WO2020228766A1 - 基于实景建模与智能识别的目标跟踪方法、系统及介质 - Google Patents

基于实景建模与智能识别的目标跟踪方法、系统及介质 Download PDF

Info

Publication number
WO2020228766A1
WO2020228766A1 PCT/CN2020/090217 CN2020090217W WO2020228766A1 WO 2020228766 A1 WO2020228766 A1 WO 2020228766A1 CN 2020090217 W CN2020090217 W CN 2020090217W WO 2020228766 A1 WO2020228766 A1 WO 2020228766A1
Authority
WO
WIPO (PCT)
Prior art keywords
scene
real
dimensional
intelligent recognition
modeling
Prior art date
Application number
PCT/CN2020/090217
Other languages
English (en)
French (fr)
Inventor
李新福
Original Assignee
广东康云科技有限公司
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 广东康云科技有限公司 filed Critical 广东康云科技有限公司
Publication of WO2020228766A1 publication Critical patent/WO2020228766A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T19/00—Manipulating 3D models or images for computer graphics
    • G06T19/003—Navigation within 3D models or images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/285—Analysis of motion using a sequence of stereo image pairs
    • 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
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence
    • G06T2207/10021—Stereoscopic video; Stereoscopic image sequence
    • 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
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning

Definitions

  • the invention relates to the field of three-dimensional modeling and pattern recognition, in particular to a target tracking method, system and medium based on real scene modeling and intelligent recognition.
  • Three-dimensional modeling can enable computers and other equipment to do this.
  • Three-dimensional modeling is to use three-dimensional data to reconstruct real three-dimensional objects or scenes in computers and other equipment, and finally realize the simulation of real three-dimensional objects or scenes on computers and other equipment.
  • the three-dimensional data is the data collected by various three-dimensional data acquisition equipment, which records various physical parameters of the finite body surface at discrete points.
  • Most of the current 3D modeling technologies can only reconstruct the static 3D model of the scene or object, but the real scene often contains dynamic information such as the movement of the object.
  • the existing 3D modeling technology is difficult to display these dynamics in the reconstructed 3D model. information.
  • CCTV Cellular Circuit Television Monitoring System
  • other video shooting devices are generally installed in streets, intersections, stations, and important buildings in cities at all levels.
  • the main method of target tracking is to identify the target object by analyzing the content captured by the camera.
  • most of the current target tracking methods can only provide a 2D tracking screen, and the 2D tracking screen fails to give the user a real feeling of being there.
  • the purpose of the embodiments of the present invention is to provide a target tracking method, system and medium based on real scene modeling and intelligent recognition.
  • the target tracking method based on real scene modeling and intelligent recognition includes the following steps:
  • the tracking target is obtained from the 3D real scene of the scene, and the tracking target is tracked in the 3D real scene of the scene.
  • the step of acquiring the three-dimensional data of the scene and the real-time video stream of the scene and performing real-scene three-dimensional modeling to obtain the three-dimensional real scene of the scene specifically includes:
  • the scanning device including a space scanner, an aerial scanner, an object scanner, and a human body scanner;
  • the three-dimensional reconstruction includes model repair, editing, cropping, surface reduction, mold reduction, compression, material processing, texture processing, lighting processing and compression Rendering
  • the real-time video stream of the scene is merged into the 3D model of the scene to obtain the 3D real scene of the scene.
  • the step of intelligently identifying the real-time video stream of the scene specifically includes:
  • step of inputting the real-time video stream of the three-dimensional real scene of the scene into the trained intelligent recognition model to obtain the intelligent recognition result is specifically:
  • the real-time video stream of the three-dimensional real scene of the scene is input into the trained intelligent recognition model, and the first result of the recognition is obtained.
  • the first result includes the type and name of the object, the attribute of the object, and the behavior of the object.
  • the type of the object includes a person , Animals and objects, the attributes of the objects include color, style, gender, age and model.
  • the step of acquiring the tracking target from the three-dimensional real scene of the scene according to the result of intelligent recognition, and tracking the tracking target in the three-dimensional real scene of the scene specifically includes:
  • the labeled tracking target is automatically tracked in the three-dimensional real scene of the scene, so as to obtain the moving path or trajectory of the tracking target.
  • the step of acquiring the tracking target from the 3D real scene of the scene according to the result of the intelligent recognition, and tracking the tracking target in the 3D real scene of the scene also specifically includes:
  • the target tracking system based on real scene modeling and intelligent recognition includes the following modules:
  • the real-scene 3D modeling module is used to obtain the 3D data of the scene and the real-time video stream of the scene and perform real-scene 3D modeling to obtain the 3D real scene of the scene.
  • the real 3D real scene of the scene displays the real-time video stream of the scene in the 3D model of the scene ;
  • the intelligent recognition module is used to intelligently recognize the three-dimensional real scene of the scene, and the intelligent recognition includes segmentation and intelligent recognition of the three-dimensional model of the scene, and intelligent recognition of the real-time video stream of the scene;
  • the target tracking module is used to obtain the tracking target from the 3D real scene of the scene according to the result of the intelligent recognition, and track the tracking target in the 3D real scene of the scene.
  • the real scene 3D modeling module specifically includes:
  • the scanning unit is used to obtain three-dimensional data of the scene by scanning with a scanning device, the scanning device including a space scanner, an aerial scanner, an object scanner, and a human body scanner;
  • the video acquisition unit is used to acquire the real-time video stream of the scene through the video acquisition device;
  • the 3D reconstruction unit is used to perform 3D reconstruction using artificial intelligence methods according to the 3D data of the scene to generate a 3D model of the scene.
  • the 3D reconstruction includes model repair, editing, cropping, surface reduction, model reduction, compression, material processing, and processing Mapping, lighting and compression rendering;
  • the link generation unit is used to generate the corresponding link according to the three-dimensional model of the scene
  • the fusion unit is used to fuse the real-time video stream of the scene into the 3D model of the scene to obtain the 3D real scene of the scene.
  • Target tracking system based on real scene modeling and intelligent recognition, including:
  • At least one processor At least one processor
  • At least one memory for storing at least one program
  • the at least one processor When the at least one program is executed by the at least one processor, the at least one processor implements the target tracking method based on real scene modeling and intelligent recognition according to the present invention.
  • a medium storing instructions executable by the processor, and the instructions executable by the processor are used to implement the target tracking method based on real-scene modeling and intelligent recognition according to the present invention when executed by the processor.
  • the embodiment of the present invention first performs 3D modeling of real scenes, generates 3D real scenes, and then intelligently recognizes the real-time video streams of 3D real scenes, and finally detects the three-dimensional real scenes
  • the tracking target is tracked, and the real-time video stream containing the dynamic information of the scene is implanted into the 3D model of the scene, which realizes the function of displaying dynamic information in the reconstructed 3D model; according to the result of intelligent recognition, the real-time 3D scene of the scene is checked.
  • the tracking target is tracked, and the 3D real scene tracking screen is provided through the 3D real scene fusion of the 3D model and the real-time video stream, which can give the user an immersive and more realistic experience.
  • FIG. 1 is a flowchart of a target tracking method based on real scene modeling and intelligent recognition provided by an embodiment of the present invention
  • FIG. 2 is a structural block diagram of a target tracking system based on real scene modeling and intelligent recognition provided by an embodiment of the present invention
  • Fig. 3 is another structural block diagram of a target tracking system based on real scene modeling and intelligent recognition provided by an embodiment of the present invention.
  • first, second, third, etc. may be used in this disclosure to describe various elements, these elements should not be limited to these terms. These terms are only used to distinguish elements of the same type from each other.
  • first element may also be referred to as the second element, and similarly, the second element may also be referred to as the first element.
  • second element may also be referred to as the first element.
  • the use of any and all examples or exemplary language (“such as”, “such as”, etc.) provided herein is only intended to better illustrate the embodiments of the present invention, and unless otherwise required, will not impose limitations on the scope of the present invention .
  • an embodiment of the present invention provides a target tracking method based on real scene modeling and intelligent recognition, including the following steps:
  • the scene in this embodiment may be a scene such as a certain city, a certain park, or a certain building.
  • the three-dimensional data of the scene can be two-dimensional images, point cloud data of the scene, etc., which can be collected by various manual or automatic scanning devices (such as cameras, automatic scanning robots, etc.).
  • the real-time video stream of the scene can be obtained through CCTV, camera and other video capture devices.
  • S101 Perform intelligent recognition of the three-dimensional real scene of the scene, the intelligent recognition including segmentation and intelligent recognition of the three-dimensional model of the scene, and intelligent recognition of the real-time video stream of the scene;
  • the process of segmenting and intelligently identifying the three-dimensional model of the scene in this embodiment may include the following implementation steps:
  • the three-dimensional model of the scene is composed of multiple point clouds (collections of points), it can provide corresponding point cloud data after generating the three-dimensional model of the scene to facilitate subsequent segmentation and intelligent recognition.
  • S10101 Segment the point cloud data according to the relationship between the point and the neighboring points (such as color, pixel, size, size, distance, etc.) to generate a point cloud of each object in the scene according to the point cloud data;
  • S10102. Use an artificial intelligence algorithm to intelligently identify the point cloud of each object in the scene.
  • the scene can contain multiple objects, characters and other objects.
  • the point cloud data is segmented, and the point cloud obtained for the objects in the scene also contains multiple point clouds.
  • the specific types and names of these segmented point clouds can be obtained after being recognized by artificial intelligence algorithms.
  • the intelligent recognition model pre-trained by the artificial intelligence algorithm can be used for automatic recognition to facilitate subsequent target tracking or other intelligent analysis.
  • the intelligent recognition of real-time video stream mainly recognizes the objects contained in the real-time video stream and the posture and movement of the objects.
  • S102 Obtain a tracking target from the three-dimensional real scene of the scene according to the result of the intelligent recognition, and track the tracking target in the three-dimensional real scene of the scene.
  • the tracking target can be marked out and locked in the three-dimensional real scene by means of a rectangular marking frame, and the CCTV closed-circuit television that obtains the video stream Video acquisition devices such as surveillance systems are networked, so that even if the tracking target walks from the current video acquisition device to the acquisition area of another video acquisition device, it can still be identified.
  • this embodiment embeds the real-time video stream containing the dynamic information of the scene into the three-dimensional model of the scene, and realizes the function of displaying dynamic information in the reconstructed three-dimensional model; according to the result of intelligent recognition, the three-dimensional real scene of the scene is displayed.
  • the tracking target is tracked internally, and the 3D real scene tracking screen is provided by the 3D real scene fusion of the 3D model and the real-time video stream, which can give the user an immersive and more realistic experience.
  • the step S100 of acquiring the three-dimensional data of the scene and the real-time video stream of the scene and performing real-scene three-dimensional modeling to obtain the three-dimensional real scene of the scene specifically includes:
  • S1000 Acquire three-dimensional data of the scene by scanning by a scanning device, the scanning device including a space scanner, an aerial scanner, an object scanner, and a human body scanner;
  • the scanning device is used to scan objects in the scene and upload the scanned data to the cloud or back-end server.
  • the scanning device may be an aerial scanning device, a space scanner, an object scanner, or a human body scanning device.
  • the aerial scanning equipment which can be aerial photographing equipment such as an aerial photographing plane, is used to scan the three-dimensional data of the area within the scene (such as the entire park).
  • Spatial scanning equipment used for scanning indoor environment (such as the inside of a certain building of a building) or scanning outdoor environment (such as a certain road outside a building) 3D data.
  • the space scanning device can be a handheld scanning device (such as a camera with a support frame) or other automatic scanning equipment (such as an automatic scanning robot).
  • the object scanner used to scan an object (such as apple, pen).
  • the object scanner can be a handheld scanning device (such as an RGB-D camera with a support frame, etc.).
  • the human body scanner is used to scan the three-dimensional data of the human body.
  • the human body scanner may be an existing human body scanner specifically for human body modeling.
  • Three-dimensional data includes data such as two-dimensional pictures and depth information.
  • the scanning device of this embodiment can be integrated with a GPU chip that has edge computing capabilities and can be implanted with artificial intelligence algorithms, and can perform calculations while scanning, thereby generating a three-dimensional model of the scene part, so that only the cloud or background server needs to be Just generate a 3D model of the rest of the scene, which greatly improves the efficiency of modeling.
  • S1001 Obtain a real-time video stream of the scene through a video capture device
  • the video capture device may be a surveillance camera, CCTV closed-circuit television and other devices.
  • the three-dimensional reconstruction includes model repair, editing, cropping, surface reduction, mold reduction, compression, material processing, texture processing, and lighting And compressed rendering;
  • an artificial intelligence method is used to perform three-dimensional reconstruction according to the three-dimensional data of the scene, and the process of generating a three-dimensional model of the scene can be performed in a scanning device, a cloud or a background server.
  • the scanning device, cloud or back-end server integrates AI algorithms, which can realize fully automated and rapid modeling without human involvement, significantly improving the efficiency of modeling and a high degree of intelligence.
  • this embodiment can generate links (such as URL links, etc.) to the three-dimensional model of the scene, so that any computing device (including smart phones, tablets, laptops, smart watches, smart TVs, computers, etc.) that supports browsers
  • the 3D model can be accessed through this link, eliminating the need to install the APP, which is more convenient and more versatile.
  • the location of the video capture device used to capture real-time video streams is known or can be set in advance, after generating a three-dimensional model of the scene, you only need to find out its corresponding location in the three-dimensional model and set the video
  • the real-time video stream collected by the stream collection device is superimposed on this position for continuous playback, and the 3D video stream of the scene can be dynamically displayed in the 3D model, which overcomes the defect that the 3D models obtained by traditional 3D scanning modeling technology are static.
  • Realize the real 3D monitoring of the scene. Users can access the 3D model of the scene and the 3D video stream dynamically played or displayed in the 3D model through the link corresponding to the 3D model.
  • This embodiment uses this method to truly realize the seamless integration of the 3D model and the real-time video stream, without being affected by the change of the scene and the angle, and the real-time video can still be watched in the 3D model after the scene or the angle is changed. flow.
  • the step of intelligently identifying the real-time video stream of the scene specifically includes:
  • the training samples and labels given in this embodiment can be provided by a pre-established database.
  • a face recognition model can be put into the face database through various types of faces collected in advance, and can be directly trained during training. Obtain human faces from the database as samples, and acquire corresponding face types or names (such as faces of different ages, countries, etc.) as labels.
  • This embodiment uses the artificial intelligence method to train an accurate intelligent recognition model, so that even if there is a new tag unknown data input in the real-time video stream, the intelligent recognition model can be automatically recognized, which is more efficient and more intelligent.
  • step S10110 of inputting the real-time video stream of the three-dimensional real-time scene of the scene into the trained intelligent recognition model to obtain the intelligent recognition result is specifically:
  • the real-time video stream of the three-dimensional real scene of the scene is input into the trained intelligent recognition model, and the first result of the recognition is obtained.
  • the first result includes the type and name of the object, the attribute of the object, and the behavior of the object.
  • the type of the object includes a person , Animals and objects, the attributes of the objects include color, style, gender, age and model.
  • this embodiment can accurately identify the type and name of the object, the attributes of the object, and the behavior of the object through the intelligent recognition of the real-time video stream, which is conducive to further intelligent analysis and subsequent target tracking operations.
  • one scene and one object in the three-dimensional real scene of the scene can be identified, which greatly facilitates subsequent target tracking and monitoring applications.
  • the step S102 of acquiring the tracking target from the 3D real scene of the scene according to the result of the intelligent recognition, and tracking the tracking target in the 3D real scene of the scene specifically includes:
  • S1020 Determine and mark the tracking target in the 3D real scene of the scene
  • an object in the scene (such as a criminal suspect, a suspicious vehicle, etc.) can be added by adding a rectangular frame and adding an overlay according to actual tracking needs.
  • the irregular shape of the object outline is marked out and locked as the tracking target.
  • the target tracking solution of this embodiment provides a real-time video stream of the scene while also providing a three-dimensional model of the location of the video stream (360 degrees without dead angles) Roaming), realizing the 3D real scene tracking of video stream + 3D model, which is more three-dimensional and real.
  • the step S102 of acquiring the tracking target from the three-dimensional real scene of the scene according to the result of the intelligent recognition, and tracking the tracking target in the three-dimensional real scene of the scene further specifically includes:
  • S1023 Recognizing the marked movement posture of the tracking target in the 3D real scene of the scene.
  • the recognition model trained and learned by artificial intelligence algorithms can also recognize the motion posture of the tracking target (such as whether the person is running or walking, whether the person is holding something or wearing sunglasses, etc.), which can provide richer detailed information , Is conducive to target tracking operation, more intelligent and convenient.
  • the embodiment of the present invention also provides a target tracking system based on real scene modeling and intelligent recognition, including the following modules:
  • the real-scene three-dimensional modeling module 201 is used to obtain the three-dimensional data of the scene and the real-time video stream of the scene and perform real-scene three-dimensional modeling to obtain the three-dimensional real scene of the scene, and the real-time video of the scene is displayed in the three-dimensional model of the scene. flow;
  • the intelligent recognition module 202 is used to intelligently recognize the three-dimensional real scene of the scene, and the intelligent recognition includes segmentation and intelligent recognition of the three-dimensional model of the scene and intelligent recognition of the real-time video stream of the scene;
  • the target tracking module 203 is configured to obtain a tracking target from the three-dimensional real scene of the scene according to the result of intelligent recognition, and track the tracking target in the three-dimensional real scene of the scene.
  • the real-scene three-dimensional modeling module 201 specifically includes:
  • the scanning unit 2011 is configured to obtain three-dimensional data of the scene by scanning with a scanning device, the scanning device including a space scanner, an aerial scanner, an object scanner, and a human body scanner;
  • the video acquisition unit 2012 is used to acquire the real-time video stream of the scene through the video acquisition device;
  • the three-dimensional reconstruction unit 2013 is used to perform three-dimensional reconstruction using artificial intelligence methods according to the three-dimensional data of the scene to generate a three-dimensional model of the scene.
  • the three-dimensional reconstruction includes model repair, editing, cropping, surface reduction, mold reduction, compression, material processing, Processing textures, processing lighting and compression rendering;
  • the link generation unit 2014 is configured to generate corresponding links according to the three-dimensional model of the scene
  • the fusion unit 2015 is used for fusing the real-time video stream of the scene into the three-dimensional model of the scene to obtain the three-dimensional real scene of the scene.
  • the embodiment of the present invention also provides a target tracking system based on real scene modeling and intelligent recognition, including:
  • At least one processor 301 At least one processor 301;
  • the at least one processor 301 realizes the target tracking method based on real scene modeling and intelligent recognition according to the present invention.
  • the embodiment of the present invention also provides a medium in which instructions executable by the processor are stored. When the instructions executable by the processor are executed by the processor, they are used to implement the reality-based modeling and intelligent recognition of the present invention. Target tracking method.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Remote Sensing (AREA)
  • Computer Graphics (AREA)
  • Computer Hardware Design (AREA)
  • General Engineering & Computer Science (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Analysis (AREA)

Abstract

一种基于实景建模与智能识别的目标跟踪方法、系统及介质，方法包括：获取场景的三维数据和场景的实时视频流并进行实景三维建模（S100）；对场景的三维实景进行智能识别（S101）；根据智能识别的结果从场景的三维实景中获取跟踪目标，并在场景的三维实景内对跟踪目标进行跟踪（S102）。通过将包含有场景的动态信息的实时视频流植入场景的三维模型，实现了在重建出的三维模型中展示动态信息的功能；根据智能识别的结果在场景的三维实景内对跟踪目标进行跟踪，通过融合三维模型和实时视频流的三维实景提供了3D实景追踪画面，能给予用户身临其境的感受，更加真实。该方法可广泛应用于三维建模与模式识别领域。

Description

基于实景建模与智能识别的目标跟踪方法、系统及介质 技术领域
本发明涉及三维建模与模式识别领域，尤其是一种基于实景建模与智能识别的目标跟踪方法、系统及介质。
背景技术
随着计算机等设备在各行各业的广泛应用，人们开始不满足于计算机等设备仅能显示二维的图像，更希望计算机等设备能表达出具有强烈真实感的现实三维世界。三维建模可以使计算机等设备做到这一点。三维建模，就是利用三维数据将现实中的三维物体或场景在计算机等设备中进行重建，最终实现在计算机等设备上模拟出真实的三维物体或场景。而三维数据就是使用各种三维数据采集设备采集得到的数据，它记录了有限体表面在离散点上的各种物理参量。目前的三维建模技术大多只能重建出场景或物体的静态三维模型，但真实的场景往往包含有物体的运动等动态信息，现有三维建模技术难以在重建出的三维模型中展示这些动态信息。
随着视频监控和网络传输技术的快速发展，在各级城市的街道、路口、车站、重要建筑物等地点普遍安装了CCTV(闭路电视监控系统)等视频拍摄装置。通过分析拍摄装置拍摄的内容以进行目标对象的识别是目前目标追踪的主要方法。然而，目前的目标追踪方法大多只能提供2D的追踪画面，2D追踪画面未能给予用户身临其境的真实感受。
目前，在重建出的三维模型中融入实时动态信息生成3D实景并用于目标追踪的方案尚未见诸报道。
发明内容
为解决上述技术问题，本发明实施例的目的在于：提供一种基于实景建模与智能识别的目标跟踪方法、系统及介质。
本发明实施例所采取的第一技术方案是：
基于实景建模与智能识别的目标跟踪方法，包括以下步骤：
获取场景的三维数据和场景的实时视频流并进行实景三维建模，得到场景的三维实景，所述场景的三维实景在场景的三维模型内展示场景的实时视频流；
对场景的三维实景进行智能识别，所述智能识别包括对场景的三维模型进行分割与智能识别以及对场景的实时视频流进行智能识别；
根据智能识别的结果从场景的三维实景中获取跟踪目标，并在场景的三维实景内对跟踪目标进行跟踪。
进一步，所述获取场景的三维数据和场景的实时视频流并进行实景三维建模，得到场景的三维实景这一步骤，具体包括：
通过扫描设备扫描的方式获取场景的三维数据，所述扫描设备包括空间扫描仪、航拍扫描仪、物体扫描仪和人体扫描仪；
通过视频采集设备获取场景的实时视频流；
根据场景的三维数据采用人工智能的方法进行三维重建，生成场景的三维模型，其中，三维重建包括模型修复、剪辑、裁剪、减面、减模、压缩、处理材质、处理贴图、处理灯光和压缩渲染；
根据场景的三维模型生成对应的链接；
将场景的实时视频流融合至场景的三维模型中，得到场景的三维实景。
进一步，所述对场景的实时视频流进行智能识别这一步骤，具体包括：
根据给定的训练样本和标签采用人工智能的方法训练智能识别模型；
将场景的三维实景的实时视频流输入训练好的智能识别模型，得到智能识别结果。
进一步，所述将场景的三维实景的实时视频流输入训练好的智能识别模型，得到智能识别结果这一步骤，具体为：
将场景的三维实景的实时视频流输入训练好的智能识别模型，识别得到第一结果，所述第一结果包括对象的类型和名称、对象的属性以及对象的行为，所述对象的类型包括人、动物和物体，所述对象的属性包括颜色、款式、性别、年龄和型号。
进一步，所述根据智能识别的结果从场景的三维实景中获取跟踪目标，并在场景的三维实景内对跟踪目标进行跟踪这一步骤，具体包括：
在场景的三维实景中确定并标注跟踪目标；
在场景的三维实景内自动对标注后的跟踪目标进行跟踪，从而获得跟踪目标的运动路径或轨迹。
进一步，所述根据智能识别的结果从场景的三维实景中获取跟踪目标，并在场景的三维实景内对跟踪目标进行跟踪这一步骤，还具体包括：
在场景的三维实景内识别标注后的跟踪目标的运动姿态。
本发明实施例所采取的第二技术方案是：
基于实景建模与智能识别的目标跟踪系统，包括以下模块：
实景三维建模模块，用于获取场景的三维数据和场景的实时视频流并进行实景三维建模，得到场景的三维实景，所述场景的三维实景在场景的三维模型内展示场景的实时视频流；
智能识别模块，用于对场景的三维实景进行智能识别，所述智能识别包括对场景的三维模型进行分割与智能识别以及对场景的实时视频流进行智能识别；
目标跟踪模块，用于根据智能识别的结果从场景的三维实景中获取跟踪目标，并在场景的三维实景内对跟踪目标进行跟踪。
进一步，所述实景三维建模模块具体包括：
扫描单元，用于通过扫描设备扫描的方式获取场景的三维数据，所述扫描设备包括空间扫描仪、航拍扫描仪、物体扫描仪和人体扫描仪；
视频采集单元，用于通过视频采集设备获取场景的实时视频流；
三维重建单元，用于根据场景的三维数据采用人工智能的方法进行三维重建，生成场景的三维模型，其中，三维重建包括模型修复、剪辑、裁剪、减面、减模、压缩、处理材质、处理贴图、处理灯光和压缩渲染；
链接生成单元，用于根据场景的三维模型生成对应的链接；
融合单元，用于将场景的实时视频流融合至场景的三维模型中，得到场景的三维实景。
本发明实施例所采取的第三技术方案是：
基于实景建模与智能识别的目标跟踪系统，包括：
至少一个处理器；
至少一个存储器，用于存储至少一个程序；
当所述至少一个程序被所述至少一个处理器执行，使得所述至少一个处理器实现本发明所述的基于实景建模与智能识别的目标跟踪方法。
本发明实施例所采取的第四技术方案是：
介质，其中存储有处理器可执行的指令，所述处理器可执行的指令在由处理器执行时用于实现本发明所述的基于实景建模与智能识别的目标跟踪方法。
上述本发明实施例中的一个或多个技术方案具有如下优点：本发明实施例先进行实景三维建模，生成三维实景，再对三维实景的实时视频流进行智能识别，最后对三维实景内的跟踪目标进行跟踪，将包含有场景的动态信息的实时视频流植入场景的三维模型，实现了在重建出的三维模型中展示动态信息的功能；根据智能识别的结果在场景的三维实景内对跟踪目标进行跟踪，通过融合三维模型和实时视频流的三维实景提供了3D实景追踪画面，能给予用户身临其境的沉浸式感受，更加真实。
附图说明
图1为本发明实施例提供的基于实景建模与智能识别的目标跟踪方法流程图；
图2为本发明实施例提供的基于实景建模与智能识别的目标跟踪系统一种结构框图；
图3为本发明实施例提供的基于实景建模与智能识别的目标跟踪系统另一种结构框图。
具体实施方式
以下将结合实施例和附图对本发明的构思、具体结构及产生的技术效果进行清楚、完整的描述，以充分地理解本发明的目的、方案和效果。
需要说明的是，如无特殊说明，当某一特征被称为“固定”、“连接”在另一个特征，它可以直接固定、连接在另一个特征上，也可以间接地固定、连接在另一个特征上。此外，本公开中所使用的上、下、左、右等描述仅仅是相对于附图中本公开各组成部分的相互位置关系来说的。在本公开中所使用的单数形式的“一种”、“所述”和“该”也旨在包括多数形式，除非上下文清楚地表示其他含义。此外，除非另有定义，本文所使用的所有的技术和科学术语与本技术领域的技术人员通常理解的含义相同。本文说明书中所使用的术语只是为了描述具体的实施例，而不是为了限制本发明。本文所使用的术语“和/或”包括一个或多个相关的所列项目的任意的组合。
应当理解，尽管在本公开可能采用术语第一、第二、第三等来描述各种元件，但这些元件不应限于这些术语。这些术语仅用来将同一类型的元件彼此区分开。例如，在不脱离本公开范围的情况下，第一元件也可以被称为第二元件，类似地，第二元件也可以被称为第一元件。本文所提供的任何以及所有实例或示例性语言(“例如”、“如”等)的使用仅意图更好地说明本发明的实施例，并且除非另外要求，否则不会对本发明的范围施加限制。
参照图1，本发明实施例提供了一种基于实景建模与智能识别的目标跟踪方法，包括以下步骤：
S100、获取场景的三维数据和场景的实时视频流并进行实景三维建模，得到场景的三维实景，所述场景的三维实景在场景的三维模型内展示场景的实时视频流；
具体地，本实施例的场景可以是某个城市、某个园区、某个建筑物等场景。场景的三维数据可以是二维的图像、场景的点云数据等，其可通过各手动或自动的扫描设备(如相机、自动扫描机器人等)来采集。
场景的实时视频流可以通过CCTV、摄像头等视频采集装置来获取。
S101、对场景的三维实景进行智能识别，所述智能识别包括对场景的三维模型进行分割与智能识别以及对场景的实时视频流进行智能识别；
具体地，本实施例对场景的三维模型进行分割与智能识别这一过程，可包括以下实现步骤：
S10100、从场景的三维模型获取点云数据；
由于场景的三维模型是由多个点云(点的集合)组成的，故其可以在生成场景的三维模型后提供对应的点云数据，以便于后续的分割与智能识别。
S10101、根据点与邻近点的关系(如颜色、像素、大小、尺寸、距离等)对点云数据进行分割，以根据点云数据生成场景内各个对象的点云；
S10102、采用人工智能算法智能识别场景内各个对象的点云。
场景内可包含多个物体、人物等对象。相应地，对点云数据进行分割，得到场景内对象的点云也包含了多个点云，这些分割出的点云的具体类型和名称可经人工智能算法识别后即可得出。
而对场景的实时视频流进行智能识别时，可通过人工智能算法预先训练的智能识别模型进行自动识别，以便于后续的目标跟踪或其他智能分析。实时视频流的智能识别主要识别的是实时视频流所包含的对象及对象的姿态、动作等信息。
S102、根据智能识别的结果从场景的三维实景中获取跟踪目标，并在场景的三维实景内对跟踪目标进行跟踪。
具体地，在对场景的三维实景内的三维模型和实时视频流识别完成后，可以在该三维实景内通过矩形标注框等方式将跟踪目标标注出来并锁定，而由于获取视频流的CCTV闭路电视监控系统等视频获取装置是联网的，这样即使跟踪目标从当前视频获取装置走到另外一个视频获取装置的采集区域仍能被识别。
由此可见，本实施例将包含有场景的动态信息的实时视频流植入场景的三维模型，实现了在重建出的三维模型中展示动态信息的功能；根据智能识别的结果在场景的三维实景内对跟踪目标进行跟踪，通过融合三维模型和实时视频流的三维实景提供了3D实景追踪画面，能给予用户身临其境的沉浸式感受，更加真实。
进一步作为优选的实施方式，所述获取场景的三维数据和场景的实时视频流并进行实景三维建模，得到场景的三维实景这一步骤S100，具体包括：
S1000、通过扫描设备扫描的方式获取场景的三维数据，所述扫描设备包括空间扫描仪、航拍扫描仪、物体扫描仪和人体扫描仪；
具体地，扫描设备，用于对场景内的对象进行扫描，并将扫描的数据上传给云端或后台服务器。扫描设备可以是航拍扫描设备、空间扫描仪、物体扫描仪或人体扫描设备。航拍扫 描设备，可以是航拍飞机等航拍设备，用于扫描场景内区域范围(如整个园区)的三维数据。空间扫描设备，用于扫描室内环境(如某栋建筑某层楼的内部)或扫描室外环境(如某栋建筑外的某条马路等)的三维数据。空间扫描设备，可以是手持扫描设备(如带支撑架的相机)或其他自动扫描设备(如自动扫描机器人)。物体扫描仪，用于对某个物体(如苹果、笔)进行扫描。物体扫描仪，可以是手持的扫描设备(如带支撑架的RGB-D摄像机等)。人体扫描仪，用于扫描人体的三维数据。人体扫描仪，可以是现有专门针对人体建模的人体扫描仪。
三维数据包括二维图片和深度信息等数据。
优选地，本实施例的扫描设备可集成有具有边缘计算能力且可以植入人工智能算法的GPU芯片，能在扫描的同时进行计算，从而生成场景部分的三维模型，这样云端或后台服务器只需生成场景余下部分的三维模型即可，大大提升了建模的效率。
S1001、通过视频采集设备获取场景的实时视频流；
具体地，视频采集设备可以是监控摄像头、CCTV闭路电视等装置。
S1002、根据场景的三维数据采用人工智能的方法进行三维重建，生成场景的三维模型，其中，三维重建包括模型修复、剪辑、裁剪、减面、减模、压缩、处理材质、处理贴图、处理灯光和压缩渲染；
具体地，根据场景的三维数据采用人工智能的方法进行三维重建，生成场景的三维模型这一过程可在扫描设备、云端或后台服务器中进行。扫描设备、云端或后台服务器集成了AI算法，能实现完全自动化的快速建模，无需人工的参与，显著提升了建模的效率且智能化程度高。
S1003、根据场景的三维模型生成对应的链接；
具体地，本实施例可生成场景的三维模型的链接(如URL链接等)，这样任何支持浏览器的计算设备(包括智能手机、平板电脑、笔记本电脑、智能手表、智能电视、计算机等)都可以通过该链接访问该三维模型，省去了装APP的过程，更加方便且通用性更强。
S1004、将场景的实时视频流融合至场景的三维模型中，得到场景的三维实景。
具体地，由于用于采集实时视频流的视频采集装置的位置是已知道或可以预先设定的，所以生成场景的三维模型后，只需在该三维模型中找出其对应的位置并将视频流采集装置采集的实时视频流视叠加在该位置进行持续播放，即可在三维模型内动态展示场景的3D视频流，克服了传统3D扫描建模技术得到的3D模型都是静态的缺陷，真正实现了场景的实景3D监控。用户通过三维模型对应的链接即可访问场景的三维模型以及在三维模型内动态播放或展示的3D视频流。本实施例利用这一方式真正实现3D模型与实时视频流的无缝融合，不 受场景的改变和角度的改变的影响，在场景改变或角度改变后仍能在3D模型中观看到该实时视频流。
进一步作为优选的实施方式，所述对场景的实时视频流进行智能识别这一步骤，具体包括：
S10110、根据给定的训练样本和标签采用人工智能的方法训练智能识别模型；
S10111、将场景的三维实景的实时视频流输入训练好的智能识别模型，得到智能识别结果。
具体地，本实施例给定的训练样本和标签可以由预先建立的数据库来提供，例如人脸识别模型则可以通过预先收集的各种不同类型的人脸放入人脸数据库，训练时可直接从该数据库中获取人脸作为样本，获取对应的人脸的类型或名称(如不同年龄、国家等的人脸)作为标签。本实施例应用人工智能的方法，可以训练出精确的智能识别模型，这样实时视频流中即使有新的标签未知的数据输入，也能自动利用该智能识别模型识别出来，效率高且更加智能。
进一步作为优选的实施方式，所述将场景的三维实景的实时视频流输入训练好的智能识别模型，得到智能识别结果这一步骤S10110，具体为：
将场景的三维实景的实时视频流输入训练好的智能识别模型，识别得到第一结果，所述第一结果包括对象的类型和名称、对象的属性以及对象的行为，所述对象的类型包括人、动物和物体，所述对象的属性包括颜色、款式、性别、年龄和型号。
具体地，本实施例通过对实时视频流的智能识别，可以精确识别出对象的类型和名称、对象的属性以及对象的行为等内容，有利于进一步的智能分析、后续的目标跟踪等操作。
例如，通过对实时视频流的智能识别，可以识别出视频内的猫猪狗等动物、车牌是什么牌照的车牌，车牌的颜色是什么，人是什么人(男女老少)，人穿什么衣服，车的款式型号是什么，人脸是什么人脸等等。
本实施例通过对实时视频流的智能识别，配合场景的三维模型的智能识别，可以识别出场景的三维实景内的一景一物，极大地方便了后续的目标跟踪、监控等应用。
进一步作为优选的实施方式，所述根据智能识别的结果从场景的三维实景中获取跟踪目标，并在场景的三维实景内对跟踪目标进行跟踪这一步骤S102，具体包括：
S1020、在场景的三维实景中确定并标注跟踪目标；
具体地，本实施例可以在完成三维场景的智能识别后，根据实际的跟踪需要将场景内的某个对象(如某个犯罪嫌疑人、某辆可疑的车辆等)通过添加矩形框、添加覆盖对象轮廓的 不规则图形等方式标注出来并锁定为跟踪目标。
S1021、在场景的三维实景内自动对标注后的跟踪目标进行跟踪，从而获得跟踪目标的运动路径或轨迹。
具体地，在对跟踪目标进行跟踪时，由于三维实景的视频流是实时且采集视频流的视频采集装置是通过局域网、互联网、工控网络等互联的，这样通过分析三维实景内多个视频采集装置的视频流即可自动得到跟踪目标的运动路径或轨迹，十分方便和高效。而且与现有目标跟踪方案只能提供2D视频监控画面不同的是，本实施例的目标跟踪方案在提供场景的实时视频流的同时也提供该视频流所在位置的三维模型(可360度无死角漫游)，真正实现了视频流+3D模型的3D实景跟踪，更加立体和真实。
进一步作为优选的实施方式，所述根据智能识别的结果从场景的三维实景中获取跟踪目标，并在场景的三维实景内对跟踪目标进行跟踪这一步骤S102，还具体包括：
S1023、在场景的三维实景内识别标注后的跟踪目标的运动姿态。
具体地，本实施例通过人工智能算法训练学习的识别模型，还可以识别跟踪目标的运动姿态(如人是跑步还是步行，人是否有拿东西或带墨镜等)，能提供更丰富的细节信息，有利于目标跟踪操作，更加智能和方便。
如图2所示，本发明实施例还提供了一种基于实景建模与智能识别的目标跟踪系统，包括以下模块：
实景三维建模模块201，用于获取场景的三维数据和场景的实时视频流并进行实景三维建模，得到场景的三维实景，所述场景的三维实景在场景的三维模型内展示场景的实时视频流；
智能识别模块202，用于对场景的三维实景进行智能识别，所述智能识别包括对场景的三维模型进行分割与智能识别以及对场景的实时视频流进行智能识别；
目标跟踪模块203，用于根据智能识别的结果从场景的三维实景中获取跟踪目标，并在场景的三维实景内对跟踪目标进行跟踪。
如图2所示，进一步作为优选的实施方式，所述实景三维建模模块201具体包括：
扫描单元2011，用于通过扫描设备扫描的方式获取场景的三维数据，所述扫描设备包括空间扫描仪、航拍扫描仪、物体扫描仪和人体扫描仪；
视频采集单元2012，用于通过视频采集设备获取场景的实时视频流；
三维重建单元2013，用于根据场景的三维数据采用人工智能的方法进行三维重建，生成场景的三维模型，其中，三维重建包括模型修复、剪辑、裁剪、减面、减模、压缩、处理 材质、处理贴图、处理灯光和压缩渲染；
链接生成单元2014，用于根据场景的三维模型生成对应的链接；
融合单元2015，用于将场景的实时视频流融合至场景的三维模型中，得到场景的三维实景。
上述方法实施例中的内容均适用于本系统实施例中，本系统实施例所具体实现的功能与上述方法实施例相同，并且达到的有益效果与上述方法实施例所达到的有益效果也相同。
如图3所示，本发明实施例还提供了一种基于实景建模与智能识别的目标跟踪系统，包括：
至少一个处理器301；
至少一个存储器302，用于存储至少一个程序；
当所述至少一个程序被所述至少一个处理器执行，使得所述至少一个处理器301实现本发明所述的基于实景建模与智能识别的目标跟踪方法。
上述方法实施例中的内容均适用于本系统实施例中，本系统实施例所具体实现的功能与上述方法实施例相同，并且达到的有益效果与上述方法实施例所达到的有益效果也相同。
本发明实施例还提供了一种介质，其中存储有处理器可执行的指令，所述处理器可执行的指令在由处理器执行时用于实现本发明所述的基于实景建模与智能识别的目标跟踪方法。
以上是对本发明的较佳实施进行了具体说明，但本发明并不限于所述实施例，熟悉本领域的技术人员在不违背本发明精神的前提下还可做作出种种的等同变形或替换，这些等同的变形或替换均包含在本申请权利要求所限定的范围内。

Claims (10)

  1. 基于实景建模与智能识别的目标跟踪方法，其特征在于：包括以下步骤：
    获取场景的三维数据和场景的实时视频流并进行实景三维建模，得到场景的三维实景，所述场景的三维实景在场景的三维模型内展示场景的实时视频流；
    对场景的三维实景进行智能识别，所述智能识别包括对场景的三维模型进行分割与智能识别以及对场景的实时视频流进行智能识别；
    根据智能识别的结果从场景的三维实景中获取跟踪目标，并在场景的三维实景内对跟踪目标进行跟踪。
  2. 根据权利要求1所述的基于实景建模与智能识别的目标跟踪方法，其特征在于：所述获取场景的三维数据和场景的实时视频流并进行实景三维建模，得到场景的三维实景这一步骤，具体包括：
    通过扫描设备扫描的方式获取场景的三维数据，所述扫描设备包括空间扫描仪、航拍扫描仪、物体扫描仪和人体扫描仪；
    通过视频采集设备获取场景的实时视频流；
    根据场景的三维数据采用人工智能的方法进行三维重建，生成场景的三维模型，其中，三维重建包括模型修复、剪辑、裁剪、减面、减模、压缩、处理材质、处理贴图、处理灯光和压缩渲染；
    根据场景的三维模型生成对应的链接；
    将场景的实时视频流融合至场景的三维模型中，得到场景的三维实景。
  3. 根据权利要求1所述的基于实景建模与智能识别的目标跟踪方法，其特征在于：所述对场景的实时视频流进行智能识别这一步骤，具体包括：
    根据给定的训练样本和标签采用人工智能的方法训练智能识别模型；
    将场景的三维实景的实时视频流输入训练好的智能识别模型，得到智能识别结果。
  4. 根据权利要求3所述的基于实景建模与智能识别的目标跟踪方法，其特征在于：所述将场景的三维实景的实时视频流输入训练好的智能识别模型，得到智能识别结果这一步骤，具体为：
    将场景的三维实景的实时视频流输入训练好的智能识别模型，识别得到第一结果，所述第一结果包括对象的类型和名称、对象的属性以及对象的行为，所述对象的类型包括人、动物和物体，所述对象的属性包括颜色、款式、性别、年龄和型号。
  5. 根据权利要求1所述的基于实景建模与智能识别的目标跟踪方法，其特征在于：所述根据智能识别的结果从场景的三维实景中获取跟踪目标，并在场景的三维实景内对跟踪目标进 行跟踪这一步骤，具体包括：
    在场景的三维实景中确定并标注跟踪目标；
    在场景的三维实景内自动对标注后的跟踪目标进行跟踪，从而获得跟踪目标的运动路径或轨迹。
  6. 根据权利要求5所述的基于实景建模与智能识别的目标跟踪方法，其特征在于：所述根据智能识别的结果从场景的三维实景中获取跟踪目标，并在场景的三维实景内对跟踪目标进行跟踪这一步骤，还具体包括：
    在场景的三维实景内识别标注后的跟踪目标的运动姿态。
  7. 基于实景建模与智能识别的目标跟踪系统，其特征在于：包括以下模块：
    实景三维建模模块，用于获取场景的三维数据和场景的实时视频流并进行实景三维建模，得到场景的三维实景，所述场景的三维实景在场景的三维模型内展示场景的实时视频流；
    智能识别模块，用于对场景的三维实景进行智能识别，所述智能识别包括对场景的三维模型进行分割与智能识别以及对场景的实时视频流进行智能识别；
    目标跟踪模块，用于根据智能识别的结果从场景的三维实景中获取跟踪目标，并在场景的三维实景内对跟踪目标进行跟踪。
  8. 根据权利要求7所述的基于实景建模与智能识别的目标跟踪系统，其特征在于：所述实景三维建模模块具体包括：
    扫描单元，用于通过扫描设备扫描的方式获取场景的三维数据，所述扫描设备包括空间扫描仪、航拍扫描仪、物体扫描仪和人体扫描仪；
    视频采集单元，用于通过视频采集设备获取场景的实时视频流；
    三维重建单元，用于根据场景的三维数据采用人工智能的方法进行三维重建，生成场景的三维模型，其中，三维重建包括模型修复、剪辑、裁剪、减面、减模、压缩、处理材质、处理贴图、处理灯光和压缩渲染；
    链接生成单元，用于根据场景的三维模型生成对应的链接；
    融合单元，用于将场景的实时视频流融合至场景的三维模型中，得到场景的三维实景。
  9. 基于实景建模与智能识别的目标跟踪系统，其特征在于：包括：
    至少一个处理器；
    至少一个存储器，用于存储至少一个程序；
    当所述至少一个程序被所述至少一个处理器执行，使得所述至少一个处理器实现如权利要求1-6任一项所述的基于实景建模与智能识别的目标跟踪方法。
  10. 介质，其中存储有处理器可执行的指令，其特征在于：所述处理器可执行的指令在由处理器执行时用于实现如权利要求1-6任一项所述的基于实景建模与智能识别的目标跟踪方法。
PCT/CN2020/090217 2019-05-14 2020-05-14 基于实景建模与智能识别的目标跟踪方法、系统及介质 WO2020228766A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910398620.8A CN110310306A (zh) 2019-05-14 2019-05-14 基于实景建模与智能识别的目标跟踪方法、系统及介质
CN201910398620.8 2019-05-14

Publications (1)

Publication Number Publication Date
WO2020228766A1 true WO2020228766A1 (zh) 2020-11-19

Family

ID=68075438

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/090217 WO2020228766A1 (zh) 2019-05-14 2020-05-14 基于实景建模与智能识别的目标跟踪方法、系统及介质

Country Status (2)

Country Link
CN (1) CN110310306A (zh)
WO (1) WO2020228766A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP4148691A1 (en) * 2021-09-10 2023-03-15 Milestone Systems A/S A method of training a machine learning algorithm to identify objects or activities in video surveillance data

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110310306A (zh) * 2019-05-14 2019-10-08 广东康云科技有限公司 基于实景建模与智能识别的目标跟踪方法、系统及介质
CN111080799A (zh) * 2019-12-04 2020-04-28 广东康云科技有限公司 基于三维建模的场景漫游方法、系统、装置和存储介质
CN112985432B (zh) * 2019-12-16 2022-02-15 湖北亿咖通科技有限公司 车辆导航方法、装置、电子设备及存储介质
CN111445574B (zh) * 2020-03-26 2023-07-07 众趣（北京）科技有限公司 一种视频监控设备布控方法、装置及系统
CN111696193B (zh) * 2020-05-06 2023-08-25 广东康云科技有限公司 基于三维场景的物联网控制方法、系统、装置及存储介质
CN111680402B (zh) * 2020-05-19 2023-07-07 广东交通职业技术学院 一种屏蔽门故障可视化方法、系统、装置及存储介质
CN112182286B (zh) * 2020-09-04 2022-11-18 中国电子科技集团公司电子科学研究院 一种基于三维实景地图的智能视频管控方法
CN112037264B (zh) * 2020-11-03 2021-02-05 浙江清鹤科技有限公司 一种视频融合系统及方法
CN113033412B (zh) * 2021-03-26 2023-08-22 平安国际智慧城市科技股份有限公司 基于实景融合的交通态势监控方法、装置、设备及介质
CN113256801A (zh) * 2021-06-11 2021-08-13 山东能之源核电科技有限公司 一种辐射场三维逆向模型可视化系统
CN116402984B (zh) * 2023-02-28 2024-04-16 神力视界(深圳)文化科技有限公司 三维模型处理方法、装置及电子设备

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040105573A1 (en) * 2002-10-15 2004-06-03 Ulrich Neumann Augmented virtual environments
CN101951502A (zh) * 2010-10-19 2011-01-19 北京硅盾安全技术有限公司 一种三维智能视频监控方法
CN103500452A (zh) * 2013-10-12 2014-01-08 杭州师范大学 基于空间关系和图像分析的景区景物移动增强现实方法
CN104021586A (zh) * 2014-05-05 2014-09-03 深圳市城市管理监督指挥中心 基于北斗定位的空地一体城市生态文明管理系统及方法
CN109889777A (zh) * 2019-01-23 2019-06-14 广东康云科技有限公司 3d实景视觉监控的切换展示方法及系统
CN110310306A (zh) * 2019-05-14 2019-10-08 广东康云科技有限公司 基于实景建模与智能识别的目标跟踪方法、系统及介质

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101646067B (zh) * 2009-05-26 2011-06-15 华中师范大学 数字化全空间智能监控系统及方法
CN106919895B (zh) * 2016-07-01 2020-03-27 湖南拓视觉信息技术有限公司 用于运动目标的跟踪方法和系统
US20190120965A1 (en) * 2017-10-25 2019-04-25 Bae Systems Information And Electronic Systems Integration Inc. Method and system of digital light processing and light detection and ranging for guided autonomous vehicles
CN109085966B (zh) * 2018-06-15 2020-09-08 广东康云多维视觉智能科技有限公司 一种基于云计算的三维展示系统及方法
CN109345510A (zh) * 2018-09-07 2019-02-15 百度在线网络技术（北京）有限公司 物体检测方法、装置、设备、存储介质及车辆

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040105573A1 (en) * 2002-10-15 2004-06-03 Ulrich Neumann Augmented virtual environments
CN101951502A (zh) * 2010-10-19 2011-01-19 北京硅盾安全技术有限公司 一种三维智能视频监控方法
CN103500452A (zh) * 2013-10-12 2014-01-08 杭州师范大学 基于空间关系和图像分析的景区景物移动增强现实方法
CN104021586A (zh) * 2014-05-05 2014-09-03 深圳市城市管理监督指挥中心 基于北斗定位的空地一体城市生态文明管理系统及方法
CN109889777A (zh) * 2019-01-23 2019-06-14 广东康云科技有限公司 3d实景视觉监控的切换展示方法及系统
CN110310306A (zh) * 2019-05-14 2019-10-08 广东康云科技有限公司 基于实景建模与智能识别的目标跟踪方法、系统及介质

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP4148691A1 (en) * 2021-09-10 2023-03-15 Milestone Systems A/S A method of training a machine learning algorithm to identify objects or activities in video surveillance data
AU2022215283B2 (en) * 2021-09-10 2023-12-21 Milestone Systems A/S A method of training a machine learning algorithm to identify objects or activities in video surveillance data

Also Published As

Publication number Publication date
CN110310306A (zh) 2019-10-08

Similar Documents

Publication Publication Date Title
WO2020228766A1 (zh) 基于实景建模与智能识别的目标跟踪方法、系统及介质
Sahu et al. Artificial intelligence (AI) in augmented reality (AR)-assisted manufacturing applications: a review
Gilbert et al. Fusing visual and inertial sensors with semantics for 3d human pose estimation
US20180012411A1 (en) Augmented Reality Methods and Devices
GB2583676A (en) Augmenting detected regions in image or video data
CN101542538B (zh) 用来对光进行建模的方法与系统
CN112053446A (zh) 一种基于三维gis的实时监控视频与三维场景融合方法
CN110210276A (zh) 一种移动轨迹获取方法及其设备、存储介质、终端
WO2020228767A1 (zh) 基于视频融合的场景动态模拟方法、系统及存储介质
WO2021190321A1 (zh) 图像处理方法和装置
Bešić et al. Dynamic object removal and spatio-temporal RGB-D inpainting via geometry-aware adversarial learning
WO2020211427A1 (zh) 基于扫描点云数据的分割与识别方法、系统及存储介质
CN112419233B (zh) 一种数据标注方法、装置、设备以及计算机可读存储介质
Elhayek et al. Fully automatic multi-person human motion capture for vr applications
Shalaby et al. Algorithms and applications of structure from motion (SFM): A survey
Dai et al. Indoor 3D human trajectory reconstruction using surveillance camera videos and point clouds
CN113065506A (zh) 一种人体姿态识别方法及系统
CN117392289A (zh) 基于ai语音自动生成案发现场视频的方法和系统
CN117115917A (zh) 基于多模态特征融合的教师行为识别方法、设备以及介质
Shah Guest introduction: the changing shape of computer vision in the twenty-first century
Liu et al. Estimation of sunlight direction using 3D object models
Zoellner et al. Reality Filtering: A Visual Time Machine in Augmented Reality.
US11461956B2 (en) 3D representation reconstruction from images using volumic probability data
Kosmopoulos et al. Vision-based production of personalized video
McClean An Augmented Reality System for Urban Environments using a Planar Building Fa cade Model

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: 20805858

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20805858

Country of ref document: EP

Kind code of ref document: A1