Vehicle-mounted monocular camera motion scale estimation method and device, computer equipment and medium

Abstract

The embodiment of the invention provides a method for estimating the motion scale of a vehicle-mounted monocular camera, and relates to the technical field of vehicle-mounted driving, the vehicle-mounted monocular camera obtains a static vanishing point, a front blind area line and the conversion Tcv from a self-vehicle coordinate system to a vehicle-mounted monocular camera coordinate system in the installation and calibration process, and the method comprises the following steps: carrying out feature extraction on ground features; a driving area is divided; outputting a camera attitude, and determining a self-vehicle attitude of the self-vehicle at the nth frame according to the camera attitude; drawing a feature map of the driving area, obtaining ground feature points according to the posture of the vehicle at the nth frame, and obtaining a ground point set; and calculating to obtain the physical scale of the pose change of the current frame. According to the scale estimation method, a feature extraction module of the odometer is optimized, a feature map provided by the odometer is reused, ground points are extracted from the feature map by utilizing prior information such as a spatial position relationship, and an accurate motion scale is estimated by combining camera height information, so that the system complexity is reduced, and the computing power overhead generated by an additionally developed framework is saved.

Description

Technical Field

[0001] The present invention relates to the field of vehicle-mounted driving technology, and in particular to a method, device, computer equipment and medium for estimating the motion scale of a vehicle-mounted monocular camera. Background Art

[0002] One of the technical difficulties of monocular vision-based simultaneous localization and mapping (SLAM) is scale uncertainty: the monocular vision SLAM system cannot directly obtain the absolute scale information of the object, so when building the map, the system cannot determine the actual size of the scene.

[0003] In the field of intelligent driving, the system needs to understand the distance and size of objects in the real world. The scale-free maps and trajectories obtained by the monocular vision SLAM system cannot provide accurate location information and environmental understanding in the field of intelligent driving.

[0004] In order to estimate the scale, we can add IMU sensors or wheel speed meters to the single-purpose sensors to obtain scale information, or build a SLAM system based on sensors that directly obtain scale information such as binocular cameras, depth cameras, and lidars. These solutions will not only increase the hardware cost, but also dramatically increase the system complexity.

[0005] Therefore, it is necessary to provide a method for realizing scale estimation in driving scenarios using only a monocular camera to improve the stability of existing intelligent driving systems. Summary of the invention

[0006] In view of this, an embodiment of the present invention provides a method for estimating the motion scale of a vehicle-mounted monocular camera to solve the problem in the prior art that the scale of an object cannot be constructed using only a monocular vehicle-mounted camera, so that the actual size of the scene cannot be determined. The method includes:

[0007] Extract ground features from images obtained by the on-board monocular camera and divide the driving area;

[0008] Output the camera posture, and determine the posture of the vehicle at the current n-th frame image according to the camera posture;

[0009] In the camera coordinate system of the initial frame image, a feature map of the driving area is drawn, and ground feature points are obtained on the feature map according to the posture of the vehicle at the current n-th frame image. The ground feature points form a ground point set in the current n-th frame self-vehicle coordinate system;

[0010] According to the ground points in the ground point set, the physical height of the camera in the real world and the scale of the current n-th frame image, the physical scale of the posture change of the current n-th frame image is calculated.

[0011] The method for estimating the motion scale of a vehicle-mounted monocular camera provided by the present invention also has the following characteristics: extracting the ground features in the image obtained by the vehicle-mounted monocular camera comprises:

[0012] The feature point extraction is performed by using any method selected from the feature point method, the direct method and the semi-direct method.

[0013] The method for estimating the motion scale of a vehicle-mounted monocular camera provided by the present invention also has the following characteristics: the driving area is represented by a triangle, and the three points of the triangle are A1, A2 and A3, respectively, wherein A1 is the static vanishing point of the image, and A2 and A3 are two points where the front blind spot line of the car extends to the left and right edges of the image, respectively, and the static vanishing point and the front blind spot line are obtained during the installation and calibration process of the vehicle-mounted monocular camera.

[0014] The vehicle-mounted monocular camera motion scale estimation method provided by the present invention also has the following characteristics: the vehicle posture at the current nth frame image is determined according to the camera posture. include:

[0015]

[0016] Among them, T 0n is the camera pose of the nth frame image, T cv is the transformation matrix from the vehicle coordinate system to the vehicle monocular camera coordinate system.

[0017] The method for estimating the motion scale of a vehicle-mounted monocular camera provided by the present invention also has the following characteristics: in the camera coordinate system of the initial frame image, a feature map of the driving area is drawn, and ground feature points are obtained on the feature map according to the posture of the vehicle at the current n-th frame image, and the ground feature points form a ground point set in the vehicle coordinate system of the current n-th frame image, including:

[0018] According to the ego-vehicle pose of the current n-th frame image and the set of local feature points in the camera coordinate system of the current n-th frame image, each set of local feature points in the ego-vehicle coordinate system of the current n-th frame image is obtained, wherein the origin of the camera coordinate system coincides with the origin of the ego-vehicle coordinate system;

[0019] Determine whether each local feature point is a ground point, remove non-ground points, and obtain the ground point set in the vehicle coordinate system of the current n-th frame image.

[0020] The vehicle-mounted monocular camera motion scale estimation method provided by the present invention also has the following characteristics: the step of determining whether each local feature point is a ground point comprises:

[0021] Determine whether the Z coordinate of the local feature point is less than 0. If it is less than 0, keep the local feature point. If it is greater than or equal to 0, the local feature point is a non-ground point and is deleted.

[0022] Take the mean Z coordinate of the retained local feature points to get the mean Z mean ;

[0023] Draw a circle with the origin of the vehicle coordinate system of the current n-th frame image as the center and the vehicle width as the diameter, and select local feature points within the circle;

[0024] Using mean Z mean The noise points are removed from the local feature points in the circular region to obtain the ground point set in the vehicle coordinate system of the current n-th frame image.

[0025] The method for estimating the motion scale of a vehicle-mounted monocular camera provided by the present invention also has the following characteristics: the physical scale of the pose change of the current n-th frame image is calculated based on the ground points in the ground point set, the physical height of the camera in the real world, and the scale of the current n-th frame image, including:

[0026] According to the ground point set in the vehicle coordinate system of the current n-th frame image, the camera height in the map scale is obtained;

[0027] According to the camera height in the map scale and the physical height of the camera in the real world, the scale value of the current n-th frame image is obtained;

[0028] Using the scale value of the current n-th frame image, we can obtain the relationship function between the spatial distance Sreal in the physical world and the spatial distance Smap in the map:

[0029] Sreal=scale*Smap.

[0030] The second object of the present invention is to provide a vehicle-mounted monocular camera motion scale estimation device, comprising:

[0031] A feature extraction module is used to extract ground features from images obtained by the vehicle-mounted monocular camera and perform regional division;

[0032] The ego vehicle posture estimation module is used to determine the ego vehicle posture at the current nth frame image;

[0033] The ground feature recognition module is used to draw a feature map of the driving area in the camera coordinate system of the initial frame image, and obtain ground feature points on the feature map according to the posture of the vehicle at the current n-th frame image. The ground feature points form a ground point set in the current n-th frame coordinate system of the vehicle;

[0034] The scale estimation module calculates the physical scale of the pose change of the current n-th frame image based on the ground points in the ground point set, the physical height of the camera in the real world, and the scale of the current n-th frame image.

[0035] The third object of the present invention is to provide a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method for estimating the motion scale of a vehicle-mounted monocular camera as described in any one of the preceding items when executing the computer program.

[0036] A fourth object of the present invention is to provide a computer-readable storage medium, wherein the computer-readable storage medium stores a method for estimating motion scale of a vehicle-mounted monocular camera as described in any of the above.

[0037] Compared with the prior art, the beneficial effects that can be achieved by at least one of the above-mentioned technical solutions adopted in the embodiments of this specification include at least the following: the technical solution provided by the present invention does not develop an additional framework to restore the monocular scale, but optimizes the feature extraction module of the odometer, and reuses the feature map provided by the odometer, and extracts ground points from it using prior information such as spatial position relationships, and estimates the accurate motion scale in combination with the camera height information, thereby reducing the system complexity and saving the computing power overhead generated by the additionally developed framework; the technical solution cleverly uses the spatial position constraint relationship between the map points and the vehicle in the driving scenario, and can achieve the estimation of ground points and the recovery of motion scale with very little computing power resource overhead; the technical solution is adapted to monocular visual odometer frameworks including feature point method, direct method, semi-direct method, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments 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 creative work.

[0039] Figure 1 is a flow chart of a motion scale estimation method provided by an embodiment of the present invention;

[0040] Figure 2 is a schematic diagram of a driving area in an embodiment of the present invention;

[0041] Figure 3 This is a visualization example of visual odometer mapping mentioned in an embodiment of the present invention;

[0042] Figure 4 It is a schematic diagram of generating feature points of a visual odometer map mentioned in an embodiment of the present invention;

[0043] Figure 5 is a visualization diagram of ground feature point recognition in an embodiment of the present invention;

[0044] Figure 6is a comparison diagram of the real trajectory and the scale-restored trajectory in an embodiment of the present invention;

[0045] Figure 7 is a structural block diagram of a computer device provided by an embodiment of the present invention;

[0046] Figure 8 The present invention is a structural block diagram of a vehicle-mounted camera motion scale estimation device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0048] The following describes the implementation methods of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The present application can also be implemented or applied through other different specific implementation methods, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that, in the absence of conflict, the following embodiments and the features in the embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work belong to the scope of protection of the present application.

[0049] In the embodiment of the present invention, Figure 1 As shown, a method for estimating motion scale of a vehicle-mounted monocular camera is provided, the method comprising:

[0050] S1: Extract ground features from the image obtained by the on-board monocular camera and divide the driving area;

[0051] S2: Output the camera posture, and determine the posture of the vehicle at the current nth frame image according to the camera posture;

[0052] S3: In the camera coordinate system of the initial frame image, a feature map of the driving area is drawn, and ground feature points are obtained on the feature map according to the posture of the vehicle at the current n-th frame image, and the ground feature points form a ground point set in the current n-th frame vehicle coordinate system;

[0053] S4: Calculate the physical scale of the posture change of the current n-th frame image according to the ground points in the ground point set, the physical height of the camera in the real world, and the scale of the current n-th frame image.

[0054] In some embodiments, the feature extraction of ground features in the image obtained by the vehicle-mounted monocular camera includes: extracting feature points using any method selected from a feature point method, a direct method, and a semi-direct method.

[0055] In the above embodiment, the technical solution provided is a plug-in solution, which is implemented based on the visual mileage framework, and the framework is not limited to the feature point method (representative frameworks include ORB-SLAM), the direct method (representative frameworks include DSO), and the semi-direct method (representative frameworks include SVO). In the above frameworks, the feature point method extracts feature points in the image; the direct method does not need to extract features, and directly calculates the pose by matching pixels based on minimizing the photometric error, and the semi-direct method combines the feature point method and the direct method.

[0056] In some embodiments, full use is made of ground features, such as Figure 2 As shown in the figure, the imaging screen of the vehicle-mounted camera can be roughly divided into four parts: the upper part is the sky, the left and right parts are the grass, facilities or buildings outside the roadside, and the lower part is the driving area. For the ground part, a more relaxed threshold is set when extracting features to extract more feature points in one go. Specifically, for ORB-SLAM, the FAST corner point threshold is set to 7 in the driving area, and the threshold for other parts is set to 20. At the same time, more ground points are retained when homogenizing features; for the direct method, the image is usually evenly scattered. At this time, a scattering density of 1 / 10 is used when scattering points in the ground area, and a scattering density of 1 / 20 is used for other parts.

[0057] like Figure 2 As shown, the driving area is represented by a triangle, and the three points of the triangle are A1, A2 and A3, respectively, wherein A1 is the static vanishing point of the image, A2 and A3 are two points where the front blind spot line of the car extends to the left and right edges of the image, respectively, and the static vanishing point and the front blind spot line are obtained during the installation and calibration process of the vehicle-mounted monocular camera.

[0058] In some embodiments, the visual odometer framework directly provides the output of the camera pose: in this case, the camera pose of the initial frame is defined as T0=I(4), which represents a unit matrix with a dimension of 4, and the camera pose of the nth frame is T 0n , which represents the transformation of the map point from the current n-th frame camera coordinate system to the initial frame camera coordinate system. Determine the ego vehicle's ego vehicle posture at the current n-th frame image based on the camera posture include:

[0059]

[0060] Among them, T 0n is the camera pose of the nth frame image, T cv is the transformation matrix from the vehicle coordinate system to the vehicle monocular camera coordinate system.

[0061] like Figure 3The figure below shows an example of visualization of the monocular visual odometer map constructed by the open source framework ORB-SLAM, which consists of feature points, camera frames with poses, and camera trajectories. The visual odometer generates a feature map during operation. The entire map is drawn in the camera coordinate system of the initial frame and is measured in the relative scale at the time of monocular initialization.

[0062] like Figure 4 As shown in the figure, the map feature points (triangles in the figure) within the range of the vehicle body in the current frame have been generated in the past frames during the driving process. In theory, they are ground points, because the map feature points covered by the vehicle body cannot be suspended in the air or sunk below the ground. This is the main basis for ground feature extraction.

[0063] In some embodiments, the number of global map feature points is usually large, the number of local map feature points is small, and the feature points covered by the current vehicle body are completely covered. Therefore, when considering saving computing resources and rationality, it is sufficient to select local map feature points. In the camera coordinate system of the initial frame image, a feature map of the driving area is drawn, and ground feature points are obtained on the feature map according to the vehicle posture at the current n-th frame image. The ground feature points form a ground point set in the vehicle coordinate system of the current n-th frame image, including:

[0064] According to the current n-th frame image, the vehicle's position The local feature point set in the camera coordinate system of the current n-th frame image Get each local feature point in the current n-th frame image in the vehicle coordinate system Then the overall local feature point set is Among them, the origin of the camera coordinate system coincides with the origin of the vehicle coordinate system;

[0065] The vehicle coordinate system uses the front-left-up (FLU) coordinate system, whose origin coincides with the origin of the camera coordinate system. In this coordinate system, the normal vector of the ground is n g =(0,0,1) T ;

[0066] Determine whether each local feature point is a ground point, remove non-ground points, and obtain the ground point set in the vehicle coordinate system of the current n-th frame image.

[0067] In some embodiments, determining whether each local feature point is a ground point includes:

[0068] Determine whether the Z coordinate of the local feature point is less than 0. If it is less than 0, keep the local feature point. If it is greater than or equal to 0, the local feature point is a non-ground point and is deleted.

[0069] Take the mean Z coordinate of the retained local feature points to get the mean Zmean ;

[0070] Draw a circle with the origin of the vehicle coordinate system of the current n-th frame image as the center and the vehicle width as the diameter, and select local feature points within the circle;

[0071] Using mean Z mean The noise points are removed from the local feature points in the circular region to obtain the ground point set in the vehicle coordinate system of the current n-th frame image.

[0072] In the above embodiment, the determination of the Z coordinate ensures that the points to be screened and retained are located below the height of the camera.

[0073] In some embodiments, a circle is drawn with the origin of the vehicle coordinate system of the current frame as the center and the vehicle width as the diameter, and screening whether the local feature points are within the circle includes: determining whether the following conditions are met:

[0074]

[0075] Wherein, W is the width of the vehicle and H is the height of the camera installed from the ground.

[0076] like Figure 5 is a map constructed during a driving period, where the small black dots are the original map feature points. The large black dots are the estimated ground points, and the black lines are the camera trajectories. The final points obtained after screening are the ground points, which are converted to the current frame self-vehicle coordinate system to obtain the ground point set P = {p k |k=0,1,...,K}.

[0077] In some embodiments, the step of calculating the physical scale of the change in the posture of the current n-th frame image according to the ground points in the ground point set, the physical height of the camera in the real world, and the scale of the current n-th frame image includes:

[0078] According to the ground point set in the vehicle coordinate system of the current n-th frame image, the camera height in the map scale is obtained;

[0079] According to the camera height in the map scale and the physical height of the camera in the real world, the scale value of the current n-th frame image is obtained;

[0080] Using the scale value of the current n-th frame image, we can obtain the relationship function between the spatial distance Sreal in the physical world and the spatial distance Smap in the map:

[0081] Sreal=scale*Smap.

[0082] The scale value is the ratio of the camera height in the physical world (Hreal) to the camera height in the map scale (Hmap), that is, scale = Hreal / Hmap.

[0083] In some embodiments, for a ground point p k =(x k ,y k ,z k ) T , -z k is the camera height in the map scale, and the physical height of the camera in the real world is H, so the scale of the current frame Using the estimated scale value as a multiplier, the physical scale of the current frame pose change can be estimated.

[0084] like Figure 6 As shown, it is an example comparison of the scale estimation method provided by any of the above embodiments on the self-collected data set. It can be seen that compared with the real trajectory, the final scale estimation effect shows a good effect.

[0085] In some embodiments, a computer device is provided, such as Figure 7 As shown, it includes a memory 201, a processor 202, and a computer program stored in the memory 201 and executable on the processor 202. When the processor 202 executes the computer program, the method for estimating the motion scale of a vehicle-mounted monocular camera of any of the above embodiments is implemented.

[0086] Specifically, the computer device may be a computer terminal, a server or a similar computing device.

[0087] In some embodiments, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a method for estimating motion scale of a vehicle-mounted monocular camera described in any of the above embodiments.

[0088] Specifically, computer-readable storage media include permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer-readable storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable storage media does not include transitory media such as modulated data signals and carrier waves.

[0089] Based on the same inventive concept, an embodiment of the present invention also provides a vehicle-mounted monocular camera motion scale estimation device, as described in the following embodiments. Since the principle of solving the problem by a vehicle-mounted monocular camera motion scale estimation device is similar to that of a vehicle-mounted monocular camera motion scale estimation, the implementation of a vehicle-mounted monocular camera motion scale estimation device can refer to the implementation of a vehicle-mounted monocular camera motion scale estimation method, and the repeated parts will not be repeated. As used below, the term "unit" or "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0090] A vehicle-mounted monocular camera motion scale estimation device, the structural block diagram is as follows Figure 8 As shown, it includes a feature extraction module, a vehicle posture estimation module, a ground feature recognition module and a scale estimation module.

[0091] The feature extraction module 301 is used to extract the ground features in the image obtained by the vehicle-mounted monocular camera and perform region division;

[0092] The ego vehicle posture estimation module 302 is used to determine the ego vehicle posture at the current n-th frame image;

[0093] The ground feature recognition module 303 is used to draw a feature map of the driving area in the camera coordinate system of the initial frame image, and obtain ground feature points on the feature map according to the posture of the vehicle at the current n-th frame image, and the ground feature points form a ground point set in the current n-th frame vehicle coordinate system;

[0094] The scale estimation module 304 calculates the physical scale of the posture change of the current n-th frame image according to the ground points in the ground point set, the physical height of the camera in the real world, and the scale of the current n-th frame image.

[0095] Obviously, those skilled in the art should understand that the modules or steps of the above-mentioned embodiments of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. In this way, the embodiments of the present invention are not limited to any specific combination of hardware and software.

[0096] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the embodiments of the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for estimating motion scale of a vehicle-mounted monocular camera, characterized in that: The method comprises: Extract ground features from images obtained by the on-board monocular camera and divide the driving area; Output the camera posture, and determine the posture of the vehicle at the current n-th frame image according to the camera posture; In the camera coordinate system of the initial frame image, a feature map of the driving area is drawn, and ground feature points are obtained on the feature map according to the posture of the vehicle at the current n-th frame image. The ground feature points form a ground point set in the current n-th frame self-vehicle coordinate system; According to the ground points in the ground point set, the physical height of the camera in the real world and the scale of the current n-th frame image, the physical scale of the posture change of the current n-th frame image is calculated.

2. The method for estimating motion scale of a vehicle-mounted monocular camera according to claim 1, characterized in that: The feature extraction of ground features in the image obtained by the vehicle-mounted monocular camera includes: The feature point extraction is performed using any one of the feature point method, direct method and semi-direct method.

3. The method for estimating motion scale of a vehicle-mounted monocular camera according to claim 1, characterized in that: The driving area is represented by a triangle, and the three points of the triangle are A1, A2 and A3, respectively, wherein A1 is the static vanishing point of the image, and A2 and A3 are two points where the front blind spot line of the car extends to the left and right edges of the image, respectively. The static vanishing point and the front blind spot line are obtained during the installation and calibration process of the vehicle-mounted monocular camera.

4. The method for estimating motion scale of a vehicle-mounted monocular camera according to claim 1, characterized in that: Determine the posture of the vehicle at the current nth frame image based on the camera posture include: Among them, T 0n is the camera pose of the nth frame image, T cv is the transformation matrix from the vehicle coordinate system to the vehicle monocular camera coordinate system.

5. The method for estimating motion scale of a vehicle-mounted monocular camera according to claim 1, characterized in that: The method of drawing a feature map of the driving area in the camera coordinate system of the initial frame image, obtaining ground feature points on the feature map according to the vehicle posture at the current n-th frame image, and forming a ground point set in the vehicle coordinate system at the current n-th frame image by the ground feature points includes: According to the ego-vehicle pose of the current n-th frame image and the set of local feature points in the camera coordinate system of the current n-th frame image, each set of local feature points in the ego-vehicle coordinate system of the current n-th frame image is obtained, wherein the origin of the camera coordinate system coincides with the origin of the ego-vehicle coordinate system; Determine whether each local feature point is a ground point, remove non-ground points, and obtain the ground point set in the vehicle coordinate system of the current n-th frame image.

6. The method for estimating motion scale of a vehicle-mounted monocular camera according to claim 5, characterized in that: The step of determining whether each local feature point is a ground point comprises: Determine whether the Z coordinate of the local feature point is less than 0. If it is less than 0, keep the local feature point. If it is greater than or equal to 0, the local feature point is a non-ground point and is deleted. Take the mean Z coordinate of the retained local feature points to get the mean Z mean ; Draw a circle with the origin of the vehicle coordinate system of the current n-th frame image as the center and the vehicle width as the diameter, and select local feature points within the circle; Using mean Z mean The noise points are removed from the local feature points in the circular region to obtain the ground point set in the vehicle coordinate system of the current n-th frame image.

7. The method for estimating motion scale of a vehicle-mounted monocular camera according to claim 1, characterized in that: The method of calculating the physical scale of the posture change of the current n-th frame image according to the ground points in the ground point set, the physical height of the camera in the real world, and the scale of the current n-th frame image includes: According to the ground point set in the vehicle coordinate system of the current n-th frame image, the camera height in the map scale is obtained; According to the camera height in the map scale and the physical height of the camera in the real world, the scale value of the current n-th frame image is obtained; Using the scale value of the current n-th frame image, we can obtain the relationship function between the spatial distance Sreal in the physical world and the spatial distance Smap in the map: Sreal=scale*Smap.

8. A vehicle-mounted monocular camera motion scale estimation device, characterized in that: include: A feature extraction module is used to extract ground features from images obtained by the vehicle-mounted monocular camera and perform regional division; The ego vehicle posture estimation module is used to determine the ego vehicle posture at the current nth frame image; The ground feature recognition module is used to draw a feature map of the driving area in the camera coordinate system of the initial frame image, and obtain ground feature points on the feature map according to the posture of the vehicle at the current n-th frame image. The ground feature points form a ground point set in the current n-th frame coordinate system of the vehicle; The scale estimation module calculates the physical scale of the pose change of the current n-th frame image based on the ground points in the ground point set, the physical height of the camera in the real world, and the scale of the current n-th frame image.

9. A computer 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 method for estimating motion scale of a vehicle-mounted monocular camera according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a method for estimating motion scale of a vehicle-mounted monocular camera according to any one of claims 1 to 7.

Images