A monocular picture-based high-position camera equivalent camera parameter acquisition method and system

Abstract

The application discloses a kind of high bit camera equivalent camera parameter acquisition method and system based on monocular picture, it is related to intelligent traffic management field, including: according to the picture that has been collected, screening out reference picture and reference target, and based on the directed enclosing frame key point detection result and geometric dimension information of reference target, initial camera equivalent parameter and the ground depth map of extended imaging plane area can be acquired, then according to initial camera equivalent parameter, the ground depth map of extended imaging plane area, and parking space geometric structure features, the bird's eye view coordinate point of each labeled parking space and reference target is generated, and based on the bird's eye view coordinate point of each labeled parking space and reference target, the ground depth map of the extended imaging plane area and initial camera equivalent parameter are updated, the accuracy of equivalent camera parameter acquisition can be improved, and the cost of manual labeling and the complexity of implementation in special application scenarios are also reduced, and the data automatic processing rate is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent traffic management, in particular to a high camera equivalent camera parameter acquisition method and system based on monocular pictures. BACKGROUND

[0002] In the field of intelligent traffic, using high cameras for automatic monitoring of static traffic has become an important technical means of urban management, and the structured reconstruction of video content is an important basis for information analysis. However, due to the intrinsic perspective deformation of two-dimensional images, the structured deformation will cause the structured noise in the data itself, which is not conducive to the design and use of subsequent information analysis algorithms. Corresponding to a single camera, how to recover the equivalent parameters of the camera through existing business data is a necessary basis for the structured reconstruction of picture content.

[0003] Currently, the calculation of camera extrinsic parameters mainly involves image and real object key point registration through known geometric size markers, and then calculating the camera extrinsic parameters through PnP method. However, based on the actual situation of the business scene, the camera will be randomly skewed due to the influence of external wind, but personnel cannot be arranged to set up markers to complete the new calibration of the camera in time; there are no stable size markers available in the field scene. Therefore, how to use the existing business data to perform real-time camera equivalent is another way to use the length and width size parameters of the parking space, but since the parking space marking has certain local conditions and randomness, neither the specific parameters nor the length-width ratio has certainty, so in actual use, it will cause a certain proportional distortion of the reconstructed structure, resulting in a large error in the acquisition of equivalent camera parameters. SUMMARY

[0004] To solve the above technical problems, the present application provides a high camera equivalent camera parameter acquisition method and system based on monocular pictures, which can solve the problems of large error and low efficiency in the acquisition of existing equivalent camera parameters.

[0005] To achieve the above purpose, on the one hand, the present application provides a high camera equivalent camera parameter acquisition method based on monocular pictures, which comprises:

[0006] obtaining a reference picture according to a preset picture quality condition, an average target occlusion rate in each picture, and preset parking space annotation information;

[0007] detecting key points according to each target in the reference picture to obtain reference targets, and obtaining initial camera equivalent parameters of the high camera according to geometric size information of the reference targets;

[0008] According to the initial camera equivalent parameters and preset related estimation values corresponding to the reference targets, a ground depth map of an extended imaging plane region is obtained;

[0009] According to the initial camera equivalent parameters, the ground depth map of the extended imaging plane region, and the parking space geometric structure features, bird's-eye view coordinate points of each labeled parking space and the reference targets are generated;

[0010] According to the parking space labeling points in the picture, the key points of the reference targets, and the bird's-eye view coordinate points of each labeled parking space and the reference targets, the ground depth map of the extended imaging plane region and the initial camera equivalent parameters are updated.

[0011] Further, the step of obtaining the initial camera equivalent parameters of the high-position camera according to the geometric size information of the reference targets comprises:

[0012] A distance deviation value of each reference target bottom center point to the labeled parking space center point in a preset picture coordinate system is obtained.

[0013] Local pose information corresponding to the target with the centered deviation value is taken as the initial camera equivalent parameters.

[0014] Further, the step of obtaining the ground depth map of the extended imaging plane region according to the initial camera equivalent parameters and the preset related estimation values corresponding to the reference targets comprises:

[0015] In a preset world coordinate system, the reference target bottom center point is taken as the origin, the local pose information is projected in the camera coordinate system and the picture coordinate system, the depth value of the regular grid point based on the preset world coordinate system in the camera coordinate system and the picture projection coordinate are obtained, and the depth value of each pixel point in the picture is further obtained by an interpolation method to form the ground depth map.

[0016] The depth value of each picture coordinate point is calculated by the multi-dimensional vector to form the ground depth map.

[0017] Further, the step of generating the bird's-eye view coordinate points of each labeled parking space and the reference targets according to the initial camera equivalent parameters, the ground depth map of the extended imaging plane region, and the parking space geometric structure features comprises:

[0018] According to the ground depth map of the extended imaging plane region and the initial camera equivalent parameters, a depth value corresponding to the labeled parking space point coordinate is obtained, and a value of a parking space vertex in a relative world coordinate system is calculated;

[0019] According to the geometric structure features of the parking space, the value of the parking space vertex in the relative world coordinate system is geometrically corrected to obtain initial bird's-eye view coordinates of the parking space and the reference targets:

[0020] The initial bird's-eye view coordinates of the berth and the reference target are simultaneously rotated to obtain the optimized corrected bird's-eye view coordinates.

[0021] Further, the step of updating the ground depth map of the extended imaging plane region and the initial camera equivalent parameters according to the berth annotation points, the key points of the reference target, and the bird's-eye view coordinate points of each annotated berth and reference target in the picture comprises:

[0022] According to the berth annotation points, the key points of the reference target, the bird's-eye view coordinates in the world coordinate system corresponding to each annotated berth and reference target, and the height value in the target size information of the reference target, the three-dimensional oriented bounding box coordinates of the reference target are obtained, and a 2D-3D registration point group between the berth and the reference target is generated.

[0023] According to the 2D-3D registration point group between the berth and the reference target, the ground depth map of the extended imaging plane region and the initial camera equivalent parameters are updated.

[0024] In another aspect, the application provides a monocular picture-based high-position camera equivalent camera parameter acquisition system, which comprises: an acquisition unit for acquiring a reference picture according to preset picture quality conditions, target average occlusion rate in each picture, and preset berth annotation information;

[0025] The acquisition unit is further configured to detect key points to obtain reference targets according to each target in the reference picture, and to obtain initial camera equivalent parameters of the high-position camera according to the geometric size information of the reference targets.

[0026] The acquisition unit is further configured to obtain a ground depth map of an extended imaging plane region according to the initial camera equivalent parameters and the preset related estimated values corresponding to the reference targets.

[0027] A generation unit is configured to generate bird's-eye view coordinate points of each annotated berth and reference target according to the initial camera equivalent parameters, the ground depth map of the extended imaging plane region, and the berth geometric structure features.

[0028] An updating unit is configured to update the ground depth map of the extended imaging plane region and the initial camera equivalent parameters according to the berth annotation points, the key points of the reference target, and the bird's-eye view coordinate points of each annotated berth and reference target in the picture.

[0029] Further, the acquisition unit is specifically configured to obtain a distance deviation value from a center point of a bottom surface of each reference target to a center point of an annotated berth in a preset picture coordinate system; and to take local pose information corresponding to a target with a centered deviation value as the initial camera equivalent parameters.

[0030] Further, the acquisition unit is specifically further configured to project the local pose information in a camera coordinate system and in a picture coordinate system with the center point of the reference target bottom surface as the origin in a preset world coordinate system, to acquire a depth value of a regular grid point based on the preset world coordinate system in the camera coordinate system and a picture projection coordinate, and to further acquire a depth value of each pixel point of the picture by an interpolation method to form a ground depth map.

[0031] Further, the generation unit is specifically configured to acquire a depth value corresponding to a labeled parking space point coordinate and a value of a parking space vertex in a relative world coordinate system according to the ground depth map of the extended imaging plane area and the initial camera equivalent parameter; perform geometric correction on the value of the parking space vertex in the relative world coordinate system according to a geometric structure feature of the parking space, to obtain an initial bird's eye view coordinate of the parking space and the reference target; and perform coordinate rotation on the initial bird's eye view coordinate of the parking space and the reference target at the same time, to obtain an optimized and corrected bird's eye view coordinate.

[0032] Further, the update unit is specifically configured to acquire a three-dimensional oriented bounding box coordinate of the reference target according to a labeled point of the picture, a key point of the reference target, a bird's eye view coordinate in a world coordinate system corresponding to each labeled parking space and the reference target respectively, and a height value in target size information of the reference target, and to generate a 2D-3D registration point group between the parking space and the reference target; and update the ground depth map of the extended imaging plane area and the initial camera equivalent parameter according to the 2D-3D registration point group between the parking space and the reference target.

[0033] The application provides a monocular picture-based high-position camera equivalent camera parameter acquisition method and system, which screens out a reference picture and a reference target according to an existing picture, acquires initial camera equivalent parameters and a ground depth map of an extended imaging plane area based on a directed bounding box key point detection result and geometric size information of the reference target, generates bird's eye view coordinate points of each labeled parking space and the reference target according to the initial camera equivalent parameters, the ground depth map of the extended imaging plane area, and a geometric structure feature of the parking space, and updates the ground depth map of the extended imaging plane area and the initial camera equivalent parameter based on the bird's eye view coordinate points of each labeled parking space and the reference target, so that the accuracy of the equivalent camera parameter acquisition is improved, the manual labeling cost and the implementation complexity in a special application scenario are reduced, and the data automatic processing rate is improved. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 is a flowchart of a monocular picture-based high-position camera equivalent camera parameter acquisition method provided by the application;

[0035] Figure 2It is a kind of high camera equivalent camera parameter acquisition system based on monocular picture provided by the application. DETAILED DESCRIPTION

[0036] The technical solutions of the application are further described below by means of drawings and examples.

[0037] A kind of high camera equivalent camera parameter acquisition method based on monocular picture provided by the embodiment of the application, as shown in Figure 1 As shown in the figure, the method comprises the following steps:

[0038] 101, according to the preset picture quality condition, the average target blocking rate in each picture, the preset parking space marking information is obtained Reference picture.

[0039] Specifically, step 101.1: a group of sample pictures filtered according to time conditions is obtained, and the picture quality is evaluated, such as brightness, blurriness, white balance, etc. The pictures that do not meet the preset quality conditions are removed. Step 101.2: the average target blocking rate in each picture is calculated using the target detection result, and if the blocking rate is higher than the preset value, it is removed. Step 101.3: the parking center point coordinates are calculated using the given parking space marking information, the average value mean and the variance var of the distance from each target bottom center point to the parking center point in the above-mentioned remaining pictures after screening are calculated, the comprehensive evaluation parameter spara=mean+var is calculated, and the picture corresponding to the minimum comprehensive evaluation parameter value in the picture set is selected as the reference picture base_img.

[0040] 102, according to each target in the reference picture, the key point detection is carried out to obtain the reference target, and the initial camera equivalent parameters of the high camera are obtained according to the geometric size information of the reference target.

[0041] Specifically, step 102.1: For each target in the image, perform keypoint detection kp8_0, and use the initially given approximate vehicle size information base_size to calculate the 3D coordinates kp8_wrd corresponding to the vehicle keypoints with the center point of the vehicle's bottom surface as the origin. Set the scaling parameter sca. Let le be a fixed value used as the equivalent intrinsic parameter of the camera. Then, use 2D-3D point groups to perform PnP to calculate the local pose information transp of the target. Then, use the calculated pose data to project kp8_wrd back into the image coordinate system to obtain the reconstructed key point coordinates kp8_1. Calculate the projection error err_kp8=||kp8_1-kp8_0||. If the error is greater than the preset threshold thr1, it indicates that the key point estimation result is of poor quality. Remove the target as a baseline target candidate and retain the corresponding calculated parameters of the targets in the candidate set. Step 102.2: In the candidate set obtained in the above steps, calculate the distance deviation value from the center point of the bottom surface of each target to the center point of the marked berth in the image coordinate system. Select the target with the middle deviation value as the baseline target base_obj and output its corresponding calculated parameters transp, kp8_1 and kp8_wrd (where the bottom surface is bev). Use transp as the initial estimated camera extrinsic parameter.

[0042] 103. Based on the initial camera equivalent parameters and the preset correlation estimates corresponding to the reference target, obtain the ground depth map of the extended imaging plane region.

[0043] Specifically, step 103.1: In the world coordinate system, expand a finite surrounding area with the center point of the base_obj bottom surface as the origin. After uniformly dividing the expanded area into a grid, use transp to project the grid points: first project to the camera coordinate system, then project to the image coordinate system, and obtain the depth value z of each image coordinate point in the camera coordinate system, forming an n*3 dimensional vector interp3, where n represents the number of grid points; Step 103.2: Given the image area expansion amount pxel_exp, use interp3 to expand the area where -pxel_exp <= x <

[0044] 104. Based on the initial camera equivalent parameters, the ground depth map of the extended imaging plane area, and the geometric features of the berths, generate bird's-eye view coordinate points for each marked berth and reference target.

[0045] ​Specifically, step 104.1: first, the corresponding depth value zi is obtained by using the depth map depthmap for position retrieval for the labeled parking point coordinate P i_img, and the value of the parking vertex in the camera coordinate system P i_cam=(P i_img_x*zi, P i_img_y*zi, zi) is calculated, and then the initial extrinsic parameter transp is used to perform back projection on the parking vertex P i_cam in the camera coordinate system to obtain the value P i_wrd in the relative world coordinate system; step 104.2: the projected point P i_wrd is geometrically corrected by using the prior information that the geometric structure of the parking space is a rectangle: that is, the two groups of edge lengths are rotated to be parallel, and the intersecting edges are perpendicular, and the intersection points of the four edges are recalculated to obtain a new rectangle as the initial BEV coordinate of the parking space-bth_bev; step 104.3: then, the BEV coordinates of the parking space and the reference target base_obj are simultaneously rotated, so that the short side of the parking space is parallel to the x axis and the long side is parallel to the y axis in the new coordinate system, and finally the BEV coordinate point after optimization and correction is obtained.

[0046] 105. Update the ground depth map of the extended imaging plane region and the initial camera equivalent parameter according to the parking mark points in the picture, the key points of the reference target, and the bird's eye view coordinate points of each labeled parking space and reference target.

[0047] Specifically, the parking mark points P i_img and the base_kp8 of the reference target base_obj on the picture, and the corresponding BEV coordinates in the world coordinate system, and the three-dimensional directional bounding box coordinates base_kp8_wrd of the reference target are further obtained based on the height value in the target size information base_size of the reference target, forming a new 2D-3D point pair, and then the PnP algorithm is used to update and calculate the equivalent camera extrinsic parameter transp, and the equivalent relative depth map depthmap is updated according to step 4.

[0048] The application provides a monocular picture-based high-position camera equivalent camera parameter acquisition method, which can obtain initial camera equivalent parameters and a ground depth map of an extended imaging plane region according to existing target directional bounding box key point detection results and geometric size information, then generate bird's eye view coordinate points of each labeled parking space and reference target according to the initial camera equivalent parameters, the ground depth map of the extended imaging plane region, and the geometric structure characteristics of the parking space, and update the ground depth map of the extended imaging plane region and the initial camera equivalent parameters based on the bird's eye view coordinate points of each labeled parking space and reference target, which can improve the accuracy of equivalent camera parameter acquisition, reduce the artificial labeling cost and implementation complexity in special application scenarios, and improve the data automatic processing rate.

[0049] To realize the method provided by the embodiment of the present application, the embodiment of the present application provides a high-position camera equivalent camera parameter acquisition system based on monocular pictures, as shown in Figure 2 The system comprises an acquisition unit 21, a generation unit 22 and an update unit 23.

[0050] The acquisition unit 21 is configured to acquire reference pictures according to preset picture quality conditions, target average occlusion rates in each picture and preset parking space labeling information.

[0051] The acquisition unit 21 is further configured to acquire reference targets by performing key point detection on each target in the reference pictures, and acquire initial camera equivalent parameters of the high-position camera according to geometric size information of the reference targets.

[0052] The acquisition unit 21 is further configured to acquire a ground depth map of an extended imaging plane region according to the initial camera equivalent parameters and preset related estimated values corresponding to the reference targets.

[0053] The generation unit 22 is configured to generate bird's-eye view coordinate points of each labeled parking space and reference target according to the initial camera equivalent parameters, the ground depth map of the extended imaging plane region and parking space geometric structure features.

[0054] The update unit 23 is configured to update the ground depth map of the extended imaging plane region and the initial camera equivalent parameters according to parking space labeling points in the pictures, key points of the reference targets and the bird's-eye view coordinate points of each labeled parking space and reference target.

[0055] Further, the acquisition unit 21 is specifically configured to acquire distance deviation values from a bottom center point of each reference target to a center point of a labeled parking space in a preset picture coordinate system, and take local pose information corresponding to a target with a centered deviation value as the initial camera equivalent parameters.

[0056] Further, the acquisition unit 21 is specifically configured to take the bottom center point of the reference target as an origin in a preset world coordinate system, project the local pose information in a camera coordinate system and a picture coordinate system to acquire a depth value and a multi-dimensional vector of each picture coordinate point in the camera coordinate system, and calculate the depth value of each picture coordinate point through the multi-dimensional vector to form the ground depth map.

[0057] Further, the generation unit 22 is specifically configured to obtain a depth value corresponding to a labeled parking space point coordinate and a value of a parking space vertex in a relative world coordinate system according to the ground depth map of the extended imaging plane region and the initial camera equivalent parameter; perform geometric correction on the value of the parking space vertex in the relative world coordinate system according to a geometric structure feature of the parking space, to obtain initial bird's eye view coordinates of the parking space and the reference target; and perform coordinate rotation on the initial bird's eye view coordinates of the parking space and the reference target, to obtain optimized and corrected bird's eye view coordinates.

[0058] Further, the update unit 23 is specifically configured to obtain a three-dimensional oriented bounding box coordinate of the reference target according to a picture parking space labeled point, a key point of the reference target, a bird's eye view coordinate in a world coordinate system corresponding to each labeled parking space and the reference target respectively, and a height value in target size information of the reference target, and generate a 2D-3D registration point group between the parking space and the reference target; and update the ground depth map of the extended imaging plane region and the initial camera equivalent parameter according to the 2D-3D registration point group between the parking space and the reference target.

[0059] The application provides a high-position camera equivalent camera parameter acquisition system based on a monocular picture, which can acquire initial camera equivalent parameters and a ground depth map of an extended imaging plane region according to existing target oriented bounding box key point detection results and geometric size information, then generate bird's eye view coordinate points of each labeled parking space and reference target according to the initial camera equivalent parameters, the ground depth map of the extended imaging plane region, and a parking space geometric structure feature, and update the ground depth map of the extended imaging plane region and the initial camera equivalent parameters based on the bird's eye view coordinate points of each labeled parking space and reference target, so that the accuracy of equivalent camera parameter acquisition can be improved, the manual labeling cost and implementation complexity in a special application scenario are reduced, and the data automatic processing rate is improved.

[0060] It should be understood that the specific order or hierarchy of steps in the processes disclosed should not be taken as a limitation of the example methods. By way of example, the specific order or hierarchy of steps in the processes could be re-arranged or reordered without departing from the scope of the present disclosure. The accompanying method claims present elements of the various steps in a sample order, and are not meant to be limited to the specific order or hierarchy presented.

[0061] In the above detailed description, various features are grouped together in single embodiments for the purpose of streamlining the disclosure. This method of disclosure is not to be interpreted as reflecting a necessity to more features than are expressly described in each claim. Rather, inventive embodiments are defined solely by those parts expressly recited in the claims. Thus, the claims are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate preferred embodiment.

[0062] The disclosed embodiments are to be considered merely illustrative of the principles of the application, and various modifications can be made by those skilled in the art to the application without departing from the scope and spirit of the application as described in the following claims.

[0063] The above description includes exemplary embodiments of the present application. Of course, not all possible combinations of components or method steps are described, but one of ordinary skill in the art having the benefit of this disclosure will recognize that many other combinations and permutations of the described elements are possible. It is intended that the scope of the application be defined by the claims appended hereto rather than the examples described herein. Further, to the extent that the term "includes" is used in either the detailed description or the claims, such term is intended to be interpreted as "including but not limited to." Additionally, to the extent that the term "or" is used in either the detailed description or the claims (such as the following claims), such term is intended to be interpreted as "exclusive or". That is, "A or B" means "A or B, but not both".

[0064] Those of skill would further appreciate that the various illustrative logical blocks, modules, and steps described in connection with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans can implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present embodiments.

[0065] The various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the embodiments disclosed herein can be implemented as electronic hardware, computer software, or combinations of both. The various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality, without limitation. Depending upon the implementation, the various illustrative blocks, modules, circuits, and steps could be implemented in digital electronic circuitry, or in computer software, firmware, or hardware, including the structures disclosed in this specification and their structural equivalents, or in combinations of them. The software implemented as a computer program, software, or firmware instructions can be stored in any suitable

[0066] The steps of a method or algorithm described in connection with the embodiments disclosed herein can be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module can reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium can be coupled to the processor, such that the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium can be integral to the processor. The processor and the storage medium can reside in an ASIC. The ASIC can reside in a user terminal. In the alternative, the processor and the storage medium can reside as distinct components in a user terminal. Additionally, in some embodiments, the medium can comprise computer-readable storage media or communications media.

[0067] In one or more exemplary designs, the functions described can be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions can be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media includes both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. Storage media can be any available media that can be accessed by a computer. By way of example, such computer-readable media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a general-purpose or special-purpose computer, or a general-purpose or special-purpose processor. Also, any connection is properly termed a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or other wire-based, fiber- optic based, or wireless based communications, then the coaxial cable, fiber optic cable, twisted pair, DSL, or other wire-based, fiber-optic based, or wireless based communications are included in the definition of medium. Disk and disc, as used herein, includes compact disc (CD), laser disc, optical disc, DVD, floppy disk, and Blu-ray® disc where disks usually reproduce data magnetically, while discs reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0068] The specific implementation described above is illustrative for purposes of teaching the present application. The application should not be limited, however, to these specific implementation. Any modifications, equivalents, or improvements, as long as they are within the spirit and scope of the present application are included in the scope of the present application.

Claims

1. A monocular image-based high-position camera equivalent camera parameter acquisition method, characterized in that, The method comprises: According to the preset picture quality condition, the target average occlusion rate in each picture, and the preset parking space label information, a reference picture is obtained; According to each target in the reference picture, key point detection is performed to obtain a reference target, and according to the geometric size information of the reference target, initial camera equivalent parameters of the high-position camera are obtained; According to the initial camera equivalent parameters and preset related estimated values corresponding to the reference target, a ground depth map of an extended imaging plane area is obtained; The step of obtaining the ground depth map of the extended imaging plane area according to the initial camera equivalent parameters and the preset related estimated values corresponding to the reference target comprises: In a preset world coordinate system, taking the center point of the bottom surface of the reference target as the origin, the depth value of the regular grid point in the camera coordinate system based on the preset world coordinate system and the picture projection coordinates are obtained through projection in the camera coordinate system and the picture coordinate system by using the local posture information, and the depth value of each pixel point of the picture is further obtained by using an interpolation method to form the ground depth map; The depth value of each picture coordinate point is calculated by using a multi-dimensional vector to form the ground depth map; According to the initial camera equivalent parameters, the ground depth map of the extended imaging plane area, and the geometric structure characteristics of the parking space, an overhead view coordinate point of each labeled parking space and the reference target is generated; According to the parking space label points in the picture, the key points of the reference target, and the overhead view coordinate points of each labeled parking space and the reference target, the ground depth map of the extended imaging plane area and the initial camera equivalent parameters are updated; The step of updating the ground depth map of the extended imaging plane area and the initial camera equivalent parameters according to the parking space label points in the picture, the key points of the reference target, and the overhead view coordinate points of each labeled parking space and the reference target comprises: According to the parking space label points in the picture, the key points of the reference target, the overhead view coordinates in the world coordinate system corresponding to each labeled parking space and the reference target, and the height value in the target size information of the reference target, a three-dimensional directed bounding box coordinate of the reference target is obtained, and a 2D-3D registration point group between the parking space and the reference target is generated; According to the 2D-3D registration point group between the parking space and the reference target, the ground depth map of the extended imaging plane area and the initial camera equivalent parameters are updated.

2. The method of claim 1, wherein, The step of obtaining the initial camera equivalent parameters of the high-position camera according to the geometric size information of the reference target comprises: The distance deviation value of each bottom center point of the reference target to the center point of the labeled parking space in the preset picture coordinate system is obtained; The local posture information corresponding to the target with the centered deviation value is taken as the initial camera equivalent parameters.

3. The method of claim 1, wherein, The step of generating the overhead view coordinate point of each labeled parking space and the reference target according to the initial camera equivalent parameters, the ground depth map of the extended imaging plane area, and the geometric structure characteristics of the parking space comprises: According to the ground depth map of the extended imaging plane area and the initial camera equivalent parameters, the depth value corresponding to the labeled parking space point coordinate is obtained, and the value of the parking space vertex in the relative world coordinate system is calculated; According to the geometric structure characteristics of the parking space, the value of the parking space vertex in the relative world coordinate system is geometrically corrected to obtain initial bird's-eye view coordinates of the parking space and the reference target: The initial bird's-eye view coordinates of the parking space and the reference target are simultaneously subjected to coordinate rotation to obtain optimized and corrected bird's-eye view coordinates.

4. A system for obtaining high camera equivalent camera parameters based on monocular pictures, characterized in that, The system comprises: An acquisition unit is configured to acquire a reference picture according to preset picture quality conditions, average target occlusion rates in each picture, and preset parking space annotation information; The acquisition unit is further configured to acquire reference targets by performing key point detection on each target in the reference picture, and acquire initial camera equivalent parameters of the high-position camera according to geometric size information of the reference targets; The acquisition unit is further configured to acquire a ground depth map of an extended imaging plane region according to the initial camera equivalent parameters and preset related estimated values corresponding to the reference targets; The acquisition unit is specifically further configured to take the center point of the bottom surface of the reference target as an origin in a preset world coordinate system, project local pose information in a camera coordinate system and a picture coordinate system, acquire depth values and picture projection coordinates of regular grid points in the camera coordinate system based on the preset world coordinate system, and further acquire depth values of each pixel point in the picture by an interpolation method to form the ground depth map; A generation unit is configured to generate bird's-eye view coordinate points of each annotated parking space and reference target according to the initial camera equivalent parameters, the ground depth map of the extended imaging plane region, and geometric structure characteristics of the parking space; An update unit is configured to update the ground depth map of the extended imaging plane region and the initial camera equivalent parameters according to parking space annotation points, key points of the reference targets, and the bird's-eye view coordinate points of each annotated parking space and reference target in the picture; The update unit is specifically configured to acquire three-dimensional directional bounding box coordinates of the reference targets according to parking space annotation points, key points of the reference targets, bird's-eye view coordinates in the world coordinate system corresponding to each annotated parking space and reference target, and height values in target size information of the reference targets, generate a 2D-3D registration point group between the parking space and the reference target, and update the ground depth map of the extended imaging plane region and the initial camera equivalent parameters according to the 2D-3D registration point group between the parking space and the reference target.

5. The high-position camera equivalent camera parameter acquisition system based on monocular pictures according to claim 4, wherein The acquisition unit is specifically configured to acquire distance deviation values from the center point of the bottom surface of each reference target to the center point of the annotated parking space in a preset picture coordinate system, and take local pose information corresponding to a target with a centered deviation value as the initial camera equivalent parameters.

6. The high-position camera equivalent camera parameter acquisition system based on monocular pictures according to claim 4, wherein The generation unit is specifically configured to obtain a depth value corresponding to a labeled parking space point coordinate according to the ground depth map of the extended imaging plane region and the initial camera equivalent parameter, and calculate a value of a parking space vertex in a relative world coordinate system; perform geometric correction on the value of the parking space vertex in the relative world coordinate system according to a geometric structure feature of the parking space, to obtain initial bird's-eye view coordinates of the parking space and the reference target; and simultaneously perform coordinate rotation on the initial bird's-eye view coordinates of the parking space and the reference target, to obtain optimized and corrected bird's-eye view coordinates.

Images