WO2021169625A1 - 网络翻拍照片的检测方法、装置、计算机设备及存储介质 - Google Patents

网络翻拍照片的检测方法、装置、计算机设备及存储介质 Download PDF

Info

Publication number
WO2021169625A1
WO2021169625A1 PCT/CN2021/070734 CN2021070734W WO2021169625A1 WO 2021169625 A1 WO2021169625 A1 WO 2021169625A1 CN 2021070734 W CN2021070734 W CN 2021070734W WO 2021169625 A1 WO2021169625 A1 WO 2021169625A1
Authority
WO
WIPO (PCT)
Prior art keywords
video
fourier
fourier feature
picture
feature map
Prior art date
Application number
PCT/CN2021/070734
Other languages
English (en)
French (fr)
Inventor
徐国诚
Original Assignee
深圳壹账通智能科技有限公司
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 深圳壹账通智能科技有限公司 filed Critical 深圳壹账通智能科技有限公司
Publication of WO2021169625A1 publication Critical patent/WO2021169625A1/zh

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00Arrangements for image or video recognition or understanding
    • G06V10/40Extraction of image or video features
    • G06V10/42Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
    • G06V10/431Frequency domain transformation; Autocorrelation
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/24Classification techniques
    • G06F18/241Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods

Definitions

  • This application relates to the technical field of picture recognition, and in particular to a detection method, device, computer equipment, and storage medium for network re-photographing.
  • the embodiments of the present application provide a method, device, computer equipment, and storage medium for detecting a photo retaken on the network, so as to solve the technical problem of high false recognition rate of identifying whether a video or photo is retaken in the prior art.
  • a method for detecting network photo retakes comprising:
  • the number of identified peak transformation points is at least two and at least two peak transformation points are symmetrically distributed, it is determined that the picture or video to be identified corresponding to the Fourier feature map is a remake.
  • a detection device for network photocopying comprising:
  • the extraction module is configured to extract the Fourier feature value of the picture or the Fourier feature value of the video frame picture of the video when the picture or video to be recognized is received;
  • a generating module configured to generate a Fourier feature map of the picture or the video frame picture according to the extracted Fourier feature value
  • a recognition module configured to recognize the number of peak transformation points in the Fourier feature map through a pre-trained target detection model
  • a judging module configured to judge the picture or video to be identified corresponding to the Fourier feature map when the number of identified peak transformation points is at least two and at least two of the peak transformation points are symmetrically distributed For a remake.
  • a computer device includes a memory, a processor, and computer-readable instructions that are stored in the memory and can run on the processor, wherein the processor implements the following steps when the processor executes the computer-readable instructions:
  • the number of identified peak transformation points is at least two and at least two peak transformation points are symmetrically distributed, it is determined that the picture or video to be identified corresponding to the Fourier feature map is a remake.
  • One or more readable storage media storing computer readable instructions.
  • the computer readable storage medium stores computer readable instructions.
  • the computer readable instructions execute the following steps:
  • the number of identified peak transformation points is at least two and at least two peak transformation points are symmetrically distributed, it is determined that the picture or video to be identified corresponding to the Fourier feature map is a remake.
  • the method, device, computer equipment, and storage medium for detecting a photo taken on the network provided in this application extract the Fourier feature value of the picture or extract the Fourier feature value of the video frame picture of the video, and then extract the Fourier feature value of the video frame picture according to the extracted
  • the Fourier feature value generates a Fourier feature map of the picture or the video frame picture, and then recognizes the number of peak transformation points in the Fourier feature map through a pre-trained target detection model, Finally, according to the number of identified peak transformation points, it is judged whether the corresponding picture or photo is a remake.
  • this application does not need to extract the Fourier spectrum characteristics of the RGB three-channel, but directly Generate a Fourier feature map, which reduces the amount of calculation for remake recognition, and improves the calculation speed of remake recognition.
  • the present application judges whether the corresponding picture or photo is a remake according to the number of peak transformation points identified, which improves Improve the accuracy of remake recognition.
  • FIG. 1 is a schematic diagram of an application environment of a method for detecting network photo retakes in an embodiment of the present application
  • FIG. 2 is a flowchart of a method for detecting network photo retakes in an embodiment of the present application
  • FIG. 3 is a flowchart of another method for detecting a photo retaken on the Internet in an embodiment of the present application
  • FIG. 4 is a flowchart of another method for detecting photo retakes on the Internet in an embodiment of the present application
  • Fig. 5-a is the position of the peak change point in the non-rephotographed picture or video in an embodiment of the present application
  • Fig. 5-b is the position of the peak change point in the retaken picture or video in an embodiment of the present application
  • FIG. 6 is a schematic diagram of the structure of a detection device for network photo recapture in an embodiment of the present application.
  • Fig. 7 is a schematic diagram of a computer device in an embodiment of the present application.
  • the method for detecting network photo retakes provided in this application can be applied in the application environment as shown in FIG. 1.
  • the computer equipment shown in FIG. 1 includes, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices.
  • Fig. 2 is a flowchart of a method for detecting network photo retakes in an embodiment of the present application.
  • a method for detecting network photo retakes is provided, and the method is applied in Fig. 1
  • the computer device in is taken as an example for description, including the following steps S101 to S104.
  • this step further includes:
  • the received picture or video to be recognized includes the original picture or video, and also includes the compressed picture or video.
  • the video frame picture may be any frame or multiple frames in the video to be recognized.
  • S102 Generate a Fourier feature map of the picture or the video frame picture according to the extracted Fourier feature value.
  • this step further includes:
  • a Fourier feature map of each video frame picture in the same video to be identified is generated.
  • the target detection model may be SSD (Single Shot MultiBox Detector (Single Shot MultiBox Detector) target detection model is an end-to-end single-stage target detection network, which treats the detection target on the picture as a discretized synthesis of many default boxes.
  • SSD Single Shot MultiBox Detector
  • the images to be detected will contain targets with different scales and resolutions.
  • SSD has expanded its feature selection range.
  • this step further includes:
  • Figure 5-a is the position of the peak transition point in the non-rephotographed picture or video in an embodiment of the present application
  • Figure 5-b is the position of the peak transition point in the reproduced picture or video in an embodiment of the present application, such as As shown in Figure 5-a and 5-b, it is found through multiple experiments that there is only one peak transition point in a non-rephotographed picture or video, while there are generally many peak transition points in a re-photographed picture or video, so it can be By extracting the Fourier feature map of the picture or video to be recognized, identifying the number of peak transformation points in the Fourier feature map through a pre-trained target detection model, and judging the corresponding according to the number of peak transformation points Whether the picture or video of is a remake.
  • the Fourier feature value of the picture is extracted or the Fourier feature value of the video frame picture of the video is extracted, and then the picture or the video frame is generated according to the extracted Fourier feature value
  • the Fourier feature map of the picture, and then the pre-trained target detection model is used to identify the number of peak transformation points in the Fourier feature map, and finally the corresponding picture is determined according to the number of identified peak transformation points
  • this application does not need to extract the Fourier spectrum characteristics of the RGB (red, green and blue) three channels, but directly generates the Fourier feature map, which reduces the remake recognition
  • the amount of calculation increases the calculation speed of the remake recognition.
  • the present application judges whether the corresponding picture or photo is a remake according to the number of the identified peak transformation points, which improves the accuracy of the remake recognition.
  • FIG. 3 is a flowchart of another method for detecting a photo retaken by a network in an embodiment of the present application.
  • the steps of training the target detection model according to this embodiment are shown in FIG. 3, wherein the training of the target detection model
  • the steps include the following steps S301 to S304.
  • S301 Extract the Fourier feature value of the re-photographed sample picture, and generate a re-photographed Fourier feature map according to the Fourier feature value of the re-photographed sample picture.
  • the extracted Fourier feature values of the re-photographed sample image are mainly the texture features of the sample image.
  • the texture feature of the re-photographed image is more prominent than the non-re-photographed texture feature, which is directly re-photographed.
  • the re-photographed sample pictures include re-photographed pictures of finished products, re-photographed pictures displayed on a screen interface, re-photographed pictures of video screens played on the screen, and the like.
  • the Fourier characteristic map is the right picture in Fig. 5-a or Fig. 5-b.
  • the first labeling area and the second labeling area are areas input by the user for the machine to learn which shape belongs to the peak transformation point.
  • the training process is the process of letting the machine learn what shape belongs to the "peak transformation point".
  • the machine adjusts the parameters of the target detection model by continuously learning the area of the peak transformation point marked by the user to obtain a trained target detection model , Can detect the peak transform point in the unknown Fourier characteristic map by itself.
  • the target detection model can use SSD (Single Shot MultiBox Detector, a target detection algorithm based on deep learning) model.
  • the Fourier feature map picture with the first label area/the second label area can be input into the fifth convolution block in the vgg_16 neural network of the target detection model SSD.
  • Three convolutional layers (conv5_3), get the feature map feature_map(h*w*c), take the feature map into several windows, calculate the receptive field of the window, and generate the corresponding default box through the calculated receptive field, Traverse all the first label area and the second label area, learn and adjust the position offset of the convolution sampling so that the predicted peak transformation point area predict box includes the actual peak transformation point label area default box, until the output is one The largest prediction range IOU (Intersection over Union), complete the training.
  • IOU Intersection over Union
  • This embodiment proposes a method for training the target detection model.
  • the recognition speed of the picture or video to be recognized can be improved. , Making the recognition more efficient.
  • FIG. 4 is a flowchart of another method for detecting network photo retakes in an embodiment of the present application.
  • the method for detecting network photo retakes according to an embodiment of the present application will be described in detail below with reference to FIG. 4.
  • the method In addition to the above steps S101 and S102, the following steps S401 to S403 are also included.
  • S401 Identify the number and positions of peak transformation points in the Fourier feature map through a pre-trained target detection model.
  • the peak transformation point located at the center of the Fourier characteristic map may be determined as the central peak transformation point, and all other peak transformation points may be determined as non-peak transformation points.
  • the method for detecting network photo retakes further includes:
  • each Fourier feature map in the same video to be identified is at least two, identify each Fourier feature map according to the position of the peak transformation point The central peak change point and the non-central peak change point.
  • the step of judging that the picture or video to be recognized corresponding to the Fourier feature map is a remake in the foregoing step further includes:
  • the object to be recognized is a video
  • the positions of the peak transform points in the Fourier feature map are performed on each frame picture belonging to the same video to be recognized Identification, judgment of the number of peak transformation points, judgment of whether the non-central peak transformation points are symmetrically distributed around the central peak transformation point, when it is judged that the non-peak transformation points in each of the Fourier characteristic maps surround the corresponding If the center peak change points of the center peak are symmetrically distributed, it is judged that the corresponding video to be recognized is a remake, so as to improve the accuracy of the recognition of the remake video.
  • FIG. 6 is a schematic structural diagram of a detection device for network photo recapture in an embodiment of the present application.
  • the following describes in detail the detection device for network photo recapture according to an embodiment of the present application in conjunction with FIG. 6, as shown in FIG. 6
  • a detection device which corresponds to the detection method of network photo recapture in the above-mentioned embodiment in a one-to-one correspondence.
  • the device 100 for detecting a photo taken over the network includes an extraction module 11, a generation module 12, an identification module 13, and a judgment module 14.
  • the extraction module 11 is configured to extract the Fourier feature value of the picture or the Fourier feature value of the video frame picture of the video when the picture or video to be recognized is received.
  • the received picture or video to be recognized includes the original picture or video, and also includes the compressed picture or video.
  • the video frame picture may be any frame or multiple frames in the video to be recognized.
  • the generating module 12 is configured to generate a Fourier feature map of the picture or the video frame picture according to the extracted Fourier feature value.
  • the recognition module 13 is configured to recognize the number of peak transformation points in the Fourier feature map through a pre-trained target detection model.
  • the target detection model may be SSD (Single Shot MultiBox Detector (Single Shot MultiBox Detector) target detection model is an end-to-end single-stage target detection network, which treats the detection target on the picture as a discretized synthesis of many default boxes.
  • SSD Single Shot MultiBox Detector
  • the images to be detected will contain targets with different scales and resolutions.
  • SSD has expanded its feature selection range.
  • the judging module 14 is used for judging the picture to be identified corresponding to the Fourier feature map or when the number of identified peak transformation points is at least two and at least two of the peak transformation points are symmetrically distributed.
  • the video is a remake.
  • Figure 5-a is the position of the peak transition point in the non-rephotographed picture or video in an embodiment of the present application
  • Figure 5-b is the position of the peak transition point in the reproduced picture or video in an embodiment of the present application, such as As shown in Figure 5-a and 5-b, it is found through multiple experiments that there is only one peak transition point in a non-rephotographed picture or video, while there are generally many peak transition points in a re-photographed picture or video, so it can be
  • the Fourier feature map of the picture or video to be recognized is extracted by the extraction module 11, and the recognition module 13 recognizes the number of peak transformation points in the Fourier feature map through the pre-trained target detection model, and finally passes the judgment module 14 Determine whether the corresponding picture or video is a remake according to the number of peak change points.
  • the device 100 for detecting network photo retakes further includes:
  • the first generating unit is configured to extract the Fourier feature values of the re-photographed sample pictures, and generate the re-photographed Fourier feature maps according to the Fourier characteristic values of the re-photographed sample pictures;
  • the second generating unit is used to extract the Fourier feature values of the non-rephotographed sample pictures, and generate the non-rephotographed Fourier feature maps according to the Fourier feature values of the non-rephotographed sample pictures;
  • An annotation module configured to receive the first annotation area of the peak transformation point in the reproduced Fourier feature map, and receive the second annotation area of the peak transformation point in the non-reproduced Fourier feature map;
  • a training module configured to input the re-photographed Fourier feature map, the first labeled region, the non-re-photographed Fourier feature map, and the second labeled region into the target detection model for training , To obtain the trained target detection model.
  • the device for detecting network photo retakes further includes:
  • the position identification module is used to identify the position of the peak transformation point in the Fourier feature map through the target detection model; and is also used to identify the peak transformation point according to the number of identified peak transformation points at least two.
  • the position of the peak change point identifies the central peak change point and the non-central peak change point;
  • the judging module is specifically configured to judge whether the non-peak transformation points are symmetrically distributed around the central peak transformation point, and if so, judge that the picture or video to be identified corresponding to the Fourier feature map is a remake; otherwise, judge The corresponding picture or video to be recognized is a non-remake.
  • the extraction module is specifically configured to arbitrarily extract multiple video frame pictures from the video to be recognized, and is also configured to extract the Fourier feature value of each video frame picture.
  • the generating module is specifically configured to generate a Fourier feature map of each of the video frame pictures in the same video to be recognized according to the extracted Fourier feature values.
  • the identification module is specifically used to identify the number of peak transformation points in each Fourier feature map in the same video to be identified.
  • the determining module is specifically configured to determine the corresponding video to be recognized when the number of peak transform points of the corresponding Fourier feature map in each video frame picture included in the same video to be recognized is at least two For a remake.
  • the object to be recognized is a video
  • the positions of the peak transform points in the Fourier feature map are performed on each frame picture belonging to the same video to be recognized Identification, judgment of the number of peak transformation points, judgment of whether the non-central peak transformation points are symmetrically distributed around the central peak transformation point, when it is judged that the non-peak transformation points in each of the Fourier characteristic maps surround the corresponding If the center peak change points of the center peak are symmetrically distributed, it is judged that the corresponding video to be recognized is a remake, so as to improve the accuracy of the recognition of the remake video.
  • the various modules in the above-mentioned network photo recapture detection device can be implemented in whole or in part by software, hardware, and a combination thereof.
  • the above-mentioned modules may be embedded in the form of hardware or independent of the processor in the computer equipment, or may be stored in the memory of the computer equipment in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned modules.
  • a computer device is provided.
  • the computer device may be a terminal, and its internal structure diagram may be as shown in FIG. 7.
  • the computer equipment includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus.
  • the processor of the computer device is used to provide calculation and control capabilities.
  • the memory of the computer device includes a readable storage medium and an internal memory.
  • the readable storage medium stores an operating system and computer readable instructions.
  • the internal memory provides an environment for the operation of the operating system and computer readable instructions in the readable storage medium.
  • the network interface of the computer device is used to communicate with an external server through a network connection. When the computer-readable instruction is executed by the processor, a method for detecting network photo retakes is realized.
  • the readable storage medium may be a non-volatile readable storage medium or a volatile readable storage medium.
  • a computer device including a memory, a processor, and computer-readable instructions stored in the memory and running on the processor.
  • the processor executes the computer-readable instructions
  • the network in the foregoing embodiment is implemented.
  • the steps of the method for detecting a photo retake are, for example, step 101 to step 104 shown in FIG. 2.
  • the processor executes the computer-readable instructions
  • the functions of the modules/units of the network photo-recapture detection apparatus in the foregoing embodiment are implemented, for example, the functions of the modules 11 to 14 shown in FIG. 7. To avoid repetition, I won’t repeat them here.
  • one or more readable storage media storing computer readable instructions are provided.
  • the computer readable storage medium stores computer readable instructions, wherein the computer readable instructions are stored by one or more
  • the processor is executed, the one or more processors are executed to implement the steps of the method for detecting network photo retakes in the foregoing embodiment, for example, step 101 to step 104 shown in FIG. 2.
  • the functions of the modules/units of the network photo-recapture detection apparatus in the foregoing embodiment are realized, for example, the functions of the modules 11 to 14 shown in FIG. 7. To avoid repetition, I won’t repeat them here.
  • Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
  • Volatile memory may include random access memory (RAM) or external cache memory.
  • RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Channel (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
  • the method, device, computer equipment, and storage medium for detecting network photo retakes provided in this embodiment extract the Fourier feature value of the picture or extract the Fourier feature value of the video frame picture of the video, and then extract the Fourier feature value according to the extracted The Fourier feature value generates the Fourier feature map of the picture or the video frame picture, and then recognizes the number of peak transform points in the Fourier feature map through a pre-trained target detection model Finally, it is determined whether the corresponding picture or photo is a remake according to the number of the identified peak transformation points.
  • this application does not need to extract the RGB (red, green, and blue) three-channel Fourier Instead of generating Fourier feature maps directly, the Fourier feature map is directly generated, which reduces the amount of calculation for remake recognition and improves the calculation speed of remake recognition. On the other hand, this application judges the corresponding value based on the number of peak transformation points identified Whether the picture or photo is a remake improves the accuracy of remake recognition.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)

Abstract

本申请公开了一种网络翻拍照片的检测方法、装置、计算机设备及存储介质,应用于图片识别技术领域,用于解决现有技术识别视频或照片是否为翻拍的误识别率高的技术问题。本申请提供的网络翻拍照片的检测方法包括:接收到待识别的图片或者视频时,提取所述图片的傅里叶特征值或者提取所述视频的视频帧图片的傅里叶特征值;根据提取的所述傅里叶特征值生成所述图片或者所述视频帧图片的傅里叶特征图谱;通过预先训练好的目标检测模型识别所述傅里叶特征图谱中的峰值变换点的个数;当识别的所述峰值变换点的个数至少为两个且至少两个所述峰值变换点呈对称分布时,判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍。

Description

网络翻拍照片的检测方法、装置、计算机设备及存储介质
本申请要求于2020年02月28日提交中国专利局、申请号为202010127392.3,发明名称为“网络翻拍照片的检测方法、装置、计算机设备及存储介质”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本申请涉及图片识别技术领域,尤其涉及网络翻拍照片的检测方法、装置、计算机设备及存储介质。
背景技术
基于某些场景的需要,需要识别图片或者视频是否为翻拍的图片或视频。人眼识别某个图片或者视频是否为翻拍存在较大的误差,现智能识别图片或者视频是否为翻拍的现有技术主要包括以下两种方案:
1 )通过对屏幕翻拍的照片生成的图片或者视频帧,然后对生成的图片或者视频帧进行分割后提取特征值识别像素点和摩尔纹,根据特征值识别像素点和摩尔纹判断对应的图片或者视频是否为翻拍。发明人意识到这种方案的缺点在于需要原始的翻拍图像,图像尺寸很大使得数据运算量很大、处理速度慢,且拍摄角度和距离有一定的要求,使得如果未按要求拍摄会使得有比较大概率的像素点和摩尔纹不明显导致识别不到,使得误识别率较高,且如果被拍摄的带有大量的类似像素点的物体会被识别为翻拍照片,会提高使得误识别率。
2 )提取待识别文件的 RGB 三通道的傅里叶频谱特征,将提取的特征输入到分类模型中,通过该分类模型确定待识别图片或者视频是否为翻拍。发明人意识到这种识别方法需要提取待识别图片或者视频帧的 RGB 三通道的傅里叶频谱特征,使得运算量较大,如果被拍摄的带有大量的类似像素点的物体,这种方案也会将待识别的图片或者视频识别为翻拍,使得识别结果不准确。
技术问题
本申请实施例提供一种网络翻拍照片的检测方法、装置、计算机设备及存储介质,以解决现有技术识别视频或照片是否为翻拍的误识别率高的技术问题。
技术解决方案
一种网络翻拍照片的检测方法,所述方法包括:
接收到待识别的图片或者视频时,提取所述图片的傅里叶特征值或者提取所述视频的视频帧图片的傅里叶特征值;
根据提取的所述傅里叶特征值生成所述图片或者所述视频帧图片的傅里叶特征图谱;
通过预先训练好的目标检测模型识别所述傅里叶特征图谱中的峰值变换点的个数;
当识别的所述峰值变换点的个数至少为两个且至少两个所述峰值变换点呈对称分布时,判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍。
一种网络翻拍照片的检测装置,所述装置包括:
提取模块,用于接收到待识别的图片或者视频时,提取所述图片的傅里叶特征值或者提取所述视频的视频帧图片的傅里叶特征值;
生成模块,用于根据提取的所述傅里叶特征值生成所述图片或者所述视频帧图片的傅里叶特征图谱;
识别模块,用于通过预先训练好的目标检测模型识别所述傅里叶特征图谱中的峰值变换点的个数;
判断模块,用于当识别的所述峰值变换点的个数至少为两个且至少两个所述峰值变换点呈对称分布时,判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍。
一种计算机设备,包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机可读指令,其中,所述处理器执行所述计算机可读指令时实现如下步骤:
接收到待识别的图片或者视频时,提取所述图片的傅里叶特征值或者提取所述视频的视频帧图片的傅里叶特征值;
根据提取的所述傅里叶特征值生成所述图片或者所述视频帧图片的傅里叶特征图谱;
通过预先训练好的目标检测模型识别所述傅里叶特征图谱中的峰值变换点的个数;
当识别的所述峰值变换点的个数至少为两个且至少两个所述峰值变换点呈对称分布时,判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍。
一个或多个存储有计算机可读指令的可读存储介质,所述计算机可读存储介质存储有计算机可读指令,其中,所述计算机可读指令被一个或多个处理器执行时,使得所述一个或多个处理器执行如下步骤:
接收到待识别的图片或者视频时,提取所述图片的傅里叶特征值或者提取所述视频的视频帧图片的傅里叶特征值;
根据提取的所述傅里叶特征值生成所述图片或者所述视频帧图片的傅里叶特征图谱;
通过预先训练好的目标检测模型识别所述傅里叶特征图谱中的峰值变换点的个数;
当识别的所述峰值变换点的个数至少为两个且至少两个所述峰值变换点呈对称分布时,判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍。
有益效果
本申请提供的网络翻拍照片的检测方法、装置、计算机设备及存储介质,通过提取所述图片的傅里叶特征值或者提取所述视频的视频帧图片的傅里叶特征值,然后根据提取的所述傅里叶特征值生成所述图片或者所述视频帧图片的傅里叶特征图谱,再通过预先训练好的目标检测模型识别所述傅里叶特征图谱中的峰值变换点的个数,最后根据识别的所述峰值变换点的个数判断对应的图片或者照片是否为翻拍,相比于现有的翻拍识别技术,本申请不需要提取RGB三通道的傅里叶频谱特征,而是直接生成傅里叶特征图谱,减少了翻拍识别的计算量,提高了翻拍识别的运算速度,另一方面本申请根据识别的所述峰值变换点的个数判断对应的图片或者照片是否为翻拍,提高了翻拍识别的准确率。
附图说明
为了更清楚地说明本申请实施例的技术方案,下面将对本申请实施例的描述中所需要使用的附图作简单地介绍,显而易见地,下面描述中的附图仅仅是本申请的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动性的前提下,还可以根据这些附图获得其他的附图。
图1是本申请一实施例中网络翻拍照片的检测方法的一应用环境示意图;
图2是本申请一实施例中网络翻拍照片的检测方法的一流程图;
图3是本申请一实施例中又一网络翻拍照片的检测方法的一流程图;
图4是本申请一实施例中另一网络翻拍照片的检测方法的一流程图;
图5-a是本申请一实施例中非翻拍的图片或者视频中的峰值变换点的位置;
图5-b是本申请一实施例中翻拍的图片或者视频中的峰值变换点的位置;
图6是本申请一实施例中网络翻拍照片的检测装置的结构示意图;
图7是本申请一实施例中计算机设备的一示意图。
具体实施方式
下面将结合本申请实施例中的附图,对本申请实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例是本申请一部分实施例,而不是全部的实施例。基于本申请中的实施例,本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例,都属于本申请保护的范围。
本申请提供的网络翻拍照片的检测方法,可应用在如图1的应用环境中。其中,图1所示的计算机设备包括但不限于各种个人计算机、笔记本电脑、智能手机、平板电脑和便携式可穿戴设备。
图2是本申请一实施例中网络翻拍照片的检测方法的一流程图,在一实施例中,如图2所示,提供一种网络翻拍照片的检测方法,以该方法应用在图1中的计算机设备为例进行说明,包括如下步骤S101至S104。
S101、接收到待识别的图片或者视频时,提取所述图片的傅里叶特征值或者提取所述视频的视频帧图片的傅里叶特征值。
在其中一个实施例中,当接收到待识别的视频时,该步骤进一步包括:
从所述待识别的视频中任意提取多个视频帧图片;
提取每个所述视频帧图片的傅里叶特征值。
在其中一个实施例中,接收到待识别的图片或者视频包括原始的图片或者视频,还包括压缩的图片或者视频。
在该实施例中,所述视频帧图片可以是待识别视频中的任意一帧或者任意多帧。
S102、根据提取的所述傅里叶特征值生成所述图片或者所述视频帧图片的傅里叶特征图谱。
在其中一个实施例中,当接收到待识别的视频时,该步骤进一步包括:
根据提取的所述傅里叶特征值生成同一待识别视频中每个所述视频帧图片的傅里叶特征图谱。
S103、通过预先训练好的目标检测模型识别所述傅里叶特征图谱中的峰值变换点的个数。
在其中一个实施例中,该目标检测模型可以是SSD(Single Shot MultiBox Detector,单镜头多盒检测器)目标检测模型,是一种端到端的单阶段目标检测网络,它将图片之上的检测目标视为是许多个default boxes默认框的离散化合成。
一般欲检测的图片当中会包含有着不同尺度大小及分辨率的目标。为了有效地涵盖这种种不同尺度、分辨率的目标检测,SSD扩大了其特征选取的范围。
S104、当识别的所述峰值变换点的个数至少为两个且至少两个所述峰值变换点呈对称分布时,判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍。
在其中一个实施例中,当接收到待识别的视频时,该步骤进一步包括:
当同一待识别的视频包括的每个所述视频帧图片中相应的傅里叶特征图谱的峰值变换点的个数均至少为两个时,判断对应的待识别视频为翻拍。
图5-a是本申请一实施例中非翻拍的图片或者视频中的峰值变换点的位置,图5-b是本申请一实施例中翻拍的图片或者视频中的峰值变换点的位置,如图5-a对照5-b所示,通过多次试验发现,非翻拍的图片或视频中的峰值变换点为1个,而翻拍的图片或视频中的峰值变换点一般有很多个,故可以通过提取该待识别的图片或视频的傅里叶特征图谱,通过预先训练好的目标检测模型识别该傅里叶特征图谱中的峰值变换点的个数,依据该峰值变换点的个数判断对应的图片或者视频是否为翻拍。
本实施例通过提取所述图片的傅里叶特征值或者提取所述视频的视频帧图片的傅里叶特征值,然后根据提取的所述傅里叶特征值生成所述图片或者所述视频帧图片的傅里叶特征图谱,再通过预先训练好的目标检测模型识别所述傅里叶特征图谱中的峰值变换点的个数,最后根据识别的所述峰值变换点的个数判断对应的图片或者照片是否为翻拍,相比于现有的翻拍识别技术,本申请不需要提取RGB(红绿蓝)三通道的傅里叶频谱特征,而是直接生成傅里叶特征图谱,减少了翻拍识别的计算量,提高了翻拍识别的运算速度,另一方面本申请根据识别的所述峰值变换点的个数判断对应的图片或者照片是否为翻拍,提高了翻拍识别的准确率。
图3是本申请一实施例中又一网络翻拍照片的检测方法的一流程图,根据本实施例的训练所述目标检测模型的步骤如图3所示,其中,训练所述目标检测模型的步骤包括以下步骤S301至S304。
S301、提取翻拍的样本图片的傅里叶特征值,根据翻拍的样本图片的傅里叶特征值生成翻拍的傅里叶特征图谱。
在其中一个实施例中,提取的翻拍的样本图片的傅里叶特征值主要是样本图片的纹理特征,翻拍的图片的纹理特征相比非翻拍的纹理特征更为突出,由其是直接翻拍的手机、电脑等显示屏上的图片的纹理。
其中,所述翻拍的样本图片包括翻拍成品相片的图片、翻拍屏幕界面显示的图片、翻拍屏幕内播放的视频画面的图片等。
S302、提取非翻拍的样本图片的傅里叶特征值,根据非翻拍的样本图片的傅里叶特征值生成非翻拍的傅里叶特征图谱。
在该实施例中,所述傅里叶特征图谱如图5-a或图5-b中右侧的图片。
S303、接收所述翻拍的傅里叶特征图谱中峰值变换点的第一标注区域,接收所述非翻拍的傅里叶特征图谱中峰值变换点的第二标注区域。
在其中一个实施例中,所述第一标注区域和所述第二标注区域为用户输入的供机器学习哪种形状属于峰值变换点的区域。
S304、将所述翻拍的傅里叶特征图谱、所述第一标注区域、所述非翻拍的傅里叶特征图谱和所述第二标注区域输入到所述目标检测模型中进行训练,得到所述训练好的目标检测模型。
训练的过程是让机器学习什么样的形状属于“峰值变换点”的过程,训练过程中机器通过不断地学习用户标注的峰值变换点的区域调整目标检测模型的参数,得到训练好的目标检测模型,能够自行检测出未知傅里叶特征图谱中的峰值变换点。
在其中一个实施例中,该目标检测模型可以选用SSD(Single Shot MultiBox Detector,基于深度学习的目标检测算法)模型。
根据本实施例的一个训练场景例如,可以将带有第一标注区域/第二标注区域的傅里叶特征图谱图片输入到目标检测模型SSD的vgg_16神经网络中第五个卷积block里面的第三个卷积层(conv5_3),获取特征图feature_map(h*w*c),将该特征图取成若干个窗口,计算该窗口的感受野,通过计算得到的感受野生成对应的default box,遍历所有的第一标注区域和第二标注区域,学习和调整卷积采样的位置偏移量offset使得预测的峰值变换点区域predict box包括实际的峰值变换点标注区域default box,直到输出得出一个最大的预测范围IOU(Intersection over Union),完成训练。
本实施例提出了训练该目标检测模型的方法,通过预先训练自动识别出峰值变换点的目标检测模型对傅里叶图谱中的峰值变换点进行识别,可以提高待识别的图片或视频的识别速度,使得识别效率更高。
图4是本申请一实施例中另一网络翻拍照片的检测方法的一流程图,下面结合图4详细描述根据本申请以实施例的网络翻拍照片的检测方法,如图4所示,该方法在包括上述步骤S101及S102的基础上还包括以下步骤S401至S403。
S401、通过预先训练好的目标检测模型识别所述傅里叶特征图谱中的峰值变换点的个数和位置。
S402、当识别的所述峰值变换点的个数至少为两个时,根据所述峰值变换点的位置识别中心峰值变换点及非中心峰值变换点。
在其中一个实施例中,可以将位于该傅里叶特征图谱中的中心位置的峰值变换点确定为该中心峰值变换点,将其他所有的峰值变换点确定为非峰值变换点。
S403、判断所述非峰值变换点是否围绕所述中心峰值变换点对称分布,若是,则判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍,否则,判断对应的待识别的图片或者视频为非翻拍。
在其中一个实施例中,如图5-a对照图5-b所示,进一步分析图5-b示出的翻拍的图片或者视频中的峰值变换点的位置可知,非中心峰值变换点围绕该中心峰值变换点呈中心对称分布,因此,可以将这一特征作为判断对应的待识别的图片或者视频是否为翻拍的进一步限定条件,以进一步提高翻拍识别的准确性。
在其中一个实施例中,当接收到待识别的视频时,该网络翻拍照片的检测方法还包括:
通过所述目标检测模型识别同一待识别视频中每个所述傅里叶特征图谱中的峰值变换点的位置;
当识别的同一待识别视频中每个所述傅里叶特征图谱中的峰值变换点的个数均至少为两个时,根据所述峰值变换点的位置识别每个所述傅里叶特征图谱的中心峰值变换点及非中心峰值变换点。
在该实施例中,上述步骤判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍的步骤进一步包括:
判断每个所述傅里叶特征图谱中的所述非峰值变换点是否均围绕对应的中心峰值变换点呈对称分布,若是,则判断对应待识别的视频为翻拍,否则,判断对应的待识别的视频为非翻拍。
本实施例限定待识别的对象为视频时,通过从该视频中任意提取多个视频帧图片,对属于同一待识别视频中的各个帧图片均进行傅里叶特征图谱中的峰值变换点的位置识别、峰值变换点的个数判断、非中心峰值变换点是否围绕该中心峰值变换点呈对称分布的判断,当判断每个所述傅里叶特征图谱中的所述非峰值变换点均围绕对应的中心峰值变换点呈对称分布则判断对应待识别的视频为翻拍,以提高翻拍视频识别准确性。
应理解,上述实施例中各步骤的序号的大小并不意味着执行顺序的先后,各过程的执行顺序应以其功能和内在逻辑确定,而不应对本申请实施例的实施过程构成任何限定。
图6是本申请一实施例中网络翻拍照片的检测装置的结构示意图,下面结合图6详细描述根据本申请一实施例中网络翻拍照片的检测装置,如图6所示提供的网络翻拍照片的检测装置,该网络翻拍照片的检测装置与上述实施例中网络翻拍照片的检测方法一一对应。如图6所示,该网络翻拍照片的检测装置100包括提取模块11、生成模块12、识别模块13和判断模块14。
提取模块11,用于接收到待识别的图片或者视频时,提取所述图片的傅里叶特征值或者提取所述视频的视频帧图片的傅里叶特征值。
在其中一个实施例中,接收到待识别的图片或者视频包括原始的图片或者视频,还包括压缩的图片或者视频。
在该实施例中,所述视频帧图片可以是待识别视频中的任意一帧或者任意多帧。
生成模块12,用于根据提取的所述傅里叶特征值生成所述图片或者所述视频帧图片的傅里叶特征图谱。
识别模块13,用于通过预先训练好的目标检测模型识别所述傅里叶特征图谱中的峰值变换点的个数。
在其中一个实施例中,该目标检测模型可以是SSD(Single Shot MultiBox Detector,单镜头多盒检测器)目标检测模型,是一种端到端的单阶段目标检测网络,它将图片之上的检测目标视为是许多个default boxes默认框的离散化合成。
一般欲检测的图片当中会包含有着不同尺度大小及分辨率的目标。为了有效地涵盖这种种不同尺度、分辨率的目标检测,SSD扩大了其特征选取的范围。
判断模块14,用于当识别的所述峰值变换点的个数至少为两个且至少两个所述峰值变换点呈对称分布时,判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍。
图5-a是本申请一实施例中非翻拍的图片或者视频中的峰值变换点的位置,图5-b是本申请一实施例中翻拍的图片或者视频中的峰值变换点的位置,如图5-a对照5-b所示,通过多次试验发现,非翻拍的图片或视频中的峰值变换点为1个,而翻拍的图片或视频中的峰值变换点一般有很多个,故可以通过提取模块11提取该待识别的图片或视频的傅里叶特征图谱,识别模块13通过预先训练好的目标检测模型识别该傅里叶特征图谱中的峰值变换点的个数,最后通过判断模块14依据该峰值变换点的个数判断对应的图片或者视频是否为翻拍。
在其中一个实施例中,所述网络翻拍照片的检测装置100还包括:
第一生成单元,用于提取翻拍的样本图片的傅里叶特征值,根据翻拍的样本图片的傅里叶特征值生成翻拍的傅里叶特征图谱;
第二生成单元,用于提取非翻拍的样本图片的傅里叶特征值,根据非翻拍的样本图片的傅里叶特征值生成非翻拍的傅里叶特征图谱;
标注模块,用于接收所述翻拍的傅里叶特征图谱中峰值变换点的第一标注区域,接收所述非翻拍的傅里叶特征图谱中峰值变换点的第二标注区域;
训练模块,用于将所述翻拍的傅里叶特征图谱、所述第一标注区域、所述非翻拍的傅里叶特征图谱和所述第二标注区域输入到所述目标检测模型中进行训练,得到所述训练好的目标检测模型。
进一步地,在其中一个实施例中,所述网络翻拍照片的检测装置还包括:
位置识别模块,用于通过所述目标检测模型识别所述傅里叶特征图谱中的峰值变换点的位置;还用于当识别的所述峰值变换点的个数至少为两个时,根据所述峰值变换点的位置识别中心峰值变换点及非中心峰值变换点;
所述判断模块具体用于判断所述非峰值变换点是否围绕所述中心峰值变换点对称分布,若是,则判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍,否则,判断对应的待识别的图片或者视频为非翻拍。
在该实施例中,如图5-a对照图5-b所示,进一步分析图5-b示出的翻拍的图片或者视频中的峰值变换点的位置可知,非中心峰值变换点围绕该中心峰值变换点呈中心对称分布,因此,可以将这一特征作为判断对应的待识别的图片或者视频是否为翻拍的进一步限定条件,以进一步提高翻拍识别的准确性。
在其中一个实施例中,当接收到待识别的视频时:
所述提取模块具体用于从所述待识别的视频中任意提取多个视频帧图片,还用于提取每个所述视频帧图片的傅里叶特征值。
所述生成模块具体用于根据提取的所述傅里叶特征值生成同一待识别视频中每个所述视频帧图片的傅里叶特征图谱。
所述识别模块具体用于识别同一待识别视频中每个傅里叶特征图谱中的峰值变换点的个数。
所述判断模块具体用于当同一待识别的视频包括的每个所述视频帧图片中相应的傅里叶特征图谱的峰值变换点的个数均至少为两个时,判断对应的待识别视频为翻拍。
本实施例限定待识别的对象为视频时,通过从该视频中任意提取多个视频帧图片,对属于同一待识别视频中的各个帧图片均进行傅里叶特征图谱中的峰值变换点的位置识别、峰值变换点的个数判断、非中心峰值变换点是否围绕该中心峰值变换点呈对称分布的判断,当判断每个所述傅里叶特征图谱中的所述非峰值变换点均围绕对应的中心峰值变换点呈对称分布则判断对应待识别的视频为翻拍,以提高翻拍视频识别准确性。
 
关于网络翻拍照片的检测装置的具体限定可以参见上文中对于网络翻拍照片的检测方法的限定,在此不再赘述。上述网络翻拍照片的检测装置中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于计算机设备中的处理器中,也可以以软件形式存储于计算机设备中的存储器中,以便于处理器调用执行以上各个模块对应的操作。
在一个实施例中,提供了一种计算机设备,该计算机设备可以是终端,其内部结构图可以如图7所示。该计算机设备包括通过系统总线连接的处理器、存储器、网络接口、显示屏和输入装置。其中,该计算机设备的处理器用于提供计算和控制能力。该计算机设备的存储器包括可读存储介质、内存储器。该可读存储介质存储有操作系统和计算机可读指令。该内存储器为可读存储介质中的操作系统和计算机可读指令的运行提供环境。该计算机设备的网络接口用于与外部服务器通过网络连接通信。该计算机可读指令被处理器执行时以实现一种网络翻拍照片的检测方法。本示例中,可读存储介质可以是非易失性可读存储介质,也可以是易失性可读存储介质。
在一个实施例中,提供了一种计算机设备,包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机可读指令,处理器执行计算机可读指令时实现上述实施例中网络翻拍照片的检测方法的步骤,例如图2所示的步骤101至步骤104。或者,处理器执行计算机可读指令时实现上述实施例中网络翻拍照片的检测装置的各模块/单元的功能,例如图7所示模块11至模块14的功能。为避免重复,这里不再赘述。
在一个实施例中,提一个或多个存储有计算机可读指令的可读存储介质,所述计算机可读存储介质存储有计算机可读指令,其中,所述计算机可读指令被一个或多个处理器执行时,使得所述一个或多个处理器执行时实现上述实施例中网络翻拍照片的检测方法的步骤,例如图2所示的步骤101至步骤104。或者,计算机可读指令被处理器执行时实现上述实施例中网络翻拍照片的检测装置的各模块/单元的功能,例如图7所示模块11至模块14的功能。为避免重复,这里不再赘述。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程,是可以通过计算机可读指令来指令相关的硬件来完成,所述的计算机可读指令可存储于一非易失性计算机可读取存储介质中,该计算机可读指令在执行时,可包括如上述各方法的实施例的流程。其中,本申请所提供的各实施例中所使用的对存储器、存储、数据库或其它介质的任何引用,均可包括非易失性和/或易失性存储器。非易失性存储器可包括只读存储器(ROM)、可编程ROM(PROM)、电可编程ROM(EPROM)、电可擦除可编程ROM(EEPROM)或闪存。易失性存储器可包括随机存取存储器(RAM)或者外部高速缓冲存储器。作为说明而非局限,RAM以多种形式可得,诸如静态RAM(SRAM)、动态RAM(DRAM)、同步DRAM(SDRAM)、双数据率SDRAM(DDRSDRAM)、增强型SDRAM(ESDRAM)、同步链路(Synchlink) DRAM(SLDRAM)、存储器总线(Rambus)直接RAM(RDRAM)、直接存储器总线动态RAM(DRDRAM)、以及存储器总线动态RAM(RDRAM)等。
本实施例提供的网络翻拍照片的检测方法、装置、计算机设备及存储介质,通过提取所述图片的傅里叶特征值或者提取所述视频的视频帧图片的傅里叶特征值,然后根据提取的所述傅里叶特征值生成所述图片或者所述视频帧图片的傅里叶特征图谱,再通过预先训练好的目标检测模型识别所述傅里叶特征图谱中的峰值变换点的个数,最后根据识别的所述峰值变换点的个数判断对应的图片或者照片是否为翻拍,相比于现有的翻拍识别技术,本申请不需要提取RGB(红、绿、蓝)三通道的傅里叶频谱特征,而是直接生成傅里叶特征图谱,减少了翻拍识别的计算量,提高了翻拍识别的运算速度,另一方面本申请根据识别的所述峰值变换点的个数判断对应的图片或者照片是否为翻拍,提高了翻拍识别的准确率。
所属领域的技术人员可以清楚地了解到,为了描述的方便和简洁,仅以上述各功能单元、模块的划分进行举例说明,实际应用中,可以根据需要而将上述功能分配由不同的功能单元、模块完成,即将所述装置的内部结构划分成不同的功能单元或模块,以完成以上描述的全部或者部分功能。
以上所述实施例仅用以说明本申请的技术方案,而非对其限制;尽管参照前述实施例对本申请进行了详细的说明,本领域的普通技术人员应当理解:其依然可以对前述各实施例所记载的技术方案进行修改,或者对其中部分技术特征进行等同替换;而这些修改或者替换,并不使相应技术方案的本质脱离本申请各实施例技术方案的精神和范围,均应包含在本申请的保护范围之内。

Claims (20)

  1. 一种网络翻拍照片的检测方法,其中,所述方法包括:
    接收到待识别的图片或者视频时,提取所述图片的傅里叶特征值或者提取所述视频的视频帧图片的傅里叶特征值;
    根据提取的所述傅里叶特征值生成所述图片或者所述视频帧图片的傅里叶特征图谱;
    通过预先训练好的目标检测模型识别所述傅里叶特征图谱中的峰值变换点的个数;
    当识别的所述峰值变换点的个数至少为两个且至少两个所述峰值变换点呈对称分布时,判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍。
  2. 根据权利要求1所述的网络翻拍照片的检测方法,其中,训练所述目标检测模型的步骤包括:
    提取翻拍的样本图片的傅里叶特征值,根据翻拍的样本图片的傅里叶特征值生成翻拍的傅里叶特征图谱;
    提取非翻拍的样本图片的傅里叶特征值,根据非翻拍的样本图片的傅里叶特征值生成非翻拍的傅里叶特征图谱;
    接收所述翻拍的傅里叶特征图谱中峰值变换点的第一标注区域,接收所述非翻拍的傅里叶特征图谱中峰值变换点的第二标注区域;
    将所述翻拍的傅里叶特征图谱、所述第一标注区域、所述非翻拍的傅里叶特征图谱和所述第二标注区域输入到所述目标检测模型中进行训练,得到所述训练好的目标检测模型。
  3. 根据权利要求1所述的网络翻拍照片的检测方法,其中,所述方法还包括:
    通过所述目标检测模型识别所述傅里叶特征图谱中的峰值变换点的位置;
    所述判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍的步骤进一步包括:
    当识别的所述峰值变换点的个数至少为两个时,根据所述峰值变换点的位置识别中心峰值变换点及非中心峰值变换点;
    判断所述非峰值变换点是否围绕所述中心峰值变换点对称分布,若是,则判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍,否则,判断对应的待识别的图片或者视频为非翻拍。
  4. 根据权利要求1所述的网络翻拍照片的检测方法,其中,当接收到待识别的视频时,所述提取所述图片的傅里叶特征值或者提取所述视频的视频帧图片的傅里叶特征值的步骤包括:
    从所述待识别的视频中任意提取多个视频帧图片;
    提取每个所述视频帧图片的傅里叶特征值;
    所述根据提取的所述傅里叶特征值生成所述图片或者所述视频帧图片的傅里叶特征图谱的步骤进一步包括:
    根据提取的所述傅里叶特征值生成同一待识别视频中每个所述视频帧图片的傅里叶特征图谱;
    所述当识别的所述峰值变换点的个数至少为两个时,判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍的步骤进一步包括:
    当同一待识别的视频包括的每个所述视频帧图片中相应的傅里叶特征图谱的峰值变换点的个数均至少为两个时,判断对应的待识别视频为翻拍。
  5. 根据权利要求4所述的网络翻拍照片的检测方法,其中,所述方法还包括:
    通过所述目标检测模型识别同一待识别视频中每个所述傅里叶特征图谱中的峰值变换点的位置;
    当识别的同一待识别视频中每个所述傅里叶特征图谱中的峰值变换点的个数均至少为两个时,根据所述峰值变换点的位置识别每个所述傅里叶特征图谱的中心峰值变换点及非中心峰值变换点;
    所述判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍的步骤进一步包括:
    判断每个所述傅里叶特征图谱中的所述非峰值变换点是否均围绕对应的中心峰值变换点呈对称分布,若是,则判断对应待识别的视频为翻拍,否则,判断对应的待识别的图片或者视频为非翻拍。
  6. 一种网络翻拍照片的检测装置,其中,所述装置包括:
    提取模块,用于接收到待识别的图片或者视频时,提取所述图片的傅里叶特征值或者提取所述视频的视频帧图片的傅里叶特征值;
    生成模块,用于根据提取的所述傅里叶特征值生成所述图片或者所述视频帧图片的傅里叶特征图谱;
    识别模块,用于通过预先训练好的目标检测模型识别所述傅里叶特征图谱中的峰值变换点的个数;
    判断模块,用于当识别的所述峰值变换点的个数至少为两个且至少两个所述峰值变换点呈对称分布时,判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍。
  7. 根据权利要求6所述的网络翻拍照片的检测装置,其中,所述网络翻拍照片的检测装置还包括:
    第一生成单元,用于提取翻拍的样本图片的傅里叶特征值,根据翻拍的样本图片的傅里叶特征值生成翻拍的傅里叶特征图谱;
    第二生成单元,用于提取非翻拍的样本图片的傅里叶特征值,根据非翻拍的样本图片的傅里叶特征值生成非翻拍的傅里叶特征图谱;
    标注模块,用于接收所述翻拍的傅里叶特征图谱中峰值变换点的第一标注区域,接收所述非翻拍的傅里叶特征图谱中峰值变换点的第二标注区域;
    训练模块,用于将所述翻拍的傅里叶特征图谱、所述第一标注区域、所述非翻拍的傅里叶特征图谱和所述第二标注区域输入到所述目标检测模型中进行训练,得到所述训练好的目标检测模型。
  8. 根据权利要求6所述的网络翻拍照片的检测装置,其中,所述网络翻拍照片的检测装置还包括:
    位置识别模块,用于通过所述目标检测模型识别所述傅里叶特征图谱中的峰值变换点的位置,还用于当识别的所述峰值变换点的个数至少为两个时,根据所述峰值变换点的位置识别中心峰值变换点及非中心峰值变换点;
    所述判断模块具体用于判断所述非峰值变换点是否围绕所述中心峰值变换点对称分布,若是,则判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍,否则,判断对应的待识别的图片或者视频为非翻拍。
  9. 根据权利要求6所述的网络翻拍照片的检测装置,其中,当接收到待识别的视频时,
    所述提取模块,用于从所述待识别的视频中任意提取多个视频帧图片;还用于提取每个所述视频帧图片的傅里叶特征值;
    所述生成模块,用于根据提取的所述傅里叶特征值生成同一待识别视频中每个所述视频帧图片的傅里叶特征图谱;
    所述判断模块,用于当同一待识别的视频包括的每个所述视频帧图片中相应的傅里叶特征图谱的峰值变换点的个数均至少为两个时,判断对应的待识别视频为翻拍。
  10. 根据权利要求9所述的网络翻拍照片的检测装置,其中,所述网络翻拍照片的检测装置还包括:
    位置识别模块,用于通过所述目标检测模型识别同一待识别视频中每个所述傅里叶特征图谱中的峰值变换点的位置;当识别的同一待识别视频中每个所述傅里叶特征图谱中的峰值变换点的个数均至少为两个时,根据所述峰值变换点的位置识别每个所述傅里叶特征图谱的中心峰值变换点及非中心峰值变换点;
    所述判断模块,用于判断每个所述傅里叶特征图谱中的所述非峰值变换点是否均围绕对应的中心峰值变换点呈对称分布,若是,则判断对应待识别的视频为翻拍,否则,判断对应的待识别的图片或者视频为非翻拍。
  11. 一种计算机设备,包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机可读指令,其中,所述处理器执行所述计算机可读指令时实现如下步骤:
    接收到待识别的图片或者视频时,提取所述图片的傅里叶特征值或者提取所述视频的视频帧图片的傅里叶特征值;
    根据提取的所述傅里叶特征值生成所述图片或者所述视频帧图片的傅里叶特征图谱;
    通过预先训练好的目标检测模型识别所述傅里叶特征图谱中的峰值变换点的个数;
    当识别的所述峰值变换点的个数至少为两个且至少两个所述峰值变换点呈对称分布时,判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍。
  12. 根据权利要求11所述的计算机设备,其中,训练所述目标检测模型的步骤包括:
    提取翻拍的样本图片的傅里叶特征值,根据翻拍的样本图片的傅里叶特征值生成翻拍的傅里叶特征图谱;
    提取非翻拍的样本图片的傅里叶特征值,根据非翻拍的样本图片的傅里叶特征值生成非翻拍的傅里叶特征图谱;
    接收所述翻拍的傅里叶特征图谱中峰值变换点的第一标注区域,接收所述非翻拍的傅里叶特征图谱中峰值变换点的第二标注区域;
    将所述翻拍的傅里叶特征图谱、所述第一标注区域、所述非翻拍的傅里叶特征图谱和所述第二标注区域输入到所述目标检测模型中进行训练,得到所述训练好的目标检测模型。
  13. 根据权利要求11所述的计算机设备,其中,所述处理器执行所述计算机可读指令时还实现如下步骤:
    通过所述目标检测模型识别所述傅里叶特征图谱中的峰值变换点的位置;
    所述判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍的步骤进一步包括:
    当识别的所述峰值变换点的个数至少为两个时,根据所述峰值变换点的位置识别中心峰值变换点及非中心峰值变换点;
    判断所述非峰值变换点是否围绕所述中心峰值变换点对称分布,若是,则判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍,否则,判断对应的待识别的图片或者视频为非翻拍。
  14. 根据权利要求11所述的计算机设备,其中,当接收到待识别的视频时,所述提取所述图片的傅里叶特征值或者提取所述视频的视频帧图片的傅里叶特征值的步骤包括:
    从所述待识别的视频中任意提取多个视频帧图片;
    提取每个所述视频帧图片的傅里叶特征值;
    所述根据提取的所述傅里叶特征值生成所述图片或者所述视频帧图片的傅里叶特征图谱的步骤进一步包括:
    根据提取的所述傅里叶特征值生成同一待识别视频中每个所述视频帧图片的傅里叶特征图谱;
    所述当识别的所述峰值变换点的个数至少为两个时,判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍的步骤进一步包括:
    当同一待识别的视频包括的每个所述视频帧图片中相应的傅里叶特征图谱的峰值变换点的个数均至少为两个时,判断对应的待识别视频为翻拍。
  15. 根据权利要求14所述的计算机设备,其中,所述处理器执行所述计算机可读指令时还实现如下步骤:
    通过所述目标检测模型识别同一待识别视频中每个所述傅里叶特征图谱中的峰值变换点的位置;
    当识别的同一待识别视频中每个所述傅里叶特征图谱中的峰值变换点的个数均至少为两个时,根据所述峰值变换点的位置识别每个所述傅里叶特征图谱的中心峰值变换点及非中心峰值变换点;
    所述判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍的步骤进一步包括:
    判断每个所述傅里叶特征图谱中的所述非峰值变换点是否均围绕对应的中心峰值变换点呈对称分布,若是,则判断对应待识别的视频为翻拍,否则,判断对应的待识别的图片或者视频为非翻拍。
  16. 一个或多个存储有计算机可读指令的可读存储介质,所述计算机可读存储介质存储有计算机可读指令,其中,所述计算机可读指令被一个或多个处理器执行时,使得所述一个或多个处理器执行如下步骤:
    接收到待识别的图片或者视频时,提取所述图片的傅里叶特征值或者提取所述视频的视频帧图片的傅里叶特征值;
    根据提取的所述傅里叶特征值生成所述图片或者所述视频帧图片的傅里叶特征图谱;
    通过预先训练好的目标检测模型识别所述傅里叶特征图谱中的峰值变换点的个数;
    当识别的所述峰值变换点的个数至少为两个且至少两个所述峰值变换点呈对称分布时,判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍。
  17. 根据权利要求16所述的可读存储介质,其中,训练所述目标检测模型的步骤包括:
    提取翻拍的样本图片的傅里叶特征值,根据翻拍的样本图片的傅里叶特征值生成翻拍的傅里叶特征图谱;
    提取非翻拍的样本图片的傅里叶特征值,根据非翻拍的样本图片的傅里叶特征值生成非翻拍的傅里叶特征图谱;
    接收所述翻拍的傅里叶特征图谱中峰值变换点的第一标注区域,接收所述非翻拍的傅里叶特征图谱中峰值变换点的第二标注区域;
    将所述翻拍的傅里叶特征图谱、所述第一标注区域、所述非翻拍的傅里叶特征图谱和所述第二标注区域输入到所述目标检测模型中进行训练,得到所述训练好的目标检测模型。
  18. 根据权利要求16所述的可读存储介质,其中,所述计算机可读指令被一个或多个处理器执行时,使得所述一个或多个处理器还执行如下步骤:
    通过所述目标检测模型识别所述傅里叶特征图谱中的峰值变换点的位置;
    所述判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍的步骤进一步包括:
    当识别的所述峰值变换点的个数至少为两个时,根据所述峰值变换点的位置识别中心峰值变换点及非中心峰值变换点;
    判断所述非峰值变换点是否围绕所述中心峰值变换点对称分布,若是,则判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍,否则,判断对应的待识别的图片或者视频为非翻拍。
  19. 根据权利要求16所述的可读存储介质,其中,当接收到待识别的视频时,所述提取所述图片的傅里叶特征值或者提取所述视频的视频帧图片的傅里叶特征值的步骤包括:
    从所述待识别的视频中任意提取多个视频帧图片;
    提取每个所述视频帧图片的傅里叶特征值;
    所述根据提取的所述傅里叶特征值生成所述图片或者所述视频帧图片的傅里叶特征图谱的步骤进一步包括:
    根据提取的所述傅里叶特征值生成同一待识别视频中每个所述视频帧图片的傅里叶特征图谱;
    所述当识别的所述峰值变换点的个数至少为两个时,判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍的步骤进一步包括:
    当同一待识别的视频包括的每个所述视频帧图片中相应的傅里叶特征图谱的峰值变换点的个数均至少为两个时,判断对应的待识别视频为翻拍。
  20. 根据权利要求19所述的可读存储介质,其中,所述计算机可读指令被一个或多个处理器执行时,使得所述一个或多个处理器还执行如下步骤:
    通过所述目标检测模型识别同一待识别视频中每个所述傅里叶特征图谱中的峰值变换点的位置;
    当识别的同一待识别视频中每个所述傅里叶特征图谱中的峰值变换点的个数均至少为两个时,根据所述峰值变换点的位置识别每个所述傅里叶特征图谱的中心峰值变换点及非中心峰值变换点;
    所述判断所述傅里叶特征图谱对应的待识别的图片或者视频为翻拍的步骤进一步包括:
    判断每个所述傅里叶特征图谱中的所述非峰值变换点是否均围绕对应的中心峰值变换点呈对称分布,若是,则判断对应待识别的视频为翻拍,否则,判断对应的待识别的图片或者视频为非翻拍。
     
PCT/CN2021/070734 2020-02-28 2021-01-08 网络翻拍照片的检测方法、装置、计算机设备及存储介质 WO2021169625A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010127392.3A CN111428740A (zh) 2020-02-28 2020-02-28 网络翻拍照片的检测方法、装置、计算机设备及存储介质
CN202010127392.3 2020-02-28

Publications (1)

Publication Number Publication Date
WO2021169625A1 true WO2021169625A1 (zh) 2021-09-02

Family

ID=71547808

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/070734 WO2021169625A1 (zh) 2020-02-28 2021-01-08 网络翻拍照片的检测方法、装置、计算机设备及存储介质

Country Status (2)

Country Link
CN (1) CN111428740A (zh)
WO (1) WO2021169625A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113222952A (zh) * 2021-05-20 2021-08-06 支付宝(杭州)信息技术有限公司 翻拍图像的识别方法及装置

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111428740A (zh) * 2020-02-28 2020-07-17 深圳壹账通智能科技有限公司 网络翻拍照片的检测方法、装置、计算机设备及存储介质
CN112364856A (zh) * 2020-11-13 2021-02-12 润联软件系统(深圳)有限公司 翻拍图像识别方法、装置、计算机设备及存储介质

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105957080A (zh) * 2016-04-28 2016-09-21 王超维 基于频域对屏幕拍摄身份证照的识别方法
CN107154024A (zh) * 2017-05-19 2017-09-12 南京理工大学 基于深度特征核相关滤波器的尺度自适应目标跟踪方法
CN109472768A (zh) * 2018-09-19 2019-03-15 上海泛洲信息科技有限公司 一种使用频谱分析区别实物与非实物平面影像的方法
CN111428740A (zh) * 2020-02-28 2020-07-17 深圳壹账通智能科技有限公司 网络翻拍照片的检测方法、装置、计算机设备及存储介质

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105957080A (zh) * 2016-04-28 2016-09-21 王超维 基于频域对屏幕拍摄身份证照的识别方法
CN107154024A (zh) * 2017-05-19 2017-09-12 南京理工大学 基于深度特征核相关滤波器的尺度自适应目标跟踪方法
CN109472768A (zh) * 2018-09-19 2019-03-15 上海泛洲信息科技有限公司 一种使用频谱分析区别实物与非实物平面影像的方法
CN111428740A (zh) * 2020-02-28 2020-07-17 深圳壹账通智能科技有限公司 网络翻拍照片的检测方法、装置、计算机设备及存储介质

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113222952A (zh) * 2021-05-20 2021-08-06 支付宝(杭州)信息技术有限公司 翻拍图像的识别方法及装置
CN113222952B (zh) * 2021-05-20 2022-05-24 蚂蚁胜信(上海)信息技术有限公司 翻拍图像的识别方法及装置

Also Published As

Publication number Publication date
CN111428740A (zh) 2020-07-17

Similar Documents

Publication Publication Date Title
WO2021169625A1 (zh) 网络翻拍照片的检测方法、装置、计算机设备及存储介质
WO2020037898A1 (zh) 人脸特征点检测方法、装置、计算机设备及存储介质
WO2020151329A1 (zh) 基于目标检测的识别框确定方法、装置及终端设备
WO2018233038A1 (zh) 基于深度学习的车牌识别方法、装置、设备及存储介质
EP2864933B1 (en) Method, apparatus and computer program product for human-face features extraction
US8750573B2 (en) Hand gesture detection
WO2021164306A1 (zh) 图像分类模型的训练方法、装置、计算机设备及存储介质
CN109344727B (zh) 身份证文本信息检测方法及装置、可读存储介质和终端
CN112052831B (zh) 人脸检测的方法、装置和计算机存储介质
US10713470B2 (en) Method of determining image background, device for determining image background, and a non-transitory medium for same
US20120027252A1 (en) Hand gesture detection
US11900676B2 (en) Method and apparatus for detecting target in video, computing device, and storage medium
US11151365B2 (en) People detection system with feature space enhancement
CN111310826B (zh) 样本集的标注异常检测方法、装置及电子设备
WO2020211396A1 (zh) 静默式活体图片识别方法、装置、计算机设备和存储介质
WO2020001219A1 (zh) 图像处理方法和装置、存储介质、电子设备
WO2019076187A1 (zh) 视频遮蔽区域选取方法、装置、电子设备及系统
US20180047271A1 (en) Fire detection method, fire detection apparatus and electronic equipment
WO2021169616A1 (zh) 非活体人脸的检测方法、装置、计算机设备及存储介质
WO2023284608A1 (zh) 字符识别模型生成方法、装置、计算机设备和存储介质
WO2022194079A1 (zh) 天空区域分割方法、装置、计算机设备和存储介质
WO2022171011A1 (zh) 视频审核模型训练方法、视频审核方法及相关装置
JP2022133378A (ja) 顔生体検出方法、装置、電子機器、及び記憶媒体
WO2022257314A1 (zh) 图像检测方法和相关训练方法及相关装置、设备及介质
US20200272808A1 (en) Method and system for face detection

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: 21761729

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

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 1205A DATED 12.01.2023)

122 Ep: pct application non-entry in european phase

Ref document number: 21761729

Country of ref document: EP

Kind code of ref document: A1