Video tracing fingerprint processing method and system and medium

Abstract

The invention provides a video tracing fingerprint processing method and system and a medium. The method comprises the following steps: determining scene main body distribution and key frame positioning according to a target video, selecting a watermark embedding candidate region, embedding a watermark sequence of the target video into a key frame and carrying out adaptive intensity static watermark embedding, carrying out dynamic watermark embedding on a dynamic watermark embedding region according to embedding density, and carrying out dynamic watermark embedding on a dynamic watermark embedding region according to embedding density. Performing multi-dimensional feature integration on the target video according to a preset feature fusion algorithm to generate feature fingerprint data, then embedding the check code and the feature fingerprint data into an audio stream data area of a target embedding area and a filling bit of a video stream respectively, and segmenting and embedding the feature fingerprint to obtain an embedded feature fingerprint; monitoring an object video, verifying static and dynamic watermarks and decoding feature fingerprints, judging and calibrating an infringement video, and reporting the infringement video; therefore, by adding the static and dynamic watermarks to the video and embedding the feature fingerprint information in the video data stream, accurate tracing of the video is realized, and the video network security is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of video copyright protection and traceability, in particular to a video traceability fingerprint processing method, system and medium. BACKGROUND

[0002] With the rapid growth of digital media, the compliant propagation and tamper-proofing of network video content are particularly important. The tracing of the source of video data, the tracking of the propagation, and the identification of the playback end affect network security. For example, in terms of copyright protection, it is difficult to accurately track the source of video leakage, and in the field of video monitoring, it is impossible to accurately determine the specific links of video propagation and related client information. Traditional single watermarking technology is difficult to effectively meet the demand for accurate traceability in the face of complex video processing and transmission environment. Existing traceability methods have problems such as easily noticeable watermark, insufficient robustness, and limited traceability information, and cannot perform deep tracking and accurate determination on the complete propagation chain of the video.

[0003] In view of the above problems, there is an urgent need for effective technical solutions. SUMMARY

[0004] The purpose of the present application is to provide a video traceability fingerprint processing method, system and medium, which can realize accurate traceability of video by adding static and dynamic watermarks to video and embedding feature fingerprint information in video data stream, thereby effectively protecting video network security.

[0005] The first aspect of the present application provides a video traceability fingerprint processing method, comprising the following steps: Obtaining attribute information and feature information of a target video region, positioning and marking key frames according to scene subject distribution information and key frame positioning information, and selecting watermark embedding candidate regions according to key frame texture recognition degree; Obtaining the watermark sequence of the target video, and embedding the key frames of the target embedding region screened out, and adaptively embedding the watermark according to the embedding strength; Obtaining the dynamic watermark factor of the target video and encrypting to generate a dynamic watermark pattern sequence, and dynamically embedding the dynamic watermark in the dynamic watermark embedding region screened out from the watermark embedding candidate region according to the embedding density; According to a preset feature fusion algorithm, the metadata, playback record and processing log of the target video are fused in multiple dimensions to generate feature fingerprint data and a feature fingerprint data sequence; According to the feature fingerprint data sequence, a check code is generated, the check code is embedded in the audio stream data region of the target embedding region according to the encoding embedding density, the feature fingerprint data is embedded in the padding bits of the video stream, the feature fingerprint is segmented and regularly embedded; The object video is dynamically monitored, and static watermark, dynamic watermark and decoding feature fingerprint are extracted and compared and verified, so as to judge and report the infringing warning video.

[0006] Optionally, in the video source fingerprint processing method, the attribute information and the feature information of the target video are obtained, the key frame is positioned and marked according to the scene main body distribution information and the key frame positioning information, and the watermark embedding candidate region is selected according to the key frame texture recognition degree, and the method comprises the following steps: The target video is preprocessed to obtain video distribution information, including resolution, frame rate and bit rate; The video distribution information is processed by the trained preset feature extraction model to obtain the attribute information and the feature information of the video region; The attribute information includes object category and scene category, and the feature information includes edge information and texture information; The feature information is processed according to the attribute information matched with the corresponding key frame recognition model to obtain scene main body distribution information and key frame positioning information, and the key frame of each scene main body is positioned and marked; The texture information corresponding to the marked key frame is processed according to the preset texture analysis model to obtain texture feature data, including contrast, correlation and energy value, and the texture recognition degree is obtained by weighted processing, and the watermark embedding candidate region is selected.

[0007] Optionally, in the video source fingerprint processing method, the watermark sequence of the target video is obtained, and the key frame of the target embedding region is embedded, and the watermark is adaptively embedded according to the embedding strength, and the method comprises the following steps: The source identification information of the target video is obtained, including production code, copyright ID and resource identification code; The source identification information is processed according to the preset watermark generation algorithm to generate a watermark sequence; The frame rate of the key frame of the watermark embedding candidate region is processed to obtain a frequency distribution coefficient, and the corresponding target embedding region is selected according to the low frequency coefficient; The generated watermark sequence is embedded in the low frequency coefficient of the key frame of the target embedding region by the preset frequency domain method; The brightness and contrast are extracted according to the texture information of the key frame, and the video sensitivity of the target embedding region is obtained, and the corresponding embedding strength is obtained according to the video sensitivity table; The target embedding region is adaptively embedded according to the embedding strength.

[0008] Optionally, in the video source fingerprint processing method provided in the application, the dynamic watermark factor of the target video is obtained and encrypted to generate a dynamic watermark pattern sequence, and the dynamic watermark embedding region screened out from the watermark embedding candidate region according to the embedding density is subjected to dynamic watermark embedding, comprising: The timestamp of the target video and the key frame hash value of the watermark embedding candidate region are obtained, and a random number generated in real time is obtained; The dynamic watermark factor is generated according to the key frame hash value, the timestamp and the random number; The dynamic watermark factor is encrypted according to a preset encryption algorithm to generate a dynamic watermark pattern sequence; The time interval and content change rate of the key frame of the watermark embedding candidate region are obtained, and compared with the preset interval threshold and change rate threshold, if both are greater than the threshold, it is determined as a dynamic watermark embedding region; The corresponding embedding density is obtained according to the time interval and content change rate, and the dynamic watermark embedding region is subjected to dynamic watermark embedding according to the embedding density.

[0009] Optionally, in the video source fingerprint processing method provided in the application, the metadata, playback record and processing log of the target video are subjected to multi-dimensional feature weight fusion according to a preset feature fusion algorithm to generate feature fingerprint data and feature fingerprint data sequence, comprising: The metadata, playback record and processing log of the target video are obtained; The metadata includes encoding format, encoding level and sampling rate, the playback record includes playback identification, playback time and playback platform, and the processing log includes editing time, encoding parameter and special effect type; The metadata, playback record and processing log of the target video are subjected to multi-dimensional feature weight fusion according to a preset feature fusion algorithm to generate feature fingerprint data and feature fingerprint data sequence.

[0010] Optionally, in the video source fingerprint processing method provided in the application, the feature fingerprint data sequence is generated according to the feature fingerprint data sequence, the audio stream data region of the target embedding region is embedded with the check code according to the encoding embedding density, the feature fingerprint data is embedded into the padding bits of the video stream, the feature fingerprint is segmented and regularly embedded, comprising: The feature fingerprint data sequence is encoded to generate a check code; The corresponding encoding embedding density is obtained according to the video sensitivity and capacity of the target embedding region; The check code is embedded into the audio stream data region of the target embedding region according to the encoding embedding density; embedding the feature fingerprint data into padding bits of a video stream of the target embedding region according to a preset randomization embedding strategy; segmenting the feature fingerprint according to a preset block embedding chroma component, and embedding according to a preset random mask rule.

[0011] Optionally, in the video source tracking fingerprint processing method described in the present application, the dynamic monitoring of the object video and the extraction of the static watermark, the dynamic watermark, and the decoded feature fingerprint for comparison and verification, the judgment of the infringement warning video, and the reporting, comprise: dynamically monitoring the object video through a pre-deployed distributed monitoring node; tracking the object video when an abnormal propagation behavior is monitored or a tracking request is received; extracting the static watermark, the dynamic watermark, and the decoded feature fingerprint of the object video; comparing the similarity of the extracted static watermark, the dynamic watermark, and the corresponding watermark of the target video through a similarity algorithm, and marking the object video that does not meet the similarity comparison requirement as a suspected infringement video; comparing and verifying the feature fingerprint according to the feature fingerprint information of the target video, marking the suspected infringement video that fails the verification as an infringement warning video, and reporting to a tracking database.

[0012] In a second aspect, the present application provides a video source tracking fingerprint processing system, which comprises: a video preprocessing module for identifying the scene subject distribution of the target video video region, positioning the key frame, and screening the watermark embedding candidate region; a static watermark generation and embedding module for generating a watermark sequence according to the target video, and adaptively embedding according to the embedding strength; a dynamic watermark generation and embedding module for generating a dynamic watermark pattern sequence and dynamically embedding the dynamic watermark in the dynamic watermark embedding region according to the embedding density; a feature fingerprint information extraction and integration module for multi-dimensional feature integration of the metadata, the play record, and the processing log of the target video, generating feature fingerprint data and a feature fingerprint data sequence; a feature fingerprint information embedding module for embedding the feature fingerprint data into the padding bits of the video stream according to the encoding embedding density, embedding the feature fingerprint data into the padding bits of the video stream, and segmenting and embedding the feature fingerprint; a video monitoring and tracking identification module for dynamically monitoring the object video and comparing and verifying according to the watermark and the decoded feature fingerprint, judging and marking the infringement video, and reporting.

[0013] Optionally, in the video source fingerprint processing system provided in the application, the system further comprises a memory and a processor, the memory comprises a video source fingerprint processing method program, and the video source fingerprint processing method program is executed by the processor to implement the following steps: obtaining attribute information and feature information of a video region of the target video, positioning and marking the key frame according to the scene subject distribution information and the key frame positioning information, and selecting a watermark embedding candidate region according to a key frame texture recognition degree; obtaining a watermark sequence of the target video, and embedding the key frame in the target embedding region screened out, and adaptively embedding the watermark according to an embedding intensity; obtaining a dynamic watermark factor of the target video, and generating a dynamic watermark pattern sequence by encryption, and dynamically embedding the dynamic watermark in the dynamic watermark embedding region screened out from the watermark embedding candidate region according to an embedding density; performing multi-dimensional feature weight fusion on metadata, playback records and processing logs of the target video according to a preset feature fusion algorithm, generating feature fingerprint data and a feature fingerprint data sequence; generating a check code according to the feature fingerprint data sequence, embedding the check code in an audio stream data region of the target embedding region according to an encoding embedding density, embedding the feature fingerprint data in a padding bit of a video stream, segmenting and regularly embedding the feature fingerprint; performing dynamic monitoring on the object video, extracting static watermarks, dynamic watermarks and decoded feature fingerprints, and comparing and verifying the extracted static watermarks, dynamic watermarks and decoded feature fingerprints, judging a target infringement alarm video, and reporting the target infringement alarm video.

[0014] In a third aspect, the application further provides a computer readable storage medium, wherein a video source fingerprint processing method program is stored in the computer readable storage medium, and the video source fingerprint processing method program is executed by a processor to implement the steps of the video source fingerprint processing method according to any one of the above.

[0015] As can be seen from the above, the video source fingerprint processing method, system and medium provided in the application determine scene subject distribution and key frame positioning according to a target video, select a watermark embedding candidate region, embed a watermark sequence of the target video in a key frame and perform adaptive intensity static watermark embedding, perform dynamic watermark embedding on a dynamic watermark embedding region according to an embedding density, perform multi-dimensional feature integration on the target video according to a preset feature fusion algorithm to generate feature fingerprint data, embed a check code and the feature fingerprint data in an audio stream data region of a target embedding region and a padding bit of a video stream respectively, segment and embed the feature fingerprint, monitor an object video, and verify static watermarks, dynamic watermarks and decoded feature fingerprints, judge a target infringement video, and report the target infringement video; thus, by adding static watermarks and dynamic watermarks to a video and embedding feature fingerprint information in a video data stream, precise video source tracing is achieved, and video network security is ensured.

[0016] Other features and advantages of the present application will be set forth in the following description, and in part will be apparent from the description, or can be learned by practice of the application. The objects and other advantages of the present application will be realized and attained by the structure particularly pointed out in the written description and claims thereof as well as the appended drawings. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments of the present application. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be considered as a limitation to the scope, and for those skilled in the art, other related drawings can also be obtained without creative labor.

[0018] Figure 1 The flow chart of the video source fingerprint processing method provided by the embodiments of the present application; Figure 2 The flow chart of positioning and marking key frames and selecting watermark embedding candidate regions of the video source fingerprint processing method provided by the embodiments of the present application; Figure 3 The flow chart of static watermark adaptive strength embedding of the video source fingerprint processing method provided by the embodiments of the present application; Figure 4 The system diagram of the video source fingerprint processing system provided by the embodiments of the present application. DETAILED DESCRIPTION

[0019] The technical solutions of the embodiments of the present application will be described clearly and completely in the following with reference to the drawings of the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. The components of the embodiments of the present application described and shown in the drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of the present application.

[0020] It should be noted that similar reference numerals and letters indicate similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. Meanwhile, in the description of the present application, the terms “first”, “second”, etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0021] In the first aspect,Figure 1 , Figure 1 is a flowchart of a video source fingerprint processing method in some embodiments of the present application. The video source fingerprint processing method is used in a terminal device, such as a computer, a mobile phone terminal, etc. The video source fingerprint processing method includes the following steps: S11, attribute information and feature information of a target video region are obtained, key frames are positioned and marked according to scene subject distribution information and key frame positioning information, and a watermark embedding candidate region is selected according to key frame texture recognition degree; S12, a watermark sequence of the target video is obtained, and the key frames of the target embedding region are embedded, and the watermark is adaptively embedded according to embedding strength; S13, dynamic watermark factors of the target video are obtained and a dynamic watermark pattern sequence is generated by encryption, and dynamic watermark embedding is performed on the dynamic watermark embedding region screened from the watermark embedding candidate region according to embedding density; S14, according to a preset feature fusion algorithm, multi-dimensional feature weight fusion is performed on the metadata, playback records and processing logs of the target video to generate feature fingerprint data and a feature fingerprint data sequence; S15, a check code is generated according to the feature fingerprint data sequence, the check code is embedded in the audio stream data region of the target embedding region according to the encoding embedding density, the feature fingerprint data is embedded in the padding bits of the video stream, the feature fingerprint is segmented and regularly embedded; S16, the target video is dynamically monitored, and static watermark, dynamic watermark and decoded feature fingerprint are compared and verified, and the infringement warning video is judged and reported.

[0022] In order to realize accurate video tracing and effectively protect video network security, static and dynamic watermarks are added to the video, and feature fingerprint information is embedded in the video data stream to realize video security information embedding and anti-infringement tracing means. The scene subject distribution and key frame positioning are determined according to the target video, the watermark embedding candidate region is selected, the key frames are embedded according to the watermark sequence of the target video and adaptively embedded with static watermark, the dynamic watermark embedding region is dynamically embedded according to the embedding density, the feature fingerprint data is generated by multi-dimensional feature integration of the target video according to the preset feature fusion algorithm, and the check code and the feature fingerprint data are respectively embedded in the audio stream data region of the target embedding region and the padding bits of the video stream. The feature fingerprint is segmented and embedded, the target video is monitored, and the static and dynamic watermarks and the decoded feature fingerprint are verified, the infringement video is judged and reported, and the network application security supervision technology of the video is realized.

[0023] Please refer to Figure 2 , Figure 2is a flowchart of positioning and marking key frames and selecting watermark embedding candidate regions in a video forgery fingerprint processing method in some embodiments of the present application. According to an embodiment of the present application, the attribute information and feature information of the video region of the target video are obtained, the key frames are positioned and marked according to the scene subject distribution information and key frame positioning information, and the watermark embedding candidate region is selected according to the key frame texture recognition degree, including: S21, pre-processing the target video to obtain video distribution information, including resolution, frame rate and bit rate; S22, processing the video distribution information by using the trained preset feature extraction model to obtain attribute information and feature information of the video region; S23, the attribute information includes object category and scene category, and the feature information includes edge information and texture information; S24, processing the feature information according to the attribute information matched corresponding key frame recognition model to obtain scene subject distribution information and key frame positioning information, and positioning and marking the key frames of each scene subject; S25, processing the texture information corresponding to the marked key frames according to the preset texture analysis model to obtain texture feature data including contrast, correlation and energy value, and weighting processing to obtain texture recognition degree, and selecting watermark embedding candidate region.

[0024] Among them, the feature extraction model suitable for video feature extraction task is generated by learning model such as PyTorch and ResNet, the video distribution information of each distribution region in the video is input into the feature extraction model for processing and recognition, the attribute feature information and feature information of the video region are obtained, including object category, scene category, edge information and texture information, the matching key frame recognition model is selected according to the scene category of the video region, such as event scene, the object detection model with subject processing function such as YOLO can be used, and then the edge and texture features of the video are detected and positioned by using the matching model, the position distribution of the key frame where the foreground subject is located is obtained, that is, the distribution relationship between the scene subject and the key frame positioning is obtained, the key frame of the scene subject is positioned and marked, the key frame position of the scene subject of the video region is obtained, and then the texture feature data of the key frame is recognized by using the local binary algorithm of texture analysis, the contrast, correlation and energy value are obtained, and the texture recognition degree is obtained by using the preset weighting method, and the watermark embedding region is selected according to the recognition degree, so as to improve the hiding effect of the watermark.

[0025] Please refer to Figure 3 , Figure 3is a flowchart of adaptive strength embedding of static watermark in a video source fingerprint processing method in some embodiments of the present application. According to an embodiment of the present application, the watermark sequence of the target video is obtained, and the key frame of the target embedding area is embedded. The watermark is adaptively embedded according to the embedding strength, including: S31, obtaining the source identification information of the target video, including the producer code, the copyright party ID and the resource identification code; S32, processing the source identification information according to a preset watermark generation algorithm to generate a watermark sequence; S33, processing according to the frame rate of the key frame of the watermark embedding candidate area to obtain the frequency distribution coefficient, and screening the corresponding target embedding area of the qualified low frequency coefficient; S34, embedding the generated watermark sequence into the low frequency coefficient of the key frame of the target embedding area by a preset frequency domain method; S35, extracting brightness and contrast according to the texture information of the key frame, obtaining the video sensitivity of the target embedding area, and obtaining the corresponding embedding strength according to the video sensitivity; S36, embedding the watermark in the target embedding area according to the embedding strength.

[0026] Among them, the watermark generation algorithm based on pseudo-random sequence is used to identify and encode the source identification information to obtain a binary watermark sequence. The frequency distribution coefficient is obtained by multiplying the frame rate of the key frame of the watermark embedding candidate area and the sampling rate of each frame. The threshold is set to screen out the qualified low frequency coefficient and the corresponding target embedding area. The watermark sequence is embedded into the low frequency coefficient of the key frame of the target embedding area by using the frequency domain method of discrete cosine transform. These coefficients are relatively stable in video compression, so the robustness of the watermark can be ensured. The brightness and contrast extracted from the texture information of the key frame are processed to obtain the video sensitivity of the target embedding area. The corresponding embedding strength is obtained according to the video sensitivity table. The embedding strength is adaptively adjusted according to the visual sensitivity to improve the detectability of the watermark signal in the high sensitivity area of the watermark invisibility, and the static watermark embedding is realized.

[0027] According to an embodiment of the present application, the dynamic watermark factor of the target video is obtained, and a dynamic watermark pattern sequence is generated by encryption. The dynamic watermark embedding area screened out from the watermark embedding candidate area is dynamically embedded according to the embedding density, including: Obtaining the timestamp of the target video and the key frame hash value of the watermark embedding candidate area, and obtaining the random number generated in real time; Generating a dynamic watermark factor according to the key frame hash value, the timestamp and the random number; Encrypting the dynamic watermark factor according to a preset encryption algorithm to generate a dynamic watermark pattern sequence; The time interval and content change rate of the key frame of the watermark embedding candidate region are acquired and compared with preset interval threshold and change rate threshold, and if both comparison relations are greater than the threshold, the dynamic watermark embedding region is determined. The corresponding embedding density is obtained according to the time interval and content change rate, and the dynamic watermark embedding region is dynamically embedded according to the embedding density.

[0028] The hash value of the key frame is calculated by acquiring the current timestamp of the target video and using the hash algorithm MD5, a random number is obtained by using a random number generator such as Java, and the three are concatenated into a string and encoded as a watermark factor by using a preset sorting concatenation method, or a combination value is obtained by using a preset operation rule as a watermark value factor, the watermark value factor is encrypted by using a preset encryption algorithm such as AES algorithm to generate a unique dynamic watermark pattern sequence, and then the time interval and content change rate of the key frame of the watermark embedding candidate region are used to dynamically embed the watermark, if both are greater than the corresponding threshold, the dynamic embedding position of the watermark is determined, and then the corresponding embedding density is obtained according to the table lookup result to realize the real-time adaptive embedding of the dynamic watermark.

[0029] According to the embodiment of the application, the metadata, the playback record and the processing log of the target video are fused according to a preset feature fusion algorithm to generate feature fingerprint data and a feature fingerprint data sequence, including: The metadata, the playback record and the processing log of the target video are acquired. The metadata includes encoding format, encoding level and sampling rate, the playback record includes playback identification, playback time and playback platform, and the processing log includes editing time, encoding parameter and special effect type. The metadata, the playback record and the processing log of the target video are fused according to a preset feature fusion algorithm to generate feature fingerprint data and a feature fingerprint data sequence.

[0030] The metadata, playback records, and processing logs of the target video are obtained by a preset reading extraction tool, for example, the static features such as the encoding format, the encoding level, and the sampling rate are obtained by using a reading video tool, the playback records including the playback identifier, the playback time, and the playback platform are extracted, the dynamic features such as the playback frequency, the time length, and the platform distribution are calculated by using time sequence analysis, the clip time, the encoding parameter, and the special effect type are extracted by using a log analysis tool, the above scattered features are integrated by using a preset feature fusion algorithm according to a preset data structure, the feature fingerprint data sequence is formed, the comprehensive feature fingerprint is formed by combining the features according to a certain weight through a feature engineering method, the feature fingerprint data is encoded in a JSON format, and the feature fingerprint data sequence is generated, so that the readability and the expandability of the data are ensured.

[0031] According to the embodiment of the application, the feature fingerprint data sequence is encoded to generate a check code, the check code is embedded into the audio stream data region of the target embedding region according to the encoding embedding density, the feature fingerprint data is embedded into the padding bits of the video stream according to a preset randomization embedding strategy, and the feature fingerprint is segmented according to a preset block embedding chroma component and is embedded according to a preset random mask rule. The feature fingerprint data sequence is encoded to generate a check code; The corresponding encoding embedding density is obtained by combining the video sensitivity of the target embedding region with the capacity lookup table; The check code is embedded into the audio stream data region of the target embedding region according to the encoding embedding density; The feature fingerprint data is embedded into the padding bits of the video stream of the target embedding region according to a preset randomization embedding strategy; The feature fingerprint is segmented according to a preset block embedding chroma component and is embedded according to a preset random mask rule.

[0032] According to the embodiment of the application, the feature fingerprint data is respectively embedded into the data region of the audio stream and the padding bits of the video stream of the target embedding region by using a data hiding technology, the feature fingerprint is segmented according to an embedding chroma component and is embedded, the feature fingerprint data sequence is encoded to generate a check code, the matching embedding density is determined according to the video sensitivity and the region capacity of the target embedding region, the encoded check code is embedded, the feature fingerprint data is distributed and embedded into the bits of the video stream by using a randomization embedding strategy, the feature fingerprint is segmented into blocks of MxN according to the block embedding chroma component, and the data is mixed and embedded under the random mask rule to ensure the visual invisibility of the embedded data, so that the feature fingerprint is combined with the video data by optimizing the embedding, and the high integrity and accuracy of the feature fingerprint can be maintained after the video is subjected to multiple transcoding, editing, conversion, and other operations.

[0033] According to the embodiment of the present application, the object video is dynamically monitored and static watermark, dynamic watermark and decoding feature fingerprint are extracted and compared and verified, the infringing warning video is judged and reported, comprising: The object video is dynamically monitored by the pre-deployed distributed monitoring node; The object video is traced when abnormal propagation behavior is monitored or a trace request is received; The static watermark, dynamic watermark and decoding feature fingerprint of the object video are extracted; The similarity of the extracted static watermark and dynamic watermark is compared with the corresponding watermark of the target video by a similarity algorithm, and the object video that does not meet the similarity comparison requirement is marked as a suspected infringing video; The feature fingerprint is compared and verified according to the feature fingerprint information of the target video, and the suspected infringing video that fails to pass the verification is marked as an infringing warning video and reported to the trace database.

[0034] In the process of monitoring the network video, the distributed monitoring node distributed in the video transmission network uses real-time video stream analysis technology to dynamically monitor each object video, and when abnormal propagation behavior (such as unauthorized sharing and content tampering) is detected or a trace request is received, the trace is automatically started, the static watermark, dynamic watermark and decoding feature fingerprint in the object video are extracted, the similarity of the static and dynamic watermarks is compared by using the similarity measurement algorithm of the watermark to screen out the suspected infringing video, and then the feature fingerprint information is combined for verification, the suspected video that fails to pass the verification is marked as an infringing warning video and reported to the trace database, so that the detection and trace of the video according to the watermark and the feature fingerprint are realized, the network security supervision of the whole link of the video propagation is realized, and the video network security is improved.

[0035] Please refer to Figure 4 , Figure 4 is a system diagram of a video trace fingerprint processing system in some embodiments of the present application.

[0036] In a second aspect, the present application also discloses a video trace fingerprint processing system 4, which comprises: A video preprocessing module 401 is configured to identify the scene subject distribution of the target video video area, locate the key frame, and screen the watermark embedding candidate area; A static watermark generation and embedding module 402 is configured to generate a watermark sequence according to the target video, and adaptively embed the target embedding area according to the embedding strength; A dynamic watermark generation and embedding module 403 is configured to generate a dynamic watermark pattern sequence and dynamically embed the dynamic watermark embedding area according to the embedding density; The feature fingerprint information extraction and integration module 404 is configured to perform multi-dimensional feature integration on the metadata, the play record and the processing log extracted from the target video, and generate feature fingerprint data and a feature fingerprint data sequence. The feature fingerprint information embedding module 405 is configured to embed the feature fingerprint data into the padding bits of the video stream according to the encoding embedding density, and perform segmented embedding on the feature fingerprint. The video monitoring and traceability identification module 406 is configured to perform dynamic monitoring on the object video, and perform comparison and verification on the static watermark, the dynamic watermark and the decoded feature fingerprint, and determine and report the infringing video.

[0037] According to the embodiment of the present application, the video traceability fingerprint processing system further comprises a memory and a processor, the memory comprises a video traceability fingerprint processing method program, and the video traceability fingerprint processing method program is executed by the processor to realize the following steps: Attribute information and feature information of a video region of a target video are acquired, key frames are positioned and marked according to scene subject distribution information and key frame positioning information, and a watermark embedding candidate region is selected according to key frame texture recognition degree; A watermark sequence of the target video is acquired, and the key frames of the selected target embedding region are embedded, and the watermark is adaptively embedded according to embedding strength; Dynamic watermark factors of the target video are acquired, and a dynamic watermark pattern sequence is generated by encryption, and dynamic watermark embedding is performed on the dynamic watermark embedding region selected from the watermark embedding candidate region according to embedding density; Multi-dimensional feature weight fusion is performed on the metadata, the play record and the processing log of the target video according to a preset feature fusion algorithm, and feature fingerprint data and a feature fingerprint data sequence are generated; A check code is generated according to the feature fingerprint data sequence, the check code is embedded into an audio stream data region of the target embedding region according to encoding embedding density, the feature fingerprint data is embedded into padding bits of a video stream, and the feature fingerprint is segmented and regularly embedded; The object video is dynamically monitored, and comparison and verification are performed on the static watermark, the dynamic watermark and the decoded feature fingerprint, and the infringing alarm video is determined and reported.

[0038] In order to realize accurate tracing of the video, effectively guarantee the network security of the video, the safety information embedding and anti-infringement tracing means of the video are realized by adding static and dynamic watermarks to the video and embedding feature fingerprint information in the video data stream, the scene subject distribution and key frame positioning of the target video are determined, the watermark embedding candidate area is selected, the key frame is embedded according to the watermark sequence of the target video and adaptive intensity static watermark embedding is carried out, the dynamic watermark embedding area is embedded according to the embedding density, the multi-dimensional feature integration of the target video is carried out according to the preset feature fusion algorithm to generate feature fingerprint data, the check code and the feature fingerprint data are embedded in the audio stream data area and the filling bits of the video stream of the target embedding area respectively, the feature fingerprint is segmented and embedded, the object video is monitored and the static and dynamic watermarks and the decoded feature fingerprint are verified, the infringing video is judged and reported, and the network application safety supervision technology of the video is realized.

[0039] According to the embodiment of the application, the attribute information and the feature information of the target video video area are acquired, the key frame is positioned and marked according to the scene subject distribution information and the key frame positioning information, and the watermark embedding candidate area is selected according to the texture recognition degree of the key frame. The target video is preprocessed to acquire video distribution information, including resolution, frame rate and bit rate; The video distribution information is processed by the trained preset feature extraction model to obtain the attribute information and the feature information of the video area; The attribute information includes object category and scene category, and the feature information includes edge information and texture information; The feature information is processed according to the attribute information matching the corresponding key frame recognition model to obtain the scene subject distribution information and the key frame positioning information, and the key frame of each scene subject is positioned and marked; The texture information corresponding to the marked key frame is processed according to the preset texture analysis model to obtain texture feature data, including contrast, correlation and energy value, and the texture recognition degree is obtained by weighted processing, and the watermark embedding candidate area is selected.

[0040] The video distribution information of each distribution area in the video is input into the feature extraction model for processing and recognition, attribute feature information and feature information of the video area are obtained, including object categories, scene categories, edge information and texture information, a matching key frame recognition model is selected according to the scene category of the video area, for example, an event scene can use a target detection model such as YOLO which has a main processing function for objects and faces, then the edge and texture features of the video are detected and positioned by using the matching model, the position distribution of the key frame where the foreground subject is located is obtained, that is, the distribution relationship between the subject of the scene and the key frame positioning is obtained, the key frame of the scene subject is positionally marked and positioned, the key frame position of the scene subject of the video area is obtained, the texture of the key frame is identified according to the local binary algorithm of texture analysis, the contrast, correlation and ability value are obtained, the texture recognition degree is obtained by processing according to the preset weighting method, and the recognition degree is used to select a watermark embedding area, so as to improve the hiding effect of the watermark.

[0041] According to the embodiment of the application, the watermark sequence of the target video is obtained, and the key frame of the selected target embedding area is embedded, and the watermark is adaptively embedded according to the embedding strength, including: The source identification information of the target video is obtained, including a production code, a copyright ID and a resource identification code; The source identification information is processed according to a preset watermark generation algorithm to generate a watermark sequence; The frame rate of the key frame of the watermark embedding candidate area is processed to obtain a frequency distribution coefficient, and the corresponding target embedding area is selected according to the medium and low frequency coefficients; The generated watermark sequence is embedded into the medium and low frequency coefficients of the key frame of the target embedding area by using a preset frequency domain method; The brightness and contrast are extracted according to the texture information of the key frame, the video sensitivity of the target embedding area is obtained, and the corresponding embedding strength is obtained according to the video sensitivity; The watermark is adaptively embedded into the target embedding area according to the embedding strength.

[0042] The watermark generation algorithm based on the pseudo-random sequence is adopted to identify and encode the source identification information to obtain a binary watermark sequence, and the frame rate of the key frame of the watermark embedding candidate area and the sampling rate of each frame are multiplied to obtain a frequency distribution coefficient, and a threshold is set to screen out the required medium and low frequency coefficients and the corresponding target embedding area, and the watermark sequence is embedded into the medium and low frequency coefficients of the key frame of the target embedding area by using the frequency domain method of discrete cosine transform, these coefficients are relatively stable in video compression, so the robustness of the watermark can be ensured, the brightness and contrast of the key frame are processed according to the texture information to obtain the video sensitivity of the target embedding area, the corresponding embedding strength is obtained according to the video sensitivity, and the embedding strength is adaptively adjusted according to the visual sensitivity, so that the detectability of the watermark signal is improved in the high sensitivity area of the watermark invisibility, and static watermark embedding is realized.

[0043] According to the embodiment of the application, the dynamic watermark factor of the target video is obtained, and a dynamic watermark pattern sequence is generated by encryption, and the dynamic watermark embedding area screened out from the watermark embedding candidate area according to the embedding density is dynamically embedded with the dynamic watermark, comprising: The timestamp of the target video and the key frame hash value of the watermark embedding candidate area are obtained, and a random number generated in real time is obtained; The dynamic watermark factor is generated according to the key frame hash value, the timestamp and the random number; The dynamic watermark factor is encrypted according to a preset encryption algorithm to generate a dynamic watermark pattern sequence; The time interval and content change rate of the key frame of the watermark embedding candidate area are obtained, and compared with the preset interval threshold and change rate threshold, if both are greater than the threshold, it is determined as a dynamic watermark embedding area; The corresponding embedding density is obtained according to the time interval and the content change rate, and the dynamic watermark embedding area is dynamically embedded with the dynamic watermark according to the embedding density.

[0044] The current timestamp of the target video is obtained, the hash value of the key frame is calculated by using the hash algorithm MD5, the random number is obtained by using the random number generator such as Java, and the three are concatenated into a string by using the preset sorting splicing method and encoding, which is used as the watermark factor, or the combination value is obtained by using the preset operation rule, which is used as the watermark value factor, the watermark value factor is encrypted by using the preset encryption algorithm such as AES algorithm, a unique dynamic watermark pattern sequence is generated, and then the time interval and the content change rate of the key frame of the watermark embedding candidate area are used for dynamic watermark embedding judgment, if both are greater than the corresponding threshold, it is determined as the dynamic embedding position of the watermark, and then the corresponding embedding density is obtained according to the lookup table result, and the real-time adaptability embedding of the dynamic watermark is realized.

[0045] According to the embodiment of the present application, the multi-dimensional feature weight fusion of the metadata, the play record and the processing log of the target video according to the preset feature fusion algorithm generates the feature fingerprint data and the feature fingerprint data sequence, which comprises: Obtaining the metadata, the play record and the processing log of the target video; The metadata comprises the encoding format, the encoding level and the sampling rate, the play record comprises the play identification, the play time and the play platform, and the processing log comprises the clip time, the encoding parameter and the special effect type; The multi-dimensional feature weight fusion of the metadata, the play record and the processing log of the target video according to the preset feature fusion algorithm generates the feature fingerprint data and the feature fingerprint data sequence.

[0046] Among them, the metadata, the play record and the processing log of the target video are obtained through the preset reading extraction tool, such as the static features of the encoding format, the encoding level and the sampling rate obtained by using the video reading tool, the dynamic features of the play frequency, the time length and the platform distribution calculated by extracting the play record including the play identification, the play time and the play platform, and the clip time, the encoding parameter and the special effect type extracted by using the log analysis tool, and then the above scattered features are integrated according to the preset data structure according to the preset feature fusion algorithm, forming the feature fingerprint data sequence, such as the comprehensive feature fingerprint combined according to a certain weight by the feature engineering method, and then the feature fingerprint data is encoded in the JSON format to generate the feature fingerprint data sequence, which can ensure the readability and scalability of the data.

[0047] According to the embodiment of the present application, the feature fingerprint data sequence is encoded to generate a check code, the check code is embedded into the audio stream data area of the target embedding area according to the encoding embedding density, the feature fingerprint data is embedded into the padding bits of the video stream, the feature fingerprint is segmented and embedded according to the preset rule, which comprises: Encoding the feature fingerprint data sequence to generate a check code; According to the video sensitivity combined capacity table of the target embedding area, the corresponding encoding embedding density is obtained; According to the encoding embedding density, the check code is embedded into the audio stream data area of the target embedding area; According to the preset randomization embedding strategy, the feature fingerprint data is embedded into the padding bits of the video stream of the target embedding area; According to the preset block embedding chroma component, the feature fingerprint is segmented, and embedded according to the preset random mask rule.

[0048] In the feature fingerprint, a data hiding technology is used to embed feature fingerprint data into a data area of an audio stream and a padding bit of a video stream of a target embedding area, and the feature fingerprint is embedded after being segmented according to an embedded chroma component. First, a feature fingerprint data sequence is encoded using a BCH code to generate a check code, a matching embedding density is determined according to a video sensitivity and an area capacity of the target embedding area, the encoded check code is embedded, a randomization embedding strategy is used to distribute and embed the feature fingerprint data into a bit of the video stream, then the feature fingerprint is embedded in the chroma component according to a pixel that can be written at will, the feature fingerprint is segmented into each block as MxN according to the chroma component, and the data is mixed and embedded under a random mask rule to ensure visual invisibility of the embedded data, so that the feature fingerprint is combined with the video data through optimized embedding, and high integrity and accuracy of the feature fingerprint can be maintained after the video is subjected to multiple transcoding, editing and conversion operations.

[0049] According to the embodiment of the present application, the object video is dynamically monitored, and static watermark, dynamic watermark and decoded feature fingerprint are extracted and compared and verified, and the infringing alarm video is judged and reported, comprising: The object video is dynamically monitored by the pre-deployed distributed monitoring node; The object video is traced when an abnormal propagation behavior is monitored or a trace request is received; The static watermark, dynamic watermark and decoded feature fingerprint of the object video are extracted; The similarity of the extracted static watermark, dynamic watermark and the corresponding watermark of the target video is compared by a similarity algorithm, and the object video that does not meet the similarity comparison requirement is marked as a suspected infringing video; The feature fingerprint is compared and verified according to the feature fingerprint information of the target video, and the suspected infringing video that fails to pass the verification is marked as an infringing alarm video and reported to the trace database.

[0050] In the process of monitoring the network video, the distributed monitoring node distributed in the video transmission network uses real-time video stream analysis technology to dynamically monitor each object video, and when an abnormal propagation behavior (such as unauthorized sharing, content tampering) is detected or a trace request is received, the trace is automatically started, the static watermark, dynamic watermark and decoded feature fingerprint in the object video are extracted, the suspected infringing video is screened out by comparing the similarity of the static and dynamic watermarks using a similarity measurement algorithm of the watermark, and then the suspected video that fails to pass the verification is marked as an infringing alarm video and reported to the trace database, so that the video is detected and traced according to the watermark and the feature fingerprint, the network security supervision of the whole link of the video propagation is realized, and the video network security is improved.

[0051] The third aspect of the present application provides a readable storage medium, wherein a video source fingerprint processing method program is stored in the readable storage medium, and when the video source fingerprint processing method program is executed by a processor, the steps of the video source fingerprint processing method according to any one of the preceding aspects are implemented.

[0052] The video source fingerprint processing method, system and medium disclosed in the present application determine scene subject distribution and key frame positioning according to a target video, select a watermark embedding candidate region, embed a watermark sequence in a key frame according to the target video and perform adaptive intensity static watermark embedding, perform dynamic watermark embedding on a dynamic watermark embedding region according to embedding density, perform multi-dimensional feature integration on the target video according to a preset feature fusion algorithm to generate feature fingerprint data, embed a check code and the feature fingerprint data in an audio stream data region and a padding bit of a video stream of a target embedding region respectively, perform feature fingerprint segmentation embedding, monitor a target video and perform static and dynamic watermark verification and decoding feature fingerprint verification, and judge and report a marked infringement video; thus, by adding static and dynamic watermarks to the video and embedding feature fingerprint information in a video data stream, accurate video source tracing is achieved, and video network security is ensured.

[0053] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other manners. The above described device embodiments are only schematic. For example, the division of the units is only a logical function division. There can be another division manner for the actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed coupling, or direct coupling or communication connection between the components can be indirect coupling or communication connection through some interfaces, or electrical, mechanical or other forms.

[0054] The units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units; they can be located in one place, or distributed on multiple network units; and some or all of the units can be selected according to actual needs to achieve the purpose of the present embodiment.

[0055] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be realized in the form of hardware, or in the form of hardware plus software functional unit.

[0056] Those skilled in the art can understand that all or part of the steps of the above-mentioned method embodiments can be completed by relevant hardware of program instructions, and the foregoing program can be stored in a readable storage medium, and the program executes the steps of the above-mentioned method embodiments when executed; and the foregoing storage medium includes various media capable of storing program codes, such as a mobile storage device, a read-only memory (ROM), a random access memory (RAM), a magnetic disc or an optical disc.

[0057] Alternatively, the integrated unit of the present application can also be stored in a readable storage medium if it is realized in the form of a software function module and sold or used as an independent product. Based on such understanding, the technical solutions of the embodiments of the present application can be embodied in the form of a software product in essence or in the form of a part of the prior art that makes a contribution, and the software product is stored in a storage medium, including a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in the embodiments of the present application. The foregoing storage medium includes various media capable of storing program codes, such as a mobile storage device, a ROM, a RAM, a magnetic disc or an optical disc.

Claims

1. A method of video origin spoof fingerprinting, the method comprising: The method comprises the following steps: Obtain attribute information and feature information of a target video region, mark key frames according to scene subject distribution information and key frame positioning information, and select watermark embedding candidate regions according to key frame texture recognition degree; Obtain a watermark sequence of the target video, and embed the key frames of the target embedding region, and adaptively embed the watermark according to embedding strength; Obtain dynamic watermark factors of the target video and generate a dynamic watermark pattern sequence by encryption, and dynamically embed the dynamic watermark in the dynamic watermark embedding region selected from the watermark embedding candidate region according to embedding density; According to a preset feature fusion algorithm, multi-dimensional feature weight fusion is performed on the metadata, playback records and processing logs of the target video to generate feature fingerprint data and a feature fingerprint data sequence; According to the feature fingerprint data sequence, a check code is generated, the check code is embedded in the audio stream data region of the target embedding region according to the encoding embedding density, the feature fingerprint data is embedded in the padding bits of the video stream, and the feature fingerprint is segmented and regularly embedded; The object video is dynamically monitored, and static watermark, dynamic watermark and decoded feature fingerprint are compared and verified to determine the infringement alarm video and report.

2. The method of claim 1, wherein, The method comprises the following steps: Preprocess the target video to obtain video distribution information, including resolution, frame rate and bit rate; Process the video distribution information through a trained preset feature extraction model to obtain attribute information and feature information of the video region; The attribute information includes object category and scene category, and the feature information includes edge information and texture information; Process the feature information according to the attribute information matched with the corresponding key frame recognition model to obtain scene subject distribution information and key frame positioning information, and mark the key frames of each scene subject; Process the texture information of the marked key frames according to a preset texture analysis model to obtain texture feature data, including contrast, correlation and energy value, and obtain texture recognition degree by weighted processing, and select watermark embedding candidate regions.

3. The method of claim 2, wherein, The method comprises the following steps: Obtain source identification information of the target video, including producer code, copyright party ID and resource identification code; Process the source identification information according to a preset watermark generation algorithm to generate a watermark sequence; Process the frame rate of the key frames of the watermark embedding candidate region to obtain frequency distribution coefficients, and select the corresponding target embedding region according to the low-frequency coefficients; Embed the generated watermark sequence into the low-frequency coefficients of the key frames of the target embedding region by a preset frequency domain method; Extract brightness and contrast according to the texture information of the key frames, obtain video sensitivity of the target embedding region, and obtain the corresponding embedding strength according to the video sensitivity. According to the embedding intensity, the target embedding region is subjected to watermark adaptive intensity embedding.

4. The method of claim 3, wherein, The method further includes: obtaining a timestamp of the target video and a key frame hash value of the watermark embedding candidate region, and obtaining a real-time generated random number; generating a dynamic watermark factor according to the key frame hash value, the timestamp, and the random number; encrypting the dynamic watermark factor according to a preset encryption algorithm to generate a dynamic watermark pattern sequence; obtaining a time interval and a content change rate of a key frame of the watermark embedding candidate region, and comparing the time interval and the content change rate with a preset interval threshold and a change rate threshold; if both the comparison results are greater than the threshold, determining that the watermark embedding candidate region is a dynamic watermark embedding region; and embedding a dynamic watermark in the dynamic watermark embedding region according to an embedding density corresponding to the time interval and the content change rate. The method further includes: obtaining metadata, a play record, and a processing log of the target video; the metadata includes an encoding format, an encoding level, and a sampling rate; the play record includes a play identifier, a play time, and a play platform; and the processing log includes a clipping time, an encoding parameter, and an effect type. The method further includes: obtaining metadata, a play record, and a processing log of the target video; the metadata includes an encoding format, an encoding level, and a sampling rate; the play record includes a play identifier, a play time, and a play platform; and the processing log includes a clipping time, an encoding parameter, and an effect type. The method further includes: encoding the feature fingerprint data sequence to generate a check code; obtaining a corresponding encoding embedding density according to the video sensitivity and the capacity of the target embedding region; embedding the check code into an audio stream data region of the target embedding region according to the encoding embedding density; embedding the feature fingerprint data into a padding bit of a video stream of the target embedding region according to a preset randomization embedding strategy; and segmenting the feature fingerprint according to a preset block embedding chroma component, and embedding the feature fingerprint according to a preset random mask rule. The method further includes: dynamically monitoring the object video through a pre-deployed distributed monitoring node; tracing the object video when an abnormal propagation behavior is monitored or a trace request is received; extracting a static watermark, a dynamic watermark, and a decoded feature fingerprint of the object video; comparing the extracted static watermark and dynamic watermark with corresponding watermarks of the target video through a similarity algorithm, and marking an object video that does not meet a similarity comparison requirement as a suspected infringing video; and determining and reporting an infringing alarm video. ​ 5. The method of claim 4, wherein, ​ ​ ​ ​ 6. The method of claim 5, wherein, ​ ​ ​ ​ ​ ​ 7. The method of claim 6, wherein, ​ ​ ​ ​ ​ The feature fingerprint is compared and verified according to the feature fingerprint information of the target video, a suspected infringement video that fails the verification is marked as an infringement warning video, and is reported to a traceability database.

8. A video origin fingerprinting system characterized by, The system comprises: A video preprocessing module for identifying scene subject distribution of a target video region, locating key frames, and screening watermark embedding candidate regions; A static watermark generation and embedding module for generating a watermark sequence according to the target video, and adaptively embedding the target embedding region according to embedding intensity; A dynamic watermark generation and embedding module for generating a dynamic watermark pattern sequence and dynamically embedding a dynamic watermark embedding region according to embedding density; A feature fingerprint information extraction and integration module for multi-dimensional feature integration of metadata, playback records, and processing logs of the target video, generating feature fingerprint data and a feature fingerprint data sequence; A feature fingerprint information embedding module for embedding a check code of the feature fingerprint data sequence into an audio stream data region according to encoding embedding density, embedding the feature fingerprint data into padding bits of a video stream, and segmenting and embedding the feature fingerprint; A video monitoring and traceability identification module for dynamic monitoring of the target video and comparison and verification of static watermarks, dynamic watermarks, and decoded feature fingerprints, marking and reporting infringement warning videos.

9. The video watermarking system of claim 8, wherein, The system further comprises a memory and a processor, wherein the memory comprises a video traceability fingerprint processing method program, and the video traceability fingerprint processing method program is executed by the processor to implement the following steps: Obtain attribute information and feature information of a target video region, mark key frames according to scene subject distribution information and key frame positioning information, and select watermark embedding candidate regions according to key frame texture recognition degree; Obtain a watermark sequence of the target video, and embed the key frames of the selected target embedding region according to embedding intensity to adaptively embed the watermark; Obtain dynamic watermark factors of the target video, and encrypt to generate a dynamic watermark pattern sequence, and dynamically embed a dynamic watermark embedding region in the watermark embedding candidate region according to embedding density; Integrate multi-dimensional feature weights of metadata, playback records, and processing logs of the target video according to a preset feature fusion algorithm, generate feature fingerprint data and a feature fingerprint data sequence; Generate a check code according to the feature fingerprint data sequence, embed the check code into an audio stream data region of the target embedding region according to encoding embedding density, embed the feature fingerprint data into padding bits of a video stream, and segment and regularly embed the feature fingerprint; Dynamically monitor the target video, extract static watermarks, dynamic watermarks, and decoded feature fingerprints, compare and verify them, mark and report infringement warning videos.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium comprises a video traceability fingerprint processing method program, and the video traceability fingerprint processing method program is executed by the processor to implement the steps of the video traceability fingerprint processing method according to any one of claims 1 to 7.

Images