WO2022056924A1 - 一种目标定位方法、装置和计算机可读介质 - Google Patents

一种目标定位方法、装置和计算机可读介质 Download PDF

Info

Publication number
WO2022056924A1
WO2022056924A1 PCT/CN2020/116575 CN2020116575W WO2022056924A1 WO 2022056924 A1 WO2022056924 A1 WO 2022056924A1 CN 2020116575 W CN2020116575 W CN 2020116575W WO 2022056924 A1 WO2022056924 A1 WO 2022056924A1
Authority
WO
WIPO (PCT)
Prior art keywords
target object
physical environment
area
frame
picture
Prior art date
Application number
PCT/CN2020/116575
Other languages
English (en)
French (fr)
Inventor
李远哲
闵捷
Original Assignee
西门子（中国）有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 西门子（中国）有限公司 filed Critical 西门子（中国）有限公司
Priority to EP20953782.8A priority Critical patent/EP4195150A4/en
Priority to US18/044,475 priority patent/US20230316566A1/en
Priority to PCT/CN2020/116575 priority patent/WO2022056924A1/zh
Priority to CN202080103037.0A priority patent/CN115867937A/zh
Publication of WO2022056924A1 publication Critical patent/WO2022056924A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
    • G06T7/74—Determining position or orientation of objects or cameras using feature-based methods involving reference images or patches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G06V20/54—Surveillance or monitoring of activities, e.g. for recognising suspicious objects of traffic, e.g. cars on the road, trains or boats
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
    • G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
    • G06V20/586—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of parking space
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
    • G06V20/588—Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30204—Marker
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30232—Surveillance

Definitions

  • Embodiments of the present invention relate to the technical field of computer vision, and in particular, to a target positioning method, apparatus, and computer-readable medium.
  • GPS positioning In applications such as parking monitoring, vehicle tracking, personnel positioning, etc., it is necessary to locate the tracked target.
  • target positioning There are many methods of target positioning, such as: Global Positioning System GPS positioning.
  • GPS positioning requires the tracked target to upload its GPS location information, which often involves personal privacy, so it cannot be promoted in applications such as parking monitoring.
  • a localization method is to use a camera to take an image of the target object, and to locate the target object through image processing and target recognition. However, through target recognition, only the position of the target object in the picture can be determined, and then the position of the target object in the picture can be mapped to the position in the physical environment to realize the positioning of the target object.
  • the coordinates of each marker in the physical environment and the pixel coordinates of these markers in the picture of the physical environment captured by the camera can be obtained, and curve fitting can be performed to obtain the position between the position in the picture captured by the camera and the position in the physical environment. corresponding function.
  • curve fitting has limitations, and the obtained function may be inaccurate. In this case, the position in the physical environment determined from the fitted curve may be inaccurate.
  • Embodiments of the present invention provide a target positioning method, apparatus, and computer-readable medium, so as to determine the position of a target object in a real physical environment.
  • a target positioning method is provided.
  • at least one marker in the physical environment is determined, the physical environment is divided into at least two first regions according to the at least one marker, and the at least one A marker, which divides the physical environment in the picture captured by the first camera into at least two second areas according to the same division method of the at least two first areas, and then determines the at least two first areas There is a one-to-one correspondence between the at least two second regions.
  • the first frame of picture is obtained from the first camera, the target object is identified from the first frame of picture, and the second area of the target object in the first frame of picture is determined, and determining the first area corresponding to the second area where the target object is located according to the corresponding relationship.
  • a target positioning device may include:
  • a location mapping module configured to: determine at least one marker in the physical environment, divide the physical environment into at least two first areas according to the at least one marker, and take pictures of the physical environment from the first camera Identify the at least one mark in the at least two first areas, divide the physical environment in the picture captured by the first camera into at least two second areas according to the same division method of the at least two first areas, and determine the at least two second areas.
  • the picture processing module is configured to obtain the first frame of pictures from the first camera;
  • a target recognition module configured to recognize a target object from the first frame of pictures
  • the position mapping module is further configured to determine a second area of the target object in the first frame of pictures, and determine a first area corresponding to the second area where the target object is located according to the corresponding relationship.
  • a target positioning device comprising: at least one memory configured to store computer-readable codes; at least one processor configured to invoke the computer-readable codes to perform the steps provided in the first aspect .
  • a computer-readable medium storing computer-readable instructions on the computer-readable medium, the computer-readable instructions, when executed by a processor, cause the processor to execute the method provided in the first aspect. step.
  • the corresponding relationship between a limited number of areas is used as the position relationship, which is simple to implement and avoids complex curve fitting.
  • the at least two first areas may be different parking spaces.
  • coarse-grained features of the target object may be extracted, and the coarse-grained features include color, shape, outline, and sign. At least one: determine the information of the movement vector of the target object; determine whether the target object appears in the pictures taken by other cameras other than the first camera according to the coarse-grained features of the target object and the information of the movement vector .
  • the simulated physical environment may also be displayed; the target object is displayed in the simulated physical environment at the first region where the determined target object is located. It has the advantage of being intuitive and clear.
  • the at least two first regions may also be displayed in the simulated physical environment. By displaying each first area, the observer can observe the position of the target object more conveniently and clearly.
  • information about a group of positions of the target object in the physical environment may also be received, wherein the group of positions is determined by a group of temporally continuous pictures of the identified target object, respectively.
  • information about the first area of the target object in the physical environment identified in the second frame of picture may also be obtained, where the second frame of picture and the first The shooting time of a frame of pictures is the same; and according to the respective first regions of the target object in the physical environment identified in the first frame of pictures and the second frame of pictures, update the target object in the physical environment. a first area in the physical environment.
  • the second area where the target object is located in the first frame picture may be compared with the second area where the target object is located in the second frame picture.
  • the size of the second area; and the first area corresponding to the larger second area is used as the updated first area. Since a larger second area represents a closer target object in the camera, the accuracy of target recognition is usually higher.
  • the target object is a vehicle
  • the at least two first areas are different parking spaces.
  • FIG. 1 is a schematic structural diagram of a target positioning system provided by an embodiment of the present invention.
  • FIG. 2 is a schematic structural diagram of a target positioning apparatus provided by an embodiment of the present invention.
  • FIG. 3 is a schematic diagram of a relationship between a main node and a sub-node of a target positioning apparatus in a target positioning system according to an embodiment of the present invention.
  • FIG. 4 is a flowchart of a target positioning method provided by an embodiment of the present invention.
  • 5A to 5E illustrate a process of tracking and positioning a target object according to an embodiment of the present invention.
  • Target Positioning System 10 Camera 11: Target positioning device 11a: Main target positioning device 11b: Sub-target positioning device 20: Target Locator 111: At least one memory 112: at least one processor 113: Communication module 201: Location Mapping Module 202: Image processing module 203: target recognition module 204: Tracking Module 205: Motion detection module 206: Location update module 400: Targeting method S401 ⁇ S416: method steps
  • the term “including” and variations thereof represent open-ended terms meaning “including but not limited to”.
  • the term “based on” means “based at least in part on”.
  • the terms “one embodiment” and “an embodiment” mean “at least one embodiment.”
  • the term “another embodiment” means “at least one other embodiment.”
  • the terms “first”, “second”, etc. may refer to different or the same objects. Other definitions, whether explicit or implicit, may be included below. The definition of a term is consistent throughout the specification unless the context clearly dictates otherwise.
  • FIG. 1 shows a target positioning system 100 provided by an embodiment of the present invention, which includes at least one camera 10 for taking pictures of at least one target object (such as the vehicle shown in FIG. 1, but also pedestrians, objects, etc.), The obtained picture is sent to the target positioning device 11, and the target positioning device 11 performs target identification and positioning on the received picture.
  • at least one camera 10 for taking pictures of at least one target object (such as the vehicle shown in FIG. 1, but also pedestrians, objects, etc.)
  • the obtained picture is sent to the target positioning device 11, and the target positioning device 11 performs target identification and positioning on the received picture.
  • the target positioning device 11 can be deployed on the edge side, such as on the side of the road, parking lot, school gate, etc., and the pictures collected by the camera 10 can be processed in real time on the edge side, avoiding the transmission of a large amount of data.
  • one target positioning device 11 may be connected to one or more cameras 10 to process the pictures collected by the connected cameras.
  • the target positioning device 11 can also be integrated with one or more cameras 10 in the same physical device and deployed on the edge side.
  • the target positioning device 11 can also be deployed in the cloud, and the pictures collected by the camera 10 on the edge side are transmitted to the target positioning device 11 in the cloud for further target identification and positioning.
  • Time synchronization can be performed between the cameras 10. Taking a vehicle as an example, the same vehicle may appear in the pictures captured by the two cameras 10 at the same time. The location in the physical environment of the same vehicle in each image is the same.
  • What each camera 10 collects is a frame of pictures arranged in time sequence.
  • the target object is positioned in each frame of pictures that appear, and the moving trajectory of the target object can be obtained according to the time sequence of each frame of pictures, that is, the target object is tracked.
  • the target positioning system 100 includes multiple cameras 10, and different cameras 10 monitor different areas, target tracking in a large scene across cameras can be implemented.
  • the target positioning apparatus 11 provided in the embodiment of the present invention may be implemented as a network of computer processors, so as to execute the target positioning method 400 in the embodiment of the present invention.
  • the target locating device 11 may also be a single computer, single board or chip as shown in FIG. 2, including at least one memory 111 including a computer readable medium such as random access memory (RAM).
  • Apparatus 11 also includes at least one processor 112 coupled with at least one memory 111 .
  • Computer-executable instructions are stored in at least one memory 111 and, when executed by at least one processor 112, can cause at least one processor 112 to perform the steps described herein.
  • the at least one processor 112 may include a microprocessor, an application specific integrated circuit (ASIC), a digital signal processor (DSP), a central processing unit (CPU), a graphics processing unit (GPU), a state machine, and the like.
  • ASIC application specific integrated circuit
  • DSP digital signal processor
  • CPU central processing unit
  • GPU graphics processing unit
  • Examples of computer readable media include, but are not limited to, floppy disks, CD-ROMs, magnetic disks, memory chips, ROM, RAM, ASICs, configured processors, all-optical media, all magnetic tapes or other magnetic media, or from which a computer processor can Any other medium from which to read instructions.
  • various other forms of computer-readable media can transmit or carry instructions to a computer, including routers, private or public networks, or other wired and wireless transmission devices or channels.
  • the target positioning apparatus 11 may further include a communication module 113 respectively coupled with at least one memory 111 and at least one processor 112 , for implementing communication between the target positioning apparatus 11 and external devices, such as receiving pictures from the camera 10 .
  • the at least one memory 111 shown in FIG. 1 may contain the target positioning program 20, so that the at least one processor 112 executes the method for target positioning 400 described in the embodiment of the present invention.
  • the target location program 20 may include:
  • the position mapping module 201 is configured to determine the relationship between the position in the physical environment and the position in the picture taken by the camera 10 of the physical environment. As mentioned earlier, the functional relationship obtained by curve fitting may be inaccurate. In this embodiment of the present invention, the location mapping module 201 first determines at least one marker in the physical environment (for example, in a traffic management scenario, the centerline of the lane may be used as a marker or some markers are manually set), and according to the at least one marker, the physical The environment is divided into at least two first regions.
  • the location mapping module 201 identifies the above at least one mark from a picture taken by a camera (to distinguish it from other cameras, referred to as a "first camera” here) on the physical environment, and divides the at least two first areas according to the same division.
  • the physical environment in the picture captured by the first camera is divided into at least two second areas, and then a one-to-one correspondence between the at least two first areas and the at least two second areas is determined.
  • the physical environment is the environment where the target object is located, such as the road where the vehicle is located (a two-dimensional plane) or the space where the vehicle is located (a three-dimensional space).
  • the physical environment is divided into at least two A region, and for a three-dimensional space, the physical environment is divided into at least two spaces, where the first region can be broadly understood as a plane or space.
  • the implementation is simple, and complex curve fitting is avoided.
  • it is simple to determine whether a vehicle is parked in a certain parking space.
  • Accurate advantages for a parking lot, the at least two first areas may be different parking spaces.
  • the target positioning program 20 may also include:
  • the picture processing module 202 is configured to obtain a first frame of picture from the first camera 10 (different from pictures obtained by other cameras 10 described later), and perform encoding and decoding processing on the picture. It can use graphics processing unit (GPU) resources to accelerate image processing to meet real-time requirements.
  • GPU graphics processing unit
  • the target recognition module 203 configured to recognize the target object from the first frame of pictures; to recognize the target object in real time, and optionally, to extract the features of the target object.
  • the position mapping module 201 is further configured to determine the second area of the target object in the first frame of pictures, and to determine the first area corresponding to the second area where the target object is located according to the correspondence obtained above.
  • the object location program 20 may further include a tracking module 204 configured to track the object under the connected camera.
  • a tracking module 204 configured to track the object under the connected camera.
  • the tracking module 204 may also display the simulated physical environment, and display the target object at the first area determined by the location mapping module 203 in the simulated physical environment.
  • the method of linear fitting can be used to generate a three-dimensional physical environment and display the target object.
  • the target positioning program 20 may further include an action calculation module 205, which is configured to determine the action information of the target object, such as: moving, stopping and The direction of movement of the target object.
  • an action calculation module 205 which is configured to determine the action information of the target object, such as: moving, stopping and The direction of movement of the target object.
  • the target positioning apparatus 11 can be divided into a main node 11a and each sub-node 11b, and one main node 11a is connected to each sub-node 11b.
  • One object positioning device 11 can be connected with one or more cameras. Consider that the same target object may appear in the pictures taken by the cameras connected to two or more target positioning devices 11 at the same time.
  • each child node 11b sends the result of its positioning to the master node 11a.
  • the target positioning program 20 of the master node 11a further includes a location update module 206, which is configured to slave
  • Each sub-node 11b receives the information, and determines the position of the target object in the physical environment according to the positions of different sub-nodes 11b of the same target object at the same time.
  • wireless data transmission may be used between the sub-node 11b and the master node 11a, for example, using a 4G network for transmission.
  • embodiments of the present invention may include apparatuses having architectures different from those shown in FIG. 2 .
  • the above architecture is only exemplary, and is used to explain the method 400 provided by the embodiment of the present invention.
  • the above modules can also be regarded as various functional modules implemented by hardware, which are used to realize various functions involved in the target positioning device 11 when executing the target positioning method, such as the control logic of each process involved in the method in advance. Burned into a chip such as a Field-Programmable Gate Array (FPGA) or a Complex Programmable Logic Device (CPLD), and these chips or devices perform the functions of the above modules.
  • FPGA Field-Programmable Gate Array
  • CPLD Complex Programmable Logic Device
  • an exemplary method 400 includes the following steps:
  • -S403 Identify at least one marker from a picture taken by the first camera of the physical environment
  • - S405 Determine a one-to-one correspondence between the at least two first regions and the at least two second regions.
  • steps S401 to S405 the physical environment is divided into several first areas according to the marks in the physical environment.
  • the marks in the physical environment are identified from the photographed pictures, and the same method is used as follows:
  • the marker divides the physical environment in the picture into a number of second regions.
  • several second areas and several first areas are corresponding to form the corresponding relationship between the first areas and the second areas.
  • the first area where the target object is located in the physical environment is determined according to the second area where the target object is located in the picture, so as to achieve the purpose of locating the target object.
  • the identification and positioning of the target object are realized through the following steps S406-S409:
  • an implementation method is to identify each target object from the pictures taken by different cameras, and perform feature matching on all the identified target objects, in which a complex neural network and other complex methods are used.
  • the deep learning algorithm has high requirements on the computing power of the device, so it can only be implemented by the server in the cloud.
  • surveillance cameras are intensively deployed. For example, in the application scenario of on-street parking management, a camera is deployed next to a parking space and is dedicated to monitoring the parking status of the parking space. These densely deployed cameras continuously collect passing vehicles in chronological order, generating a large amount of redundant information, wasting the computing power of the equipment, and increasing the cost of the entire system.
  • the tracking of the target object across the cameras can be performed.
  • target recognition and tracking across cameras can be performed according to the coarse-grained features of the target object, such as color, shape, outline, marker light information, and information about the movement vector of the target object. Therefore, the above step S407 may specifically include the following sub-steps:
  • the coarse-grained features including at least one of color, shape, outline, and logo;
  • a complex deep learning algorithm is used to perform feature matching across cameras. It is also possible to use the same camera to cover multiple monitored areas, avoiding the waste of edge devices.
  • the localization of the target object can also be implemented in a simulated physical environment.
  • the method 400 may further include the following steps:
  • the method 400 may further include:
  • -S412 Display at least two first regions in the simulated physical environment.
  • the observer can observe the position of the target object more conveniently and clearly.
  • the target object can also be tracked in a simulated physical environment, and the movement trajectory of the target object can be displayed.
  • the method 400 may further include the following steps:
  • -S413 Receive information about a group of positions of the target object in the physical environment, wherein a group of positions is each first position of the target object in the physical environment determined by a group of temporally consecutive pictures of the identified target object respectively area;
  • Steps S410 to S414 can also be performed by a server in the cloud, the server displays the physical environment, and the target positioning device 11 on the edge side sends the positioning information of the target object to the server, and the server displays the target object in the environment according to the received positioning information.
  • the real physical environment is shown on the left side of each figure, and the physical environment is shown on the right side.
  • the road centerline is used as a marker to divide the physical environment into each first area.
  • the road centerline is also displayed as a mark, and the second regions corresponding to the respective first regions are also displayed.
  • 5A to 5E are arranged in chronological order.
  • the monitored target vehicle has not yet appeared, and other vehicles are located in different second areas of the left lane respectively. Centerline”), the position of each vehicle can be clearly seen from the physical environment on the right, and the position of each vehicle in the physical environment can be determined intuitively.
  • the monitored target vehicle is approaching from the right lane from far to near; in Fig. 5C, the monitored vehicle is about to reverse into the parking space.
  • the target vehicle In the simulated physical environment, the target vehicle is located at the front of the vehicle. Right; in Figure 5D, most of the monitored target vehicle body (the part contained in the bounding box) has entered the parking space behind the frontmost vehicle, so in the simulated physical environment, the monitored target vehicle The parking space is judged to be located behind the frontmost vehicle; in Figure 5E, since the target vehicle is adjusted in the parking space, the distance between the target vehicle and the frontmost vehicle increases compared to Figure 5D, which is shown in Figure 5E The physical environment on the right can be clearly shown.
  • the same target object may appear in the pictures captured by the cameras connected to two or more target positioning devices 11 at the same time.
  • the final position of the target object in the physical environment can be determined according to the multiple positioned positions. Therefore, the method 400 may further include the following steps:
  • -S415 Obtain the information of the first area of the target object identified in the second frame of picture in the physical environment, wherein the second frame of picture and the first frame of picture are taken at the same time;
  • step S416 may further include the following sub-steps:
  • the embodiments of the present invention further provide a computer-readable medium, where computer-readable instructions are stored on the computer-readable medium, and when the computer-readable instructions are executed by the processor, the processor executes the foregoing target positioning method .
  • Examples of computer-readable media include floppy disks, hard disks, magneto-optical disks, optical disks (eg, CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), magnetic tape, non- Volatile memory cards and ROMs.
  • the computer readable instructions may be downloaded from a server computer or cloud over a communications network.
  • the embodiments of the present invention provide a target positioning method, apparatus, system, and computer-readable medium.
  • the markers in the physical environment are used to divide the area, and the corresponding relationship between the area in the physical environment and the area in the picture is used to locate the target object, which is different from the previous method of using curve fitting to determine the physical environment and camera shooting.
  • the error of function fitting can be effectively avoided.
  • the recognition and tracking of the same target object between different cameras is realized according to the coarse-grained features of the target object and the information of the target object's movement vector.
  • it has the advantages of simple implementation and lower requirements on equipment computing power, so it can realize real-time target tracking across cameras on the edge side. It not only makes full use of the resources of multiple cameras, but also avoids complex algorithms. It can be applied to static and dynamic traffic management in large scenes.
  • the same target object may appear in pictures taken by different cameras at the same time.
  • the position of the target object in the physical environment is finally determined according to the positioning results of different cameras, and the continuous cross-camera detection of the target object is realized. tracking and accurate positioning.
  • the physical environment is simulated and the target object is displayed according to the localized area in the simulated physical environment, and further, the target object can be tracked, so that the observer can monitor the target object intuitively and clearly.
  • the main node in the target positioning device collects the positioning information of the sub-nodes, realizes the cross-camera tracking of the target object, and updates the position of the target object.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Analysis (AREA)

Abstract

一种目标定位方法、装置和计算机可读介质，涉及计算机视觉技术，用以确定目标物体在真实的物理环境中的位置。该方法包括：从第一摄像头处获取(S406)第一帧图片；从所述第一帧图片中识别(S407)目标物体；确定(S408)所述目标物体在所述第一帧图片中的第二区域；确定(S409)所述目标物体所在第二区域对应的第一区域。避免了曲线拟合得到摄像头拍摄的图片中的位置与物理环境中的位置之间函数不准确的问题。将有限个区域之间的对应关系作为位置关系，相比于复杂的曲线拟合，实现简单，对于车辆定位精细度要求不高的场景，比如停车场管理，判断车辆是否停入某一车位具有简单、准确的优点。

Description

一种目标定位方法、装置和计算机可读介质 技术领域
本发明实施例涉及计算机视觉技术领域，尤其涉及一种目标定位方法、装置和计算机可读介质。
背景技术
在诸如停车监视、车辆跟踪、人员定位等应用中，需要对所跟踪的目标进行定位。目标定位的方法有很多，比如：全球定位***GPS定位。但GPS定位需要被跟踪目标上传其GPS位置信息，这往往涉及个人隐私，因而在诸如停车监视等应用中无法得以推广。
一种定位的方法是使用摄像头拍摄目标物体的图像，通过图像处理、目标识别来定位目标物体。但通过目标识别仅能确定目标物体在图片中的位置，然后将目标物体在图片中的位置映射到在物理环境中的位置，以实现对目标物体的定位。
其中，可获取物理环境中各个标记的坐标，以及摄像头所拍摄的物理环境的图片中这些标记的像素坐标，进行曲线拟合，得到摄像头拍摄得到的图片中的位置与物理环境中的位置之间对应关系的的函数。但曲线拟合具有局限性，得到的函数可能不准确，此时，根据拟合的曲线确定的物理环境中的位置可能不准确。
发明内容
本发明实施例实施例提供了一种目标定位方法、装置和计算机可读介质，用以确定目标物体在真实的物理环境中的位置。
第一方面，提供一种目标定位方法。该方法中，确定物理环境中的至少一个标记，根据所述至少一个标记将所述物理环境划分为至少两个第一区域，从第一摄像头对所述物理环境拍摄的图片中识别所述至少一个标记，按照所述至少两个第一区域同样的划分方式，将所述第一摄像头拍摄图片中的所述物理环境划分为至少两个第二区域，进而确定所述至少两个第一区域与所述至少两个第二区域之间的一一对应关系。在对目标物体定位时，从所述第一摄像头处获取第一帧图片，从所述第一帧图片中识别目标物体，确定所述目标物体在所述第一帧图片中的第二区域，并根据所述对应关系确定所述目标物体所在第二区域对应的第一区域。
第二方面，提供一种目标定位装置，该装置可包括：
-位置映射模块，被配置为：确定物理环境中的至少一个标记，根据所述至少一个标记 将所述物理环境划分为至少两个第一区域，从第一摄像头对所述物理环境拍摄的图片中识别所述至少一个标记，按照所述至少两个第一区域同样的划分方式，将所述第一摄像头拍摄图片中的所述物理环境划分为至少两个第二区域，确定所述至少两个第一区域与所述至少两个第二区域之间的一一对应关系，图片处理模块，被配置为从所述第一摄像头处获取第一帧图片；
-目标识别模块，被配置为从所述第一帧图片中识别目标物体；
-所述位置映射模块，还被配置为确定所述目标物体在所述第一帧图片中的第二区域，以及根据所述对应关系确定所述目标物体所在第二区域对应的第一区域。
第三方面，提供一种目标定位装置，包括：至少一个存储器，被配置为存储计算机可读代码；至少一个处理器，被配置为调用所述计算机可读代码，执行第一方面所提供的步骤。
第四方面，一种计算机可读介质，所述计算机可读介质上存储有计算机可读指令，所述计算机可读指令在被处理器执行时，使所述处理器执行第一方面所提供的步骤。
其中，将有限个区域之间的对应关系作为位置关系，实现简单，避免了复杂的曲线拟合，对于车辆定位精细度要求不高的场景，比如停车场管理，判断车辆是否停入某一车位具有简单、准确的优点。可选地，对于停车场，上述至少两个第一区域可以是不同的停车位。
对于上述任一方面，可选地，在从所述第一帧图片中识别目标物体时，可提取所述目标物体的粗粒度特征，所述粗粒度特征包括颜色、形状、轮廓、标志中的至少一种；确定所述目标物体的移动向量的信息；根据所述目标物体的粗粒度特征和移动向量的信息确定所述第一摄像头之外的其他摄像头拍摄的图片中是否出现所述目标物体。无需采用复杂的深度学习的算法进行跨摄像头的特征匹配，也能够使用同一个摄像头覆盖多个监控的区域，避免边缘设备的浪费。
对于上述任一方面，可选地，还可以显示模拟的所述物理环境；在模拟的所述物理环境中的、确定的所述目标物体所在的第一区域处显示所述目标物体。具有直观、清晰的优点。
进一步，可选地，还可在模拟的所述物理环境中显示所述至少两个第一区域。通过显示各个第一区域，观察者可更方便、清楚地观察目标物体的位置。
并且，还可以接收所述目标物体在所述物理环境中的一组位置的信息，其中，所述一组位置是由识别出的所述目标物体的一组时间上连续的图片分别确定的所述目标物体在所述物理环境中的各个第一区域；在模拟的所述物理环境中按照时间顺序在所述一组位置所对应的每一个第一区域中显示所述目标物体。实现在模拟的物理环境中对目标物体的跟踪。
对于上述任一方面，可选地，还可获取在第二帧图片识别出的所述目标物体在所述物理环境中的第一区域的信息，其中，所述第二帧图片与所述第一帧图片的拍摄时间相同；并根 据在所述第一帧图片和所述第二帧图片分别识别出的所述目标物体在所述物理环境中各自的第一区域，更新所述目标物体在所述物理环境中的第一区域。当同一个目标物体同时出现在不同摄像头拍摄的图片中时，可通过该可选实现方式实现目标物体的准确定位。
进一步地，在更新所述目标物体在所述物理环境中的第一区域时，可比较所述目标物体在所述第一帧图片中所在的第二区域与所述第二帧图片中所在的第二区域的大小；并将较大的第二区域所对应的第一区域作为更新后的第一区域。由于较大的第二区域代表摄像头里目标物体更近，通常目标识别的准确率会更高。
对于上述任一方面，可选地，所述目标物体为车辆，所述至少两个第一区域是不同的停车位。
附图说明
图1为本发明实施例提供的目标定位***的结构示意图。
图2为本发明实施例提供的目标定位装置的结构示意图。
图3为本发明实施例提供的目标定位***中目标定位装置的主节点和子节点之间的关系示意图。
图4为本发明实施例提供的目标定位方法的流程图。
图5A～图5E示出了本发明实施例对目标物体进行跟踪定位的过程。
附图标记列表：
100：目标定位*** 10：摄像头 11：目标定位装置
11a：主目标定位装置 11b：子目标定位装置  
20：目标定位程序 111：至少一个存储器 112：至少一个处理器
113：通信模块 201：位置映射模块 202：图片处理模块
203：目标识别模块 204：跟踪模块 205：动作检测模块
206：位置更新模块 400：目标定位方法 S401～S416：方法步骤
具体实施方式
现在将参考示例实施方式讨论本文描述的主题。应该理解，讨论这些实施方式只是为了使得本领域技术人员能够更好地理解从而实现本文描述的主题，并非是对权利要求书中所阐述的保护范围、适用性或者示例的限制。可以在不脱离本发明实施例内容的保护范围的情况下，对所讨论的元素的功能和排列进行改变。各个示例可以根据需要，省略、替代或者添加 各种过程或组件。例如，所描述的方法可以按照与所描述的顺序不同的顺序来执行，以及各个步骤可以被添加、省略或者组合。另外，相对一些示例所描述的特征在其它例子中也可以进行组合。
如本文中使用的，术语“包括”及其变型表示开放的术语，含义是“包括但不限于”。术语“基于”表示“至少部分地基于”。术语“一个实施例”和“一实施例”表示“至少一个实施例”。术语“另一个实施例”表示“至少一个其他实施例”。术语“第一”、“第二”等可以指代不同的或相同的对象。下面可以包括其他的定义，无论是明确的还是隐含的。除非上下文中明确地指明，否则一个术语的定义在整个说明书中是一致的。
下面结合附图对本发明实施例进行详细说明。
图1示出了本发明实施例提供的目标定位***100，其中包括至少一个摄像头10，用于对至少一个目标物体进行拍照(比如图1所示的车辆，也可以是行人、物品等)，获得的图片发送至目标定位装置11，目标定位装置11对收到的图片进行目标识别和定位。
这里，目标定位装置11可部署在边缘侧，比如在道路边、停车场、学校门口等，摄像头10采集的图片可在边缘侧进行实时处理，避免了大量数据的传输。其中，一个目标定位装置11可连接一个或多个摄像头10，对所连接的摄像头采集的图片进行处理。目标定位装置11也可与一个或多个摄像头10集成在同一个物理设备中，部署在边缘侧。此外，目标定位装置11也可部署在云端，位于边缘侧的摄像头10采集的图片传输至云端的目标定位装置11进行进一步的目标识别与定位。
各个摄像头10之间可进行时间同步，以车辆为例，同一个车辆可能在同一时刻出现在两个摄像头10所拍摄的图片中，由于摄像头10之间是时间同步的，因此可认为出现在两个图片中的同一个车辆所在物理环境中的位置是同一个。
各摄像头10采集的是按照时间顺序排列的一帧帧图片。将目标物体在出现的每一帧图片中定位，按照各帧图片的时间顺序，可得到目标物体移动的轨迹，即对目标物体进行跟踪。当目标定位***100中包括多个摄像头10，不同摄像头10监视不同的区域，可实现跨摄像头的大场景的目标跟踪。
本发明实施例提供的目标定位装置11可以实现为计算机处理器的网络，以执行本发明实施例中的目标定位方法400。目标定位装置11也可以是如图2所示的单台计算机、单板或芯片，包括至少一个存储器111，其包括计算机可读介质，例如随机存取存储器(RAM)。装置11还包括与至少一个存储器111耦合的至少一个处理器112。计算机可执行指令存储在至少 一个存储器111中，并且当由至少一个处理器112执行时，可以使至少一个处理器112执行本文所述的步骤。至少一个处理器112可以包括微处理器、专用集成电路(ASIC)、数字信号处理器(DSP)、中央处理单元(CPU)、图形处理单元(GPU)、状态机等。计算机可读介质的实施例包括但不限于软盘、CD-ROM、磁盘，存储器芯片、ROM、RAM、ASIC、配置的处理器、全光介质、所有磁带或其他磁性介质，或计算机处理器可以从中读取指令的任何其他介质。此外，各种其它形式的计算机可读介质可以向计算机发送或携带指令，包括路由器、专用或公用网络、或其它有线和无线传输设备或信道。指令可以包括任何计算机编程语言的代码，包括C、C++、C语言、Visual Basic、java和JavaScript。此外，目标定位装置11还可包括与至少一个存储器111和至少一个处理器112分别耦合的通信模块113，用于实现目标定位装置11与外部设备的通信，比如从摄像头10处接收图片。
当由至少一个处理器112执行时，图1中所示的至少一个存储器111可以包含目标定位程序20，使得至少一个处理器112执行本发明实施例中所述的用于目标定位方法400。目标定位程序20可以包括：
-位置映射模块201，被配置为确定物理环境中的位置和摄像头10对物理环境所拍摄图片中的位置的关系。如前所述，曲线拟合得到的函数关系可能不准确。本发明实施例中，位置映射模块201首先确定物理环境中的至少一个标记(比如在交通管理场景中，可利用车道中心线作为标记或者人为设置一些标记等)，并根据该至少一个标记将物理环境划分为至少两个第一区域。此外，位置映射模块201从一个摄像头(为区别于其他摄像头，这里称为“第一摄像头”)对物理环境拍摄的图片中识别上述至少一个标记，并按照上述至少两个第一区域同样的划分方式，将第一摄像头拍摄图片中的物理环境划分为至少两个第二区域，然后确定至少两个第一区域与至少两个第二区域之间的一一对应关系。其中，物理环境是目标物体所在的环境，比如车辆所在的道路(为二维的平面)或者车辆所在的空间(为三维的空间)，对于二维的平面，物理环境被划分为至少两个第一区域，而对于三维的空间，物理环境则被划分为至少两个空间，这里第一区域可广义地理解为平面或空间。
将有限个区域之间的对应关系作为位置关系，实现简单，避免了复杂的曲线拟合，对于车辆定位精细度要求不高的场景，比如停车场管理，判断车辆是否停入某一车位具有简单、准确的优点。可选地，对于停车场，上述至少两个第一区域可以是不同的停车位。
此外，目标定位程序20还可包括：
-图片处理模块202，被配置为从第一摄像头10处获取第一帧图片(区别于后续描述的其他摄像头10获取的图片)，并对图片进行编解码处理。其可采用图形处理单元(GPU)的资源加速图片处理，以满足实时要求。
-目标识别模块203，被配置为从第一帧图片中识别出目标物体；对目标物体进行实时识别，可选地，还可提取目标物体的特征。
-位置映射模块201，还被配置为确定目标物体在第一帧图片中的第二区域，以及根据上述得到的对应关系确定目标物所在第二区域对应的第一区域。
此外，目标定位程序20还可包括跟踪模块204，被配置为跟踪所连接的摄像头下的目标物体。对于主节点11a，还需要对整个目标定位***100中的目标物体进行跨摄像头跟踪。可选地，跟踪模块204还可显示模拟的物理环境，以及在模拟的物理环境中按照位置映射模块203确定的第一区域处显示目标物体。其中，可采用线性拟合的方法生成三维的物理环境并显示目标物体。
可选地，目标定位程序20还可包括动作计算模块205，被配置为根据目标物体的定位的位置随时间的变化，即目标物体的轨迹，确定目标物体的动作信息，比如：移动、停止和目标物体的运动方向。
如图3所示，目标定位装置11可分为主节点11a和各个子节点11b，一个主节点11a连接各个子节点11b。一个目标定位装置11可连接一个或多个摄像头。考虑在同一时刻同一目标物体可能出现在两个或更多个目标定位装置11连接的摄像头所拍摄的图片中。在对目标物体进行定位时，需要依据多个定位的位置确定最终的目标物体在物理环境中的位置。因此，本发明实施例中，各个子节点11b将自身定位的结果发送至主节点11a，主机点11a与各个子节点11b相比，目标定位程序20还包括一个位置更新模块206，被配置为从各个子节点11b接收信息，根据同一个目标物体同一时刻不同子节点11b所定位的位置确定目标物体在物理环境中的位置。可选地，子节点11b与主节点11a之间可采用无线数据传输，比如利用4G网络进行传输。
应当提及的是，本发明实施例可以包括具有不同于图2所示架构的装置。上述架构仅仅是示例性的，用于解释本发明实施例提供的方法400。
此外，上述各模块还也可视为由硬件实现的各个功能模块，用于实现目标定位装置11在执行目标定位方法时涉及的各种功能，比如预先将该方法中涉及的各流程的控制逻辑烧制到诸如现场可编程门阵列(Field-Programmable Gate Array，FPGA)芯片或复杂可编程逻辑器件(Complex Programmable Logic Device，CPLD)中，而由这些芯片或器件执行上述各模块的功能，具体实现方式可依工程实践而定。
上述各模块的其他可选实现方式可参考方法400中的描述。
如图4所示，根据本发明实施例的一个示例性方法400包括以下步骤：
-S401：确定物理环境中的至少一个标记；
-S402：根据至少一个标记将物理环境划分为至少两个第一区域；
-S403：从第一摄像头对物理环境拍摄的图片中识别至少一个标记；
-S404：按照至少两个第一区域同样的划分方式，将第一摄像头拍摄图片中的物理环境划分为至少两个第二区域；
-S405：确定至少两个第一区域与至少两个第二区域之间的一一对应关系。
步骤S401～S405中，根据物理环境中的标记将物理环境划分为若干个第一区域，第一摄像头对物理环境拍照后，从拍摄的图片中识别出物理环境中标记，并采用同样的方法按照标记对图片中的物理环境划分为若干个第二区域。然后按照这些标记，将若干个第二区域和若干个第一区域对应起来，形成第一区域与第二区域的对应关系。利用该对应关系，当第一摄像头对物理环境拍照，并对拍摄的图片进行目标识别和定位时，由于摄像头的视野与物理环境之间的位置关系不变，则可利用之前确定的上述对应关系，在下述的步骤S406～S409中根据图片中目标物***于的第二区域确定目标物体在物理环境中所处的第一区域，从而达到对目标物体定位的目的。
方法400中，通过下述的步骤S406～S409实现对目标物体的识别和定位：
-S406：从第一摄像头处获取第一帧图片；
-S407：从第一帧图片中识别目标物体；
-S408：确定目标物体在第一帧图片中的第二区域；
-S409：根据对应关系确定目标物体所在第二区域对应的第一区域。
在以往的目标识别和跟踪的方法中，一种实现方式是：从不同摄像头所拍摄的图片分别识别出各个目标物体，并对识别出的所有目标物体进行特征匹配，其中采用神经网络等复杂的深度学习算法，对设备的算力要求很高，因此仅能通过云端的服务器实现，一方面提高了***的成本，另一方面引入了数据传输、处理的延时，无法满足实时性要求。另一种实现方式中，密集部署监控摄像头，比如对于路边停车管理的应用场景，一个停车位旁边部署一个摄像头，专用于该停车位的停车状态监控。这些密集部署的摄像头按照时间顺序持续采集路过的车辆，产生大量的冗余信息，浪费设备的算力，提高了整个***的成本。
本发明实施例中，可进行跨摄像头的目标物体的跟踪。具体地，可根据目标物体的粗粒度特征，比如：颜色、形状、轮廓、标志灯信息，以及目标物体的移动向量的信息，来进行跨摄像头的目标识别和跟踪。因此，上述步骤S407具体可包括如下子步骤：
-提取目标物体的粗粒度特征，粗粒度特征包括颜色、形状、轮廓、标志中的至少一种；
-确定目标物体的移动向量的信息；
-根据目标物体的粗粒度特征和移动向量的信息确定第一摄像头之外的其他摄像头拍摄的图片中是否出现目标物体。
这样一方面避免了前述的第一种实现方式中，采用复杂的深度学习的算法进行跨摄像头的特征匹配。也能够使用同一个摄像头覆盖多个监控的区域，避免边缘设备的浪费。
可选地，还可在模拟的物理环境中实现对目标物体的定位。具体地，方法400还可包括如下步骤：
-S410：显示的物理环境；
-S411：在模拟的物理环境中的、确定的目标物体所在的第一区域处显示目标物体。
并且，可选地，方法400还可包括：
-S412：在模拟的物理环境中显示至少两个第一区域。
通过显示各个第一区域，观察者可更方便、清楚地观察目标物体的位置。
可选地，还可在模拟的物理环境跟踪目标物体，显示目标物体的移动轨迹，具体地，方法400还可包括如下步骤：
-S413：接收目标物体在物理环境中的一组位置的信息，其中，一组位置是由识别出的目标物体的一组时间上连续的图片分别确定的目标物体在物理环境中的各个第一区域；
-S414：在模拟的物理环境中按照时间顺序在一组位置所对应的每一个第一区域中显示目标物体。
步骤S410～S414也可由云端的服务器执行，服务器并显示物理环境，位于边缘侧的目标定位装置11将目标物体的定位信息发送至服务器，由服务器在环境中按照接收到的定位信息显示目标物体。
如图5A～图5E所示，每一个图中左侧示出了真实的物理环境，右侧示出了的物理环境。在真实的物理环境中，道路中心线作为标记，将物理环境换分为各个第一区域。在对应的的物理环境中，道路中心线作为标记也被显示出来，此外与各个第一区域分别对应的第二区域也显示出来。图5A～图5E是按照时间顺序排列的。图5A中，被监测的目标车辆还没有出现，其他车辆分别位于左侧车道不同的第二区域中，由第二区域与第一区域的对应关系，以及各个示出的各个标记(即“道路中心线”)，从右侧的物理环境中可清楚地看到各个车辆的位置，并可直观地确定各个车辆在物理环境中的位置。图5B中，被监测的目标车辆由远及近从右侧的车道驶来；图5C中，被监测的车辆即将倒车入停车位，在模拟的物理环境中，目标车辆位于最前面的车辆的右侧；图5D中，被监测的目标车辆车身的大部分(包围盒bounding box中包含的部分)已进入最前面的车辆后的停车位，因此在模拟的物理环境中，被监测的目标车辆被判断为位于最前面车辆后面的停车位；图5E中，由于目标车辆在车位中调整了位置， 与图5D相比，目标车辆与最前面的车辆之间的距离增大，这在图5E右侧的的物理环境中可以清楚地显示出来。
如前所述，在同一时刻同一目标物体可能出现在两个或更多个目标定位装置11连接的摄像头所拍摄的图片中。在对目标物体进行定位时，可依据多个定位的位置确定最终的目标物体在物理环境中的位置。因此，方法400还可进一步包括如下步骤：
-S415：获取在第二帧图片识别出的目标物体在物理环境中的第一区域的信息，其中，第二帧图片与第一帧图片的拍摄时间相同；
-S416：根据在第一帧图片和第二帧图片分别识别出的目标物体在物理环境中各自的第一区域，更新目标物体在物理环境中的第一区域。
其中，步骤S416可进一步包括如下子步骤：
-S416a：比较目标物体在第一帧图片中所在的第二区域与第二帧图片中所在的第二区域的大小；
-S416b：将较大的第二区域所对应的第一区域作为更新后的第一区域。这是由于较大的第二区域代表摄像头里目标物体更近，通常目标识别的准确率会更高。
此外，本发明实施例实施例还提供一种计算机可读介质，该计算机可读介质上存储有计算机可读指令，计算机可读指令在被处理器执行时，使处理器执行前述的目标定位方法。计算机可读介质的实施例包括软盘、硬盘、磁光盘、光盘(如CD-ROM、CD-R、CD-RW、DVD-ROM、DVD-RAM、DVD-RW、DVD+RW)、磁带、非易失性存储卡和ROM。可选地，可以由通信网络从服务器计算机上或云上下载计算机可读指令。
综上，本发明实施例提供一种目标定位方法、装置、***和计算机可读介质。其中，利用物理环境中的标记进行区域划分，通过物理环境中的区域和图片中区域之间的对应关系，实现对目标物体的定位，与以往的采用曲线拟合的方式确定物理环境和摄像头拍摄图片之间的位置对应关系的函数相比，可有效避免函数拟合的错误。
此外，通过多摄像头融合，根据目标物体的粗粒度特征和目标物体的移动向量的信息，实现不同摄像头之间对同一目标物体的识别和跟踪。与采用深度学习的算法在多个摄像头采集的各图片中进行特征匹配相比，具有实现简单，对设备算力要求较低，因此可实现边缘侧的跨摄像头的实时目标跟踪。既充分利用了多摄像头的资源，同时又避免了复杂的算法。可应用于大场景的静态和动态的交通管理。
跨摄像头时，同一目标物体可能同时出现在不同摄像头拍摄的图片中，本发明实施例中， 根据不同摄像头定位的结果最终确定目标物体在物理环境中的位置，实现了对目标物体跨摄像头的连续跟踪，且定位准确。
此外，模拟物理环境并在模拟的物理环境中按照定位的区域显示目标物体，进一步地，还可跟踪目标物体，使得观察着能够直观、清楚地监视目标物体。
在***结构方面，通过目标定位装置中的主节点收集子节点的定位信息，实现对目标物体的跨摄像头跟踪，以及目标物***置的更新，处理不依赖于云端的服务器，具有实时、节省算力的优点。
需要说明的是，上述各流程和各***结构图中不是所有的步骤和模块都是必须的，可以根据实际的需要忽略某些步骤或模块。各步骤的执行顺序不是固定的，可以根据需要进行调整。上述各实施例中描述的***结构可以是物理结构，也可以是逻辑结构，即，有些模块可能由同一物理实体实现，或者，有些模块可能分由多个物理实体实现，或者，可以由多个独立设备中的某些部件共同实现。

Claims (17)

  1. 一种目标定位方法(400)，其特征在于，包括：
    -确定(S401)物理环境中的至少一个标记；
    -根据所述至少一个标记将所述物理环境划分(S402)为至少两个第一区域；
    -从第一摄像头对所述物理环境拍摄的图片中识别(S403)所述至少一个标记；
    -按照所述至少两个第一区域同样的划分方式，将所述第一摄像头拍摄图片中的所述物理环境划分(S404)为至少两个第二区域；
    -确定(S405)所述至少两个第一区域与所述至少两个第二区域之间的一一对应关系；
    -从所述第一摄像头处获取(S406)第一帧图片；
    -从所述第一帧图片中识别(S407)目标物体；
    -确定(S408)所述目标物体在所述第一帧图片中的第二区域；
    -根据所述对应关系确定(S409)所述目标物体所在第二区域对应的第一区域。
  2. 如权利要求1所述的方法，其特征在于，从所述第一帧图片中识别(S407)目标物体，包括：
    -提取所述目标物体的粗粒度特征，所述粗粒度特征包括颜色、形状、轮廓、标志中的至少一种；
    -确定所述目标物体的移动向量的信息；
    -根据所述目标物体的粗粒度特征和移动向量的信息确定所述第一摄像头之外的其他摄像头拍摄的图片中是否出现所述目标物体。
  3. 如权利要求1所述的方法，其特征在于，还包括：
    -显示(S410)模拟的所述物理环境；
    -在模拟的所述物理环境中的、确定的所述目标物体所在的第一区域处显示(S411)所述目标物体。
  4. 如权利要求3所述的方法，其特征在于，还包括：
    -在模拟的所述物理环境中显示(S412)所述至少两个第一区域。
  5. 如权利要求3或4所述的方法，其特征在于，还包括：
    -接收(S413)所述目标物体在所述物理环境中的一组位置的信息，其中，所述一组位置是由识别出的所述目标物体的一组时间上连续的图片分别确定的所述目标物体在所述物理 环境中的各个第一区域；
    -在模拟的所述物理环境中按照时间顺序在所述一组位置所对应的每一个第一区域中显示(S414)所述目标物体。
  6. 如权利要求1所述的方法，其特征在于，还包括：
    -获取(S415)在第二帧图片识别出的所述目标物体在所述物理环境中的第一区域的信息，其中，所述第二帧图片与所述第一帧图片的拍摄时间相同；
    -根据在所述第一帧图片和所述第二帧图片分别识别出的所述目标物体在所述物理环境中各自的第一区域，更新(S416)所述目标物体在所述物理环境中的第一区域。
  7. 如权利要求6所述的方法，其特征在于，更新(S416)所述目标物体在所述物理环境中的第一区域，包括：
    -比较(S416a)所述目标物体在所述第一帧图片中所在的第二区域与所述第二帧图片中所在的第二区域的大小；
    -将较大的第二区域所对应的第一区域作为(S416b)更新后的第一区域。
  8. 一种目标定位装置(11)，其特征在于，包括：
    -位置映射模块(201)，被配置为：
    -确定物理环境中的至少一个标记；
    -根据所述至少一个标记将所述物理环境划分为至少两个第一区域；
    -从第一摄像头对所述物理环境拍摄的图片中识别所述至少一个标记；
    -按照所述至少两个第一区域同样的划分方式，将所述第一摄像头拍摄图片中的所述物理环境划分为至少两个第二区域；
    -确定所述至少两个第一区域与所述至少两个第二区域之间的一一对应关系；
    -图片处理模块(202)，被配置为从所述第一摄像头处获取第一帧图片；
    -目标识别模块(203)，被配置为从所述第一帧图片中识别目标物体；
    -所述位置映射模块(201)，还被配置为确定所述目标物体在所述第一帧图片中的第二区域，以及根据所述对应关系确定所述目标物体所在第二区域对应的第一区域。
  9. 如权利要求8所述的装置，其特征在于，所述目标识别模块(202)在从所述第一帧图片中识别目标物体时，被配置为：
    -提取所述目标物体的粗粒度特征，所述粗粒度特征包括颜色、形状、轮廓、标志中的至少一种；
    -确定所述目标物体的移动向量的信息；
    -根据所述目标物体的粗粒度特征和移动向量的信息确定所述第一摄像头之外的其他摄像头拍摄的图片中是否出现所述目标物体。
  10. 如权利要求8所述的装置，其特征在于，还包括：跟踪模块(204)，被配置为：
    -显示的所述物理环境；
    -在模拟的所述物理环境中的、确定的所述目标物体所在的第一区域处显示所述目标物体。
  11. 如权利要求10所述的装置，其特征在于，所述跟踪模块(204)，还被配置为：
    -在模拟的所述物理环境中显示所述至少两个第一区域。
  12. 如权利要求10或11所述的装置，其特征在于，所述跟踪模块(204)，还被配置为：
    -接收所述目标物体在所述物理环境中的一组位置的信息，其中，所述一组位置是由识别出的所述目标物体的一组时间上连续的图片分别确定的所述目标物体在所述物理环境中的各个第一区域；
    -在模拟的所述物理环境中按照时间顺序在所述一组位置所对应的每一个第一区域中显示所述目标物体。
  13. 如权利要求8所述的装置，其特征在于，还包括：位置更新模块(206)，被配置为：
    -获取在第二帧图片识别出的所述目标物体在所述物理环境中的第一区域的信息，其中，所述第二帧图片与所述第一帧图片的拍摄时间相同；
    -根据在所述第一帧图片和所述第二帧图片分别识别出的所述目标物体在所述物理环境中各自的第一区域，更新所述目标物体在所述物理环境中的第一区域。
  14. 如权利要求13所述的装置，其特征在于，所述位置更新模块(206)在更新所述目标物体在所述物理环境中的第一区域时，被配置为：
    -比较所述目标物体在所述第一帧图片中所在的第二区域与所述第二帧图片中所在的第二区域的大小；
    -将较大的第二区域所对应的第一区域作为更新后的第一区域。
  15. 一种目标定位装置(11)，其特征在于，包括：
    至少一个存储器(111)，被配置为存储计算机可读代码；
    至少一个处理器(112)，被配置为调用所述计算机可读代码，执行如权利要求1～7任一项所述的方法。
  16. 一种计算机可读介质，其特征在于，所述计算机可读介质上存储有计算机可读指令，所述计算机可读指令在被处理器执行时，使所述处理器执行如权利要求1～7任一项所述的方法。
  17. 如权利要求1～16所述的方法，装置或计算机可读介质，其特征在于，所述目标物体为车辆，所述至少两个第一区域是不同的停车位。
PCT/CN2020/116575 2020-09-21 2020-09-21 一种目标定位方法、装置和计算机可读介质 WO2022056924A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
EP20953782.8A EP4195150A4 (en) 2020-09-21 2020-09-21 TARGET POSITIONING METHOD AND APPARATUS AND COMPUTER-READABLE MEDIUM
US18/044,475 US20230316566A1 (en) 2020-09-21 2020-09-21 Target Positioning Method and Device, and Computer-Readable Medium
PCT/CN2020/116575 WO2022056924A1 (zh) 2020-09-21 2020-09-21 一种目标定位方法、装置和计算机可读介质
CN202080103037.0A CN115867937A (zh) 2020-09-21 2020-09-21 一种目标定位方法、装置和计算机可读介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/116575 WO2022056924A1 (zh) 2020-09-21 2020-09-21 一种目标定位方法、装置和计算机可读介质

Publications (1)

Publication Number Publication Date
WO2022056924A1 true WO2022056924A1 (zh) 2022-03-24

Family

ID=80775837

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/116575 WO2022056924A1 (zh) 2020-09-21 2020-09-21 一种目标定位方法、装置和计算机可读介质

Country Status (4)

Country Link
US (1) US20230316566A1 (zh)
EP (1) EP4195150A4 (zh)
CN (1) CN115867937A (zh)
WO (1) WO2022056924A1 (zh)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105571606A (zh) * 2014-11-04 2016-05-11 沃尔沃汽车公司 能够改进车辆定位的方法和***
CN109636828A (zh) * 2018-11-20 2019-04-16 北京京东尚科信息技术有限公司 基于视频图像的物体跟踪方法及装置
CN110443850A (zh) * 2019-08-05 2019-11-12 珠海优特电力科技股份有限公司 目标对象的定位方法及装置、存储介质、电子装置
CN110659548A (zh) * 2018-06-29 2020-01-07 比亚迪股份有限公司 车辆及其目标检测方法、装置
CN111046762A (zh) * 2019-11-29 2020-04-21 腾讯科技（深圳）有限公司 一种对象定位方法、装置电子设备及存储介质
US20200160069A1 (en) * 2018-11-19 2020-05-21 Nauto, Inc. System and method for vehicle localization

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8830322B2 (en) * 2012-08-06 2014-09-09 Cloudparc, Inc. Controlling use of a single multi-vehicle parking space and a restricted location within the single multi-vehicle parking space using multiple cameras
JP6223685B2 (ja) * 2013-01-30 2017-11-01 富士通テン株式会社 画像処理装置および画像処理方法
US20190378345A1 (en) * 2018-06-07 2019-12-12 Capital One Services, Llc Locating and bidding for a parking space via peer-to-peer communication of parking information
CN109064755B (zh) * 2018-07-24 2021-06-15 河北德冠隆电子科技有限公司 基于四维实景交通仿真路况感知管理***的路径识别方法
US11295161B2 (en) * 2019-05-22 2022-04-05 Zoox, Inc. Localization using semantically segmented images
AU2019100914A4 (en) * 2019-08-16 2019-09-26 Weifang University Method for identifying an intersection violation video based on camera cooperative relay
US11520033B2 (en) * 2019-12-12 2022-12-06 Amazon Technologies, Inc. Techniques for determining a location of a mobile object

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105571606A (zh) * 2014-11-04 2016-05-11 沃尔沃汽车公司 能够改进车辆定位的方法和***
CN110659548A (zh) * 2018-06-29 2020-01-07 比亚迪股份有限公司 车辆及其目标检测方法、装置
US20200160069A1 (en) * 2018-11-19 2020-05-21 Nauto, Inc. System and method for vehicle localization
CN109636828A (zh) * 2018-11-20 2019-04-16 北京京东尚科信息技术有限公司 基于视频图像的物体跟踪方法及装置
CN110443850A (zh) * 2019-08-05 2019-11-12 珠海优特电力科技股份有限公司 目标对象的定位方法及装置、存储介质、电子装置
CN111046762A (zh) * 2019-11-29 2020-04-21 腾讯科技（深圳）有限公司 一种对象定位方法、装置电子设备及存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP4195150A4 *

Also Published As

Publication number Publication date
US20230316566A1 (en) 2023-10-05
EP4195150A4 (en) 2024-06-26
CN115867937A (zh) 2023-03-28
EP4195150A1 (en) 2023-06-14

Similar Documents

Publication Publication Date Title
Patil et al. The h3d dataset for full-surround 3d multi-object detection and tracking in crowded urban scenes
US10403138B2 (en) Traffic accident warning method and traffic accident warning apparatus
CN108694882B (zh) 用于标注地图的方法、装置和设备
Broggi et al. Terramax vision at the urban challenge 2007
US11280630B2 (en) Updating map data
JP2005268847A (ja) 画像生成装置、画像生成方法、および画像生成プログラム
US11025865B1 (en) Contextual visual dataspaces
WO2020156923A2 (en) Map and method for creating a map
CN111353453B (zh) 用于车辆的障碍物检测方法和装置
CN112950717A (zh) 一种空间标定方法和***
Cao et al. Amateur: Augmented reality based vehicle navigation system
Vu et al. Traffic sign detection, state estimation, and identification using onboard sensors
AU2021255130A1 (en) Artificial intelligence and computer vision powered driving-performance assessment
Choi et al. Methods to detect road features for video-based in-vehicle navigation systems
JP2023064093A (ja) 交通標識検出方法及び交通標識検出モデルのトレーニング方法
KR20200013156A (ko) 간판 검출 성능 향상을 위한 방법 및 시스템
CN115004273A (zh) 交通道路的数字化重建方法、装置和***
CN110827340B (zh) 地图的更新方法、装置及存储介质
WO2022056924A1 (zh) 一种目标定位方法、装置和计算机可读介质
CN111754388B (zh) 一种建图方法及车载终端
JP2022513830A (ja) 道路の表面上の対象体を検出してモデル化する方法
Delgado et al. Virtual validation of a multi-object tracker with intercamera tracking for automotive fisheye based surround view systems
CN115457084A (zh) 一种多相机目标检测跟踪方法、装置
Lee et al. Semi-automatic framework for traffic landmark annotation
EP4078087A1 (en) Method and mobile entity for detecting feature points in an image

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20953782

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2020953782

Country of ref document: EP

Effective date: 20230308

NENP Non-entry into the national phase

Ref country code: DE