Method for detecting target in monocular image, vehicle positioning method, device and equipment

By identifying the relative feature information of the target to be detected and other targets in the monocular image and adjusting their feature information, the problem of low detection accuracy of three-dimensional object in the monocular image is solved, and efficient three-dimensional feature information recognition and data processing are achieved.

CN113255407BActive Publication Date: 2024-11-12ALIBABA GROUP HOLDING LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010089173.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-02-12
Publication Date
2024-11-12
Estimated Expiration
2040-02-12

AI Technical Summary

Technical Problem

The accuracy of three-dimensional object detection in existing monocular images is low, and the data processing is complex, making it difficult to effectively identify three-dimensional feature information, especially when the object is far away or blocked, and the detection accuracy is reduced, and the uncertainty of the prediction results cannot be effectively evaluated. .

Method used

By acquiring a monocular image, identifying the target to be detected and other targets, and determining their characteristic information in the preset coordinate system, adjusting the characteristic information of the target to be detected using relative feature information to achieve accurate three-dimensional feature information for the target in the monocular image Identify and reduce the complexity of data processing.

Benefits of technology

It realizes accurate and effective identification of targets in monocular images, reduces the complexity of data processing, and improves the accuracy and practicality of three-dimensional feature information detection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113255407B_ABST
    Figure CN113255407B_ABST
Patent Text Reader

Abstract

The embodiments of the present invention provide a method for detecting a target in a monocular image, a method for positioning a vehicle, a device and an apparatus. The method comprises: acquiring a monocular image; identifying a target to be detected included in the monocular image, and determining first feature information of the target to be detected in a preset coordinate system, wherein the preset coordinate system corresponds to the monocular image; identifying other targets included in the monocular image, and determining at least one relative feature information between the target to be detected and other targets in the preset coordinate system; adjusting the first feature information based on at least one relative feature information to obtain second feature information of the target to be detected. The technical solution provided in this embodiment adjusts the first feature information of the target to be detected based on at least one relative feature information of the target to be detected relative to other targets, thereby effectively realizing the ability to accurately and effectively identify the three-dimensional feature information of the target in the monocular image on the basis of reducing the complexity of data processing.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a method for detecting a target in a monocular image, a vehicle positioning method, a device and equipment. Background Art

[0002] A monocular image refers to image information acquired from a single camera sensor, while 3D object detection refers to detecting information such as the position, size, and orientation of objects in an image in three-dimensional space. 3D object detection is a crucial component of visual perception, with numerous applications in scenarios such as autonomous driving, robotics, and augmented reality (AR). Current object detection methods typically require the integration of lidar or binocular cameras to obtain relatively accurate scene geometry. However, these solutions incur high hardware costs and involve complex data processing. Summary of the Invention

[0003] The embodiments of the present invention provide a method for detecting a target in a monocular image, a method, an apparatus, and an equipment for positioning a vehicle, which can accurately and effectively identify the three-dimensional feature information of the target in the monocular image and effectively reduce the complexity of data processing.

[0004] In a first aspect, an embodiment of the present invention provides a method for detecting an object in a monocular image, comprising:

[0005] Acquire a monocular image;

[0006] Identifying a target to be detected included in a monocular image, and determining first feature information of the target to be detected in a preset coordinate system, wherein the preset coordinate system corresponds to the monocular image;

[0007] Identifying other targets included in the monocular image, and determining at least one relative feature information between the target to be detected and the other targets in the preset coordinate system;

[0008] The first feature information is adjusted based on the at least one relative feature information to obtain second feature information of the target to be detected.

[0009] In a second aspect, an embodiment of the present invention provides a device for detecting an object in a monocular image, comprising:

[0010] A first acquisition module is used to acquire a monocular image;

[0011] a first detection module, configured to identify a target to be detected included in a monocular image and determine first feature information of the target to be detected in a preset coordinate system, wherein the preset coordinate system corresponds to the monocular image;

[0012] The first detection module is configured to identify other targets included in the monocular image and determine at least one relative feature information between the target to be detected and the other targets in the preset coordinate system;

[0013] The first processing module is configured to adjust the first feature information based on the at least one relative feature information to obtain second feature information of the target to be detected.

[0014] In a third aspect, an embodiment of the present invention provides an electronic device comprising: a memory and a processor; wherein the memory is used to store one or more computer instructions, wherein when the one or more computer instructions are executed by the processor, the method for detecting targets in monocular images in the above-mentioned first aspect is implemented.

[0015] In a fourth aspect, an embodiment of the present invention provides a computer storage medium for storing a computer program, wherein the computer program enables a computer to implement the method for detecting a target in a monocular image in the first aspect mentioned above when executed.

[0016] In a fifth aspect, an embodiment of the present invention provides a vehicle positioning method, comprising:

[0017] Acquire a monocular image;

[0018] Identifying a to-be-identified vehicle and a reference vehicle included in the monocular image;

[0019] determining first positioning information of the vehicle to be identified in a preset coordinate system and at least one relative positioning information between the vehicle to be identified and the reference vehicle in the preset coordinate system, wherein the preset coordinate system corresponds to the monocular image;

[0020] The first positioning information is adjusted according to the at least one relative feature information to obtain second positioning information of the vehicle to be identified.

[0021] In a sixth aspect, an embodiment of the present invention provides a vehicle positioning device, comprising:

[0022] A second acquisition module is used to acquire a monocular image;

[0023] A second recognition module is used to recognize the vehicle to be recognized and the reference vehicle included in the monocular image;

[0024] a second determining module, configured to determine first positioning information of the vehicle to be identified in a preset coordinate system, and at least one relative positioning information between the vehicle to be identified and the reference vehicle in the preset coordinate system, wherein the preset coordinate system corresponds to the monocular image;

[0025] The second positioning module is used to adjust the first positioning information according to the at least one relative feature information to obtain second positioning information of the vehicle to be identified.

[0026] In the seventh aspect, an embodiment of the present invention provides an electronic device, comprising: a memory and a processor; wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions, when executed by the processor, implement the vehicle positioning method in the above-mentioned fifth aspect.

[0027] In an eighth aspect, an embodiment of the present invention provides a computer storage medium for storing a computer program, wherein the computer program enables a computer to implement the vehicle positioning method in the fifth aspect when executed.

[0028] The method for detecting a target in a monocular image, the method for positioning a vehicle, the apparatus, and the equipment provided in this embodiment acquire a monocular image, identify the target to be detected and other targets included in the monocular image, and determine first feature information of the target to be detected in a preset coordinate system and at least one relative feature information between the target to be detected and the other targets in the preset coordinate system. The first feature information is then adjusted based on the at least one relative feature information to obtain second feature information of the target to be detected. This not only achieves accurate and effective identification of the three-dimensional feature information of the target in the monocular image, but also effectively reduces the complexity of data processing, further improves the practicality of the method for detecting a target in a monocular image, and is conducive to market promotion and application. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0030] Figure 1 A schematic diagram of a flow chart of a method for detecting an object in a monocular image provided by an embodiment of the present invention;

[0031] Figure 2 A schematic diagram of an application scenario of the method for detecting an object in a monocular image provided by an embodiment of the present invention;

[0032] Figure 3 A schematic diagram of a process for identifying a target to be detected included in a monocular image and determining first feature information of the target to be detected in a preset coordinate system provided by an embodiment of the present invention;

[0033] Figure 4A schematic diagram of a process for determining projection information of the target to be detected in the monocular image provided by an embodiment of the present invention;

[0034] Figure 5 A schematic diagram of a process for identifying other targets included in the monocular image and determining at least one relative feature information between the target to be detected and the other targets in the preset coordinate system, provided by an embodiment of the present invention;

[0035] Figure 6 Schematic diagram of a process for adjusting the first feature information based on the at least one relative feature information to obtain the second feature information of the target to be detected provided in an embodiment of the present invention Figure 1 ;

[0036] Figure 7 Schematic diagram of a process for adjusting the first feature information based on the at least one relative feature information to obtain the second feature information of the target to be detected provided in an embodiment of the present invention Figure 2 ;

[0037] Figure 8 A schematic diagram of a process for adjusting the first feature information based on an adjustment parameter corresponding to the confidence information to obtain the second feature information of the target to be detected provided by an embodiment of the present invention;

[0038] Figure 9 A schematic flow chart of another method for detecting an object in a monocular image provided by an embodiment of the present invention;

[0039] Figure 10 A schematic flow chart of a method for detecting an object in a monocular image provided by an application embodiment of the present invention;

[0040] Figure 11 A schematic flow chart of a vehicle positioning method provided by an embodiment of the present invention;

[0041] Figure 12 A schematic structural diagram of a device for detecting an object in a monocular image provided by an embodiment of the present invention;

[0042] Figure 13 For Figure 12 A schematic structural diagram of an electronic device corresponding to a device for detecting a target in a monocular image provided by the illustrated embodiment;

[0043] Figure 14 A schematic structural diagram of a vehicle positioning device provided by an embodiment of the present invention;

[0044] Figure 15 For Figure 14 A schematic structural diagram of an electronic device corresponding to the vehicle positioning device provided in the illustrated embodiment. DETAILED DESCRIPTION

[0045] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0046] The terms used in the embodiments of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The singular forms "a," "the," and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, unless the context clearly indicates otherwise. "A plurality" generally includes at least two, but does not exclude the inclusion of at least one.

[0047] It should be understood that the term "and / or" as used herein is merely a description of the relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0048] As used herein, the words "if" and "if" may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to the determination" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)," depending on the context.

[0049] It should also be noted that the terms "include," "comprises," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a product or system comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such product or system. In the absence of further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the product or system comprising the element.

[0050] In addition, the step sequence in the following method embodiments is only an example and not a strict limitation.

[0051] To facilitate understanding of the technical solution of this application, the following is a brief description of the prior art: Currently, when performing three-dimensional target detection based on monocular images, the main challenge of three-dimensional target detection comes from the lack of 3D information due to the low accuracy of 3D target detection from monocular images, which leads to large uncertainty in the mapping relationship between 2D and 3D. In order to achieve accurate estimation of three-dimensional target detection, complex networks are usually used to model the 2D-3D mapping relationship. For example, a multi-stage network is used to complete the candidate, estimation, and refinement processes, while embedding multi-task (such as depth, segmentation, key points, etc.) networks to mine semantic and geometric features in the image.

[0052] However, modeling this 2D-3D mapping relationship requires multiple steps and is quite challenging, significantly increasing the complexity of 3D object detection. Furthermore, information mining is limited to the object itself. When objects are far away or occluded, insufficient information about the object itself or a high degree of inherent uncertainty significantly reduces detection accuracy. Furthermore, existing methods fail to estimate the uncertainty of prediction results, making risk assessment difficult in practice.

[0053] The following embodiments of the present invention are described in detail with reference to the accompanying drawings. In the absence of conflicts between the embodiments, the following embodiments and features therein may be combined with each other.

[0054] Figure 1 A schematic diagram of a flow chart of a method for detecting an object in a monocular image provided by an embodiment of the present invention;

[0055] Figure 2 Schematic diagram of an application scenario of the method for detecting a target in a monocular image provided by an embodiment of the present invention; Figure 1-Figure 2 As shown, this embodiment provides a method for detecting a target in a monocular image. The method may be performed by a device for detecting a target in a monocular image. It is understood that the device may be implemented as software or a combination of software and hardware. Specifically, the detection method may include:

[0056] Step S101: Acquire a monocular image.

[0057] Step S102: identifying a target to be detected included in the monocular image, and determining first feature information of the target to be detected in a preset coordinate system, wherein the preset coordinate system corresponds to the monocular image.

[0058] Step S103: identifying other targets included in the monocular image, and determining at least one relative feature information between the target to be detected and the other targets in a preset coordinate system.

[0059] Step S104: adjusting the first characteristic information based on at least one relative characteristic information to obtain second characteristic information of the target to be detected.

[0060] The following is a detailed explanation of each of the above steps:

[0061] Step S101: Acquire a monocular image.

[0062] The monocular image may refer to image information obtained from a single camera sensor. This embodiment does not limit the specific implementation method for the detection device to obtain the monocular image. Those skilled in the art can configure it according to specific application requirements and design requirements. For example, the camera sensor can be connected to the detection device for communication. After the camera sensor captures and obtains the monocular image, the detection device can obtain the monocular image through the camera sensor. Specifically, the detection device can actively obtain the monocular image obtained by the camera sensor, or the camera sensor can actively send the monocular image to the detection device, so that the detection device obtains the monocular image. Alternatively, the monocular image can be stored in a preset area, and the detection device can obtain the monocular image by accessing the preset area.

[0063] Step S102: identifying a target to be detected included in the monocular image, and determining first feature information of the target to be detected in a preset coordinate system, wherein the preset coordinate system corresponds to the monocular image.

[0064] Generally speaking, a monocular image may include an image background and at least one image target. Therefore, after obtaining the monocular image, the monocular image may be analyzed and processed to identify the target to be detected included in the monocular image, wherein the target to be detected may be one or more targets in at least one image target. For example, the target to be detected may include at least one of the following: a vehicle, a reference vehicle, an animal, a plant, a building, and the like.

[0065] In addition, when analyzing and processing the monocular image, the first characteristic information of the target to be detected in the preset coordinate system can also be determined, wherein the preset coordinate system can be a pre-established coordinate system corresponding to the monocular image, and the above-mentioned first characteristic information can include at least one of the following: position information, size information, and angle information of the target to be detected in the preset coordinate system, that is, the characteristic information of the target to be detected in the preset coordinate system can be obtained according to specific application requirements. For example, when it is necessary to obtain the position of the target to be detected, the position information of the target to be detected in the preset coordinate system can be obtained; when it is necessary to obtain the position and size of the target to be detected, the position information and size information of the target to be detected in the preset coordinate system can be obtained, and so on.

[0066] Step S103: identifying other targets included in the monocular image, and determining at least one relative feature information between the target to be detected and the other targets in a preset coordinate system.

[0067] When analyzing and processing a monocular image, other targets included in the monocular image can also be identified, where the other targets can include at least one of the following: a target object occluded by the target to be detected, a target object occluding the target to be detected, and a target object independent of the target to be detected. Furthermore, when identifying other targets included in the monocular image, at least one relative feature information between the target to be detected and other targets in a preset coordinate system can also be determined. This relative feature information is the spatial constraint information of the target to be detected relative to other targets in the preset coordinate system. Specifically, this relative feature information can include at least one of the following: relative position information, relative size information, and relative angle information between the target to be detected and other targets in the preset coordinate system.

[0068] Step S104: adjusting the first characteristic information based on at least one relative characteristic information to obtain second characteristic information of the target to be detected.

[0069] After obtaining at least one relative feature information, the first feature information can be adjusted based on the at least one relative feature information to obtain second feature information of the target to be detected, wherein the second feature information is feature information adjusted corresponding to the first feature information, and the second feature information may also include at least one of the following: position information, size information, and angle information of the target to be detected in a preset coordinate system.

[0070] For ease of understanding, the following example is used to illustrate the target to be detected as a vehicle in a monocular image and the first feature information as position information. Figure 2 As shown, after acquiring the monocular image, a corresponding preset coordinate system OXYZ can be established based on the monocular image. Specifically, the forward direction of the camera lens used to shoot the monocular image can be determined as the Z-axis direction in the coordinate system, the horizontal right direction can be determined as the X-axis direction, and the vertical downward direction can be determined as the Y-axis direction. The optical center position of the camera can be the origin position O in the preset coordinate system.

[0071] Then, the monocular image is analyzed and processed to identify the vehicles and other targets included in the monocular image (for example, the reference vehicle located in the upper left corner of the image). For ease of understanding, the reference vehicle is used as an example for explanation. It can be understood that the other targets can also be other reference objects appearing in the monocular image, such as reference vehicles, animals, plants, buildings, other mobile devices, etc. In addition, after the vehicle included in the monocular image is identified, the first feature information (X1, Y1, Z1) of the vehicle in the preset coordinate system can be determined, where X1 identifies the X-axis information of the vehicle in the preset coordinate system, Y1 identifies the Y-axis information of the vehicle in the preset coordinate system, and Z1 identifies the Z-axis information of the vehicle in the preset coordinate system. In the preset coordinate system, the relative characteristic information (|X|, |Y|, |Z|) between the vehicle and the reference vehicle can also be determined. The relative characteristic information is the relative position information of the vehicle relative to the reference vehicle in the preset coordinate system. Specifically, |X| is the X-axis information of the vehicle relative to the reference vehicle, |Y| is the Y-axis information of the vehicle relative to the reference vehicle, and |Z| is the Z-axis information of the vehicle relative to the reference vehicle.

[0072] After obtaining the first feature information (X1, Y1, Z1) and the relative feature information (|X|, |Y|, |Z|), the first feature information (X1, Y1, Z1) of the vehicle can be adjusted based on the relative feature information (|X|, |Y|, |Z|), that is, the position of the vehicle in the monocular image is adjusted so that the relative position information between the adjusted vehicle and the reference vehicle satisfies the above relative feature information, and the position information of the adjusted vehicle is determined as the second feature information (X2, Y2, Z2), where X2 can be the same as or different from X1, Y2 can be the same as or different from Y1, and Z2 can be the same as or different from Z1. Through the above process, the three-dimensional position information of the vehicle in the preset coordinate system corresponding to the monocular image can be accurately and effectively obtained.

[0073] Similarly, when the first characteristic information is size information, the implementation process is as follows: the monocular image is analyzed and processed to identify the vehicle and other targets included in the monocular image (for example, the reference vehicle located in the upper left corner of the image), and the first characteristic information (h1, w1, d1) of the vehicle in the preset coordinate system can be determined, where h1 identifies the height information of the vehicle in the preset coordinate system, w1 identifies the width information of the vehicle in the preset coordinate system, and d1 identifies the depth information of the vehicle in the preset coordinate system. In the preset coordinate system, the relative characteristic information (|h|, |w|, |d|) between the vehicle and the reference vehicle can also be determined. The relative characteristic information is the relative size information of the vehicle relative to the reference vehicle in the preset coordinate system. Specifically, |h| identifies the height information of the vehicle relative to the reference vehicle, |w| identifies the width information of the vehicle relative to the reference vehicle, and |d| identifies the depth information of the vehicle relative to the reference vehicle.

[0074] After obtaining the first feature information (h1, w1, d1) and the relative feature information (|h|, |w|, |d|), the first feature information (h1, w1, d1) can be adjusted based on the relative feature information (|h|, |w|, |d|). That is, the size of the vehicle in the monocular image is adjusted so that the relative size information between the adjusted vehicle and the reference vehicle satisfies the above relative feature information, and the size information of the adjusted vehicle is determined as the second feature information (h2, w2, d2), where h2 can be the same as or different from h1, w2 can be the same as or different from w1, and d2 can be the same as or different from d1. Through the above process, the size information of the vehicle in the preset coordinate system corresponding to the monocular image can be accurately and effectively obtained.

[0075] Similarly, when the first feature information is the angle information of the target to be detected in a preset coordinate system, the implementation process is as follows: the monocular image is analyzed and processed to identify the vehicle and other targets included in the monocular image (for example, the reference vehicle located in the upper left corner of the image), and the first feature information (yaw_1, pitch_1, roll_1) of the vehicle in the preset coordinate system can be determined, where yaw_1 indicates the yaw angle of the vehicle in the preset coordinate system, pitch_1 indicates the pitch angle of the vehicle in the preset coordinate system, and roll_1 indicates the roll angle of the vehicle in the preset coordinate system. In the preset coordinate system, relative feature information (|yaw|, |pitch|, |roll|) between the vehicle and the reference vehicle can also be determined. This relative feature information is the relative angle information of the vehicle relative to the reference vehicle in the preset coordinate system. Specifically, |yaw| indicates the yaw angle of the vehicle relative to the reference vehicle, |pitch| indicates the pitch angle of the vehicle relative to the reference vehicle, and |roll| indicates the roll angle of the vehicle relative to the reference vehicle.

[0076] After obtaining the first feature information (yaw_1, pitch_1, roll_1) and the relative feature information (|yaw|, |pitch|, |roll|), the first feature information (yaw_1, pitch_1, roll_1) can be adjusted based on the relative feature information (|yaw|, |pitch|, |roll|). That is, the angle of the vehicle in the monocular image is adjusted so that the relative angle information between the adjusted vehicle and the reference vehicle satisfies the above relative feature information, and the adjusted angle information of the vehicle is determined as the second feature information (yaw_2, pitch_2, roll_2), where yaw_2 can be the same as or different from yaw_1, pitch_2 can be the same as or different from pitch_1, and roll_2 can be the same as or different from roll_1. Through the above process, the angle information of the vehicle in the preset coordinate system corresponding to the monocular image can be accurately and effectively obtained.

[0077] It is understandable that those skilled in the art can detect different three-dimensional features of the target in the monocular image according to specific application requirements and design requirements. For example, the position feature, size feature or angle feature of the target in the monocular image can be detected separately, or at least two of the position feature, size feature or angle feature of the target in the monocular image can be detected, etc. The specific implementation principle can be referred to the above statement and will not be repeated here.

[0078] The method for detecting targets in monocular images provided in this embodiment obtains a monocular image, identifies the target to be detected and other targets included in the monocular image, and determines first feature information of the target to be detected in a preset coordinate system and at least one relative feature information between the target to be detected and the other targets in the preset coordinate system. Then, the first feature information is adjusted based on the at least one relative feature information to obtain second feature information of the target to be detected. This not only achieves accurate and effective identification of the three-dimensional feature information of the target in the monocular image, but also effectively reduces the complexity of data processing, further improves the practicality of the method for detecting targets in monocular images, and is conducive to market promotion and application.

[0079] Figure 3 A schematic diagram of a process for identifying a target to be detected in a monocular image and determining first feature information of the target to be detected in a preset coordinate system provided by an embodiment of the present invention; based on the above embodiment, further reference is made to the attached Figure 3 As shown, this embodiment does not limit the specific implementation method for identifying the target to be detected included in the monocular image and determining the first feature information of the target to be detected in the preset coordinate system. Those skilled in the art can make settings according to specific application requirements and design requirements. Preferably, in this embodiment, identifying the target to be detected included in the monocular image and determining the first feature information of the target to be detected in the preset coordinate system includes:

[0080] Step S301: Acquire a feature image corresponding to a monocular image.

[0081] After obtaining the monocular image, the monocular image may be analyzed and processed to obtain a feature image corresponding to the monocular image. Specifically, obtaining the feature image corresponding to the monocular image may include:

[0082] Step S3011: Analyze and process the monocular image using the second machine learning model to obtain a feature image corresponding to the monocular image, wherein the second machine learning model is trained to determine the feature image corresponding to the monocular image.

[0083] The second machine learning model may be pre-trained to determine a feature image corresponding to a monocular image. The feature image may include at least one of the following: pixel feature information, color feature information, brightness feature information, texture feature information, edge feature information, shape feature information, and semantic feature information. Furthermore, when training the machine learning model, the machine learning model may be established based on a convolutional neural network. Therefore, after obtaining a monocular image, the second machine learning model may be used to analyze and process the monocular image, thereby obtaining a feature image corresponding to the monocular image.

[0084] In this embodiment, the monocular image is analyzed and processed by the trained second machine learning model to obtain a feature image corresponding to the monocular image, which not only effectively ensures the accuracy and reliability of the feature image acquisition, but also ensures the quality and efficiency of the feature image acquisition, further improving the quality and efficiency of the application of the detection method for targets in monocular images.

[0085] Step S302: identifying a target to be detected included in the monocular image according to the feature image, and determining first feature information of the target to be detected in a preset coordinate system.

[0086] After obtaining the feature image, the feature image can be analyzed and processed to identify the target to be detected included in the monocular image and determine the first feature information of the target to be detected in the preset coordinate system. Specifically, identifying the target to be detected included in the monocular image based on the feature image and determining the first feature information of the target to be detected in the preset coordinate system may include:

[0087] Step S3021: Identify the target to be detected based on the feature image, and determine the projection information of the target to be detected in the monocular image.

[0088] Step S3022: Determine first feature information of the target to be detected in the preset coordinate system according to the projection information.

[0089] Specifically, after acquiring the feature image, the feature image can be analyzed and processed to identify the target to be detected included in the monocular image, and projection information of the target to be detected in the monocular image can be determined. The projection information can include information about at least one projection point of the target to be detected in the monocular image. After acquiring the projection information, first feature information of the target to be detected in a preset coordinate system can be determined based on the projection information.

[0090] It is understandable that those skilled in the art may also adopt other methods to identify the target to be detected included in the monocular image based on the feature image, and determine the first feature information of the target to be detected in the preset coordinate system, as long as the accuracy of identifying the target to be detected and the accuracy and reliability of obtaining the first feature information can be guaranteed, and no further details will be given here.

[0091] In this embodiment, by acquiring a feature image corresponding to a monocular image, identifying the target to be detected included in the monocular image based on the feature image, and determining the first feature information of the target to be detected in a preset coordinate system, the accuracy and reliability of the acquisition of the first feature information are effectively achieved, and the accuracy of detecting the three-dimensional features of the target in the monocular image is further improved.

[0092] Figure 4A schematic diagram of a process for determining projection information of a target to be detected in a monocular image according to an embodiment of the present invention; based on the above embodiment, further reference is made to the attached Figure 4 As shown, this embodiment does not limit the specific implementation method for determining the projection information of the target to be detected in the monocular image. Those skilled in the art can set it according to specific application requirements and design requirements. Preferably, in this embodiment, determining the projection information of the target to be detected in the monocular image may include:

[0093] Step S401: Determine the position information of the target to be detected in the monocular image.

[0094] After obtaining the feature image corresponding to the monocular image, the feature image may be analyzed and processed to determine the position information of the target to be detected in the monocular image. Specifically, determining the position information of the target to be detected in the monocular image may include:

[0095] Step S4011: Process the feature image using the first machine learning model to obtain position information of the target to be detected in the monocular image, wherein the first machine learning model is trained to identify the position information of the target in the monocular image based on the feature image.

[0096] The first machine learning model may be pre-trained to determine, based on a feature image, the location information of a target in a monocular image corresponding to the feature image. Specifically, the feature image may include at least one of the following: color features, brightness features, texture features, edge features, basic shape features, semantic features, etc. Furthermore, when training the machine learning model, the machine learning model may be established based on a convolutional neural network.

[0097] In this embodiment, the feature image is analyzed and processed by the trained first machine learning model to obtain the position information of the target to be detected in the monocular image, which effectively ensures the quality and efficiency of obtaining the position information of the target to be detected in the monocular image, and further improves the quality and efficiency of the application of the detection method of the target in the monocular image.

[0098] Step S402: Determine the projection information of the target to be detected in the monocular image according to the feature image and the position information.

[0099] After obtaining the position information of the target to be detected in the monocular image, the feature image and the position information can be analyzed and processed to determine the projection information of the target to be detected in the monocular image. Specifically, determining the projection information of the target to be detected in the monocular image based on the feature image and the position information can include:

[0100] Step S4021: Determine the projection offset information of the target to be detected in the monocular image according to the feature image.

[0101] Step S4022: Determine the projection information of the target to be detected in the monocular image based on the projection offset information and the position information.

[0102] Among them, after obtaining the feature image, the feature image can be analyzed and processed. Specifically, the feature image can be analyzed and processed using a pre-trained third machine learning model, so that the projection offset information of the target to be detected in the monocular image can be obtained. The above-mentioned third machine learning model is trained to determine the projection offset information of the target to be detected in the monocular image based on the feature image. The projection offset information is used to identify the offset of the projection information of the target to be detected in the monocular image relative to the position information of the target to be detected in the monocular image. Therefore, after obtaining the projection offset information, the projection offset information and the position information can be analyzed and processed, so that the projection information of the target to be detected in the monocular image can be determined.

[0103] In this embodiment, by determining the position information of the target to be detected in the monocular image, the projection information of the target to be detected in the monocular image is determined based on the feature image and the position information, thereby achieving accurate and effective acquisition of the projection information of the target to be detected in the monocular image, further improving the accuracy and reliability of detecting the three-dimensional features of the target in the monocular image.

[0104] Figure 5 A schematic diagram of a process for identifying other targets included in a monocular image and determining at least one relative feature information between the target to be detected and the other targets in a preset coordinate system is provided in an embodiment of the present invention; based on the above embodiment, refer to the attached Figure 5 As shown, in this embodiment, the specific implementation method for identifying other targets included in the monocular image and determining at least one relative feature information between the target to be detected and the other targets in the preset coordinate system is not limited. Those skilled in the art can, based on specific application requirements and setting requirements, more preferably, in this embodiment, identifying other targets included in the monocular image and determining at least one relative feature information between the target to be detected and the other targets in the preset coordinate system may include:

[0105] Step S501: Acquire a feature image corresponding to a monocular image.

[0106] After obtaining the monocular image, the monocular image may be analyzed and processed to obtain a feature image corresponding to the monocular image. Specifically, obtaining the feature image corresponding to the monocular image may include:

[0107] Step S5011: Analyze and process the monocular image using the second machine learning model to obtain a feature image corresponding to the monocular image, wherein the second machine learning model is trained to determine the feature image corresponding to the monocular image.

[0108] The second machine learning model may be pre-trained to determine a feature image corresponding to a monocular image. The feature image may include at least one of the following: pixel feature information, color feature information, brightness feature information, texture feature information, edge feature information, shape feature information, and semantic feature information. Furthermore, when training the machine learning model, the machine learning model may be established based on a convolutional neural network. Therefore, after obtaining a monocular image, the second machine learning model may be used to analyze and process the monocular image, thereby obtaining a feature image corresponding to the monocular image.

[0109] In this embodiment, the monocular image is analyzed and processed by the trained second machine learning model to obtain a feature image corresponding to the monocular image, which not only effectively ensures the accuracy and reliability of the feature image acquisition, but also ensures the quality and efficiency of the feature image acquisition, further improving the quality and efficiency of the application of the detection method for targets in monocular images.

[0110] Step S502: identifying other targets included in the monocular image based on the feature image, and determining at least one relative feature information between the target to be detected and the other targets.

[0111] After obtaining the feature image, the feature image may be analyzed and processed to identify other targets included in the monocular image and determine at least one relative feature information between the target to be detected and the other targets. Specifically, identifying other targets included in the monocular image based on the feature image and determining at least one relative feature information between the target to be detected and the other targets may include:

[0112] Step S5021: Use the third machine learning model to process the feature image to obtain other targets included in the monocular image corresponding to the feature image, and determine at least one relative feature information between the target to be detected and the other targets, wherein the third machine learning model is trained to identify and determine the targets included in the monocular image, and determine the relative feature information between any two targets.

[0113] Among them, the third machine learning model can be pre-trained to be used to identify and determine the targets included in the monocular image, and to determine the relative feature information between any two targets. Specifically, when learning and training the machine learning model, a machine learning model can be established based on a convolutional neural network. Therefore, after obtaining the feature image, the third machine learning model can be used to analyze and process the feature image, so that other targets included in the monocular image corresponding to the feature image (that is, all target objects except the target to be detected) can be obtained, and at least one relative feature information between the target to be detected and other targets can be determined. The relative feature information may include at least one of the following: relative position information, relative size information, relative angle information of the target to be detected relative to other targets, and so on.

[0114] In this embodiment, the monocular image is analyzed and processed by a trained third machine learning model, and the feature image is processed using the third machine learning model to obtain other targets included in the monocular image corresponding to the feature image, and at least one relative feature information between the target to be detected and the other targets is determined, thereby effectively ensuring the quality and efficiency of obtaining at least one relative feature information, and further improving the quality and efficiency of applying the detection method for targets in monocular images.

[0115] Figure 6 A flow chart of adjusting the first feature information based on at least one relative feature information to obtain the second feature information of the target to be detected is provided in an embodiment of the present invention; based on the above embodiment, continue to refer to the attached Figure 6 As shown, this embodiment does not limit the specific implementation method for adjusting the first feature information based on at least one relative feature information to obtain the second feature information of the target to be detected. Those skilled in the art can set it according to specific application requirements and design requirements. One implementation method is that in this embodiment, adjusting the first feature information based on at least one relative feature information to obtain the second feature information of the target to be detected may include:

[0116] Step S601: Determine uncertainty information corresponding to any one relative feature information in at least one relative feature information.

[0117] Step S602: When the uncertainty information satisfies the preset condition, the first feature information is adjusted based on the relative feature information corresponding to the uncertainty information to obtain the second feature information of the target to be detected. Or,

[0118] Step S603: When the uncertainty information does not meet the preset condition, the relative feature information corresponding to the uncertainty information is ignored.

[0119] After obtaining at least one relative feature information, uncertainty information corresponding to any relative feature information can be determined. The uncertainty information is used to identify the uncertainty of the obtained relative feature information. Specifically, a pre-trained machine learning model can be used to analyze and process the at least one relative feature information to obtain the uncertainty information corresponding to any relative feature information.

[0120] After obtaining the uncertainty information corresponding to any relative feature information, the uncertainty information can be analyzed and matched with the preset conditions, wherein the preset conditions can be a preset maximum limit value used to determine whether the uncertainty information of the relative feature information meets the preset requirements. It can be understood that different application scenarios and application requirements can correspond to different preset conditions, and the preset conditions can also have different forms of expression, for example: when the uncertainty information is expressed in the form of a score, the preset conditions can also be expressed in the form of a score; or, when the uncertainty information is expressed in the form of probability, the preset conditions can also be expressed in the form of probability.

[0121] If the result of analyzing and matching the uncertainty information with the preset conditions is that the uncertainty information meets the preset conditions, then the relative feature information corresponding to the uncertainty information has a low uncertainty and a high confidence level, and thus the first feature information can be adjusted based on the relative feature information to obtain the second feature information of the target to be detected. If the result of analyzing and matching is that the uncertainty information does not meet the preset conditions, then the relative feature information corresponding to the uncertainty information has a high uncertainty and a low confidence level, and thus the relative feature information corresponding to the uncertainty information can be ignored.

[0122] In this embodiment, by determining the uncertainty information corresponding to any one relative feature information in at least one relative feature information, when the uncertainty information meets the preset conditions, the first feature information is adjusted based on the relative feature information corresponding to the uncertainty information to obtain the second feature information of the target to be detected; when the uncertainty information does not meet the preset conditions, the relative feature information corresponding to the uncertainty information is ignored; thereby effectively achieving that when the confidence level of the relative feature information is high, the first feature information is adjusted using the relative feature information; when the confidence level of the relative feature information is low, the first feature information will not be adjusted using the relative feature information, thereby effectively ensuring the accuracy and reliability of obtaining the second feature information.

[0123] Figure 7 Schematic diagram of a process for adjusting first feature information based on at least one relative feature information to obtain second feature information of a target to be detected according to an embodiment of the present invention Figure 2Based on the above embodiment, continue to refer to the attached Figure 7 As shown, this embodiment does not limit the specific implementation method for adjusting the first feature information based on at least one relative feature information to obtain the second feature information of the target to be detected. Those skilled in the art can set it according to specific application requirements and design requirements. Another implementation method is that in this embodiment, adjusting the first feature information based on at least one relative feature information to obtain the second feature information of the target to be detected may include:

[0124] Step S701: generating at least one adjustment parameter for adjusting first feature information based on at least one relative feature information.

[0125] Step S702: Determine confidence information of at least one adjustment parameter.

[0126] Step S703: When the confidence information is greater than or equal to the preset threshold, the first feature information is adjusted based on the adjustment parameter corresponding to the confidence information to obtain the second feature information of the target to be detected. Or,

[0127] Step S704: when the confidence information is less than a preset threshold, the adjustment parameter corresponding to the confidence information is ignored.

[0128] Among them, after obtaining at least one relative feature information, at least one adjustment parameter for adjusting the first feature information can be generated based on the at least one relative feature information. Specifically, a pre-trained machine learning model can be used to analyze and process the at least one relative feature information, so as to generate at least one adjustment parameter for adjusting the first feature information.

[0129] After obtaining at least one adjustment parameter for adjusting the first feature information, the confidence information of the at least one adjustment parameter can be determined. Specifically, the adjustment parameter can be analyzed and identified using a preset machine learning model, thereby obtaining at least one confidence information corresponding to the at least one adjustment parameter. The confidence information can then be analyzed and matched with a preset threshold, wherein the preset threshold can be a preset limit information used to determine whether the confidence information meets the requirements. It is understandable that different application scenarios and application requirements can correspond to different preset thresholds, and the preset threshold can also have different display forms. For example, when the confidence information is expressed in the form of a score, the preset threshold can also be expressed in the form of a score; when the confidence information is expressed in the form of a probability, the preset threshold can also be expressed in the form of a probability.

[0130] When the result of analyzing and matching the confidence information with the preset threshold is that the confidence information is greater than or equal to the preset threshold, it means that the confidence of the adjustment parameter corresponding to the confidence information is high, and thus the first feature information can be adjusted based on the adjustment parameter corresponding to the confidence information. When the matching result is that the confidence information is less than the preset threshold, it means that the confidence of the adjustment parameter corresponding to the confidence information is low, and thus the adjustment parameter corresponding to the confidence information can be ignored.

[0131] In this embodiment, at least one adjustment parameter for adjusting the first feature information is generated based on at least one relative feature information, and confidence information for the at least one adjustment parameter is determined. When the confidence information is greater than or equal to a preset threshold, the first feature information is adjusted based on the adjustment parameter corresponding to the confidence information to obtain second feature information of the target to be detected. When the confidence information is less than the preset threshold, the adjustment parameter corresponding to the confidence information is ignored. This effectively ensures that when the confidence level of the adjustment parameter is high, the adjustment parameter can be used to adjust the first feature information; when the confidence level of the adjustment parameter is low, the adjustment parameter is not used to adjust the first feature information, thereby effectively ensuring the accurate and reliable acquisition of the second feature information.

[0132] Figure 8 The present invention provides a flow chart of adjusting the first feature information based on the adjustment parameter corresponding to the confidence information to obtain the second feature information of the target to be detected; based on the above embodiment, continue to refer to the attached Figure 8 As shown, in this embodiment, adjusting the first feature information based on the adjustment parameter corresponding to the confidence information to obtain the second feature information of the target to be detected may include:

[0133] Step S801: adjusting the first feature information based on the adjustment parameter corresponding to the confidence information to obtain at least one adjusted relative feature information between the target to be detected and other targets.

[0134] Step S802: determining an information offset between at least one adjusted relative feature information and corresponding relative feature information.

[0135] Step S803: When the information offset is less than or equal to the preset offset threshold, the second feature information of the target to be detected is obtained.

[0136] Among them, after obtaining the adjustment parameters corresponding to the confidence information, the first feature information can be adjusted based on the adjustment parameters, and then at least one adjusted relative feature information between the target to be detected and other targets can be obtained. After that, the adjusted relative feature information and the relative feature information can be analyzed and matched, so as to determine the information offset between at least one adjusted relative feature information and the corresponding relative feature information.

[0137] After obtaining the information offset, the information offset can be analyzed and compared with a preset offset threshold, where the preset offset threshold is a preset maximum offset used to determine whether the adjusted relative feature information is close to the relative feature information. If the result of the analysis and comparison of the information offset with the preset offset threshold is that the information offset is less than or equal to the preset offset threshold, then the adjustment operation on the first feature information has met the preset requirements, and the adjusted first feature information can be determined as the second feature information, thereby effectively obtaining the second feature information of the target to be detected. The second feature information is the three-dimensional feature information of the target to be detected in the preset three-dimensional coordinate system obtained based on the monocular image.

[0138] In this embodiment, the first characteristic information is adjusted based on the adjustment parameter corresponding to the confidence information, and at least one adjusted relative characteristic information between the target to be detected and other targets is obtained. The information offset between the at least one adjusted relative characteristic information and the corresponding relative characteristic information is determined. When the information offset is less than or equal to the preset offset threshold, the second characteristic information of the target to be detected is obtained, thereby effectively achieving the stable reliability of the adjustment of the first characteristic information and further improving the accuracy of obtaining the second characteristic information.

[0139] Figure 9 A flow chart of another method for detecting a target in a monocular image provided by an embodiment of the present invention; based on the above embodiment, further reference is made to the attached Figure 9 As shown, after obtaining the second feature information of the target to be detected, the method in this embodiment may further include:

[0140] Step S901: Generate a labeling box corresponding to the target to be detected based on the second feature information.

[0141] Step S902: Use a labeling box to mark the target to be detected.

[0142] Specifically, after obtaining the second feature information of the target to be detected, a labeling box corresponding to the target to be detected can be generated based on the second feature information, and then the labeling box can be used to label the target to be detected, so that the user can directly obtain the three-dimensional feature information of the target to be detected in the monocular image through the labeling box, further improving the practicality of the detection method.

[0143] For specific applications, refer to the attached Figure 2 and attached Figure 10 As shown, this application embodiment provides a method for detecting targets in a monocular image. The detection method can be performed by a detection system, which can include a backbone network, a two-dimensional detection branch network, a three-dimensional detection branch network, and a paired detection branch network. Specifically, the detection method may include the following steps:

[0144] Step 1: Obtain a monocular image, input the monocular image into the backbone network, and obtain the feature image corresponding to the monocular image.

[0145] A monocular image is obtained by a camera sensor and then fed into a backbone network. The backbone network is a pre-trained machine learning model that is used to obtain a feature image corresponding to the monocular image. The feature image may include at least one of the following: color feature information, brightness feature information, edge feature information, basic shape feature information, descriptive feature information for a preset area, semantic feature information, and the like.

[0146] Step 2: Input the feature image into three detection branch networks respectively. Each branch network is as follows:

[0147] Step 2.1: Input the feature image into the two-dimensional detection branch network to obtain the two-dimensional detection box of the target to be detected in the monocular image.

[0148] The two-dimensional detection frame can be a regularly shaped identification frame used to identify the characteristic information of the target to be detected in the monocular image. This characteristic information may include position information, size information, angle information, etc. For example, after inputting the feature image into the two-dimensional detection branch network, the center point and size information of the two-dimensional detection frame can be obtained. The center point and size information of the two-dimensional detection frame can be used to identify the position information of the target to be detected in the monocular image.

[0149] Step 2.2: Input the feature image into the 3D detection branch network to obtain the 3D detection frame of the target to be detected in the monocular image.

[0150] Among them, the three-dimensional detection frame can be an irregularly shaped identification frame, which is used to identify the three-dimensional feature information of the target to be detected in the preset coordinate system. The three-dimensional feature information may include position information, size information, angle information, etc. It should be noted that when the three-dimensional feature information includes angle information, the angle information can be the rotation angle of the target to be detected in the preset plane, or the angle information can also be the spatial angle of the target to be detected in the preset coordinate system.

[0151] In addition, the above-mentioned preset coordinate system corresponds to the monocular image. Specifically, the forward direction of the camera lens can be determined as the Z-axis direction in the coordinate system, the horizontal right direction can be determined as the X-axis direction, and the vertical downward direction can be determined as the Y-axis direction. Among them, the optical center position of the camera can be the origin position in the preset coordinate system.

[0152] Step 2.3: Input the feature image into the pair detection branch network to determine at least one three-dimensional spatial constraint information between the target to be detected and other targets in the preset coordinate system.

[0153] Among them, other targets include at least one of the following: a target object blocked by the target to be detected, a target object blocking the target to be detected, and a target object independent of the target to be detected, such as Figure 2 As shown, when the target to be detected is a vehicle, the other target can be a reference vehicle that is independent of the vehicle and located in the upper left corner of the monocular image. In addition, the at least one three-dimensional spatial constraint information obtained between the target to be detected and the other targets can also be called relative feature information. Specifically, the relative feature information can include at least one of the following: relative position information, relative size information, and relative angle information between the target to be detected and the other targets in a preset coordinate system. After obtaining at least one relative feature information, the target feature information of the target to be detected can be obtained through the relative feature information. This effectively solves the problem of reducing the accuracy of detecting the feature information of the target to be detected when the target to be detected is occluded due to incomplete or partial missing information of the target to be detected.

[0154] Step 3: Based on at least one three-dimensional spatial constraint information, the feature information corresponding to the three-dimensional detection frame can be optimized and adjusted to obtain target feature information corresponding to the target to be detected.

[0155] Among them, based on at least one three-dimensional spatial constraint information, the feature information corresponding to the three-dimensional detection frame can be optimized and adjusted so that the relative feature relationship between the target to be detected and other targets satisfies the above-mentioned at least one three-dimensional spatial constraint information, that is, the relative feature relationship between the target to be detected and other targets is consistent with at least one three-dimensional spatial constraint information.

[0156] Specifically, when optimizing and adjusting the feature information corresponding to the three-dimensional detection frame based on at least one three-dimensional spatial constraint information, all objects (targets to be detected and other targets) on the monocular image are adjusted based on the at least one three-dimensional spatial constraint information, or some objects on the monocular image are adjusted based on the at least one three-dimensional spatial constraint information, so that the relative feature information of all objects can satisfy the above-mentioned at least one three-dimensional spatial constraint information, and then the target feature information corresponding to the target to be detected can be obtained.

[0157] Step 4: The adjusted feature information is determined as the final output 3D object detection frame.

[0158] The final output three-dimensional object detection frame may include the position information, angle information, and size information of the target to be detected in the preset coordinate system.

[0159] This application embodiment proposes a concise, efficient and practical method for detecting three-dimensional targets in monocular images. Specifically, the geometric constraint relationship between multiple targets can be predicted through a pre-trained machine learning model, and then the prediction result of a single target is globally optimized using the obtained geometric constraint relationship, thereby effectively realizing the detection estimation of the features of the target to be detected through the relative feature information between the target to be detected and other targets, that is, by mining the constraint relationship between targets, the target information is accurately predicted. In this way, when it is difficult to detect a single object, the feature information of the target to be detected can be optimized by using the relative feature information between the surrounding objects that are easy to detect, thereby overcoming the defect in the prior art that when the target object is small or the target object is blocked, it is impossible to accurately detect the target. Therefore, the method in this embodiment can effectively improve the accuracy of detecting the three-dimensional feature information of the target in the monocular image, and further improve the practicality of the method.

[0160] Figure 11 A flow chart of a vehicle positioning method provided by an embodiment of the present invention; Figure 11 As shown, this embodiment provides a vehicle positioning method. The execution subject of the vehicle positioning method can be a vehicle positioning device. The vehicle positioning device can be applied to a vehicle monitoring system, a road monitoring system, etc. Specifically, the vehicle positioning method may include:

[0161] S1101: Acquire a monocular image.

[0162] Among them, the monocular image can refer to the image information obtained from a single camera sensor or a vehicle monitoring video. This embodiment does not limit the specific implementation method of the vehicle positioning device to obtain the monocular image. Those skilled in the art can set it according to specific application requirements and design requirements. For example, the camera sensor or the vehicle monitoring device can be connected to the vehicle positioning device for communication. After the camera sensor captures the monocular image or the vehicle monitoring device obtains the vehicle monitoring video, the vehicle positioning device can obtain the monocular image through the camera sensor, or the monocular image can be extracted from the vehicle monitoring video. Specifically, the vehicle positioning device can actively obtain the monocular image obtained by the camera sensor, or the camera sensor can actively send the monocular image to the detection device, so that the detection device obtains the monocular image. Alternatively, the monocular image can be stored in a preset area, and the detection device can obtain the monocular image by accessing the preset area.

[0163] S1102: Identify the vehicle to be identified and the reference vehicle included in the monocular image.

[0164] After obtaining the monocular image, the monocular image can be analyzed and processed to identify the vehicle to be identified and the reference vehicle included in the monocular image, wherein the positional relationship between the reference vehicle and the vehicle to be identified may include the following situations: the vehicle to be identified and the reference vehicle are independent of each other in the monocular image, the vehicle to be identified is at least partially obscured by the reference vehicle in the monocular image, and the vehicle to be identified obscures at least part of the reference vehicle in the monocular image.

[0165] S1103: Determine first positioning information of the vehicle to be identified in a preset coordinate system, and at least one relative positioning information between the vehicle to be identified and the reference vehicle in the preset coordinate system, wherein the preset coordinate system corresponds to the monocular image.

[0166] After determining the vehicle to be identified and the reference vehicle, the first positioning information of the vehicle to be identified in the preset coordinate system and at least one relative positioning information between the vehicle to be identified and the reference vehicle in the preset coordinate system can be determined based on the barrage image. It can be understood that the first positioning information may include at least one of the following: position information, size information, angle information of the target to be detected in the preset coordinate system. Similarly, the relative positioning information may include at least one of the following: relative position information, relative size information, relative angle information of the target to be detected in the preset coordinate system.

[0167] S1104: Adjust the first positioning information according to the at least one relative feature information to obtain second positioning information of the vehicle to be identified.

[0168] It should be noted that the execution process and technical effects of the above steps S1101-S1104 in this embodiment are the same as those in the above Figure 1 The execution process and technical effects in the illustrated embodiments are similar, and for details, please refer to the above statements, which will not be repeated here.

[0169] Figure 12 A schematic diagram of a device for detecting a target in a monocular image according to an embodiment of the present invention; Figure 12 As shown, this embodiment provides a device for detecting a target in a monocular image. The device can perform the above Figure 1 The method for detecting a target in a monocular image is shown. Specifically, the detection device may include:

[0170] A first acquisition module 11 is used to acquire a monocular image;

[0171] a first detection module 12, configured to identify a target to be detected included in the monocular image and determine first feature information of the target to be detected in a preset coordinate system, wherein the preset coordinate system corresponds to the monocular image;

[0172] The first detection module 12 is used to identify other targets included in the monocular image and determine at least one relative feature information between the target to be detected and other targets in a preset coordinate system;

[0173] The first processing module 13 is configured to adjust the first feature information based on at least one relative feature information to obtain second feature information of the target to be detected.

[0174] In some instances, the first feature information includes at least one of the following: position information, size information, and angle information of the target to be detected in a preset coordinate system; correspondingly, the second feature information includes at least one of the following: position information, size information, and angle information of the target to be detected in a preset coordinate system.

[0175] In some examples, the other targets include at least one of the following: a target object blocked by the target to be detected, a target object blocking the target to be detected, and a target object independent of the target to be detected.

[0176] In some examples, the relative feature information includes at least one of the following: relative position information, relative size information, and relative angle information between the target to be detected and other targets in a preset coordinate system.

[0177] In some instances, when the first detection module 12 identifies the target to be detected included in the monocular image and determines the first feature information of the target to be detected in the preset coordinate system, the first detection module 12 can be used to perform: obtaining a feature image corresponding to the monocular image; identifying the target to be detected included in the monocular image based on the feature image, and determining the first feature information of the target to be detected in the preset coordinate system.

[0178] In some instances, when the first detection module 12 identifies the target to be detected included in the monocular image based on the feature image and determines the first feature information of the target to be detected in the preset coordinate system, the first detection module 12 can be used to perform: identifying the target to be detected based on the feature image and determining the projection information of the target to be detected in the monocular image; determining the first feature information of the target to be detected in the preset coordinate system based on the projection information.

[0179] In some instances, when the first detection module 12 determines the projection information of the target to be detected in the monocular image, the first detection module 12 can be used to perform: determining the position information of the target to be detected in the monocular image; and determining the projection information of the target to be detected in the monocular image based on the feature image and the position information.

[0180] In some instances, when the first detection module 12 determines the position information of the target to be detected in the monocular image, the first detection module 12 can be used to perform: processing the feature image using the first machine learning model to obtain the position information of the target to be detected in the monocular image, wherein the first machine learning model is trained to identify the position information of the target in the monocular image based on the feature image.

[0181] In some instances, when the first detection module 12 determines the projection information of the target to be detected in the monocular image based on the feature image and position information, the first detection module 12 can be used to perform: determining the projection offset information of the target to be detected in the monocular image based on the feature image; and determining the projection information of the target to be detected in the monocular image based on the projection offset information and the position information.

[0182] In some instances, when the first detection module 12 identifies other targets included in the monocular image and determines at least one relative feature information between the target to be detected and the other targets in a preset coordinate system, the first detection module 12 can be used to perform: obtaining a feature image corresponding to the monocular image; identifying other targets included in the monocular image based on the feature image, and determining at least one relative feature information between the target to be detected and the other targets.

[0183] In some instances, when the first detection module 12 obtains a feature image corresponding to a monocular image, or when the first detection module 12 obtains a feature image corresponding to a monocular image, the first detection module 12 can be used to perform: analyzing and processing the monocular image using a second machine learning model to obtain a feature image corresponding to the monocular image, wherein the second machine learning model is trained to determine the feature image corresponding to the monocular image.

[0184] In some examples, the feature image includes at least one of the following: pixel feature information, color feature information, edge feature information, shape feature information, and semantic feature information.

[0185] In some instances, when the first detection module 12 identifies other targets included in the monocular image based on the feature image and determines at least one relative feature information between the target to be detected and the other targets, the first detection module 12 can be used to execute: processing the feature image using a third machine learning model to obtain other targets included in the monocular image corresponding to the feature image, and determining at least one relative feature information between the target to be detected and the other targets, wherein the third machine learning model is trained to identify and determine the targets included in the monocular image and determine the relative feature information between any two targets.

[0186] In some instances, when the first processing module 13 adjusts the first feature information based on at least one relative feature information to obtain the second feature information of the target to be detected, the first processing module 13 can be used to perform: determining the uncertainty information corresponding to any one of the at least one relative feature information; when the uncertainty information meets the preset conditions, adjusting the first feature information based on the relative feature information corresponding to the uncertainty information to obtain the second feature information of the target to be detected; or, when the uncertainty information does not meet the preset conditions, ignoring the relative feature information corresponding to the uncertainty information.

[0187] In some instances, when the first processing module 13 adjusts the first feature information based on at least one relative feature information to obtain second feature information of the target to be detected, the first processing module 13 can be used to perform: generating at least one adjustment parameter for adjusting the first feature information based on at least one relative feature information; determining confidence information of at least one adjustment parameter; when the confidence information is greater than or equal to a preset threshold, adjusting the first feature information based on the adjustment parameter corresponding to the confidence information to obtain the second feature information of the target to be detected; or, when the confidence information is less than the preset threshold, ignoring the adjustment parameter corresponding to the confidence information.

[0188] In some instances, when the first processing module 13 adjusts the first feature information based on the adjustment parameters corresponding to the confidence information to obtain the second feature information of the target to be detected, the first processing module 13 can be used to perform: adjusting the first feature information based on the adjustment parameters corresponding to the confidence information to obtain at least one adjusted relative feature information between the target to be detected and other targets; determining the information offset between at least one adjusted relative feature information and the corresponding relative feature information; and when the information offset is less than or equal to a preset offset threshold, obtaining the second feature information of the target to be detected.

[0189] In some instances, after obtaining the second characteristic information of the target to be detected, the first processing module 13 in this embodiment can be used to execute: generating a labeling box corresponding to the target to be detected based on the second characteristic information; and identifying the target to be detected using the labeling box.

[0190] Figure 12 The device shown can perform Figures 1-10 For the method of the embodiment shown in FIG. 1 , reference may be made to the description of the part not described in detail in the embodiment. Figures 1-10 The implementation process and technical effects of this technical solution can be found in Figures 1-10 The description in the illustrated embodiment will not be repeated here.

[0191] In one possible design, Figure 12 The structure of the device for detecting a target in a monocular image shown in the figure can be realized as an electronic device, which can be a mobile phone, a tablet computer, a server or other devices. Figure 13 As shown, the electronic device may include: a first processor 21 and a first memory 22. The first memory 22 is used to store the corresponding electronic device to perform the above Figures 1-10 In the illustrated embodiment, the first processor 21 is configured to execute the program stored in the first memory 22 .

[0192] The program includes one or more computer instructions, wherein when the one or more computer instructions are executed by the first processor 21, the following steps can be implemented:

[0193] Acquire a monocular image;

[0194] Identifying a target to be detected included in the monocular image and determining first feature information of the target to be detected in a preset coordinate system, wherein the preset coordinate system corresponds to the monocular image;

[0195] Identify other targets included in the monocular image and determine at least one relative feature information between the target to be detected and the other targets in a preset coordinate system;

[0196] The first characteristic information is adjusted based on at least one relative characteristic information to obtain second characteristic information of the target to be detected.

[0197] Furthermore, the first processor 21 is also used to execute the aforementioned Figures 1-10 All or part of the steps in the illustrated embodiments.

[0198] The structure of the electronic device may further include a first communication interface 23 for the electronic device to communicate with other devices or a communication network.

[0199] In addition, an embodiment of the present invention provides a computer storage medium for storing computer software instructions used by electronic devices, which includes instructions for executing the above Figures 1-10 The procedures involved in the method for detecting targets in a monocular image in the method embodiment shown.

[0200] Figure 14 A schematic diagram of the structure of a vehicle positioning device provided by an embodiment of the present invention; Figure 14 As shown, this embodiment provides a vehicle positioning device, which can perform the above Figure 11 The vehicle positioning method shown. Specifically, the vehicle positioning device may include:

[0201] A second acquisition module 31 is used to acquire a monocular image;

[0202] A second recognition module 32 is configured to recognize a vehicle to be recognized and a reference vehicle included in the monocular image;

[0203] a second determining module 33, configured to determine first positioning information of the vehicle to be identified in a preset coordinate system and at least one relative positioning information between the vehicle to be identified and the reference vehicle in the preset coordinate system, wherein the preset coordinate system corresponds to the monocular image;

[0204] The second positioning module 34 is configured to adjust the first positioning information according to the at least one relative feature information to obtain second positioning information of the vehicle to be identified.

[0205] Figure 14 The device shown can also perform Figure 2-Figure 11 For the method of the embodiment shown in FIG. 1 , reference may be made to the description of the part not described in detail in the embodiment. Figure 2-Figure 11 The implementation process and technical effects of this technical solution can be found in Figure 2-Figure 11 The description in the illustrated embodiment will not be repeated here.

[0206] In one possible design, Figure 14The structure of the vehicle positioning device shown can be implemented as an electronic device, which can be a mobile phone, tablet computer, server and other devices. Figure 15 As shown, the electronic device may include: a second processor 41 and a second memory 42. The second memory 42 is used to store the corresponding electronic device to perform the above Figure 2-Figure 11 In the illustrated embodiment, the second processor 41 is configured to execute the program of the vehicle positioning method stored in the second memory 42 .

[0207] The program includes one or more computer instructions, wherein when the one or more computer instructions are executed by the second processor 41, the following steps can be implemented:

[0208] Acquire a monocular image;

[0209] Identifying a to-be-identified vehicle and a reference vehicle included in the monocular image;

[0210] determining first positioning information of the vehicle to be identified in a preset coordinate system and at least one relative positioning information between the vehicle to be identified and the reference vehicle in the preset coordinate system, wherein the preset coordinate system corresponds to the monocular image;

[0211] The first positioning information is adjusted according to the at least one relative feature information to obtain second positioning information of the vehicle to be identified.

[0212] Furthermore, the second processor 41 is also used to execute the aforementioned Figure 2-Figure 11 All or part of the steps in the illustrated embodiments.

[0213] The structure of the electronic device may further include a second communication interface 43 for the electronic device to communicate with other devices or a communication network.

[0214] In addition, an embodiment of the present invention provides a computer storage medium for storing computer software instructions used by electronic devices, which includes instructions for executing the above Figure 2-Figure 11 The procedures involved in the vehicle positioning method in the method embodiment shown.

[0215] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0216] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by adding a necessary general hardware platform, and of course can also be implemented by a combination of hardware and software. Based on this understanding, the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a computer product. The present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0217] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable device to produce a machine, so that the instructions executed by the processor of the computer or other programmable device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0218] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0219] These computer program instructions can also be loaded onto a computer or other programmable device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0220] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0221] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0222] Computer-readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0223] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for detecting a target in a monocular image, characterized in that: include: Get a monocular image; Identify a target to be detected included in a monocular image, and determine first feature information of the target to be detected in a preset coordinate system, wherein the preset coordinate system corresponds to the monocular image; Identify other targets included in the monocular image, and determine at least one relative feature information between the target to be detected and the other targets in the preset coordinate system, wherein the relative feature information is spatial constraint information of the target to be detected relative to the other targets in the preset coordinate system; The first feature information is adjusted based on the at least one relative feature information so that the relative feature information is satisfied between the adjusted target to be detected and other targets, thereby obtaining second feature information of the target to be detected.

2. The method according to claim 1, characterized in that The first feature information includes at least one of the following: position information, size information, and angle information of the target to be detected in the preset coordinate system; Correspondingly, the second feature information includes at least one of the following: position information, size information, and angle information of the target to be detected in the preset coordinate system.

3. The method according to claim 1, characterized in that: The other objectives include at least one of the following: A target object blocked by the target to be detected, a target object blocking the target to be detected, and a target object independent of the target to be detected.

4. The method according to claim 1, characterized in that: The relative feature information includes at least one of the following: relative position information, relative size information, and relative angle information between the target to be detected and other targets in the preset coordinate system.

5. The method according to claim 1, characterized in that Identifying a target to be detected included in a monocular image and determining first feature information of the target to be detected in a preset coordinate system, including: Acquire a feature image corresponding to the monocular image; The target to be detected included in the monocular image is identified according to the feature image, and first feature information of the target to be detected in the preset coordinate system is determined.

6. The method according to claim 5, characterized in that Identifying the target to be detected included in the monocular image according to the feature image, and determining first feature information of the target to be detected in the preset coordinate system, including: Identify the target to be detected according to the feature image, and determine projection information of the target to be detected in the monocular image; Determine first feature information of the target to be detected in the preset coordinate system according to the projection information.

7. The method according to claim 6, characterized in that Determining projection information of the target to be detected in the monocular image includes: Determine the position information of the target to be detected in the monocular image; Projection information of the target to be detected in the monocular image is determined according to the feature image and the position information.

8. The method according to claim 7, characterized in that Determining the position information of the target to be detected in the monocular image includes: The feature image is processed using a first machine learning model to obtain position information of the target to be detected in the monocular image, wherein the first machine learning model is trained to identify the position information of the target in the monocular image based on the feature image.

9. The method according to claim 7, characterized in that: Determining projection information of the target to be detected in the monocular image according to the feature image and the position information includes: Determine projection offset information of the target to be detected in the monocular image according to the feature image; Based on the projection offset information and the position information, projection information of the target to be detected in the monocular image is determined.

10. The method according to claim 1, characterized in that Identifying other targets included in the monocular image and determining at least one relative feature information between the target to be detected and the other targets in the preset coordinate system, including: Acquire a feature image corresponding to the monocular image; Based on the feature image, other targets included in the monocular image are identified, and at least one relative feature information between the target to be detected and the other targets is determined.

11. The method according to claim 5 or 10, characterized in that: Acquiring a feature image corresponding to the monocular image, comprising: The monocular image is analyzed and processed using a second machine learning model to obtain a feature image corresponding to the monocular image, wherein the second machine learning model is trained to determine the feature image corresponding to the monocular image.

12. The method according to claim 11, characterized in that The feature image includes at least one of the following: pixel feature information, color feature information, edge feature information, shape feature information, and semantic feature information.

13. The method according to claim 10, characterized in that Identifying other targets included in the monocular image based on the feature image, and determining at least one relative feature information between the target to be detected and the other targets, including: The feature image is processed using a third machine learning model to obtain other targets included in the monocular image corresponding to the feature image, and at least one relative feature information between the target to be detected and the other targets is determined, wherein the third machine learning model is trained to identify and determine the targets included in the monocular image, and determine the relative feature information between any two targets.

14. The method according to any one of claims 1 to 10, characterized in that: Adjusting the first feature information based on the at least one relative feature information to obtain second feature information of the target to be detected includes: Determining uncertainty information corresponding to any one of the at least one relative feature information; When the uncertainty information satisfies a preset condition, the first feature information is adjusted based on relative feature information corresponding to the uncertainty information to obtain second feature information of the target to be detected; or, When the uncertainty information does not meet the preset condition, the relative feature information corresponding to the uncertainty information is ignored.

15. The method according to any one of claims 1 to 10, characterized in that: Adjusting the first feature information based on the at least one relative feature information to obtain second feature information of the target to be detected includes: generating at least one adjustment parameter for adjusting the first feature information based on the at least one relative feature information; determining confidence information for the at least one adjustment parameter; When the confidence information is greater than or equal to a preset threshold, the first feature information is adjusted based on an adjustment parameter corresponding to the confidence information to obtain second feature information of the target to be detected; or, When the confidence information is less than a preset threshold, the adjustment parameter corresponding to the confidence information is ignored.

16. The method according to claim 15, characterized in that The method further comprises: adjusting the first feature information based on an adjustment parameter corresponding to the confidence information to obtain second feature information of the target to be detected, comprising: Adjust the first feature information based on an adjustment parameter corresponding to the confidence information to obtain at least one adjusted relative feature information between the target to be detected and other targets; Determining an information offset between the at least one adjusted relative characteristic information and the corresponding relative characteristic information; When the information offset is less than or equal to a preset offset threshold, the second feature information of the target to be detected is obtained.

17. The method according to any one of claims 1 to 10, characterized in that: After obtaining the second feature information of the target to be detected, the method further includes: Generating a labeling box corresponding to the target to be detected based on the second feature information; The target to be detected is marked using the marking frame.

18. A vehicle positioning method, characterized in that: include: Get a monocular image; Identifying a to-be-identified vehicle and a reference vehicle included in the monocular image; Determine first positioning information of the vehicle to be identified in a preset coordinate system, and at least one relative positioning information between the vehicle to be identified and the reference vehicle in the preset coordinate system, wherein the relative positioning information is spatial constraint information of the vehicle to be identified relative to the reference vehicle in the preset coordinate system, and the preset coordinate system corresponds to the monocular image; The first positioning information is adjusted according to the at least one relative positioning information so that the relative positioning information between the adjusted vehicle to be identified and the reference vehicle is satisfied, thereby obtaining the second positioning information of the vehicle to be identified.

19. A device for detecting a target in a monocular image, characterized in that: include: A first acquisition module, used for acquiring a monocular image; A first detection module, used for identifying a target to be detected included in a monocular image, and determining first feature information of the target to be detected in a preset coordinate system, wherein the preset coordinate system corresponds to the monocular image; The first detection module is used to identify other targets included in the monocular image and determine at least one relative feature information between the target to be detected and the other targets in the preset coordinate system, wherein the relative feature information is the spatial constraint information of the target to be detected relative to the other targets in the preset coordinate system; The first processing module is used to adjust the first characteristic information based on the at least one relative characteristic information so that the relative characteristic information is satisfied between the adjusted target to be detected and other targets, and obtain second characteristic information of the target to be detected.

20. An electronic device, characterized in that: include: A memory, a processor; wherein the memory is used to store one or more computer instructions, wherein when the one or more computer instructions are executed by the processor, the method for detecting a target in a monocular image as described in any one of claims 1 to 17 is implemented.

21. A vehicle positioning device, characterized in that: include: A second acquisition module is used to acquire a monocular image; A second recognition module, used to recognize the vehicle to be recognized and the reference vehicle included in the monocular image; a second determination module, configured to determine first positioning information of the vehicle to be identified in a preset coordinate system, and at least one relative positioning information between the vehicle to be identified and the reference vehicle in the preset coordinate system, wherein the relative positioning information is spatial constraint information of the vehicle to be identified relative to the reference vehicle in the preset coordinate system, and the preset coordinate system corresponds to the monocular image; The second positioning module is used to adjust the first positioning information according to the at least one relative positioning information so that the relative positioning information between the adjusted vehicle to be identified and the reference vehicle is satisfied, thereby obtaining the second positioning information of the vehicle to be identified.

22. An electronic device, characterized in that: include: A memory, a processor; wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions, when executed by the processor, implement the vehicle positioning method as described in claim 18.