Method and device for determining a distance of a vehicle to an object

Abstract

The present application relates to the field of vehicle scene recognition. The present application provides a method for determining the distance from a vehicle to an object, comprising the following steps: S1: obtaining a two-dimensional image of the environment in front of the vehicle taken by a vehicle-mounted camera, the image comprising an object and a traffic marker in front of the vehicle; S2: determining the image geometric features of the traffic marker in the image; S3: determining the image positional relationship of the object relative to the traffic marker in the image; S4: establishing a correlation between the image geometric features of the traffic marker and the pre-determined actual geometric features, and determining the distance from the vehicle to the object based on the correlation and the image positional relationship. The present application also provides a device for determining the distance from a vehicle to an object. In the present application, the positional relationship between an unknown object and a known marker is utilized to convert the distance measurement problem into a mapping problem between the image points and the actual points of the known marker, thereby providing an efficient distance measurement scheme that can be realized by using only a monocular camera.

Description

Technical Field

[0001] The present invention relates to a method for determining the distance from a vehicle to an object, and also to a device for determining the distance from a vehicle to an object. Background Technology

[0002] For vision-based autonomous driving functions, accurately detecting the distance to objects ahead is crucial. Stereo cameras generate a spatial image of the vehicle's surroundings by overlaying images from multiple cameras, enabling highly accurate determination of the 3D outline of objects in front and the distance from the vehicle to those objects. However, hardware cost and computational complexity are major challenges for stereo vision. Furthermore, the frequent calibration and coordination of multiple cameras in a stereo system is time-consuming and prone to introducing errors. Monocular cameras are typically implemented at a lower cost and in a smaller size, but because they cannot utilize stereo effects, they can only achieve coarse distance detection when the distance to objects is large or when the field of view is poor.

[0003] Currently, existing technologies propose a distance measurement method based on lane markings, which indirectly estimates the distance by identifying the number of lane markings between the vehicle and the object ahead. However, due to the discontinuity of lane markings and the occlusion of the markings by the vehicle itself, this counting process is highly inaccurate.

[0004] Another known method for measuring vehicle distance estimates the distance using the time it takes for the vehicle ahead to pass two consecutive markers and the standard distance between the two markers. This method relies on the regularity of the markers' arrangement and requires waiting for the vehicle ahead to completely pass both markers before distance measurement can be performed, thus limiting its application scenarios.

[0005] In this context, there is a need to provide an improved ranging scheme based on a monocular camera, aiming to improve the accuracy of measurement results. Summary of the Invention

[0006] The purpose of this invention is to provide a method for determining the distance from a vehicle to an object and a device for determining the distance from a vehicle to an object, so as to at least solve some of the problems in the prior art.

[0007] According to a first aspect of the present invention, a method for determining the distance from a vehicle to an object is provided, the method comprising the following steps:

[0008] S1: Acquire a two-dimensional image of the environment in front of the vehicle captured by an in-vehicle camera, the image including objects and traffic signs located in front of the vehicle;

[0009] S2: Determine the image geometric features of traffic signs in the image;

[0010] S3: Determine the image positional relationship of the object relative to traffic signs in the image; and

[0011] S4: Establish a correlation between the image geometric features of traffic signs and the predetermined actual geometric features, and determine the distance from the vehicle to the object based on the correlation and the image positional relationship.

[0012] This invention specifically includes the following technical concept: by utilizing the positional relationship between an unknown object and a known reference point, the distance detection problem between a vehicle and any object is transformed into a mapping problem between image points on the known reference point and actual points. During this process, it is not necessary to use two or more images, nor is it necessary to constrain the vehicle's own motion, thus providing an efficient distance measurement scheme that can be achieved using only a single camera.

[0013] Optionally, the method further includes the following steps: initializing the vehicle-mounted camera, wherein, with the distance from the traffic sign to the vehicle as the reference distance, the reference image geometric features of the traffic sign are determined by means of the vehicle-mounted camera, and the actual geometric features of the traffic sign, the reference image geometric features, and the reference distance are stored together as the initial calibration parameters of the vehicle-mounted camera.

[0014] This achieves the following technical advantages: multiple predetermined traffic sign reference information can be pre-entered during the initialization process, making it convenient to retrieve and use them at any time during normal driving.

[0015] Optionally, step S3 includes: constructing an imaginary projection of a traffic sign on the horizontal extension line of the object in the image using interpolation.

[0016] Optionally, step S4 includes: determining a scaling factor by comparing the image geometric features of the hypothetical projection of the traffic sign with the actual geometric features of the traffic sign; and calculating the distance from the vehicle to the object based on the scaling factor.

[0017] This achieves the following technical advantages: by constructing a hypothetical projection of traffic signs, computational conditions can be created regardless of whether the object in front precisely satisfies a specific relative positional relationship with a known reference object (e.g., exactly aligned or coplanar), and regardless of whether a special sequence of signs exists, thus enabling distance measurement at any time. This improves the flexibility and time efficiency of the entire scheme.

[0018] Optionally, calculating the distance from the vehicle to the object based on the scaling factor includes:

[0019] Determine the first image distance from the center of the object to the imaginary projection of the traffic sign along the horizontal extension line of the object, and convert the first image distance into a first actual distance using a scaling factor;

[0020] The reference distance in the initial calibration parameters of the vehicle-mounted camera is converted into a second actual distance using a scaling factor; and

[0021] The distance from the vehicle to the object is calculated based on the first actual distance and the second actual distance.

[0022] This achieves the following technical advantages: by using this scaling relationship, arbitrary geometric features of the object plane on which the hypothetical projection is located can be determined, thereby enabling the distance from the vehicle to the object in front to be obtained through simple mathematical conversions and approximations.

[0023] Optionally, if there are at least two traffic signs of the same type in the image, the imaginary projection is constructed at the intersection of the connecting line of the corresponding contour points of the at least two traffic signs and the horizontal extension line of the object, according to the proportional relationship defined by the connecting line.

[0024] Optionally, if a traffic sign is present in the image, the vanishing point is determined along the vehicle's direction of travel, a central perspective line is constructed based on the vanishing point and the outline of the traffic sign, and the imaginary projection is constructed at the intersection of the central perspective line and the horizontal extension line of the object, according to the image proportions defined by the central perspective line.

[0025] This achieves the following technical advantages: the construction of the hypothetical projection is not limited by the type and number of traffic signs, so the most efficient ranging scheme can be selected in different application scenarios while taking into account the computational overhead of different principles.

[0026] Optionally, the method further includes the following steps:

[0027] Traffic signs are identified using trained classifiers and / or artificial neural networks; and

[0028] Based on the identification results, the actual geometric features of traffic signs are retrieved from local and / or external databases of the vehicle.

[0029] This provides the following technical advantages: Specifically, multiple traffic signs of predetermined sizes can be stored in a database and retrieved when needed. Therefore, even if the vehicle-mounted camera has not been initially calibrated for a specific type of traffic sign, conversion relationships can be established at any time by combining the camera's general intrinsic and extrinsic parameters.

[0030] Optionally, the method further includes the following steps:

[0031] Determine additional image features of the traffic sign and / or at least one other type of traffic sign in the image and obtain additional actual features corresponding to the additional image features; and

[0032] The reliability of the distance from the vehicle to the object determined in step S4 is verified by comparing the additional image features and the additional actual features.

[0033] This achieves the following technical advantages: by understanding information about other reference objects, distance measurement can be performed in a redundant manner, and the final measurement result is made more accurate and reasonable through a cross-validation process.

[0034] According to a second aspect of the invention, a device for determining the distance from a vehicle to an object is provided, the device being used to perform the method according to a first aspect of the invention, the device comprising:

[0035] The acquisition module is configured to acquire a two-dimensional image of the environment in front of the vehicle captured by an onboard camera, the image including objects and traffic signs located in front of the vehicle.

[0036] A recognition module configured to determine image geometric features of traffic signs in the image, and further configured to determine the image positional relationship of an object relative to the traffic signs in the image; and

[0037] The calculation module is configured to establish a correlation between the image geometric features of traffic signs and predetermined actual geometric features, and to determine the distance from a vehicle to an object based on the correlation and the image positional relationship. Attached Figure Description

[0038] The invention will now be described in more detail with reference to the accompanying drawings, which will provide a better understanding of its principles, features, and advantages. The drawings include:

[0039] Figure 1 A block diagram of a device for determining the distance from a vehicle to an object according to an exemplary embodiment of the present invention is shown;

[0040] Figure 2 A flowchart illustrating a method for determining the distance from a vehicle to an object according to an exemplary embodiment of the present invention is shown;

[0041] Figure 3 A flowchart illustrating two steps of a method for determining the distance from a vehicle to an object according to an exemplary embodiment of the present invention is shown.

[0042] Figure 4a and Figure 4b A schematic diagram illustrating distance detection using the method according to the present invention in an exemplary application scenario is shown; and

[0043] Figure 5 A schematic diagram is shown illustrating distance detection using the method according to the invention in another exemplary application scenario. Detailed Implementation

[0044] To make the technical problems to be solved, the technical solutions, and the beneficial technical effects of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and several exemplary embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of protection of this invention.

[0045] Figure 1 A block diagram of a device for determining the distance from a vehicle to an object according to an exemplary embodiment of the present invention is shown.

[0046] like Figure 1 As shown, device 10 is, for example, arranged in a vehicle 1 with driver assistance / autonomous driving functions. Vehicle 1 also has an onboard camera 20. Onboard camera 20 is, for example, a monocular camera or a single camera in a stereo camera, used to record the road environment in front of vehicle 1 in the form of two-dimensional image data.

[0047] To achieve accurate distance detection using the vehicle-mounted camera 20, the device 10 includes an acquisition module 11, an identification module 12, and a calculation module 13. The acquisition module 11 may be configured as a communication interface and connected wirelessly or wiredly to the vehicle-mounted camera 20 to receive image data from it. Alternatively, the acquisition module 11 may be directly configured as or include the vehicle-mounted camera 20, and thus be able to directly capture images of the road surface in front of the vehicle 1.

[0048] The recognition module 12 is connected to the acquisition module 11 via communication technology to receive images of the road surface ahead. In the recognition module 12, traffic signs and objects in front of the vehicle 1 are identified in the image. For this purpose, appropriate templates of potential objects or traffic signs are pre-stored in the memory of the recognition module 12, and image evaluation and semantic segmentation are performed by the recognition module 12 using a trained artificial neural network or machine learning model. After identifying the corresponding objects, the recognition module 12 also determines the image geometric features of the traffic signs and the image positional relationship of the objects in front of the vehicle 1 relative to the traffic signs.

[0049] The calculation module 13 is used to establish a correlation between the image geometric features of traffic signs and predetermined actual geometric features, and to determine the distance from the vehicle to the object based on this correlation and the image positional relationships. For this purpose, the calculation module 13 can, for example, access a database located locally on the vehicle 1 or in the cloud to obtain predetermined information about traffic signs, such as the actual geometric features of different traffic signs. Here, the predetermined information about traffic signs is typically standardized and uniformly managed by road mapping and regulatory authorities. It is also possible that this predetermined information varies depending on the country or region. Based on the actual geometric features assigned to each traffic sign in the image, the calculation module 13 can determine the actual distance from the vehicle to these traffic signs and indirectly calculate the distance from the vehicle to the object ahead based on the relative positional relationships between the object ahead of the vehicle and these predetermined reference points.

[0050] Figure 2 A flowchart illustrating a method for determining the distance from a vehicle to an object according to an exemplary embodiment of the present invention is shown. The method exemplarily includes steps S0-S5, and can, for example, be implemented using... Figure 1 This is implemented using the device 10 shown.

[0051] In step S0, the vehicle-mounted camera is initialized. Here, for example, the vehicle-mounted camera can be calibrated using a traffic sign of known size. During the calibration phase, with the distance from the traffic sign to the vehicle as a reference distance, the reference image geometric features of the traffic sign in the captured image can be determined using the vehicle-mounted camera. Then, the predetermined actual geometric features of the traffic sign, the reference image geometric features, and the reference distance are stored together as initial calibration parameters for the vehicle-mounted camera.

[0052] In other words, this initial calibration reveals the proportional relationship between the reference distance to a traffic sign and its geometric features. Therefore, during normal vehicle operation, the current distance from the vehicle to the traffic sign can be determined by comparing the measured data of the traffic sign with the pre-stored reference image geometric features and reference distance.

[0053] In step S1, a two-dimensional image of the environment in front of the vehicle is acquired. This image includes objects located in front of the vehicle and traffic signs surrounding those objects. This two-dimensional image can be received, for example, from a monocular camera mounted on the vehicle via a suitable communication interface.

[0054] In step S2, the image geometric features of traffic signs are determined from the acquired image. In the context of this invention, traffic signs include, for example, infrastructure devices such as road signs, contour posts, noise barriers, guardrails, and lampposts, as well as guiding devices such as traffic cones. Image geometric features include, for example, the length, width, height, size, and shape of a specific object in the captured image. These image geometric features can be, for example, measured in pixels. Correspondingly, actual geometric features represent the true geometric features of a specific object in a world coordinate system; these values ​​are typically standardized and therefore known in advance. The actual geometric features of traffic signs can, for example, be stored in memory (located locally in the vehicle or in the cloud) and can be read or retrieved during object recognition.

[0055] In step S3, the image positional relationship of the object in front of the vehicle relative to traffic signs is determined in the image. In the context of this invention, the object in front of the vehicle can be any object in the road environment in front of the vehicle, such as a moving vehicle, a parked vehicle, a person crossing the road, or an object left on the road surface. Here, the image positional relationship refers to the positional relationship of the object in front of the vehicle relative to traffic signs in the captured image. Since all depth information is unknown in a two-dimensional image, in order to determine the relative positional relationship between two objects in the image, it is necessary to establish the depth relationship between the two objects, for example, by using appropriate mathematical operations and coordinate projection, which will be discussed in conjunction with the following... Figure 3 The embodiments are further illustrated.

[0056] In step S4, a correlation is established between the image geometric features of the traffic sign and the predetermined actual geometric features, and the distance from the vehicle to the object is determined based on this correlation and the image positional relationship. For example, the proportional relationship between the image geometric features and actual geometric features of the traffic sign in a defined direction can be calculated. Based on this proportional relationship, the actual distance of all objects in the object plane containing the traffic sign relative to the vehicle-mounted camera can be obtained. Based on the determined position of the object to be measured in the object plane, the distance from the vehicle to the object ahead can be estimated.

[0057] Next, in optional step S5, the reliability of the determined distance from the vehicle to the object ahead is verified using the recognition results of another type of traffic sign. Here, for example, at least one other type of traffic sign (if present) is identified in the image, and additional image geometric features and additional actual geometric features of this other type of traffic sign are determined. Based on the correlation between the two, the distance from the vehicle to the object ahead can be recalculated. Alternatively, other image geometric features of the same traffic sign can be identified, and the distance from the vehicle to the object ahead can be calculated multiple times based on comparisons with the corresponding actual geometric features. Then, by comparing or averaging these distance detection results, errors in the measurement results can be reduced to some extent, thereby improving accuracy.

[0058] Figure 3 A flowchart illustrating two steps of a method for determining the distance from a vehicle to an object according to an exemplary embodiment of the present invention is shown. In this exemplary embodiment, Figure 2 Method step S3 includes steps S31-S32, and method step S4 includes steps S41-S44.

[0059] In step S31, a horizontally extending line is constructed in the image along the transverse direction of the object in front of the vehicle. In the context of this invention, "longitudinal" is understood as the direction extending along the vehicle's driving direction, and "transverse" is understood as a horizontal direction perpendicular to the vehicle's driving direction.

[0060] In step S32, an imaginary projection of the traffic sign is constructed on the horizontal extension line of the object using interpolation. Here, depending on the number and type of traffic signs present in the current scene, different auxiliary methods can be used to construct the imaginary projection.

[0061] In step S41, a scaling factor is determined based on a comparison between the image geometric features of the imaginary projection of the traffic sign and the actual geometric features. Here, since the imaginary projection of the traffic sign is constructed in advance along the lateral direction of the object in front, the calculated scaling factor can essentially represent the scaling relationship between the image geometric features and the actual geometric features of all objects on the object plane where the imaginary projection of the traffic sign is located (i.e., the object plane where the object in front of the vehicle is located).

[0062] In step S42, a first image distance is determined in the image from the center of the object in front of the vehicle to the imaginary projection of the traffic sign in the lateral direction, and the first image distance is converted into a first actual distance using a scaling factor.

[0063] In step S43, the reference distance in the initial calibration parameters of the vehicle-mounted camera is converted into a second actual distance using a scaling factor.

[0064] In step S44, the distance from the vehicle to the object is calculated based on the first actual distance and the second actual distance (e.g., by triangulation and approximation).

[0065] Figure 4a and Figure 4b A schematic diagram illustrating distance detection using the method according to the present invention in an exemplary application scenario is shown.

[0066] exist Figure 4a In the scenario shown, while the vehicle is in motion, a vehicle 200 traveling in the same direction appears not far ahead. In order to ensure the accuracy of the automatic driving function and / or driving assistance function, it is necessary to accurately know the distance between the vehicle and the vehicle 200 ahead.

[0067] Here, for example, a two-dimensional image of the road surface in front of the vehicle has been captured using a forward-facing camera (e.g., a monocular camera) mounted on the vehicle. In this image, in addition to the vehicle 200 in front, multiple contour posts 101, 102, 104, and 105 arranged at certain intervals along the roadside can also be identified. These contour posts 101, 102, 104, and 105 are identically constructed and have a standardized height H. Due to visual imaging reasons, these contour posts 101, 102, 104, and 105 each have different image geometric features in the captured image. For example, two adjacent contour posts 101 and 102 to the right of the vehicle 200 in front have a first height h1 and a second height h2 in pixels, respectively.

[0068] To determine the distance to the vehicle ahead 200, it is also necessary to know the relative positional relationship between the vehicle ahead 200 and these contour posts that exist as reference points. Since the vehicle ahead 200 was not at the point where it was aligned horizontally with any of the contour posts at the time the image was captured, the distance from the vehicle ahead 200 cannot be directly determined based on the geometric information of the contour posts. In this case, it is necessary to establish the positional relationship between the vehicle ahead 200 and the known contour posts 101 and 102 by constructing an imaginary projection 103 of the contour posts on the horizontal extension line 301 of the vehicle ahead 200. For example, a horizontal auxiliary line 301 is drawn laterally from the point of contact between the vehicle 200 and the road surface. Simultaneously, a line 302 is drawn on the image connecting a set of corresponding contour points (e.g., points of contact with the ground) of two adjacent contour posts 101 and 102. The intersection of the horizontal auxiliary line 301 and this connecting line 302 is the position of the imaginary projection 103 of the contour post. For example, by constructing a line 303 connecting another set of corresponding contour points of contour pillars 101 and 102 (e.g., the upper endpoints of contour pillars 101 and 102), an imaginary projection 103 of the contour pillars can be constructed at that location according to the proportional relationship defined by the two sets of connecting lines 302 and 303. The height of this imaginary projection 103 in the image is h3.

[0069] Since the standard height H of the contour pillars 101 and 102 is known, the scaling factor between the actual geometric features and the image geometric features can be determined for the object plane containing the imaginary projection 103 of the contour pillars: α3 = H / h3. Here, H is the actual height of the contour pillars, and h3 is the image height of the imaginary projection 103 of the contour pillars.

[0070] Furthermore, based on the initial calibration of the vehicle-mounted camera, it can be known that when the distance from the vehicle to the contour post is the reference distance L0, the geometric features of the reference image of the contour post with an actual height of H in the image captured by the vehicle-mounted camera, namely the reference image height h0, are also known. Therefore, the reference scaling factor is also known: α0 = H / h0.

[0071] Therefore, based on the currently determined scaling factor and in conjunction with the initial calibration parameters of the vehicle-mounted camera, the actual distance L3 from the vehicle to the hypothetical projection 103 of the contour column can be calculated:

[0072]

[0073] Where L3 is the actual distance from the vehicle to the imaginary projection 103 of the contour post, L0 is the reference distance from the vehicle to the contour post, α3 is the scaling factor between the image height and the actual height of the imaginary projection 103 of the contour post, and α0 is the reference scaling factor between the reference image height and the actual height of the contour post when the distance from the vehicle to the contour post is the reference distance L0.

[0074] Furthermore, the lateral distance b between the vehicle 200 in front and the imaginary projection 103 can be determined in the image. This lateral distance b, for example, represents the image lateral distance from the midpoint of the line connecting the vehicle 200 in front and the contact point with the road surface to the projection position. Next, this image lateral distance can be converted into an actual lateral distance using a scaling factor.

[0075] B=b·α3

[0076] Where B is the actual lateral distance from the midpoint of the line connecting the vehicle 200 and the ground contact point to the imaginary projection 103, b is the image lateral distance from the midpoint of the line connecting the vehicle 200 and the road surface contact point to the imaginary projection 103, and α3 is the scaling factor between the image height and the actual height of the imaginary projection 103 of the contour column.

[0077] The following can be used as a reference Figure 4b The geometric relationships shown are used to calculate the distance from the vehicle to the vehicle in front (200 units). From... Figure 4bIt can be seen that what is currently known is: the actual distance L3 from the vehicle to the imaginary projection 103 of the profile post, and the actual lateral distance B from the vehicle in front 200 to the imaginary projection 103 of the profile post. Therefore, the distance L from the vehicle to the vehicle in front 200 satisfies the following relationship:

[0078] L 2 =L3 2 -B 2

[0079]

[0080] Figure 5 A schematic diagram is shown illustrating distance detection using the method according to the invention in another exemplary application scenario.

[0081] This exemplary scenario and Figure 4a The difference lies in the fact that instead of a sequence of multiple identical contour pillars on both sides of the road, there is only a single contour pillar 101 with an image height h1. It can be seen that the vehicle 200 in front has already passed the contour pillar 101 at the time the image was captured; therefore, in this scene, the positional relationship between the vehicle 200 in front and the reference object cannot be directly determined.

[0082] In this case, the vanishing point method can be used to construct the imaginary projection 103 of the profile column 101. In this image, for example, the vanishing point P is first determined along the vehicle's direction of travel. This vanishing point P is located at the intersection of the extension line 307 of the road surface boundary 310 and the extension line 306 of the lane marking line 320. Next, central perspective lines 304 and 305 can be constructed based on the vanishing point P and the profile points (e.g., bottom and top profile points) on the profile column 101. Similarly, a horizontal auxiliary line 301 is drawn laterally from the point of contact between the vehicle 200 and the road surface. At the intersection of the central perspective line 304 determined based on the bottom profile point and the horizontal auxiliary line 301, an imaginary projection 103 with an image height h3 is drawn according to the image scale defined by the central perspective lines 304 and 305.

[0083] Next, we can use the help of Figures 4a-4b The scaling factor is calculated in the manner described in the text, and the distance between the vehicle and the vehicle in front of it is calculated based on the geometric relationship.

[0084] According to other embodiments of the present invention, the distance from the vehicle to the object behind it can also be determined by acquiring an image of the environment behind the vehicle taken by a camera located behind the vehicle, using a similar method.

[0085] Although specific embodiments of the invention have been described in detail herein, they are given for illustrative purposes only and should not be construed as limiting the scope of the invention. Various substitutions, alterations, and modifications can be conceived without departing from the spirit and scope of the invention.

Claims

1. A method for determining the distance from a vehicle to an object (200), the method comprising the steps of: S1: Acquire a two-dimensional image of the environment in front of the vehicle taken by means of an onboard camera (20), the image including objects (200) located in front of the vehicle and traffic signs (101). S2: Determine the image geometric features of the traffic sign (101) in the image; S3: Determine the image positional relationship of the object (200) relative to the traffic sign (101) in the image; and S4: Establish a correlation between the image geometric features of the traffic sign (101) and the predetermined actual geometric features, and determine the distance from the vehicle to the object (200) based on the correlation and the image positional relationship. In the image, an imaginary projection (103) of a traffic sign (101) is constructed on the horizontal extension line (301) of the object (200) by means of interpolation. A scaling factor is determined based on the comparison between the image geometric features of the imaginary projection (103) of the traffic sign (101) and the actual geometric features of the traffic sign (101). The distance from the vehicle to the object (200) is calculated according to the scaling factor.

2. The method according to claim 1, wherein, The method further includes the following steps: The vehicle-mounted camera (20) is initialized. When the distance from the traffic sign (101) to the vehicle is the reference distance, the reference image geometric features of the traffic sign (101) are determined by the vehicle-mounted camera (20). The actual geometric features of the traffic sign (101), the reference image geometric features and the reference distance are stored together as the initial calibration parameters of the vehicle-mounted camera (20).

3. The method according to claim 1 or 2, wherein, The distance from the vehicle to the object (200) calculated based on the scaling factor includes: Determine the first image distance from the center of the object (200) to the imaginary projection (103) of the traffic sign (101) along the horizontal extension line (301) of the object (200), and convert the first image distance into a first actual distance by means of a scaling factor; The reference distance in the initial calibration parameters of the vehicle-mounted camera (20) is converted into a second actual distance using a scaling factor; and The distance from the vehicle to the object (200) is calculated based on the first actual distance and the second actual distance.

4. The method according to any one of claims 1 to 3, wherein, In the presence of at least two traffic signs (101, 102) of the same type in the image, the imaginary projection (103) is constructed at the intersection of the connecting line (302) of the corresponding contour points of the at least two traffic signs (101, 102) and the horizontal extension line (301) of the object (200) according to the proportional relationship defined by the connecting line (302).

5. The method according to any one of claims 1 to 4, wherein, In the presence of a traffic sign (101) in the image, a vanishing point (P) is determined along the direction of vehicle travel. A central perspective line (304) is constructed based on the vanishing point (P) and the outline of the traffic sign (101). At the intersection of the central perspective line (304) and the horizontal extension line (301) of the object (200), the imaginary projection (103) is constructed according to the image scale relationship limited by the central perspective line (304).

6. The method according to any one of claims 1 to 5, wherein, The method further includes the following steps: Traffic signs (101) are identified using a trained classifier and / or an artificial neural network; and Based on the identification results, the actual geometric features of the traffic marker (101) are retrieved from a database local to the vehicle and / or external to the vehicle.

7. The method according to any one of claims 1 to 6, wherein, The method further includes the following steps: Determine additional image features of the traffic sign (101) and / or at least one other type of traffic sign in the image and obtain additional actual features corresponding to the additional image features; and The reliability of the distance from the vehicle to the object (200) determined in step S4 is verified by comparing the additional image features and the additional actual features.

8. A device (10) for determining the distance from a vehicle to an object (200), the device (10) being used to perform the method according to any one of claims 1 to 7, the device (10) comprising: The acquisition module (11) is configured to acquire a two-dimensional image of the environment in front of the vehicle taken by means of an onboard camera (20), the image including objects (200) in front of the vehicle and traffic signs (101). The recognition module (12) is configured to determine the image geometric features of the traffic sign (101) in the image, and is also configured to determine the image positional relationship of the object (200) relative to the traffic sign (101) in the image; as well as The calculation module (13) is configured to establish a correlation between the image geometric features of the traffic sign (101) and the predetermined actual geometric features, and to determine the distance from the vehicle to the object (200) based on the correlation and the image positional relationship.

Images