CN111147874A - Method and system for program list accurate to second based on image recognition - Google Patents
Method and system for program list accurate to second based on image recognition Download PDFInfo
- Publication number
- CN111147874A CN111147874A CN201911361129.4A CN201911361129A CN111147874A CN 111147874 A CN111147874 A CN 111147874A CN 201911361129 A CN201911361129 A CN 201911361129A CN 111147874 A CN111147874 A CN 111147874A
- Authority
- CN
- China
- Prior art keywords
- program
- image recognition
- video
- accuracy
- tail
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/21—Server components or server architectures
- H04N21/218—Source of audio or video content, e.g. local disk arrays
- H04N21/2187—Live feed
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/44—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream, rendering scenes according to MPEG-4 scene graphs
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/44—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream, rendering scenes according to MPEG-4 scene graphs
- H04N21/44008—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream, rendering scenes according to MPEG-4 scene graphs involving operations for analysing video streams, e.g. detecting features or characteristics in the video stream
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Databases & Information Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of digital televisions, and discloses a method for accurately marking a program to the second based on image recognition, which comprises the steps of recording a program video 5 minutes before and after the program starts to play, dividing the video into frames, manually marking the start time and the name to form a sample of the head and the tail of the program, carrying out deep learning method training on the adaptation of the marks to form a characteristic library of the head and the tail of the program, then intercepting 1-2 pictures of the live video being played and comparing the pictures with the characteristic library, if the recognition is successful, modifying the start time of the initial program, if the recognition fails, outputting the picture, manually auditing and supplementing the sample of the head and the tail of the program, and manually marking the start time and the name again to form a sample of the head and the tail of the program. The invention can correct the time of the program list to the second, thereby not only improving the accuracy of the program list, but also improving the user experience.
Description
Technical Field
The invention relates to the technical field of digital televisions, in particular to a method and a system for accurately setting a program list to be in seconds based on image recognition.
Background
The television station provides a program list, and the operator makes the EPG according to the program list provided by the television station. However, the program list provided by the television station is basically inaccurate time, and time errors often exist, namely the time on the program list is inconsistent with the actual playing time of the television program.
Therefore, in order to solve the above problems, a method and system for program guide accuracy to seconds based on image recognition is urgently needed.
Disclosure of Invention
Based on the above problems, the present invention provides a method and system for program guide accurate to second based on image recognition. The invention can correct the time of the program list to the second, thereby not only improving the accuracy of the program list, but also improving the user experience.
The invention specifically adopts the following technical scheme for realizing the purpose:
a method for image recognition based program guide accuracy to seconds, comprising the steps of:
the method comprises the following steps: a preparation stage: recording the program list video 5 minutes before and after the program starts playing, and dividing the video into frames.
Step two: and manually marking the start time and the name to form a sample of the head and the tail of the program.
Step three: and (4) carrying out deep learning method training on the adaptation of the mark to form a program leader and trailer feature library.
Step four: and (3) identification: intercepting 1-2 pictures of a live video being played, comparing the pictures with a feature library, and modifying the starting time of an initial program list if the identification is successful; if the identification fails, outputting a picture, and manually checking the head and tail samples of the supplementary program, namely returning to the second step and the third step.
The system for accurately recording the program list to the second based on the image recognition comprises a program recording system, a splitting system, a training system and a program recognition system.
Further, the program recording system automatically records live programs within 5 minutes before and after playing according to the time of the program list provided by the initial television station.
Further, the video splitting system splits the recorded program list in minute units by frame, and marks the program start time and the program name by manual work.
Further, massive program picture information is obtained in a training system, the accuracy of the feature library is trained and optimized, and the feature library is stored after training is completed.
Further, the program identification system intercepts the live broadcast picture to be identified, and compares the live broadcast picture with the feature library to identify the accurate time of the program.
The working principle is as follows: splitting recorded program list videos into frames, forming a sample of a program head and a program tail after manual recording, training by a deep learning method to form a program head and tail feature library, optimizing the accuracy of the feature library and storing the feature library in the training process, intercepting a video picture and comparing the video picture with the feature library when identifying a live video, modifying the starting time of an initial program list or adding a program head and tail sample in the comparison process, and splitting the program list videos into the frames, so that the time of the program list can be corrected to the second in the comparison process, the accuracy of the program list is improved, the characteristics of the program head and tail feature library can be increased after the newly added program head and tail sample is trained by the deep learning method, the comparison is convenient, and the time of the program list is quickly corrected.
The invention has the following beneficial effects:
(1) according to the method, recorded videos are divided into frames, the start time and the name are marked manually to form a sample of the head and the tail of a program, the marked sample is trained by using deep learning to form a characteristic library of the head and the tail of the program, captured pictures of the live broadcast videos are compared with the characteristic library, the accurate time of the program is identified, and the accuracy is up to the second, so that the accuracy of the program list is improved, and the user experience is also improved.
(2) In the invention, the recorded video is divided into frames, and the default is one second and twelve frames, so that one frame is equal to one twelfth of one second, therefore, when the picture captured by the live video is compared with the feature library, the accuracy of the program list can reach the second level, so that a user can directly watch the wanted television program when watching back, and the user experience is improved.
(3) The invention realizes self-learning and self-perfection of the feature library through continuously supplementing program leader and trailer samples and training, and the recognition rate of the program leader and trailer can be continuously improved.
(4) In the invention, after the program identification system successfully identifies the live broadcast picture to be identified, the initial program list is automatically corrected, and the adjusted program list can be accurate to the second level, so that the accuracy of the program list is improved, and the user experience is also improved.
(5) In the invention, after the program identification system fails to identify the live broadcast picture to be identified, the picture is output and is supplemented to the feature library after manual verification, so that the feature library is continuously improved.
Drawings
FIG. 1 is a system block diagram of the present invention;
FIG. 2 is a flow chart for identifying program listings;
FIG. 3 is a schematic diagram of a deep learning method;
Detailed Description
For a better understanding of the present invention by those skilled in the art, the present invention will be described in further detail below with reference to the accompanying drawings and the following examples.
Example 1:
referring to fig. 1 to 3, a method for the accuracy of a program list to seconds based on image recognition comprises the following steps:
the method comprises the following steps: a preparation stage: recording the program list video 5 minutes before and after the program starts playing, and dividing the video into frames.
Step two: and manually marking the start time and the name to form a sample of the head and the tail of the program.
Step three: and (4) carrying out deep learning method training on the adaptation of the mark to form a program leader and trailer feature library.
Step four: and (3) identification: intercepting 1-2 pictures of a live video being played, comparing the pictures with a feature library, and modifying the starting time of an initial program list if the identification is successful; if the identification fails, outputting a picture, and manually checking the head and tail samples of the supplementary program, namely returning to the second step and the third step.
The system for accurately recording the program list to the second based on the image recognition comprises a program recording system, a splitting system, a training system and a program recognition system.
Further, the program recording system automatically records live programs within 5 minutes before and after playing according to the time of the program list provided by the initial television station.
Further, the video splitting system splits the recorded program list in minute units by frame, and marks the program start time and the program name by manual work.
Further, massive program picture information is obtained in a training system, the accuracy of the feature library is trained and optimized, and the feature library is stored after training is completed.
Further, the program identification system intercepts the live broadcast picture to be identified, and compares the live broadcast picture with the feature library to identify the accurate time of the program.
In this embodiment, recording a program video 5 minutes before and after the program starts to play, dividing the video into frames, manually marking the start time and the title to form a sample of the beginning and the end of the program, performing deep learning method training on the marked adaptation to form a feature library of the beginning and the end of the program, then capturing 1-2 pictures of the live video being played, comparing the pictures with the feature library, if the identification is successful, modifying the start time of the initial program, if the identification is failed, outputting the pictures, manually auditing and supplementing the sample of the beginning and the end of the program, manually marking the start time and the title again to form a sample of the beginning and the end of the program, performing deep learning method training on the marked adaptation to form a feature library of the beginning and the end of the program, and dividing the program video into frames, so that the time of the program can be corrected to seconds during the comparison, therefore, the accuracy of the program list is improved, the newly added program head and tail sample can continuously supplement the characteristics of the program head and tail characteristic library after being trained by a deep learning method, the self-learning and self-perfection of the characteristic library are realized, and the recognition rate of the program head and the program tail can be continuously improved, so that the time of the program list is quickly corrected.
Specifically, the recorded video is divided into frames, and the default is one second and twelve frames, so that one frame is equal to one twelfth of one second, when the picture captured by the live video is compared with the feature library, the accuracy of the program list can reach the second level, the accuracy of the program list is improved, a user can directly watch the wanted television program when the user watches the program list again, and the user experience is improved.
Specifically, when a live video is identified, a video picture is captured and compared with a feature library, the starting time of an initial program list is modified or a program head and tail sample is newly added in the comparison process, and after the newly added program head and tail sample is trained through a deep learning method, the features of the program head and tail feature library can be increased, so that comparison is convenient, and the time of the program list is quickly corrected.
Specifically, the specific method of deep learning is as follows:
referring to fig. 3, an image is input and convolved through three trainable filters and an applicable bias, after convolution, three feature maps are generated at a C1 level, then four pixels in each group in the feature maps are summed, weighted, and biased, and feature maps of three S2 levels are obtained through a Sigmoid function. These maps are further filtered to obtain a layer C3. This hierarchy, again, as with S2, results in S4. Finally, these pixel values are rasterized and connected into a vector input to a conventional neural network, resulting in an output.
Specifically, the layer C is a feature extraction layer, the input of each neuron is connected to the local receptive field of the previous layer, and the local feature is extracted, and once the local feature is extracted, the position relationship between the local feature and other features is determined; the S layer is a feature mapping layer, each calculation layer of the network is composed of a plurality of feature mappings, each feature mapping is a plane, and the weights of all neurons on the plane are equal. The feature mapping structure adopts a sigmoid function with small influence function kernel as an activation function of the convolution network, so that the feature mapping has displacement invariance.
Specifically, because the neurons on one mapping surface share the weight, the number of free parameters of the network is reduced, and the complexity of network parameter selection is reduced. Each feature extraction layer in the convolutional neural network is followed by a feature mapping layer for local averaging and quadratic extraction, and the specific quadratic feature extraction structure enables the network to have high distortion tolerance capability on input samples during identification.
Specifically, Z represents a layer of convolution operation, and NN represents the result of this convolution operation.
The above is an embodiment of the present invention. The embodiments and specific parameters in the embodiments are only for the purpose of clearly illustrating the verification process of the invention and are not intended to limit the scope of the invention, which is defined by the claims, and all equivalent structural changes made by using the contents of the specification and the drawings of the present invention should be covered by the scope of the present invention.
Claims (6)
1. A method for image recognition based program guide accuracy to seconds, comprising the steps of:
the method comprises the following steps: a preparation stage: recording the program list video 5 minutes before and after the program starts playing, and dividing the video into frames.
Step two: and manually marking the start time and the name to form a sample of the head and the tail of the program.
Step three: and (4) carrying out deep learning method training on the adaptation of the mark to form a program leader and trailer feature library.
Step four: and (3) identification: intercepting 1-2 pictures of a live video being played, comparing the pictures with a feature library, and modifying the starting time of an initial program list if the identification is successful; if the identification fails, outputting a picture, and manually checking the head and tail samples of the supplementary program, namely returning to the second step and the third step.
2. The image recognition based program guide to second accuracy system of claim 1, wherein: the system comprises a program recording system, a splitting system, a training system and a program identification system.
3. The method and system of claim 2 for program guide accuracy to seconds based on image recognition, wherein: and the program recording system automatically records the live programs within 5 minutes before and after playing according to the time of the program list provided by the initial television station.
4. The method and system of claim 2 for program guide accuracy to seconds based on image recognition, wherein: the video splitting system splits the recorded minute-level program list by taking a frame as a unit and marks the program starting time and the program name by manpower.
5. The method and system of claim 2 for program guide accuracy to seconds based on image recognition, wherein: and acquiring massive program picture information in the training system, training the accuracy of the optimized feature library, and storing the feature library after the training is finished.
6. The method and system of claim 2 for program guide accuracy to seconds based on image recognition, wherein: the program identification system intercepts the live broadcast picture to be identified, compares the live broadcast picture with the feature library and identifies the accurate time of the program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911361129.4A CN111147874A (en) | 2019-12-25 | 2019-12-25 | Method and system for program list accurate to second based on image recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911361129.4A CN111147874A (en) | 2019-12-25 | 2019-12-25 | Method and system for program list accurate to second based on image recognition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111147874A true CN111147874A (en) | 2020-05-12 |
Family
ID=70520251
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911361129.4A Pending CN111147874A (en) | 2019-12-25 | 2019-12-25 | Method and system for program list accurate to second based on image recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111147874A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113301401A (en) * | 2021-05-31 | 2021-08-24 | 深圳市茁壮网络股份有限公司 | Method and device for generating electronic program list |
| CN116939197A (en) * | 2023-09-15 | 2023-10-24 | 海看网络科技(山东)股份有限公司 | Live program head broadcasting and replay content consistency monitoring method based on audio and video |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103152632A (en) * | 2013-03-05 | 2013-06-12 | 天脉聚源(北京)传媒科技有限公司 | Method and device for locating multimedia program |
| CN105451068A (en) * | 2015-11-24 | 2016-03-30 | 华数传媒网络有限公司 | Electronic program list generation method and device |
| CN108737691A (en) * | 2018-08-21 | 2018-11-02 | 贵州广播电视台 | A kind of the video broadcast system and broadcasting method of automatic emergency hand-off process |
-
2019
- 2019-12-25 CN CN201911361129.4A patent/CN111147874A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103152632A (en) * | 2013-03-05 | 2013-06-12 | 天脉聚源(北京)传媒科技有限公司 | Method and device for locating multimedia program |
| CN105451068A (en) * | 2015-11-24 | 2016-03-30 | 华数传媒网络有限公司 | Electronic program list generation method and device |
| CN108737691A (en) * | 2018-08-21 | 2018-11-02 | 贵州广播电视台 | A kind of the video broadcast system and broadcasting method of automatic emergency hand-off process |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113301401A (en) * | 2021-05-31 | 2021-08-24 | 深圳市茁壮网络股份有限公司 | Method and device for generating electronic program list |
| CN116939197A (en) * | 2023-09-15 | 2023-10-24 | 海看网络科技(山东)股份有限公司 | Live program head broadcasting and replay content consistency monitoring method based on audio and video |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8553941B2 (en) | Methods, circuits, devices, apparatuses and systems for providing image composition rules, analysis and improvement | |
| CN101621628B (en) | Photographic apparatus, setting method of photography conditions, and recording medium | |
| US20170256036A1 (en) | Automatic microlens array artifact correction for light-field images | |
| CN111178120B (en) | Pest image detection method based on crop identification cascading technology | |
| US8345119B2 (en) | Image pickup device, image processing method, and storage medium storing program | |
| CN111147874A (en) | Method and system for program list accurate to second based on image recognition | |
| US11508038B2 (en) | Image processing method, storage medium, image processing apparatus, learned model manufacturing method, and image processing system | |
| CN101246541A (en) | Image-processing device and image-processing method, image-pickup device, and computer program | |
| Bianco et al. | Personalized image enhancement using neural spline color transforms | |
| CN113052868A (en) | Cutout model training and image cutout method and device | |
| WO2019120025A1 (en) | Photograph adjustment method and apparatus, storage medium and electronic device | |
| CN114862698A (en) | Method and device for correcting real overexposure image based on channel guidance | |
| CN110554991A (en) | Method for correcting and managing text picture | |
| CN111222502B (en) | Infrared small target image labeling method and system | |
| KR102192016B1 (en) | Method and Apparatus for Image Adjustment Based on Semantics-Aware | |
| CN110060208B (en) | Method for improving reconstruction performance of super-resolution algorithm | |
| CN110365897A (en) | Image correcting method and device, electronic equipment, computer readable storage medium | |
| Cao et al. | Physics-guided iso-dependent sensor noise modeling for extreme low-light photography | |
| KR102066412B1 (en) | Apparatus and method for acquiring foreground image | |
| CN116614714A (en) | Real exposure correction method and system guided by perception characteristics of camera | |
| CN101170648A (en) | Method and apparatus for taking a moving picture | |
| CN116051681A (en) | Processing method and system for generating image data based on intelligent watch | |
| CN110648404B (en) | Interactive learning method and system thereof | |
| CN108848366B (en) | Information acquisition device and method based on 3D camera | |
| CN112637573A (en) | Multi-lens switching display method and system, intelligent terminal and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200512 |
|
| RJ01 | Rejection of invention patent application after publication |