WO2021135933A1 - 一种目标的识别方法及装置、存储介质及电子装置 - Google Patents

一种目标的识别方法及装置、存储介质及电子装置 Download PDF

Info

Publication number
WO2021135933A1
WO2021135933A1 PCT/CN2020/136593 CN2020136593W WO2021135933A1 WO 2021135933 A1 WO2021135933 A1 WO 2021135933A1 CN 2020136593 W CN2020136593 W CN 2020136593W WO 2021135933 A1 WO2021135933 A1 WO 2021135933A1
Authority
WO
WIPO (PCT)
Prior art keywords
target
target image
image
recognition
information
Prior art date
Application number
PCT/CN2020/136593
Other languages
English (en)
French (fr)
Inventor
谈笑
Original Assignee
中兴通讯股份有限公司
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 中兴通讯股份有限公司 filed Critical 中兴通讯股份有限公司
Publication of WO2021135933A1 publication Critical patent/WO2021135933A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/172—Classification, e.g. identification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures

Definitions

  • the present disclosure relates to the field of communications, and in particular, to a method and device for identifying a target, a storage medium, and an electronic device.
  • the security field is paying more and more attention to video surveillance through video analysis platforms to assist public security in handling cases.
  • the video analysis platform relies on the algorithm model obtained through deep learning training to capture faces of people passing through the camera, and the results of these captures can effectively become evidence for locating suspicious persons.
  • the video analysis platform will still recognize human faces and show them to the operator as a result. These data lead to a low recognition rate, and these data are also of no value. Data, resulting in a waste of resources.
  • the embodiments of the present disclosure provide a target recognition method and device, a storage medium, and an electronic device, so as to at least solve the problem of low target recognition rate in the related art.
  • a target recognition method which includes: performing target recognition on an image to be recognized to obtain a target image; recognizing the target image according to non-target information to determine whether the target image is The target is included, wherein the non-target information includes non-target information in a non-target image that is incorrectly recognized in a historical target image obtained by historical recognition.
  • a target recognition device including: a first recognition module configured to perform target recognition on an image to be recognized to obtain a target image; and a second recognition module configured to perform target recognition based on non-target information Recognizing the target image to determine whether the target image includes the target, wherein the non-target information includes information about the non-target in the non-target image that was incorrectly recognized in the historical target image obtained by historical recognition .
  • a computer-readable storage medium in which a computer program is stored, wherein the computer program is configured to execute any of the foregoing when running. The steps in the method embodiment.
  • an electronic device including a memory and a processor, the memory is stored with a computer program, and the processor is configured to run the computer program to execute any of the above Steps in the method embodiment.
  • the target image is obtained by performing target recognition on the image to be recognized; the target image is recognized according to non-target information to determine whether the target image includes the target, wherein the non-target information Including the non-target information in the incorrectly recognized non-target image in the historical target image obtained by the historical recognition, therefore, the problem of low target recognition rate in related technologies can be solved, and the effect of improving the accuracy of target recognition can be achieved.
  • FIG. 1 is a block diagram of the hardware structure of a computer terminal of a target identification method according to an embodiment of the present disclosure
  • Fig. 2 is a flowchart of a target recognition method according to an embodiment of the present disclosure
  • Fig. 3 is a structural block diagram of a target recognition device according to an embodiment of the present disclosure.
  • FIG. 4 is a schematic flowchart of a face recognition method according to an optional embodiment of the present disclosure.
  • Fig. 5 is a schematic diagram of a non-face labeling process according to an optional embodiment of the present disclosure.
  • FIG. 1 is a hardware structural block diagram of a server in a method for identifying a target according to an embodiment of the present disclosure.
  • the server 10 may include one or more (only one is shown in FIG. 1) processor 102 (the processor 102 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA) And the memory 104 for storing data.
  • the above-mentioned server may also include a transmission device 106 and an input/output device 108 for communication functions.
  • FIG. 1 is only for illustration, and it does not limit the structure of the foregoing server.
  • the server 10 may also include more or fewer components than shown in FIG. 1, or have a different configuration from that shown in FIG.
  • the memory 104 may be used to store computer programs, for example, software programs and modules of application software, such as the computer programs corresponding to the target identification method in the embodiments of the present disclosure.
  • the processor 102 executes the computer programs stored in the memory 104 by running Various functional applications and data processing, that is, to achieve the above methods.
  • the memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory.
  • the storage 104 may further include storage remotely provided with respect to the processor 102, and these remote storages may be connected to the server 10 through a network. Examples of the aforementioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.
  • the transmission device 106 is used to receive or send data via a network.
  • the above-mentioned specific examples of the network may include a wireless network provided by the communication provider of the server 10.
  • the transmission device 106 includes a network adapter (Network Interface Controller, NIC for short), which can be connected to other network devices through a base station to communicate with the Internet.
  • the transmission device 106 may be a radio frequency (Radio Frequency, referred to as RF) module, which is used to communicate with the Internet in a wireless manner.
  • RF Radio Frequency
  • FIG. 2 is a flowchart of the method for identifying a target according to an embodiment of the present disclosure. As shown in FIG. 2, the process includes the following steps:
  • Step S202 performing target recognition on the image to be recognized to obtain a target image
  • Step S204 Recognizing the target image according to the non-target information, and determining whether the target image includes the target, wherein the non-target information includes the non-target image in the historical target image that is incorrectly recognized in the historical target image obtained by historical recognition. Target information.
  • the target image is obtained due to the target recognition of the image to be recognized; the target image is recognized according to the non-target information, and it is determined whether the target image includes the target, wherein the non-target information includes history Recognizing the non-target information in the incorrectly recognized non-target images in the obtained historical target images, therefore, the problem of low target recognition rate in related technologies can be solved, and the effect of improving the accuracy of target recognition can be achieved.
  • the method before recognizing the target image according to the non-target information, the method further includes: extracting the image feature of the non-target in the non-target image that was incorrectly recognized in the historical target image obtained by the historical recognition as The non-target information.
  • the identifying the target image according to the non-target information and determining whether the target image includes the target includes: matching the target image with the non-target information; and determining whether the target image is contained in the target image according to the matching result Include this goal.
  • matching the target image with the non-target information includes: determining the image feature of the target image; matching the image feature of the target image with the non-target information to obtain a matching result.
  • the determining whether the target image is included in the target image according to the matching result includes: the matching result indicates that the image feature in the target image is higher than the non-target image feature in the non-target information.
  • the matching result indicates that the image feature in the target image has a low similarity with the image feature of the non-target in the non-target information
  • the preset threshold it is determined that the target image includes the target.
  • identifying the target image according to the non-target information and determining whether the target image includes the target includes: analyzing the target image using a first model to obtain an analysis result, wherein the analysis result is used for Indicates whether the target is included in the target image, where the first model is obtained by training a deep neural network using multiple sets of data, and the multiple sets of data include: multiple sample non-target images, and, the multiple samples Non-target information in the non-target image.
  • the method further includes: displaying the target image obtained after the determination.
  • a device for identifying a target is also provided, which is used to implement the above-mentioned embodiments and preferred implementations, and those that have been described will not be repeated.
  • the term "module” can implement a combination of software and/or hardware with predetermined functions.
  • the devices described in the following embodiments are preferably implemented by software, implementation by hardware or a combination of software and hardware is also possible and conceived.
  • Fig. 3 is a structural block diagram of a device for identifying a target according to an embodiment of the present disclosure. As shown in Fig. 3, the device includes:
  • the first recognition module 31 is configured to perform target recognition on the image to be recognized to obtain the target image
  • the second recognition module 33 is configured to recognize the target image according to the non-target information, and determine whether the target image is included in the target image, wherein the non-target information includes the incorrectly recognized non-target image in the historical target image obtained by historical recognition. The non-target information in the target image.
  • the target image is obtained due to the target recognition of the image to be recognized; the target image is recognized according to the non-target information to determine whether the target image is included in the target image, wherein the non-target information includes the historical recognition result
  • the non-target information in the non-target image that is incorrectly recognized in the historical target image therefore, the problem of low target recognition rate in related technologies can be solved, and the effect of improving the accuracy of target recognition can be achieved.
  • the device further includes: an extraction module, configured to extract a non-target image in the non-target image that was incorrectly recognized in the historical target image obtained by the historical recognition before recognizing the target image according to the non-target information.
  • the image feature of the target serves as the non-target information.
  • the second recognition module 33 includes: a matching sub-module configured to match the target image with the non-target information; and a determining sub-module configured to determine whether the target image includes the target according to the matching result.
  • the matching submodule includes: a determining subunit configured to determine the image feature of the target image; and a matching subunit configured to match the image feature of the target image with the non-target information to obtain a matching result.
  • the matching result indicates that the similarity between the image feature in the target image and the image feature of the non-target in the non-target information is higher than a preset threshold
  • it is determined that the target image does not include The target; or, in the case where the matching result indicates that the image feature in the target image and the non-target image feature similarity in the non-target information are lower than the preset threshold, it is determined that the target image Include this goal.
  • the second recognition module 33 includes: an analysis sub-module configured to analyze the target image using the first model to obtain an analysis result, wherein the analysis result is used to indicate whether the target image includes the Target, wherein the first model is obtained by training a deep neural network using multiple sets of data, each of the multiple sets of data includes: multiple sample non-target images, and non-target images in the multiple sample non-target images information.
  • the device further includes: a display module configured to display the target image obtained after the determination after recognizing the target image according to the non-target information and determining whether the target image includes the target.
  • a display module configured to display the target image obtained after the determination after recognizing the target image according to the non-target information and determining whether the target image includes the target.
  • each of the above modules can be implemented by software or hardware.
  • it can be implemented in the following manner, but not limited to this: the above modules are all located in the same processor; or, the above modules can be combined in any combination.
  • the forms are located in different processors.
  • the embodiment of the present disclosure is a business method for generating a non-face feature library through annotation to reduce the false detection rate of face analysis tasks, mainly for security, face recognition and other scenarios, adjusting the false detection rate of the video analysis platform, Improve system performance; this disclosure particularly relates to the field of artificial intelligence (AI).
  • AI artificial intelligence
  • the following further describes the present disclosure in combination with specific face recognition scenarios. It should be noted that the solutions provided by the embodiments of the present disclosure are not limited to scenarios for face recognition, and may also be scenarios for recognizing other targets.
  • a video analysis platform capable of recognizing human faces with a supervision function can be constructed, and the non-face images can be filtered according to the labeled non-face feature library.
  • the basic principle of the business process of manually labeling the non-face feature database is to judge and label the non-face content. For example, it can be manually labeled, and its accuracy is higher than that of a video analysis platform based on a deep learning algorithm. In view of the high similarity between the features of non-face pictures, it can be used as a classification condition to distinguish between faces and non-faces.
  • a link is added to the business layer, that is, to use the comparison with the non-face feature library to realize the filtering of the non-face.
  • the analysis platform does not have a non-face feature database, and all the results of the adult face recognized by the algorithm (whether or not Is a human face), will be displayed on the front-end page as a result.
  • the recognition result can be marked at the front end, for example, the recognition result picture can be marked by means of interface buttons. Specifically, the picture (that is, the picture that has been recognized incorrectly) in the result of recognizing the human face can be marked. (Excluding a picture of a human face), you can also mark the content of the picture that is suspected of being recognized as a human face in the picture with the recognition error.
  • the annotated picture is recorded as a non-face picture, and its features will be retained in the corresponding local directory of the server to form a non-face feature library; then, if there is a non-face feature library, every subsequent A snapshot result will be compared with the features of the non-face feature library in the background.
  • the video analysis platform will perform a second time It is determined that this picture is a non-face picture, and the face recognition result is not displayed on the front-end page.
  • the operator checks the face capture result of the analysis task, manually judges that it is a non-face picture, and marks the picture as a non-face by clicking a button; the background receives a marking request and saves the characteristics of the picture To the server, the save path is configurable; the subsequent judgment on the pictures captured by the analysis task is to compare the characteristics of the pictures captured by the analyzed characters and the previously marked non-face pictures. When the result of the similarity comparison is greater than When a certain configurable threshold is reached, the system determines that the captured image is also a non-face image, and discards it so that it will not be displayed on the front-end page.
  • the business process of using the non-face feature database for secondary comparison can reduce the false detection rate of the face recognition system; the use of manual labeling can continuously optimize the non-face feature database and further reduce the face Identify the false detection rate of the system.
  • Fig. 4 is a schematic flowchart of a face recognition method according to an optional embodiment of the present disclosure, as shown in Fig. 4, including:
  • the video analysis platform obtains the video stream from the camera point, performs face recognition through the algorithm model, and obtains the recognition result;
  • Fig. 5 is a schematic diagram of a non-face labeling process according to an optional embodiment of the present disclosure, as shown in Fig. 5, including:
  • the video analysis platform obtains the video stream from the camera point, performs face recognition through the algorithm model, and obtains the recognition result;
  • this embodiment can optimize the business process of the VAP video analysis platform, and is mainly used in the field of security monitoring.
  • Front end The front end needs to add an annotation button to send a request to tell the back-end system that this picture needs to be annotated, that is, it is a non-human face; optionally, a separate first-level menu or new
  • the web page of is specially used for labeling management, that is, unified management of all non-face images labelled on the platform; this page needs to have the following functions: image source for query, undo labeling, and export.
  • the source of the picture can be the camera point. It is an effective method to classify pictures by the point position, which is conducive to subsequent queries; for some labeled samples that may be caused by misoperation, the picture is allowed to be restored; export can Export these pictures or selected pictures to the local in the form of a compressed package.
  • the back-end system needs to provide corresponding interfaces, such as an annotation interface, and all related interfaces of an annotation management page.
  • interfaces such as an annotation interface
  • all related interfaces of an annotation management page it is necessary to add an additional process before the front-end display after the face capture is generated, that is, the process of feature comparison with the non-face feature library, and the existing comparison interface needs to be called.
  • the result of the comparison exceeds the threshold, and the picture needs to be moved (to the label management page) instead of displaying the picture at the result of the analysis task.
  • Algorithm system The back-end system saves the features of the marked pictures, which need to be saved in the algorithm system, and the feature files are loaded into the memory at the same time. For the request to cancel the annotation, the signature file needs to be deleted synchronously and updated to the memory at the same time.
  • This embodiment can be applied to a scene where a fixed camera is used for face capture business display at an exhibition site, and the face recognition technology of a video analysis platform is displayed by using a fixed camera on the spot. Due to the complex situation on the scene, the layout of the cameras cannot be very standardized, resulting in a lot of non-face pictures displayed as the recognition result of the system when the scene is actually displayed. Now using the process of non-face feature library comparison, the on-site non-face captured pictures can be labeled manually by means of manual annotation, so that the system can generate a non-face feature library based on the exhibition site, such as a chair at the exhibition site, etc. , So that after labeling, this kind of non-face capture situation is no longer generated, and the recognition effect is improved.
  • This embodiment can also be applied to a scene where a non-face feature database is pre-deployed into the system in a video surveillance scene, so as to achieve the effect of reducing the false detection rate of the face.
  • a public security bureau as an example.
  • the non-face feature library collected in the previous test can be added at the same time, and the non-face feature library and the captured image can be compared twice.
  • the method can filter out most of the non-face images.
  • the non-face feature library can be customized and optimized for the data of the site camera by manual labeling, so as to further reduce the false detection rate of face recognition.
  • the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, of course, it can also be implemented by hardware, but in many cases the former is Better implementation.
  • the technical solution of the present disclosure essentially or the part that contributes to the existing technology can be embodied in the form of a software product, and the computer software product is stored in a storage medium (such as ROM/RAM, magnetic disk, The optical disc) includes several instructions to enable a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present disclosure.
  • the embodiment of the present disclosure also provides a computer-readable storage medium, and a computer program is stored in the computer-readable storage medium, wherein the computer program is configured to execute any of the above method embodiments when running. step.
  • the foregoing computer-readable storage medium may be configured to store a computer program for executing the following steps:
  • S1 Perform target recognition on the image to be recognized to obtain the target image
  • S2 Recognizing the target image according to the non-target information, and determining whether the target image includes the target, wherein the non-target information includes the non-target in the incorrectly recognized non-target image in the historical target image obtained by historical recognition Information.
  • the target image is obtained due to the target recognition of the image to be recognized; the target image is recognized according to the non-target information, and it is determined whether the target image includes the target, wherein the non-target information includes history Recognizing the non-target information in the incorrectly recognized non-target images in the obtained historical target images, therefore, the problem of low target recognition rate in related technologies can be solved, and the effect of improving the accuracy of target recognition can be achieved.
  • the above-mentioned storage medium may include but is not limited to: U disk, Read-Only Memory (Read-Only Memory, ROM for short), Random Access Memory (Random Access Memory, RAM for short), Various media that can store computer programs, such as mobile hard disks, magnetic disks, or optical disks.
  • An embodiment of the present disclosure also provides an electronic device, including a memory and a processor, the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any of the foregoing method embodiments.
  • the aforementioned electronic device may further include a transmission device and an input-output device, wherein the transmission device is connected to the aforementioned processor, and the input-output device is connected to the aforementioned processor.
  • the foregoing processor may be configured to execute the following steps through a computer program:
  • S1 Perform target recognition on the image to be recognized to obtain the target image
  • S2 Recognizing the target image according to the non-target information, and determining whether the target image includes the target, wherein the non-target information includes the non-target in the incorrectly recognized non-target image in the historical target image obtained by historical recognition Information.
  • the target image is obtained due to the target recognition of the image to be recognized; the target image is recognized according to the non-target information, and it is determined whether the target image includes the target, wherein the non-target information includes history Recognizing the non-target information in the incorrectly recognized non-target images in the obtained historical target images, therefore, the problem of low target recognition rate in related technologies can be solved, and the effect of improving the accuracy of target recognition can be achieved.
  • modules or steps of the present disclosure can be implemented by a general computing device, and they can be concentrated on a single computing device or distributed in a network composed of multiple computing devices.
  • they can be implemented with program codes executable by a computing device, so that they can be stored in a storage device for execution by the computing device, and in some cases, can be executed in a different order than here.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Human Computer Interaction (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)

Abstract

一种目标的识别方法及装置、存储介质及电子装置，目标的识别方法包括：对待识别的图像进行目标识别，得到目标图像（S202）；根据非目标信息对所述目标图像进行识别，确定所述目标图像中是否包括所述目标，其中，所述非目标信息包括历史识别所得到的历史目标图像中被错误识别的非目标图像中的非目标的信息（S204）。

Description

一种目标的识别方法及装置、存储介质及电子装置 技术领域
本公开涉及通信领域，具体而言，涉及一种目标的识别方法及装置、存储介质及电子装置。
背景技术
在人工智能的大背景下，很多领域利用图像识别手段达到相关的目的，例如安防领域越来越重视通过视频分析平台来进行视频监控辅助公安办案。视频分析平台依赖于通过深度学习训练得到的算法模型，对通过摄像头的人群进行人脸抓拍，这些抓拍的结果能够有效的成为定位可疑人员的证据。然而，受限于算法的准确性，对于一些非人脸的物体，视频分析平台仍然会将其识别成人脸，作为结果展示给操作员，这些数据导致识别率较低，并且这些数据也是没有价值的数据，造成资源的浪费。
针对相关技术中目标识别率较低的问题，尚不存在较好的解决方案。
发明内容
本公开实施例提供了一种目标的识别方法及装置、存储介质及电子装置，以至少解决相关技术中目标识别率较低的问题。
根据本公开的一个实施例，提供了一种目标的识别方法，包括：对待识别的图像进行目标识别，得到目标图像；根据非目标信息对所述目标图像进行识别，确定所述目标图像中是否包括所述目标，其中，所述非目标信息包括历史识别所得到的历史目标图像中被错误识别的非目标图像中的非目标的信息。
根据本公开的另一个实施例，提供了一种目标的识别装置，包括：第一识别模块，设置为对待识别的图像进行目标识别，得到目标图像；第二识别模块，设置为根据非目标信息对所述目标图像进行识别，确定所述目标图像中是否包括所述目标，其中，所述非目标信息包括历史识别所得到 的历史目标图像中被错误识别的非目标图像中的非目标的信息。
根据本公开的又一个实施例，还提供了一种计算机可读的存储介质，所述计算机可读的存储介质中存储有计算机程序，其中，所述计算机程序被设置为运行时执行上述任一项方法实施例中的步骤。
根据本公开的又一个实施例，还提供了一种电子装置，包括存储器和处理器，所述存储器中存储有计算机程序，所述处理器被设置为运行所述计算机程序以执行上述任一项方法实施例中的步骤。
通过本公开实施例，由于对待识别的图像进行目标识别，得到目标图像；根据非目标信息对所述目标图像进行识别，确定所述目标图像中是否包括所述目标，其中，所述非目标信息包括历史识别所得到的历史目标图像中被错误识别的非目标图像中的非目标的信息，因此，可以解决相关技术中目标识别率较低的问题，达到提高目标识别准确率的效果。
附图说明
此处所说明的附图用来提供对本公开的进一步理解，构成本申请的一部分，本公开的示意性实施例及其说明用于解释本公开，并不构成对本公开的不当限定。在附图中：
图1是本公开实施例的一种目标的识别方法的计算机终端的硬件结构框图；
图2是根据本公开实施例的目标的识别方法的流程图；
图3是根据本公开实施例的目标的识别装置的结构框图；
图4是根据本公开可选实施方式的人脸识别方法的流程示意图；
图5是根据本公开可选实施方式的非人脸标注的流程示意图。
具体实施方式
下文中将参考附图并结合实施例来详细说明本公开。需要说明的是，在不冲突的情况下，本申请中的实施例及实施例中的特征可以相互组合。
需要说明的是，本公开的说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。
实施例1
本申请实施例一所提供的方法实施例可以在服务器、服务器或者类似的运算装置中执行。以运行在服务器上为例，图1是本公开实施例的一种目标的识别方法的服务器的硬件结构框图。如图1所示，服务器10可以包括一个或多个(图1中仅示出一个)处理器102(处理器102可以包括但不限于微处理器MCU或可编程逻辑器件FPGA等的处理装置)和用于存储数据的存储器104，可选地，上述服务器还可以包括用于通信功能的传输设备106以及输入输出设备108。本领域普通技术人员可以理解，图1所示的结构仅为示意，其并不对上述服务器的结构造成限定。例如，服务器10还可包括比图1中所示更多或者更少的组件，或者具有与图1所示不同的配置。
存储器104可用于存储计算机程序，例如，应用软件的软件程序以及模块，如本公开实施例中的目标的识别方法对应的计算机程序，处理器102通过运行存储在存储器104内的计算机程序，从而执行各种功能应用以及数据处理，即实现上述的方法。存储器104可包括高速随机存储器，还可包括非易失性存储器，如一个或者多个磁性存储装置、闪存、或者其他非易失性固态存储器。在一些实例中，存储器104可进一步包括相对于处理器102远程设置的存储器，这些远程存储器可以通过网络连接至服务器10。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。
传输装置106用于经由一个网络接收或者发送数据。上述的网络具体实例可包括服务器10的通信供应商提供的无线网络。在一个实例中，传输装置106包括一个网络适配器(Network Interface Controller，简称为NIC)，其可通过基站与其他网络设备相连从而可与互联网进行通讯。在一个实例 中，传输装置106可以为射频(Radio Frequency，简称为RF)模块，其用于通过无线方式与互联网进行通讯。
在本实施例中提供了一种运行于上述运算装置的目标的识别方法，图2是根据本公开实施例的目标的识别方法的流程图，如图2所示，该流程包括如下步骤：
步骤S202，对待识别的图像进行目标识别，得到目标图像；
步骤S204，根据非目标信息对该目标图像进行识别，确定该目标图像中是否包括该目标，其中，该非目标信息包括历史识别所得到的历史目标图像中被错误识别的非目标图像中的非目标的信息。
通过上述步骤，由于对待识别的图像进行目标识别，得到目标图像；根据非目标信息对所述目标图像进行识别，确定所述目标图像中是否包括所述目标，其中，所述非目标信息包括历史识别所得到的历史目标图像中被错误识别的非目标图像中的非目标的信息，因此，可以解决相关技术中目标识别率较低的问题，达到提高目标识别准确率的效果。
可选地，在根据该非目标信息对该目标图像进行识别之前，该方法还包括：提取该历史识别所得到的历史目标图像中被错误识别的该非目标图像中的非目标的图像特征作为该非目标信息。
可选地，该根据该非目标信息对该目标图像进行识别，确定该目标图像中是否包括该目标，包括：将该目标图像与该非目标信息进行匹配；根据匹配结果确定该目标图像中是否包括该目标。
可选地，该将该目标图像与该非目标信息进行匹配，包括：确定该目标图像的图像特征；将该目标图像的图像特征与该非目标信息进行匹配，得到匹配结果。
可选地，该根据匹配结果确定该目标图像中是否包括该目标，包括：在该匹配结果指示了该目标图像中的图像特征与该非目标信息中的该非目标的图像特征相似度高于预先设定的阈值的情况下，确定该目标图像中不包括该目标；或者，在该匹配结果指示了该目标图像中的图像特征与该 非目标信息中的该非目标的图像特征相似度低于预先设定的该阈值的情况下，确定该目标图像中包括该目标。
可选地，根据该非目标信息对该目标图像进行识别，确定该目标图像中是否包括该目标，包括：使用第一模型对该目标图像进行分析，得到分析结果，其中，该分析结果用于指示该目标图像中是否包括该目标，其中，该第一模型为使用多组数据对深度神经网络进行训练得到的，该多组数据均包括：多个样本非目标图像，以及，该多个样本非目标图像中的非目标的信息。
可选地，在该根据非目标信息对该目标图像进行识别，确定该目标图像中是否包括该目标之后，该方法还包括：显示确定后所得到的目标图像。
在本实施例中还提供了一种目标的识别装置，该装置用于实现上述实施例及优选实施方式，已经进行过说明的不再赘述。如以下所使用的，术语“模块”可以实现预定功能的软件和/或硬件的组合。尽管以下实施例所描述的装置较佳地以软件来实现，但是硬件，或者软件和硬件的组合的实现也是可能并被构想的。
图3是根据本公开实施例的目标的识别装置的结构框图，如图3所示，该装置包括：
第一识别模块31，设置为对待识别的图像进行目标识别，得到目标图像；
第二识别模块33，设置为根据非目标信息对该目标图像进行识别，确定该目标图像中是否包括该目标，其中，该非目标信息包括历史识别所得到的历史目标图像中被错误识别的非目标图像中的非目标的信息。
通过上述模块，由于对待识别的图像进行目标识别，得到目标图像；根据非目标信息对该目标图像进行识别，确定该目标图像中是否包括该目标，其中，该非目标信息包括历史识别所得到的历史目标图像中被错误识别的非目标图像中的非目标的信息，因此，可以解决相关技术中目标识别率较低的问题，达到提高目标识别准确率的效果。
可选地，该装置还包括：提取模块，设置为在根据该非目标信息对该目标图像进行识别之前，提取该历史识别所得到的历史目标图像中被错误识别的该非目标图像中的非目标的图像特征作为该非目标信息。
可选地，第二识别模块33，包括：匹配子模块，设置为将该目标图像与该非目标信息进行匹配；确定子模块，设置为根据匹配结果确定该目标图像中是否包括该目标。
可选地，匹配子模块，包括：确定子单元，设置为确定该目标图像的图像特征；匹配子单元，设置为将该目标图像的图像特征与该非目标信息进行匹配，得到匹配结果。
可选地，在该匹配结果指示了该目标图像中的图像特征与该非目标信息中的该非目标的图像特征相似度高于预先设定的阈值的情况下，确定该目标图像中不包括该目标；或者，在该匹配结果指示了该目标图像中的图像特征与该非目标信息中的该非目标的图像特征相似度低于预先设定的该阈值的情况下，确定该目标图像中包括该目标。
可选地，所述第二识别模块33，包括：分析子模块，设置为使用第一模型对该目标图像进行分析，得到分析结果，其中，该分析结果用于指示该目标图像中是否包括该目标，其中，该第一模型为使用多组数据对深度神经网络进行训练得到的，该多组数据均包括：多个样本非目标图像，以及，该多个样本非目标图像中的非目标的信息。
可选地，该装置还包括：显示模块，设置为在该根据非目标信息对该目标图像进行识别，确定该目标图像中是否包括该目标之后，显示确定后所得到的目标图像。
需要说明的是，上述各个模块是可以通过软件或硬件来实现的，对于后者，可以通过以下方式实现，但不限于此：上述模块均位于同一处理器中；或者，上述各个模块以任意组合的形式分别位于不同的处理器中。
可选实施方式
本公开实施例是一种通过标注生成非人脸特征库从而降低人脸分析 任务抓拍误检率的业务手段，主要面向安防、人脸识别等场景，对视频分析平台的误检率进行调整，提升系统性能；本公开尤其涉及人工智能(AI)领域。以下结合具体的人脸识别场景对本公开进行进一步说明，需要说明的是，本公开实施例所提供的方案也不限于人脸识别的场景，也可以是识别其他目标的场景。
在本实施方式中，可以构建一个有监督功能的能够识别人脸的视频分析平台，根据标注的非人脸特征库来实现对非人脸图片的过滤。
本实施方式中，人工标注非人脸特征库的业务流程，其基本原理是对非人脸内容进行判断和标注，例如可以人工标注，其准确性高于基于深度学习算法的视频分析平台。鉴于非人脸图片的特征彼此之间存在很高的相似度，可以作为区分人脸和非人脸的分类条件。在本实施方式中，在业务层增加一个环节，即利用和非人脸特征库的比对实现非人脸的过滤。
在日常的测试工作中，申请人发现，即被错误识别出的非人脸的抓拍图片彼此之间有着很高的相似度，甚至超过了90％，这使得我们可以在短时间内无法提升算法精确度的情况下，通过对这些非人脸抓拍图片进行过滤，从而降低人脸分析任务抓拍的误检率。
本实施方式提出的方案为：首先需要说明的是，针对一个初始的相关技术中的视频分析平台，该分析平台是没有非人脸特征库的，所有通过算法所识别成人脸的结果(不论是否是人脸)，都会被作为结果展示到前端页面。在本实施方式中，可以在前端对识别出的结果进行标注，例如通过界面按钮的方式标注该识别结果图片，具体地，可以标注出前述的识别成人脸的结果中被识别错误的图片(即不包括人脸的图片)，还可以在该识别错误的图片中标注该图片中疑似被识别成人脸的内容。标注后的图片被记为非人脸图片，其特征将会被保留到服务器本地的相应目录中，形成了非人脸特征库；然后，在存在非人脸特征库的情况下，之后的每一张抓拍结果，都会在后台和非人脸特征库的特征进行比对，鉴于非人脸图片之间存在很高的相似度，当相似度超过某一个阈值后，视频分析平台就会二次 判断确定这张图片为非人脸图片，进而不在前端页面展示人脸识别结果。
示例性地，操作员查看分析任务的人脸抓拍结果，人工判断是非人脸的图片，通过点击按钮的方式将该图片标记为非人脸；后台收到标记请求，会将该图片的特征保存到服务器，保存路径可配置；后续对分析任务抓拍到的图片进行二次判断，即将分析人物抓拍到的图片和之前标记的非人脸图片的特征进行比对，当相似度比对的结果大于某一个可配置的阈值时，系统即判定这张抓拍图片也是非人脸图片，进行丢弃处理，从而不会在前端页面展示。
需要说明的是，利用非人脸特征库进行二次比对的业务流程，可以降低人脸识别系统的误检率；利用人工标注的方式，可以不断优化非人脸特征库，进一步降低人脸识别系统的误检率。
图4是根据本公开可选实施方式的人脸识别方法的流程示意图，如图4所示，包括：
视频分析平台从摄像头点位获取视频流，通过算法模型进行人脸识别，得到识别结果；
将识别结果与非人脸特征库进行特征比对；如果识别结果中的图像特征与非人脸特征库中的特征的相似度超过阈值，则表示识别结果是非人脸的结果，则页面不展示该非人脸的识别结果；如果识别结果中的图像特征与非人脸特征库中的特征的相似度小于阈值，则表明识别结果是人脸，则页面展示该人脸识别结果。
图5是根据本公开可选实施方式的非人脸标注的流程示意图，如图5所示，包括：
视频分析平台从摄像头点位获取视频流，通过算法模型进行人脸识别，得到识别结果；
将识别结果与非人脸特征库进行特征比对；如果识别结果中的图像特征与非人脸特征库中的特征的相似度超过阈值，则表示识别结果是非人脸的结果，则页面不展示该非人脸的识别结果；如果识别结果中的图像特征 与非人脸特征库中的特征的相似度小于阈值，则表明识别结果是人脸，则页面展示该人脸识别结果。
判断页面展示的识别结果，判断其中是否包括了非人脸结果，如果不包括，则不执行额外的操作，如果页面展示的识别结果是包括了非人脸结果，则标注该图片，也可以标注该非人脸图片中的非人脸特征，以对非人脸特征库进行扩充并且将该图片移至标注管理页面。
还需要说明的是，本实施方式可以针对VAP视频分析平台进行业务流程的优化，主要运用于安防监控领域。
具体实施方案按模块划分共有三个部分：
前端：前端需要增加标注的按钮，用来发送请求，告诉后端系统这张图片需要进行标注处理，即这是一张非人脸；可选地，还需要有独立的一级菜单或是新的web页面，专门用来做标注管理，即将所有在该平台标注的非人脸图片统一管理；该页面需要具备以下功能：图片来源进行查询、撤销标注、导出。其中，图片来源可以是摄像头点位，通过点位来归类图片是一种有效的方法，有利于之后的查询；对于一些可能出现误操作带来的标注样本，允许将该图片恢复；导出可以将这些图片或者是被选中的图片以压缩包的形式导出到本地。
后端系统：后端系统需要提供相应的接口，例如标注接口，以及标注管理页面的所有相关接口。可选地，在业务流程方面，需要在产生人脸抓拍之后，在前端展示之前多增加一个流程，即和非人脸特征库进行特征比对的流程，需要调用已有的比对接口，若比对的结果超过阈值，还需要将图片进行移动(到标注管理页面)，而不往分析任务的结果处展示图片。
算法系统：后端系统对标注的图片进行特征保存，需要保存到算法系统当中，特征文件同时加载到内存当中。对于撤销标注的请求，还需要将特征文件进行同步删除操作，同时更新到内存当中。
本实施方式可以应用在在展会现场使用固定摄像头进行人脸抓拍业务展示的场景中，通过使用现场的固定摄像头，来展示视频分析平台的人 脸识别技术。由于现场情况较为复杂，摄像头的布置不能做到十分规范，导致现场实际展示时，会有较多非人脸图片作为系统的识别结果展示出来。现在利用非人脸特征库比对的过程，可以通过人工标注的方式将现场的非人脸抓拍图片进行标注，从而让系统生成基于展会现场的非人脸特征库，如展会现场的一个椅子等，从而使得在标注之后，不再产生这种类似的非人脸抓拍情况，提升了识别效果。
本实施方式也可以应用在视频监控场景预先部署非人脸特征库进入系统的场景中，以达到降低人脸误检率的效果。以某公安局点为例，在开局部署时，除了正常部署业务版本之外，可以同时添加在之前测试过程中收集的非人脸特征库，通过非人脸特征库和抓拍图片二次比对的手段，可以过滤掉大部分的非人脸图片。可以通过手工标注的方式，针对该局点摄像头的数据进行非人脸特征库定制化的调优，进一步降低人脸识别的误检率。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到根据上述实施例的方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。基于这样的理解，本公开的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质(如ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台终端设备(可以是手机，计算机，服务器，或者网络设备等)执行本公开各个实施例所述的方法。
本公开的实施例还提供了一种计算机可读的存储介质，该计算机可读的存储介质中存储有计算机程序，其中，该计算机程序被设置为运行时执行上述任一项方法实施例中的步骤。
可选地，在本实施例中，上述计算机可读的存储介质可以被设置为存储用于执行以下步骤的计算机程序：
S1，对待识别的图像进行目标识别，得到目标图像；
S2，根据非目标信息对该目标图像进行识别，确定该目标图像中是否包括该目标，其中，该非目标信息包括历史识别所得到的历史目标图像中被错误识别的非目标图像中的非目标的信息。
通过上述步骤，由于对待识别的图像进行目标识别，得到目标图像；根据非目标信息对所述目标图像进行识别，确定所述目标图像中是否包括所述目标，其中，所述非目标信息包括历史识别所得到的历史目标图像中被错误识别的非目标图像中的非目标的信息，因此，可以解决相关技术中目标识别率较低的问题，达到提高目标识别准确率的效果。
可选地，本实施例中的具体示例可以参考上述实施例及可选实施方式中所描述的示例，本实施例在此不再赘述。
可选地，在本实施例中，上述存储介质可以包括但不限于：U盘、只读存储器(Read-Only Memory，简称为ROM)、随机存取存储器(Random Access Memory，简称为RAM)、移动硬盘、磁碟或者光盘等各种可以存储计算机程序的介质。
本公开的实施例还提供了一种电子装置，包括存储器和处理器，该存储器中存储有计算机程序，该处理器被设置为运行计算机程序以执行上述任一项方法实施例中的步骤。
可选地，上述电子装置还可以包括传输设备以及输入输出设备，其中，该传输设备和上述处理器连接，该输入输出设备和上述处理器连接。
可选地，在本实施例中，上述处理器可以被设置为通过计算机程序执行以下步骤：
S1，对待识别的图像进行目标识别，得到目标图像；
S2，根据非目标信息对该目标图像进行识别，确定该目标图像中是否包括该目标，其中，该非目标信息包括历史识别所得到的历史目标图像中被错误识别的非目标图像中的非目标的信息。
通过上述步骤，由于对待识别的图像进行目标识别，得到目标图像；根据非目标信息对所述目标图像进行识别，确定所述目标图像中是否包括 所述目标，其中，所述非目标信息包括历史识别所得到的历史目标图像中被错误识别的非目标图像中的非目标的信息，因此，可以解决相关技术中目标识别率较低的问题，达到提高目标识别准确率的效果。
可选地，本实施例中的具体示例可以参考上述实施例及可选实施方式中所描述的示例，本实施例在此不再赘述。
显然，本领域的技术人员应该明白，上述的本公开的各模块或各步骤可以用通用的计算装置来实现，它们可以集中在单个的计算装置上，或者分布在多个计算装置所组成的网络上，可选地，它们可以用计算装置可执行的程序代码来实现，从而，可以将它们存储在存储装置中由计算装置来执行，并且在某些情况下，可以以不同于此处的顺序执行所示出或描述的步骤，或者将它们分别制作成各个集成电路模块，或者将它们中的多个模块或步骤制作成单个集成电路模块来实现。这样，本公开不限制于任何特定的硬件和软件结合。
以上所述仅为本公开的优选实施例而已，并不用于限制本公开，对于本领域的技术人员来说，本公开可以有各种更改和变化。凡在本公开的原则之内，所作的任何修改、等同替换、改进等，均应包含在本公开的保护范围之内。

Claims (10)

  1. 一种目标的识别方法，包括：
    对待识别的图像进行目标识别，得到目标图像；
    根据非目标信息对所述目标图像进行识别，确定所述目标图像中是否包括所述目标，其中，所述非目标信息包括历史识别所得到的历史目标图像中被错误识别的非目标图像中的非目标的信息。
  2. 根据权利要求1所述的方法，其中，在根据所述非目标信息对所述目标图像进行识别之前，所述方法还包括：
    提取所述历史识别所得到的历史目标图像中被错误识别的所述非目标图像中的非目标的图像特征作为所述非目标信息。
  3. 根据权利要求1或2所述的方法，其中，所述根据所述非目标信息对所述目标图像进行识别，确定所述目标图像中是否包括所述目标，包括：
    将所述目标图像与所述非目标信息进行匹配；
    根据匹配结果确定所述目标图像中是否包括所述目标。
  4. 根据权利要求3所述的方法，其中，所述将所述目标图像与所述非目标信息进行匹配，包括：
    确定所述目标图像的图像特征；
    将所述目标图像的图像特征与所述非目标信息进行匹配，得到匹配结果。
  5. 根据权利要求4中所述的方法，其中，所述根据匹配结果确定所述目标图像中是否包括所述目标，包括：
    在所述匹配结果指示了所述目标图像中的图像特征与所述非目 标信息中的所述非目标的图像特征相似度高于预先设定的阈值的情况下，确定所述目标图像中不包括所述目标；或者，
    在所述匹配结果指示了所述目标图像中的图像特征与所述非目标信息中的所述非目标的图像特征相似度低于预先设定的所述阈值的情况下，确定所述目标图像中包括所述目标。
  6. 根据权利要求1所述的方法，其中，根据所述非目标信息对所述目标图像进行识别，确定所述目标图像中是否包括所述目标，包括：
    使用第一模型对所述目标图像进行分析，得到分析结果，其中，所述分析结果用于指示所述目标图像中是否包括所述目标，其中，所述第一模型为使用多组数据对深度神经网络进行训练得到的，所述多组数据均包括：多个样本非目标图像，以及，所述多个样本非目标图像中的非目标的信息。
  7. 根据权利要求1所述的方法，其中，在所述根据非目标信息对所述目标图像进行识别，确定所述目标图像中是否包括所述目标之后，所述方法还包括：
    显示确定后所得到的目标图像。
  8. 一种目标的识别装置，包括：
    第一识别模块，设置为对待识别的图像进行目标识别，得到目标图像；
    第二识别模块，设置为根据非目标信息对所述目标图像进行识别，确定所述目标图像中是否包括所述目标，其中，所述非目标信息包括历史识别所得到的历史目标图像中被错误识别的非目标图像中的非目标的信息。
  9. 一种计算机可读的存储介质，所述计算机可读的存储介质中存储有计算机程序，其中，所述计算机程序被设置为运行时执行所述权利要求1至7任一项中所述的方法。
  10. 一种电子装置，包括存储器和处理器，所述存储器中存储有计算机程序，所述处理器被设置为运行所述计算机程序以执行所述权利要求1至7任一项中所述的方法。
PCT/CN2020/136593 2019-12-30 2020-12-15 一种目标的识别方法及装置、存储介质及电子装置 WO2021135933A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201911399162.6 2019-12-30
CN201911399162.6A CN113128262A (zh) 2019-12-30 2019-12-30 一种目标的识别方法及装置、存储介质及电子装置

Publications (1)

Publication Number Publication Date
WO2021135933A1 true WO2021135933A1 (zh) 2021-07-08

Family

ID=76686566

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/136593 WO2021135933A1 (zh) 2019-12-30 2020-12-15 一种目标的识别方法及装置、存储介质及电子装置

Country Status (2)

Country Link
CN (1) CN113128262A (zh)
WO (1) WO2021135933A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116055338A (zh) * 2023-03-28 2023-05-02 杭州觅睿科技股份有限公司 一种误报消除方法、装置、设备及介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101075294A (zh) * 2006-05-19 2007-11-21 腾讯科技（深圳）有限公司 一种建立人脸检测器的方法
CN106326909A (zh) * 2015-07-01 2017-01-11 株式会社日立制作所 图像识别方法和图像识别装置
CN106778800A (zh) * 2016-11-14 2017-05-31 天津津航技术物理研究所 一种AdaBoost级联分类器快速检测方法
CN107368842A (zh) * 2016-05-12 2017-11-21 北京君正集成电路股份有限公司 一种训练方法及装置
CN107729796A (zh) * 2016-08-11 2018-02-23 中国移动通信集团湖南有限公司 一种人脸图片检测方法及装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101075294A (zh) * 2006-05-19 2007-11-21 腾讯科技（深圳）有限公司 一种建立人脸检测器的方法
CN106326909A (zh) * 2015-07-01 2017-01-11 株式会社日立制作所 图像识别方法和图像识别装置
CN107368842A (zh) * 2016-05-12 2017-11-21 北京君正集成电路股份有限公司 一种训练方法及装置
CN107729796A (zh) * 2016-08-11 2018-02-23 中国移动通信集团湖南有限公司 一种人脸图片检测方法及装置
CN106778800A (zh) * 2016-11-14 2017-05-31 天津津航技术物理研究所 一种AdaBoost级联分类器快速检测方法

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116055338A (zh) * 2023-03-28 2023-05-02 杭州觅睿科技股份有限公司 一种误报消除方法、装置、设备及介质
CN116055338B (zh) * 2023-03-28 2023-08-11 杭州觅睿科技股份有限公司 一种误报消除方法、装置、设备及介质

Also Published As

Publication number Publication date
CN113128262A (zh) 2021-07-16

Similar Documents

Publication Publication Date Title
CN109284729B (zh) 基于视频获取人脸识别模型训练数据的方法、装置和介质
US11210504B2 (en) Emotion detection enabled video redaction
CN110175549B (zh) 人脸图像处理方法、装置、设备及存储介质
US20220092881A1 (en) Method and apparatus for behavior analysis, electronic apparatus, storage medium, and computer program
CN109284733B (zh) 一种基于yolo和多任务卷积神经网络的导购消极行为监控方法
US20210357624A1 (en) Information processing method and device, and storage medium
WO2019033572A1 (zh) 人脸遮挡检测方法、装置及存储介质
WO2019127273A1 (zh) 一种多人脸检测方法、装置、服务器、系统及存储介质
US20210357678A1 (en) Information processing method and apparatus, and storage medium
US10990827B2 (en) Imported video analysis device and method
CN112364827B (zh) 人脸识别方法、装置、计算机设备和存储介质
US9947105B2 (en) Information processing apparatus, recording medium, and information processing method
CN110610127B (zh) 人脸识别方法、装置、存储介质及电子设备
CN108563651B (zh) 一种多视频的目标搜索方法、装置及设备
US9633272B2 (en) Real time object scanning using a mobile phone and cloud-based visual search engine
CN109857878B (zh) 物品标注方法及装置、电子设备及存储介质
CN112989950A (zh) 一种面向多模态特征语义关联特征的暴力视频识别系统
WO2021135933A1 (zh) 一种目标的识别方法及装置、存储介质及电子装置
KR101547255B1 (ko) 지능형 감시 시스템의 객체기반 검색방법
CN113470013A (zh) 一种搬移物品的检测方法及装置
CN112434049A (zh) 表格数据存储方法、装置、存储介质及电子装置
CN110738077A (zh) 一种异物检测方法及装置
Mathew et al. Person re-identification through face detection from videos using Deep Learning
JP5552946B2 (ja) 顔画像サンプル採取装置、顔画像サンプル採取方法、プログラム
KR100434907B1 (ko) 얼굴 인식 기능을 포함한 영상 감시 시스템 및 이를이용하는 방법

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: 20908753

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: 20908753

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 21.02.2023)

122 Ep: pct application non-entry in european phase

Ref document number: 20908753

Country of ref document: EP

Kind code of ref document: A1