin position of the calibration coordinate system. Then, a motion model of the vehicle is constructed in the calibration coordinate system. When the vehicle starts, the angle θ of the vehicle's motion model in the calibration coordinate system is 0, and the calibration coordinate system is kept parallel to the vehicle body.

[0083] It can be understood that when the vehicle is driving, it moves linearly. Therefore, the position of the vehicle's motion model at the current moment in the calibrated coordinate system is correlated with the position of the vehicle's motion model at the previous moment in the calibrated coordinate system. For this reason, this example will obtain the horizontal coordinate value, vertical coordinate value and angle of the vehicle's motion model at the previous moment in the calibrated coordinate system. The horizontal coordinate value, vertical coordinate value and angle of the previous moment can represent the position of the vehicle's motion model at the previous moment in the calibrated coordinate system.

[0084] In some examples, the present application will input the vehicle's driving speed and the angle at the last moment into the horizontal axis movement distance determination formula to calculate the horizontal axis movement distance of the vehicle in the calibration coordinate system between the last moment and the current moment. The horizontal axis movement distance determination formula is as follows:

[0085] Horizontal axis movement distance = ;

[0086] in, is the vehicle’s speed, is the angle of the vehicle's motion model in the calibration coordinate system at the previous moment, t is the previous moment, and t+1 is the current moment.

[0087] In some examples, the present application will input the vehicle's speed and angle into the longitudinal axis movement distance determination formula to calculate the longitudinal axis movement distance of the vehicle in the calibration coordinate system between the previous moment and the current moment. The longitudinal axis movement distance determination formula is as follows:

[0088] Vertical axis moving distance = ;

[0089] in, is the vehicle’s speed, is the angle of the vehicle in the calibration coordinate system, t is the previous moment, and t+1 is the current moment.

[0090] In some examples, the present application will input the angular velocity of the vehicle into the movement angle determination formula to calculate the angular movement value of the vehicle in the calibration coordinate system. The movement angle determination formula is as follows:

[0091] Angle shift value = ;

[0092] in, is the vehicle’s speed, is the angular velocity of the vehicle, t is the previous moment, and t+1 is the current moment.

[0093] In some examples, the present application calculates the predicted horizontal coordinate value according to the horizontal coordinate value at the previous moment and the horizontal axis movement distance; specifically, the predicted horizontal coordinate value (that is, the horizontal coordinate value at the current moment) is recorded as , the horizontal coordinate value of the previous moment is recorded as ,but ,in, is the horizontal axis movement distance, which will not be described here.

[0094] In some examples, the present application calculates the predicted ordinate value based on the ordinate value at the previous moment and the ordinate movement distance; specifically, the predicted ordinate value (that is, the ordinate value at the current moment) is recorded as , the ordinate value of the previous moment is recorded as ,but ,in, is the vertical axis movement distance, which will not be described here.

[0095] In some examples, the present application calculates the predicted angle based on the angle and the angle movement value at the previous moment; specifically, the predicted angle (that is, the angle at the current moment) is recorded as , the angle at the previous moment is recorded as ,but ,in, is the angle movement value and will not be described here.

[0096] According to the technical solution provided in the embodiment of the present application, a calibration coordinate system is constructed, and a motion model of the vehicle is constructed, wherein the motion model is parallel to the calibration coordinate system; the horizontal coordinate value, vertical coordinate value and angle of the motion model of the vehicle at the previous moment in the calibration coordinate system are obtained; the vehicle's driving speed and the angle at the previous moment are input into the horizontal axis movement distance determination formula to calculate the horizontal axis movement distance of the vehicle in the calibration coordinate system; the vehicle's driving speed and the angle at the previous moment are input into the vertical axis movement distance determination formula to calculate the vertical axis movement distance of the vehicle in the calibration coordinate system; the vehicle's angular velocity is input into the movement angle determination formula to calculate the angular movement value of the vehicle in the calibration coordinate system; based on the horizontal coordinate value and the horizontal axis movement distance at the previous moment, a predicted horizontal coordinate value is calculated; based on the vertical coordinate value and the vertical axis movement distance at the previous moment, a predicted vertical coordinate value is calculated; based on the angle and the angular movement value at the previous moment, a predicted angle is calculated, wherein the preparation for determining the vehicle odometer posture is realized by the above method.

[0097] In some embodiments, Figure 7 As shown, based on the shooting pose and the vehicle odometer pose, the error pose is calculated, including:

[0098] S701, obtaining the acquisition time of the shooting posture, and performing difference processing on multiple vehicle odometer postures acquired in sequence according to the acquisition time of the shooting posture to obtain multiple difference vehicle odometer postures;

[0099] S702, calculating a shooting posture increment between two adjacent shooting postures, and calculating a vehicle odometer posture increment between two adjacent difference vehicle odometer postures;

[0100] S703: Calculate an error pose based on multiple shooting pose increments and multiple vehicle odometer pose increments.

[0101] In some examples, due to the difference between the shooting frequency of the shooting device and the measurement frequency of the odometer posture measurement device, in order to achieve time synchronization between the two and to make an accurate comparison between the shooting posture and the vehicle odometer posture at the same moment, this example uses a linear interpolation method to obtain the shooting posture and the vehicle odometer posture at the same moment.

[0102] For example, assuming that the shooting frequency of the shooting device is usually 30 Hz, the acquisition frequency of the shooting posture is 30 Hz, and the measurement frequency of the odometer posture measurement device is 50 Hz, then the acquisition frequency of the vehicle odometer posture is 50 Hz. In order to achieve time synchronization between the two and to make an accurate comparison between the shooting posture and the vehicle odometer posture at the same time, this example uses a linear interpolation method.

[0103] In some examples, the present application obtains the acquisition time of the shooting posture, and performs difference processing on multiple vehicle odometer postures acquired in sequence according to the acquisition time of the shooting posture to obtain multiple difference vehicle odometer postures, such as Figure 8 As shown, this example obtains time, Shooting position at the moment and , get The corresponding vehicle odometer pose { },exist time, Perform linear interpolation on the vehicle odometer posture at all times to obtain and The difference vehicle odometer posture b0 and b1 corresponding to the time.

[0104] After determining the above-mentioned difference vehicle odometer postures b0 and b1, this example can obtain Moment and The vehicle odometer position increment at the time; specifically, b1-b0 can be obtained The time corresponds to the vehicle odometer posture increment. Similarly, by a1-a0, we can get The time corresponds to the shooting pose increment.

[0105] In some examples, after calculating the shooting posture increment between two adjacent shooting postures and the vehicle odometer posture increment between two adjacent difference vehicle odometer postures, the present application calculates the error posture based on multiple shooting posture increments and multiple vehicle odometer posture increments. Specifically, take the shooting posture increments in the same time period respectively. and vehicle odometer posture increment After that, the shooting pose increment is taken as the true value, and the vehicle odometer pose increment is taken as the measured value. Assume there are 10 test samples: , we need to find a 6-dimensional variable So that the loss function Minimum, obtain the local minimum.

[0106] ;

[0107] For the above linear transformation equation, it is an overdetermined equation form, thus turning to a linear least squares problem: ; Let the first-order derivative be 0, the approximate solution = . is the error pose, It can be used to calibrate the camera and the vehicle. :Convex function : Residual function, that is, the difference between the measured value and the true value. : Coefficient matrix, a 6x6 matrix; : Augmented matrix, a 6x1 matrix.

[0108] According to the technical solution provided in the embodiments of the present application, the acquisition frequency of the shooting posture and the acquisition frequency of the vehicle odometer posture are obtained; according to the acquisition frequency of the shooting posture and the acquisition frequency of the vehicle odometer posture, the multiple vehicle odometer postures acquired in sequence are subjected to difference processing to obtain multiple difference vehicle odometer postures; the shooting posture increment between two adjacent shooting postures is calculated, and the vehicle odometer posture increment between two adjacent difference vehicle odometer postures is calculated; based on the multiple shooting posture increments and the multiple vehicle odometer posture increments, the error posture is calculated.

[0109] In some examples, such as Fig. 9 As shown, after calibrating the camera device and the vehicle based on the error posture, the method further includes:

[0110] S901, obtaining shooting data of a shooting device, where the shooting data is data shot after the shooting device and the vehicle are calibrated;

[0111] S902: Determine an assisted driving strategy according to the photographing data, and perform auxiliary control on the vehicle according to the assisted driving strategy.

[0112] Specifically, after the camera and the vehicle are calibrated, the shooting data observed by the camera will be converted to the calibration coordinate system corresponding to the vehicle so that the vehicle can accurately understand and respond to the surrounding environment.

[0113] Based on the above reasons, after obtaining the shooting data of the shooting device, the present application can determine the assisted driving strategy according to the shooting data, and perform auxiliary control on the vehicle according to the assisted driving strategy, thereby accurately determining the assisted driving strategy.

[0114] In order to better understand the present application, this embodiment provides a more specific example for illustration. Fig.10As shown in step S1001 to step S1009, the present application directly obtains a high-precision map database of the calibration scene from a map vendor or generates a calibration scene through point cloud data annotation by the OEM, and classifies and stores the map elements broadcast by the map engine according to plane features and three-dimensional features to obtain a second three-dimensional feature set and a second plane feature set. The image frame captured by the shooting camera is semantically extracted from the trained map elements based on a deep learning method to obtain a first three-dimensional feature set and a first plane feature set. The two topics are also broadcast in real time according to the plane features and the three-dimensional features. The three-dimensional feature residual between the first three-dimensional feature set and the second three-dimensional feature set is calculated using a two-step LM optimization algorithm, and the plane feature residual between the first plane feature set and the second plane feature set is calculated. The motion of the shooting device between adjacent image frames is estimated as the shooting posture according to the three-dimensional feature residual and the plane feature residual. On the other hand, for vehicles, a vehicle kinematic model based on Ackerman steering can be constructed to obtain the angular velocity and driving speed according to the vehicle kinematic model, and then the vehicle odometer posture can be determined in combination with the vehicle driving speed.

[0115] The linear least square method is used to solve the error between the shooting position and the vehicle odometer posture, and then the transformation equation between the shooting posture increment and the vehicle odometer posture increment is obtained, and the approximate solution is obtained, based on which the shooting device and the vehicle are calibrated. Thus, a high-precision camera and vehicle self-calibration system that does not rely on scene assumptions and artificial mechanical parameter measurements is realized.

[0116] The present application provides a vehicle calibration method that can make full use of high-precision map data, overcome the high-precision calibration plate production based on traditional calibration methods and the high standards such as scene lighting and flatness in the calibration room, and can also show good robustness and convenience in vehicle off-line calibration and after-sales calibration.

[0117] According to the technical solution provided in the embodiment of the present application, shooting data of the shooting device is obtained, and the shooting data is the data shot after the shooting device and the vehicle are calibrated; the assisted driving strategy is determined based on the shooting data, and the vehicle is assisted in control according to the assisted driving strategy, thereby realizing accurate acquisition of the assisted driving strategy.

[0118] All the above optional technical solutions can be arbitrarily combined to form optional embodiments of the present application, which will not be described one by one here.

[0119] The following are device embodiments of the present application, which can be used to execute the method embodiments of the present application. For details not disclosed in the device embodiments of the present application, please refer to the method embodiments of the present application.

[0120] This embodiment also provides a vehicle calibration device, such as Fig.11 As shown, the device comprises:

[0121] The shooting posture acquisition module 1101 is configured to extract contour feature points of the image frame to obtain a first stereo feature set and a first plane feature set; and determine the shooting posture corresponding to the shooting device according to the first stereo feature set and the first plane feature set;

[0122] The odometer posture acquisition module 1102 is configured to acquire the angular velocity and driving speed of the vehicle, and determine the odometer posture of the vehicle according to the angular velocity and the driving speed;

[0123] The calibration module 1103 is configured to calculate an error pose based on the shooting pose and the vehicle odometer pose, and calibrate the shooting device and the vehicle based on the error pose.

[0124] In some examples, the shooting posture acquisition module 1101 is also configured to obtain a map database corresponding to the current position of the vehicle, the map database containing stereo features and plane features; construct a second stereo feature set based on the stereo features in the map database, and construct a second plane feature set based on the plane features in the map database; calculate the stereo feature residual based on the first stereo feature set and the second stereo feature set; calculate the plane feature residual based on the first plane feature set and the second plane feature set; and calculate the shooting posture based on the stereo feature residual and the plane feature residual.

[0125] In some examples, the shooting posture acquisition module 1101 is also configured to take each stereo feature point in the first stereo feature set as a target stereo feature point, traverse the second stereo feature set to determine the first stereo comparison point and the second stereo comparison point corresponding to the target stereo feature point; obtain the first stereo difference between the target stereo feature point and the first stereo comparison point, obtain the second stereo difference between the target stereo feature point and the second stereo comparison point, and obtain the third stereo difference between the first stereo comparison point and the second stereo comparison point; and calculate the stereo feature residual corresponding to the target stereo feature point based on the first stereo difference, the second stereo difference and the third stereo difference.

[0126] In some examples, the shooting posture acquisition module 1101 is also configured to take each plane feature point in the first plane feature set as a target plane feature point, traverse the second plane feature set to determine the first plane comparison point, the second plane comparison point and the third plane comparison point corresponding to the target plane feature point, the first plane comparison point, the second plane comparison point and the third plane comparison point are three points under a plane; obtain the first plane difference between the target plane feature point and the third plane comparison point, obtain the second plane difference between the second plane comparison point and the third plane comparison point, and obtain the third plane difference between the first plane comparison point and the third plane comparison point; calculate the plane feature residual corresponding to the target plane feature point according to the first plane difference, the second plane difference and the third plane difference.

[0127] In some examples, the vehicle odometer posture includes: a predicted horizontal coordinate value, a predicted vertical coordinate value, and a predicted angle; the odometer posture acquisition module 1102 is also configured to construct a calibration coordinate system and a motion model of the vehicle, which is parallel to the calibration coordinate system; obtain the horizontal coordinate value, vertical coordinate value, and angle of the motion model of the vehicle in the calibration coordinate system at the previous moment; calculate the horizontal axis movement distance of the vehicle in the calibrated coordinate system according to the vehicle's driving speed and the angle at the previous moment; calculate the vertical axis movement distance of the vehicle in the calibrated coordinate system according to the vehicle's driving speed and the angle at the previous moment; calculate the angular movement value of the vehicle in the calibrated coordinate system according to the vehicle's angular velocity; calculate the predicted horizontal coordinate value based on the horizontal coordinate value and the horizontal axis movement distance at the previous moment; calculate the predicted vertical coordinate value based on the vertical coordinate value and the vertical axis movement distance at the previous moment; calculate the predicted angle based on the angle and the angular movement value at the previous moment.

[0128] In some examples, the calibration module 1103 is also configured to obtain the acquisition time of the shooting posture, perform difference processing on multiple vehicle odometer postures acquired in sequence according to the acquisition time of the shooting posture, and obtain multiple difference vehicle odometer postures; calculate the shooting posture increment between two adjacent shooting postures, and calculate the vehicle odometer posture increment between two adjacent difference vehicle odometer postures; calculate the error posture based on multiple shooting posture increments and multiple vehicle odometer posture increments.

[0129] In some examples, the calibration module 1103 is also configured to obtain shooting data from the shooting device, where the shooting data is data shot after the shooting device and the vehicle are calibrated; determine an assisted driving strategy based on the shooting data, and perform auxiliary control on the vehicle based on the assisted driving strategy.

[0130] According to the technical solution provided in the embodiment of the present application, the vehicle calibration device provided in the embodiment extracts contour feature points of the image frame to obtain a first stereo feature set and a first plane feature set; determines the shooting posture corresponding to the shooting device according to the first stereo feature set and the first plane feature set; obtains the angular velocity and the driving speed of the vehicle, and determines the vehicle odometer posture according to the angular velocity and the driving speed; calculates the error posture based on the shooting posture and the vehicle odometer posture, and calibrates the shooting device and the vehicle based on the error posture, with beneficial effects: wherein, the above method first obtains the corresponding shooting posture and the vehicle odometer posture respectively by self-calibrating the shooting device and the vehicle, thereby ensuring high precision and automation of the calibration process; then determines the error posture through the shooting posture and the vehicle odometer posture, and the error posture can characterize the relative posture of the shooting device relative to the vehicle, and subsequently the shooting device and the vehicle coordinate system can be accurately calibrated through the error posture, thereby avoiding the problem in the related art that the camera and the vehicle coordinate system cannot be accurately calibrated, resulting in limited automatic driving function.

[0131] Fig.12 Schematic diagram of an electronic device 12 provided in an embodiment of the present application. Fig.12 As shown, the electronic device 12 of this embodiment includes: a processor 1201, a memory 1202, and a computer program 1203 stored in the memory 1202 and executable on the processor 1201. When the processor 1201 executes the computer program 1203, the steps in the above-mentioned method embodiments are implemented. Alternatively, when the processor 1201 executes the computer program 1203, the functions of the modules / units in the above-mentioned device embodiments are implemented.

[0132] The electronic device 12 may be a desktop computer, a notebook, a PDA, a cloud server, or other electronic device. The electronic device 12 may include, but is not limited to, a processor 1201 and a memory 1202. Those skilled in the art will appreciate that Fig.12 The electronic device 12 is merely an example and does not limit the electronic device 12 , and may include more or less components than those shown in the figure, or different components.

[0133] The processor 1201 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0134] The memory 1202 may be an internal storage unit of the electronic device 12, for example, a hard disk or memory of the electronic device 12. The memory 1202 may also be an external storage device of the electronic device 12, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card (FlashCard), etc. equipped on the electronic device 12. The memory 1202 may also include both an internal storage unit of the electronic device 12 and an external storage device. The memory 1202 is used to store computer programs and other programs and data required by the electronic device.

[0135] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units.

[0136] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. The computer program may include computer program code, and the computer program code may be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of regional requirements and patent practice. For example, in some regions, according to regional requirements and patent practice, the computer-readable medium does not include electric carrier signals and telecommunication signals.

[0137] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A vehicle calibration method, characterized in that: The method comprises: Extracting contour feature points from the image frame to obtain a first three-dimensional feature set and a first plane feature set; Determining a shooting posture corresponding to a shooting device according to the first three-dimensional feature set and the first planar feature set; Obtaining the angular velocity and driving speed of the vehicle, and determining the vehicle odometer posture according to the angular velocity and the driving speed; Acquire the acquisition time of the shooting posture, perform difference processing on the multiple vehicle odometer postures acquired in sequence according to the acquisition time of the shooting posture, and obtain multiple difference vehicle odometer postures; calculate the shooting posture increment between two adjacent shooting postures, and calculate the vehicle odometer posture increment between two adjacent difference vehicle odometer postures; calculate the error posture based on the multiple shooting posture increments and the multiple vehicle odometer posture increments, and calibrate the shooting device and the vehicle based on the error posture.

2. The method according to claim 1, characterized in that Determining a shooting posture corresponding to the shooting device according to the first three-dimensional feature set and the first planar feature set includes: Acquire a map database corresponding to the current position of the vehicle, wherein the map database contains three-dimensional features and two-dimensional features; Constructing a second set of three-dimensional features according to the three-dimensional features in the map database, and constructing a second set of plane features according to the plane features in the map database; Calculating a stereo feature residual according to the first stereo feature set and the second stereo feature set; Calculating a plane feature residual according to the first plane feature set and the second plane feature set; The shooting posture is calculated according to the stereo feature residual and the planar feature residual.

3. The method according to claim 2, characterized in that Calculating a stereo feature residual according to the first stereo feature set and the second stereo feature set includes: Taking each stereo feature point in the first stereo feature set as a target stereo feature point, traversing the second stereo feature set to determine a first stereo comparison point and a second stereo comparison point corresponding to the target stereo feature point; Acquire a first stereo difference between the target stereo feature point and the first stereo comparison point, acquire a second stereo difference between the target stereo feature point and the second stereo comparison point, and acquire a third stereo difference between the first stereo comparison point and the second stereo comparison point; The stereo feature residual corresponding to the target stereo feature point is calculated according to the first stereo difference, the second stereo difference and the third stereo difference.

4. The method according to claim 2, characterized in that: Calculating a plane feature residual according to the first plane feature set and the second plane feature set includes: Taking each plane feature point in the first plane feature set as a target plane feature point, traversing the second plane feature set to determine a first plane comparison point, a second plane comparison point, and a third plane comparison point corresponding to the target plane feature point, wherein the first plane comparison point, the second plane comparison point, and the third plane comparison point are three points under a plane; Acquire a first plane difference between the target plane feature point and the third plane comparison point, acquire a second plane difference between the second plane comparison point and the third plane comparison point, and acquire a third plane difference between the first plane comparison point and the third plane comparison point; The plane feature residual corresponding to the target plane feature point is calculated according to the first plane difference, the second plane difference and the third plane difference.

5. The method according to claim 1, characterized in that The vehicle odometer posture includes: a predicted horizontal coordinate value, a predicted vertical coordinate value, and a predicted angle; determining the vehicle odometer posture at the current moment according to the angular velocity and the driving speed includes: Constructing a calibration coordinate system and constructing a motion model of the vehicle, wherein the motion model is parallel to the calibration coordinate system; Obtaining the horizontal coordinate value, the vertical coordinate value and the angle of the motion model of the vehicle at the last moment in the calibration coordinate system; Calculating the horizontal axis movement distance of the vehicle in the calibrated coordinate system according to the vehicle's travel speed and the angle at the last moment; calculating the vertical axis movement distance of the vehicle in the calibrated coordinate system according to the vehicle's travel speed and the angle at the last moment; calculating the angular movement value of the vehicle in the calibrated coordinate system according to the angular velocity of the vehicle; The predicted horizontal coordinate value is calculated based on the horizontal coordinate value at the previous moment and the horizontal axis moving distance; the predicted vertical coordinate value is calculated based on the vertical coordinate value at the previous moment and the vertical axis moving distance; the predicted angle is calculated based on the angle at the previous moment and the angle moving value.

6. The method according to claim 1, characterized in that After calibrating the camera and the vehicle based on the error posture, the method further includes: Acquire shooting data of the shooting device, where the shooting data is data shot after the shooting device and the vehicle are calibrated; An assisted driving strategy is determined according to the shooting data, and the vehicle is assisted in control according to the assisted driving strategy.

7. A vehicle calibration device, characterized in that: The device comprises: A shooting posture acquisition module is configured to extract contour feature points of the image frame to obtain a first stereo feature set and a first plane feature set; and determine a shooting posture corresponding to the shooting device according to the first stereo feature set and the first plane feature set; The odometer posture acquisition module is configured to acquire the angular velocity and the driving speed of the vehicle, and determine the vehicle odometer posture according to the angular velocity and the driving speed; The calibration module is configured to obtain the acquisition time of the shooting posture, perform difference processing on the multiple vehicle odometer postures acquired in sequence according to the acquisition time of the shooting posture, and obtain multiple difference vehicle odometer postures; calculate the shooting posture increment between two adjacent shooting postures, and calculate the vehicle odometer posture increment between two adjacent difference vehicle odometer postures; calculate the error posture based on the multiple shooting posture increments and the multiple vehicle odometer posture increments, and calibrate the shooting device and the vehicle based on the error posture.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images