Method for determining real position of pedestrian in fisheye image

Abstract

The invention provides a method for determining the real position of a pedestrian in a fisheye image, and the method comprises the following steps: collecting a plurality of initial fisheye images, carrying out the spherical expansion, obtaining corresponding experiment expansion images, obtaining the position label of the pedestrian in each experiment expansion image, and obtaining the position label of the pedestrian in each experiment expansion image; the position label at least comprises the center point coordinate of the lower edge of the detection frame and the size of the detection frame, taking the expanded experiment images and the position label of the pedestrian in each expanded experiment image as the input of an initial model, and training the initial model to obtain a positioning model; collecting a target fisheye image for spherical expansion, obtaining a target expanded image, inputting the target expanded image into the positioning model, and obtaining a position label of a pedestrian in the target expanded image; and obtaining internal parameters and external parameters of the target camera, and obtaining the real position of the pedestrian in the target expanded image based on the internal parameters, the external parameters and the position label of the pedestrian in the target expanded image. According to the invention, the real position of the pedestrian is accurately calculated, and the pedestrian positioning precision is improved.

Description

Technical Field

[0001] The present application relates to the technical fields of target detection, image ranging and image processing, and in particular to a method for determining the true position of a pedestrian in a fisheye image. Background Art

[0002] With the continuous development of intelligent transportation and autonomous driving technologies, it is crucial to accurately perceive the position of pedestrians in the vehicle's surrounding environment. In practical applications, traditional image acquisition devices often have limited field of view and cannot fully capture the surrounding environment information. Fisheye cameras can obtain more environmental information due to their wide field of view, but fisheye images are severely distorted, making it very difficult to directly determine the true position of pedestrians from fisheye images. Summary of the invention

[0003] In view of the above defects or deficiencies in the prior art, the present application aims to provide a method for determining the true position of a pedestrian in a fisheye image, comprising the following steps:

[0004] Collecting a number of initial fisheye images, and performing spherical expansion on each of the initial fisheye images to obtain an experimental expansion image corresponding to each of the initial fisheye images, wherein the pedestrian is in a vertical state in the experimental expansion image;

[0005] Obtaining a location tag of a pedestrian in each of the experimental expanded images, wherein the location tag includes at least the coordinates of a center point of a detection frame and a size of the detection frame; the detection frame is used to define a range of pedestrians in the experimental expanded image;

[0006] The experimental expanded image and the position label of the pedestrian in each of the experimental expanded images are used as inputs of the initial model, and the initial model is trained to obtain a positioning model;

[0007] Collecting a target fisheye image, and performing spherical expansion on the target fisheye image to obtain a target expanded image, inputting the target expanded image into the positioning model, and obtaining a position label of the pedestrian in the target expanded image;

[0008] Acquire internal parameters and external parameters of the target camera, and obtain the real position of the pedestrian in the target expanded image based on the internal parameters, the external parameters and the position label of the pedestrian in the target expanded image;

[0009] Wherein, the target camera is a camera for collecting the target fisheye image, the internal parameters at least include the field of view size, and the external parameters at least include the installation height.

[0010] According to the technical solution provided in the embodiment of the present application, the acquisition of a plurality of initial fisheye images comprises at least the following steps:

[0011] The surround view cameras are installed around the experimental vehicle to collect several experimental fisheye images of the vehicle's surrounding environment;

[0012] The method of using the experimental expanded image and the position label of the pedestrian in each of the experimental expanded images as the input of the initial model and training the initial model specifically includes the following steps:

[0013] Dividing all the experimental expansion images into a training set, a validation set and a test set, wherein the number of the experimental expansion images included in the training set is greater than the number of the experimental expansion images included in the validation set, and the number of the experimental expansion images included in the validation set is equal to the number of the experimental expansion images included in the test set;

[0014] The experimental expanded image is enhanced to form the training data set, wherein the enhancement operation includes but is not limited to flipping, translation, and mirroring, and the training data set is used to train the initial model.

[0015] According to the technical solution provided in the embodiment of the present application, the spherical expansion of each of the initial fisheye images comprises at least the following steps:

[0016] Mapping pixels in the image coordinate system corresponding to the initial fisheye image to the world coordinate system, and establishing a mapping relationship between the world coordinate system and the virtual spherical coordinate system;

[0017] Pixels in the initial fisheye image are mapped to the virtual spherical coordinate system to perform spherical expansion on the initial fisheye image.

[0018] According to the technical solution provided in the embodiment of the present application, the initial model selects the YOLOv6 target detection network, and adjusts the parameters of the network, setting the batch size to 64, the image size to 1280, the number of epochs to 80, the learning rate to 0.001, and the momentum to 0.937.

[0019] According to the technical solution provided in the embodiment of the present application, obtaining the real position of the pedestrian in the target expanded image based on the internal parameters, the external parameters and the position label of the pedestrian in the target expanded image comprises at least the following steps:

[0020] According to the internal parameters of the target camera and the coordinates of the center point of the lower edge of the detection frame in the image coordinate system, the horizontal field of view angle and the vertical field of view angle of the center point of the lower edge of the detection frame in the world coordinate system relative to the target camera are obtained;

[0021] According to the installation height of the target camera and the vertical field of view angle, a reference distance between the target camera and the center point of the lower edge of the detection frame in the world coordinate system is obtained;

[0022] According to the reference distance and the horizontal field of view angle, the distance between the center point of the lower edge of the detection frame and the target camera in the horizontal direction / vertical direction is obtained to represent the real position of the pedestrian in the target expanded image.

[0023] According to the technical solution provided in the embodiment of the present application, the field of view size includes at least a left field of view angle, a right field of view angle, an upper field of view angle, a lower field of view angle and an image size, and the image size includes a vertical size and a horizontal size;

[0024] According to the internal parameters of the target camera and the coordinates of the center point of the lower edge of the detection frame in the image coordinate system, obtaining the horizontal field of view angle and the vertical field of view angle of the center point of the lower edge of the detection frame in the world coordinate system relative to the target camera includes at least the following steps:

[0025] Obtaining the horizontal field of view angle according to the left field of view angle, the right field of view angle, the horizontal size, and the horizontal coordinate of the coordinate of the center point of the lower edge of the detection frame in the image coordinate system;

[0026] The vertical field of view angle is obtained according to the upper field of view angle, the lower field of view angle, the vertical size, and the ordinate of the coordinates of the center point of the lower edge of the detection frame in the image coordinate system.

[0027] According to the technical solution provided in the embodiment of the present application, obtaining the reference distance between the target camera and the center point of the lower edge of the detection frame in the world coordinate system according to the installation height of the target camera and the vertical field of view angle specifically includes the following steps:

[0028] The quotient of the installation height of the target camera and the tangent value of the vertical field of view angle is used as a reference distance between the target camera and the center point of the lower edge of the detection frame in the world coordinate system.

[0029] According to the technical solution provided in the embodiment of the present application, obtaining the distance between the center point of the lower edge of the detection frame and the target camera in the horizontal direction / vertical direction according to the reference distance and the horizontal field of view angle at least includes the following steps:

[0030] The product of the reference distance and the sine value of the horizontal field of view angle is used as the distance between the center point of the lower edge of the detection frame and the target camera in the horizontal direction.

[0031] According to the technical solution provided in the embodiment of the present application, the step of obtaining the distance between the center point of the lower edge of the detection frame and the target camera in the horizontal direction / vertical direction according to the reference distance and the horizontal field of view angle also includes the following steps:

[0032] The product of the reference distance and the cosine value of the horizontal field of view angle is used as the distance between the center point of the lower edge of the detection frame and the target camera in the vertical direction.

[0033] In summary, the present application proposes a method for determining the true position of a pedestrian in a fisheye image, comprising the following steps: acquiring a number of initial fisheye images, and performing spherical expansion on each initial fisheye image to obtain an experimental expansion image corresponding to each initial fisheye image, in which the pedestrian is in a vertical state, and obtaining a position label of the pedestrian in each experimental expansion image, the position label at least including the coordinates of the center point of the detection frame and the size of the detection frame; the detection frame is used to frame the range of the pedestrian in the experimental expansion image; using the experimental expansion image and the position label of the pedestrian in each experimental expansion image as input of the initial model, training the initial model, and obtaining a positioning model; acquiring a target fisheye image, and performing spherical expansion on the target fisheye image to obtain a target expansion image, inputting the target expansion image into the positioning model, and obtaining the position label of the pedestrian in the target expansion image; obtaining the internal parameters and external parameters of the target camera, and obtaining the true position of the pedestrian in the target expansion image based on the internal parameters, the external parameters and the position label of the pedestrian in the target expansion image; wherein the target camera is a camera for acquiring the target fisheye image, the internal parameters at least include the field of view size, and the external parameters at least include the installation height.

[0034] Compared with the prior art, the beneficial effects of the present application are as follows: the present application obtains a positioning model by training the initial model, and can accurately determine the range of pedestrians after the fisheye image is expanded. Combining the internal and external parameters of the target camera, the real position of the pedestrian can be accurately calculated, thereby improving the accuracy of pedestrian positioning. At the same time, the present solution takes into account the particularity of the fisheye image, and performs a spherical expansion of the initial fisheye image, so that the pedestrian in the target expanded image is in a vertical state, adapting to the needs of pedestrian detection in different postures, and improving the adaptability of the method. In addition, the internal parameters (such as the field of view size) and external parameters (such as the installation height) of the target camera are fully considered, and a variety of factors are integrated in the process of determining the real position of the pedestrian, making the positioning result more accurate and reliable, which helps to improve the safety of intelligent transportation and autonomous driving systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A cross-sectional schematic diagram of a method for determining the true position of a pedestrian in a fisheye image provided by an embodiment of the present application;

[0036] Figure 2 A schematic diagram of the structure of a virtual spherical coordinate system provided in an embodiment of the present application;

[0037] Figure 3A schematic diagram of the structure of spherical expansion of the initial fisheye image provided in an embodiment of the present application (a person is on the right side of the target camera);

[0038] Figure 4 A schematic diagram of the structure of spherical expansion of the initial fisheye image provided in an embodiment of the present application (a person is on the left side of the target camera);

[0039] Figure 5 Provided in the embodiments of this application Figure 3 A schematic diagram of the detection results output by the positioning model;

[0040] Figure 6 Provided in the embodiments of this application Figure 4 A schematic diagram of the detection results output by the positioning model;

[0041] Figure 7 A schematic diagram of the structure of the detection frame provided in the embodiment of the present application;

[0042] Figure 8 A schematic diagram of the coordinates of the center point of the lower edge of the detection frame in the image coordinate system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0043] The present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the invention are shown in the accompanying drawings.

[0044] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0045] Example 1

[0046] As mentioned in the background technology, in order to solve the problems in the prior art, the present application proposes a method for determining the real position of a pedestrian in a fisheye image, comprising the following steps:

[0047] S1, collecting a number of initial fisheye images, and performing spherical expansion on each of the initial fisheye images to obtain an experimental expansion image corresponding to each of the initial fisheye images, in which the pedestrian is in a vertical state;

[0048] S2, obtaining a location tag of a pedestrian in each of the experimental expanded images, wherein the location tag includes at least the coordinates of the center point of a detection frame and the size of the detection frame; the detection frame is used to define the range of the pedestrian in the experimental expanded image;

[0049] Specifically, the position of the detection frame can be determined based on the coordinates of the center point of the detection frame and the size of the detection frame, and then the range of the pedestrian can be obtained. To further explain, the detection model detects pedestrians and obtains a rectangular frame. The midpoint of the lower edge of the rectangular frame (which can be regarded as the midpoint between the pedestrian's feet) is used as the coordinate for locating the pedestrian.

[0050] S3, taking the experimental expanded image and the position label of the pedestrian in each of the experimental expanded images as the input of the initial model, training the initial model, and obtaining a positioning model;

[0051] Specifically, the vertical state means that the extension direction of the pedestrian in the rectangular image is the same as the width direction of the rectangle.

[0052] Specifically, distortion correction is performed through spherical expansion, which solves the problem of fisheye images being difficult to detect and measure distance. Combined with the deep learning target detection algorithm, the position of pedestrians can be detected accurately in real time.

[0053] Furthermore, the spherical expansion of the initial fisheye image comprises at least the following steps:

[0054] Mapping pixels in the image coordinate system corresponding to the initial fisheye image to the world coordinate system, and establishing a mapping relationship between the world coordinate system and the virtual spherical coordinate system;

[0055] Pixels in the initial fisheye image are mapped to the virtual spherical coordinate system to perform spherical expansion on the initial fisheye image.

[0056] Specifically, the sphere is composed of two angles The corresponding relationship between the parameters and the 3D spherical coordinates is: Among them, the angle between the projection of a point P in the image and the line connecting the center of the sphere on the plane where the center of the sphere is located and the x-axis is θ, is the angle between the line connecting point P and the center of the sphere and the y-axis, and f is the focal length, such as Figure 2 shown.

[0057] The coordinate transformation is performed through the following mapping relationship: Where s is the scaling factor. The fisheye image is expanded spherically, and the expansion effect is as follows: Figure 3 and Figure 4 As shown in the figure, it can be seen that in the corrected image, pedestrians are all standing upright, which is conducive to subsequent detection and positioning.

[0058] S4, collecting a target fisheye image, and performing spherical expansion on the target fisheye image to obtain a target expanded image, inputting the target expanded image into the positioning model, and obtaining a position label of the pedestrian in the target expanded image;

[0059] Please refer to Figure 5 and Figure 6 As shown, Figure 5 Corresponding Figure 3 The position of the pedestrian in the middle, get the position of the detection frame, Figure 6 Corresponding Figure 4 The position of the pedestrian in the middle is used to obtain the position of the detection box.

[0060] S5, obtaining internal parameters and external parameters of the target camera, and obtaining the real position of the pedestrian in the target expanded image based on the internal parameters, the external parameters and the position label of the pedestrian in the target expanded image;

[0061] Wherein, the target camera is a camera for collecting the target fisheye image, the internal parameters at least include the field of view size, and the external parameters at least include the installation height.

[0062] Specifically, Figure 7 As shown, the target camera is located at coordinates (0, 0), the horizontal axis represents the distance of the target from the camera in the x-axis direction, and the vertical axis represents the distance of the target from the target in the y-axis direction, in meters.

[0063] The distance is measured by combining the corrected image size, field of view size and camera height, and the coordinates of the center point of the lower edge of the detection box in the expanded image are mapped to the world coordinate system to obtain the target position.

[0064] The difference between the horizontal coordinate x and the vertical coordinate y of the measurement result of each point and the true value (|true value|-|measured value|) is shown in Table 1 below. If the measured distance is closer than the actual distance, it is positive, and if the measured distance is farther than the actual distance, it is negative, and the unit is centimeter.

[0065] Table 1 Distance measurement difference table (|true value|-|measured value|)

[0066]

[0067] Among them, number 2 is the camera position; number 1 is the midpoint of the bottom edge of the detection frame, which is not in the exact center of number 1. The same situation also applies to numbers 3-6, 9, and 11-12. In summary, outside the semicircle with a radius of 1m in front of the camera with the camera as the center, the maximum ranging error is 15.6cm, and the average error is 9.86cm. The error is within an acceptable range and meets the needs of pedestrian positioning.

[0068] In a preferred embodiment, the collecting of a plurality of initial fisheye images comprises at least the following steps:

[0069] The surround view cameras are installed around the experimental vehicle to collect several experimental fisheye images of the vehicle's surrounding environment;

[0070] The method of using the experimental expanded image and the position label of the pedestrian in each of the experimental expanded images as the input of the initial model and training the initial model specifically includes the following steps:

[0071] Dividing all the experimental expansion images into a training set, a validation set and a test set, wherein the number of the experimental expansion images included in the training set is greater than the number of the experimental expansion images included in the validation set, and the number of the experimental expansion images included in the validation set is equal to the number of the experimental expansion images included in the test set;

[0072] The experimental expanded image is enhanced to form the training data set, wherein the enhancement operation includes but is not limited to flipping, translation, and mirroring, and the training data set is used to train the initial model.

[0073] Specifically, a total of 2,000 experimental fisheye images of the vehicle's surrounding environment were collected, all of which were spherically expanded to obtain corresponding experimental expanded images and pedestrian detection results in each of the experimental expanded images, 1,600 images were randomly selected as training sets, 200 images as validation sets, and 200 images as test sets. Data enhancement was performed, and operations such as flipping, translating, and mirroring were performed on the selected images to form a training data set.

[0074] In a preferred embodiment, the initial model selects the YOLOv6 target detection network, and adjusts the parameters of the network, setting the batch size to 64, the image size to 1280, the number of epochs to 80, the learning rate to 0.001, and the momentum to 0.937.

[0075] In a preferred embodiment, obtaining the real position of the pedestrian in the target expanded image based on the internal parameters, the external parameters and the position label of the pedestrian in the target expanded image comprises at least the following steps:

[0076] According to the internal parameters of the target camera and the coordinates of the center point of the lower edge of the detection frame in the image coordinate system, the horizontal field of view angle and the vertical field of view angle of the center point of the lower edge of the detection frame in the world coordinate system relative to the target camera are obtained;

[0077] According to the installation height of the target camera and the vertical field of view angle, a reference distance between the target camera and the center point of the lower edge of the detection frame in the world coordinate system is obtained;

[0078] According to the reference distance and the horizontal field of view angle, the distance between the center point of the lower edge of the detection frame and the target camera in the horizontal direction / vertical direction is obtained to represent the real position of the pedestrian in the target expanded image.

[0079] Further, the field of view size at least includes a left field of view angle, a right field of view angle, an upper field of view angle, a lower field of view angle and an image size, and the image size includes a vertical size and a horizontal size;

[0080] According to the internal parameters of the target camera and the coordinates of the center point of the lower edge of the detection frame in the image coordinate system, obtaining the horizontal field of view angle and the vertical field of view angle of the center point of the lower edge of the detection frame in the world coordinate system relative to the target camera includes at least the following steps:

[0081] Obtaining the horizontal field of view angle according to the left field of view angle, the right field of view angle, the horizontal size, and the horizontal coordinate of the coordinate of the center point of the lower edge of the detection frame in the image coordinate system;

[0082] The vertical field of view angle is obtained according to the upper field of view angle, the lower field of view angle, the vertical size, and the ordinate of the coordinates of the center point of the lower edge of the detection frame in the image coordinate system.

[0083] Furthermore, obtaining a reference distance between the target camera and the center point of the lower edge of the detection frame in the world coordinate system according to the installation height of the target camera and the vertical field of view angle specifically includes the following steps:

[0084] The quotient of the installation height of the target camera and the tangent value of the vertical field of view angle is used as a reference distance between the target camera and the center point of the lower edge of the detection frame in the world coordinate system.

[0085] Further, obtaining the distance between the center point of the lower edge of the detection frame and the target camera in the horizontal direction / vertical direction according to the reference distance and the horizontal field of view angle at least includes the following steps:

[0086] The product of the reference distance and the sine value of the horizontal field of view angle is used as the distance between the center point of the lower edge of the detection frame and the target camera in the horizontal direction.

[0087] Furthermore, obtaining the distance between the center point of the lower edge of the detection frame and the target camera in the horizontal direction / vertical direction according to the reference distance and the horizontal field of view angle also includes the following steps:

[0088] The product of the reference distance and the cosine value of the horizontal field of view angle is used as the distance between the center point of the lower edge of the detection frame and the target camera in the vertical direction.

[0089] Specifically, Figure 8As shown, the leftmost field of view angle in the expanded diagram is θ1, the rightmost field of view angle is θ2, the topmost field of view angle is α2, and the bottommost field of view angle is α1. For example, the coordinates of the center point of the lower edge of the detection frame in the image coordinate system are (x, y), and the image size is (w, h), where w is the horizontal size and h is the vertical size. Since the pixel position is proportional to the field of view angle, the horizontal field of view angle and the vertical field of view angle of the center point of the lower edge of the detection frame in the world coordinate system relative to the target camera can be obtained, which are expressed as (θ′, α′);

[0090] Furthermore, the horizontal field of view angle and the vertical field of view angle are calculated according to the following formula:

[0091]

[0092] Specifically, the distance between the camera and point p in the world coordinate system is

[0093] Specifically, the distance between the center point of the lower edge of the detection frame in the world coordinate system and the target camera in the horizontal direction is i=H·sinθ′, and the distance between the center point of the lower edge of the detection frame and the target camera in the vertical direction is j=H·cosθ′.

[0094] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and its core ideas of this application. The above is only the preferred implementation method of this application. It should be pointed out that due to the limitations of textual expression and the objective existence of infinite specific structures, ordinary technicians in this technical field can make several improvements, modifications or changes without departing from the principles of the present invention, and can also combine the above technical features in an appropriate manner; these improvements, modifications, changes or combinations, or the direct application of the inventive concept and technical solution to other occasions without improvement, should be regarded as the scope of protection of this application.

Claims

1. A method for determining the true position of a pedestrian in a fisheye image, characterized in that: The following steps are involved: Collecting a number of initial fisheye images, and performing spherical expansion on each of the initial fisheye images to obtain an experimental expansion image corresponding to each of the initial fisheye images, wherein the pedestrian is in a vertical state in the experimental expansion image; Obtaining a location tag of a pedestrian in each of the experimental expanded images, wherein the location tag includes at least the coordinates of a center point of a detection frame and a size of the detection frame; the detection frame is used to define a range of pedestrians in the experimental expanded image; The experimental expanded image and the position label of the pedestrian in each of the experimental expanded images are used as inputs of the initial model, and the initial model is trained to obtain a positioning model; Collecting a target fisheye image, and performing spherical expansion on the target fisheye image to obtain a target expanded image, inputting the target expanded image into the positioning model, and obtaining a position label of the pedestrian in the target expanded image; Acquire internal parameters and external parameters of the target camera, and obtain the real position of the pedestrian in the target expanded image based on the internal parameters, the external parameters and the position label of the pedestrian in the target expanded image; Wherein, the target camera is a camera for collecting the target fisheye image, the internal parameters at least include the field of view size, and the external parameters at least include the installation height.

2. The method for determining the real position of a pedestrian in a fisheye image according to claim 1, characterized in that: The collecting of a plurality of initial fisheye images comprises at least the following steps: The surround view cameras are installed around the experimental vehicle to collect several experimental fisheye images of the vehicle's surrounding environment; The method of using the experimental expanded image and the position label of the pedestrian in each of the experimental expanded images as the input of the initial model and training the initial model specifically includes the following steps: Dividing all the experimental expansion images into a training set, a validation set and a test set, wherein the number of the experimental expansion images included in the training set is greater than the number of the experimental expansion images included in the validation set, and the number of the experimental expansion images included in the validation set is equal to the number of the experimental expansion images included in the test set; The experimental expanded image is enhanced to form the training data set, wherein the enhancement operation includes but is not limited to flipping, translation, and mirroring, and the training data set is used to train the initial model.

3. The method for determining the real position of a pedestrian in a fisheye image according to claim 1, characterized in that: The spherical expansion of each of the initial fisheye images comprises at least the following steps: Mapping pixels in the image coordinate system corresponding to the initial fisheye image to the world coordinate system; Establish a mapping relationship between the world coordinate system and the virtual spherical coordinate system; Pixels in the initial fisheye image are mapped to the virtual spherical coordinate system to perform spherical expansion on the initial fisheye image.

4. The method for determining the real position of a pedestrian in a fisheye image according to claim 1, characterized in that: The initial model selects the YOLOv6 target detection network, and adjusts the parameters of the network, setting the batch size to 64, the image size to 1280, the number of epochs to 80, the learning rate to 0.001, and the momentum to 0.

937.

5. The method for determining the real position of a pedestrian in a fisheye image according to claim 3, characterized in that: The obtaining of the real position of the pedestrian in the target expanded image based on the internal parameter, the external parameter and the position label of the pedestrian in the target expanded image comprises at least the following steps: According to the internal parameters of the target camera and the coordinates of the center point of the lower edge of the detection frame in the image coordinate system, the horizontal field of view angle and the vertical field of view angle of the center point of the lower edge of the detection frame in the world coordinate system relative to the target camera are obtained; According to the installation height of the target camera and the vertical field of view angle, a reference distance between the target camera and the center point of the lower edge of the detection frame in the world coordinate system is obtained; According to the reference distance and the horizontal field of view angle, the distance between the center point of the lower edge of the detection frame and the target camera in the horizontal direction / vertical direction is obtained to represent the real position of the pedestrian in the target expanded image.

6. The method for determining the real position of a pedestrian in a fisheye image according to claim 5, characterized in that: The field of view size at least includes a left field of view angle, a right field of view angle, an upper field of view angle, a lower field of view angle, and an image size, and the image size includes a vertical size and a horizontal size; According to the internal parameters of the target camera and the coordinates of the center point of the lower edge of the detection frame in the image coordinate system, obtaining the horizontal field of view angle and the vertical field of view angle of the center point of the lower edge of the detection frame in the world coordinate system relative to the target camera includes at least the following steps: Obtaining the horizontal field of view angle according to the left field of view angle, the right field of view angle, the horizontal size, and the horizontal coordinate of the coordinate of the center point of the lower edge of the detection frame in the image coordinate system; The vertical field of view angle is obtained according to the upper field of view angle, the lower field of view angle, the vertical size, and the ordinate of the coordinates of the center point of the lower edge of the detection frame in the image coordinate system.

7. The method for determining the real position of a pedestrian in a fisheye image according to claim 6, characterized in that: The step of obtaining a reference distance between the target camera and the center point of the lower edge of the detection frame in a world coordinate system according to the installation height of the target camera and the vertical field of view angle specifically includes the following steps: The quotient of the installation height of the target camera and the tangent value of the vertical field of view angle is used as a reference distance between the target camera and the center point of the lower edge of the detection frame in the world coordinate system.

8. The method for determining the real position of a pedestrian in a fisheye image according to claim 5, characterized in that: The step of obtaining the distance between the center point of the lower edge of the detection frame and the target camera in the horizontal direction / vertical direction according to the reference distance and the horizontal field of view angle comprises at least the following steps: The product of the reference distance and the sine value of the horizontal field of view angle is used as the distance between the center point of the lower edge of the detection frame and the target camera in the horizontal direction.

9. The method for determining the real position of a pedestrian in a fisheye image according to claim 8, characterized in that: The step of obtaining the distance between the center point of the lower edge of the detection frame and the target camera in the horizontal direction / vertical direction according to the reference distance and the horizontal field of view angle further includes the following steps: The product of the reference distance and the cosine value of the horizontal field of view angle is used as the distance between the center point of the lower edge of the detection frame and the target camera in the vertical direction.

Images