Television set monitor supervision platform
The application is application No. is 2017111424307, and the applying date is on November 17th, 2017, and invention and created name is
The divisional application of the patent of " Intelligent TV machine monitoring platform ".
Technical field
The present invention relates to field of television more particularly to a kind of Intelligent TV machine monitoring platforms.
Background technique
Television signal system includes common signal channel, sound channel and depending on putting three parts of final circuit, their master
Acting on is that the high-frequency signal (including picture signal and audio signal) received to antenna is amplified and handled, finally glimmering
Image out is reappeared on optical screen, and sound accompaniment is restored in loudspeaker.By high-frequency amplifier, frequency mixer and local oscillator three parts
Composition.
Satellite television its effect be select and amplify by antenna receive TV great number tuner to high-frequency TV program letter
Number, sound frequency (the first intermediate frequency) signal sound surface of volume signal and 31.5MHz in the image of 38MHz is obtained by Frequency mixing processing
Effect be to form the amplitude-frequency characteristic put in image;The effect put in pre-:Amplified signal (20dB amplification quantity), the filter of compensation sound surface
Loss of the wave device to signal;Surface wave filter realizes the impedance matching between putting in great number tuner and image.ACC is (automatic to increase
Benefit control) circuit:By putting the gain with High Amplifier Circuit in control, to keep the vision signal of wave detector output AGC and ANC
Voltage amplitude is basicly stable;ANC (automatic noise suppressed) circuit:Reduce the external influence crossed signal to television set and dry of making an uproar of TV
It disturbs.
Television set in the prior art only focuses on structure design and signal processing in itself, to the user of viewing television set
Current state lack effective testing mechanism, indulging system is also only limited to be constrained by the time, mentality of designing mistake
In simple.
Summary of the invention
To solve the above-mentioned problems, the present invention provides a kind of Intelligent TV machine monitoring platform, showing for television set is transformed
There is structure, equipment will be taken on site and be arranged on the outer framework of television set, for carrying out viewing environment image data towards spectators
Acquisition, to obtain and export viewing environment image, and to viewing environment image carry out various targeted image procossings and
The image recognition of adaptive deep neural network, so as to accurately know the current state of spectators.
According to an aspect of the present invention, a kind of Intelligent TV machine monitoring platform is provided, the platform includes:
Equipment is taken on site, is arranged on the outer framework of television set, for carrying out viewing environment image data towards spectators
Acquisition, to obtain and export viewing environment image;
Brightness measurement equipment is arranged on the outer framework of television set, is taken on site near equipment, for the scene
The light luminance of environment is measured in real time where capture apparatus, to obtain and export real-time light luminance.
Lighting source is arranged on the outer framework of television set, is taken on site near equipment, with the brightness measurement equipment
Connection, for receiving the real-time light luminance, and when the real-time light luminance transfinites, for the live shooting equipment
Viewing environment image data acquiring provides floor light light.
Scene detection equipment is connect, on the integrated circuit board of television set, for receiving with the live shooting equipment
Viewing environment image obtains the channel R pixel value, the channel G pixel value and the channel B of each pixel in the viewing environment image
Pixel value determines the gradient of all directions of the channel the R pixel value of each pixel to determine each as the channel R gradient
The gradient of all directions of the channel the G pixel value of a pixel is to determine the channel B of each pixel as the channel G gradient
The gradient of all directions of pixel value is using as channel B gradient, the channel R gradient, the channel G gradient and B based on each pixel
Channel gradient determines the corresponding scene complexity of the viewing environment image.
Recognition decision equipment is connect with the scene detection equipment, for being more than or equal in the scene complexity received
When default complexity threshold, select with the training image of scene complexity corresponding number as default training quantity, scene is complicated
Degree is higher, and the quantity of training image is more, and is also used to when the scene complexity received is less than default complexity threshold,
Select the training image of fixed quantity as default training quantity;
Training image obtains equipment, connect with the recognition decision equipment, for each type scene, chooses default training
It is multiple to obtain that the image of quantity as training image, by the training image of all types scene is all transformed into YUV color space
Training color image;
Image-preprocessing device obtains equipment with training image and connect, right for receiving the multiple trained color image
The multiple trained color image executes normalized respectively to obtain fixed-size multiple standard exercise images;
Feature extracting device is connect, according to scene respectively with the scene detection equipment and described image pre-processing device
Complexity determines the input quantity type of the model of selection, carries out according to the input quantity type of selection to each standard exercise image
Feature extraction is to obtain meet the input quantity type of selection, the corresponding training characteristics amount of the standard exercise image, wherein scene
Complexity is higher, and the corresponding data processing amount of input quantity type of the model of selection is more;
Model training equipment is connect with the feature extracting device, corresponding each for receiving each standard exercise image
Each training characteristics amount is respectively outputted to complete the training of model parameter in model, wherein model packet by a training characteristics amount
Input layer, hidden layer and output layer are included, the output quantity of the output layer of model is eyes image;
Model executes equipment, connect respectively with the feature extracting device and the scene detection equipment, sees for receiving
Ambient image is seen, to viewing environment image successively YUV color space conversion, normalized and according to the input quantity class of selection
The feature extraction of type, will to obtain the corresponding identification feature amount of the viewing environment image that meet the input quantity type of selection, described
Input of the corresponding identification feature amount of the viewing environment image as the input layer of model after training, to obtain the eyes of spectators
Image, and the eyes image based on spectators the viewing environment image position and occupy ratio and the eyes image of spectators
Size itself determine the sagging amplitude of spectators' eyes.
The present invention at least has following three important inventive points:
(1) determine that the scene of image is complicated by the channel R gradient, the channel G gradient and the channel B gradient of each pixel
Degree, improves the measurement accuracy of scene complexity;
(2) training program of the neural network based on scene complexity size has been built, to ensure that neural network
The validity of parameters;
(3) hardware configuration transformation is carried out to existing television set, enriches the function of television set.
Detailed description of the invention
Embodiment of the present invention is described below with reference to attached drawing, wherein:
Fig. 1 is the structure according to the live shooting equipment of the Intelligent TV machine monitoring platform shown in embodiment of the present invention
Schematic diagram.
Fig. 2 is the structural block diagram according to the Intelligent TV machine monitoring platform shown in embodiment of the present invention.
Appended drawing reference:1 camera;2 long focal length lenses;3 focusing gear units;4 lens converters;5 focusing rotating electric machines;
6 electric-motor drive units;7 calculation processing units;21 focusing rings.
Specific embodiment
The embodiment of Intelligent TV machine monitoring platform of the invention is described in detail below with reference to accompanying drawings.
The intelligent direction of current television set is limited to the upgrading of self structure, lacks and examines to the state of the spectators of opposite
Survey mechanism.In order to overcome above-mentioned deficiency, the present invention has built a kind of Intelligent TV machine monitoring platform, and specific embodiment is such as
Under.
Fig. 1 is the structure according to the live shooting equipment of the Intelligent TV machine monitoring platform shown in embodiment of the present invention
Schematic diagram.
The live shooting equipment is made of following and part:Camera 1, long focal length lens 2, focusing gear unit 3, mirror
Head converter 4, focusing rotating electric machine 5, electric-motor drive unit 6, calculation processing unit 7.Camera 1 and camera lens 2 are turned by camera lens
Parallel operation 4 connects;Focusing gear unit 3 and focusing ring 21 and the focusing connection of rotating electric machine 5 on camera lens 2;Focusing rotating electric machine 5
It is electrically connected with electric-motor drive unit 6;Calculation processing unit 7 is connected with 6 signal of electric-motor drive unit, can pass through motor driven
The rotation of the control focusing rotating electric machine 5 of unit 6;Calculation processing unit 7 and camera 1 connect, and handle the figure from camera 1
Picture.
Fig. 2 is the structural block diagram according to the Intelligent TV machine monitoring platform shown in embodiment of the present invention, described flat
Platform includes:
Equipment is taken on site, is arranged on the outer framework of television set, for carrying out viewing environment image data towards spectators
Acquisition, to obtain and export viewing environment image;
Brightness measurement equipment is arranged on the outer framework of television set, is taken on site near equipment, for the scene
The light luminance of environment is measured in real time where capture apparatus, to obtain and export real-time light luminance.
Then, continue that the specific structure of Intelligent TV machine monitoring platform of the invention is further detailed.
Can also include in the Intelligent TV machine monitoring platform:
Lighting source is arranged on the outer framework of television set, is taken on site near equipment, with the brightness measurement equipment
Connection, for receiving the real-time light luminance, and when the real-time light luminance transfinites, for the live shooting equipment
Viewing environment image data acquiring provides floor light light.
Can also include in the Intelligent TV machine monitoring platform:
Scene detection equipment is connect, on the integrated circuit board of television set, for receiving with the live shooting equipment
Viewing environment image obtains the channel R pixel value, the channel G pixel value and the channel B of each pixel in the viewing environment image
Pixel value determines the gradient of all directions of the channel the R pixel value of each pixel to determine each as the channel R gradient
The gradient of all directions of the channel the G pixel value of a pixel is to determine the channel B of each pixel as the channel G gradient
The gradient of all directions of pixel value is using as channel B gradient, the channel R gradient, the channel G gradient and B based on each pixel
Channel gradient determines the corresponding scene complexity of the viewing environment image.
Can also include in the Intelligent TV machine monitoring platform:
Recognition decision equipment is connect with the scene detection equipment, for being more than or equal in the scene complexity received
When default complexity threshold, select with the training image of scene complexity corresponding number as default training quantity, scene is complicated
Degree is higher, and the quantity of training image is more, and is also used to when the scene complexity received is less than default complexity threshold,
Select the training image of fixed quantity as default training quantity;
Training image obtains equipment, connect with the recognition decision equipment, for each type scene, chooses default training
It is multiple to obtain that the image of quantity as training image, by the training image of all types scene is all transformed into YUV color space
Training color image;
Image-preprocessing device obtains equipment with training image and connect, right for receiving the multiple trained color image
The multiple trained color image executes normalized respectively to obtain fixed-size multiple standard exercise images;
Feature extracting device is connect, according to scene respectively with the scene detection equipment and described image pre-processing device
Complexity determines the input quantity type of the model of selection, carries out according to the input quantity type of selection to each standard exercise image
Feature extraction is to obtain meet the input quantity type of selection, the corresponding training characteristics amount of the standard exercise image, wherein scene
Complexity is higher, and the corresponding data processing amount of input quantity type of the model of selection is more;
Model training equipment is connect with the feature extracting device, corresponding each for receiving each standard exercise image
Each training characteristics amount is respectively outputted to complete the training of model parameter in model, wherein model packet by a training characteristics amount
Input layer, hidden layer and output layer are included, the output quantity of the output layer of model is eyes image;
Model executes equipment, connect respectively with the feature extracting device and the scene detection equipment, sees for receiving
Ambient image is seen, to viewing environment image successively YUV color space conversion, normalized and according to the input quantity class of selection
The feature extraction of type, will to obtain the corresponding identification feature amount of the viewing environment image that meet the input quantity type of selection, described
Input of the corresponding identification feature amount of the viewing environment image as the input layer of model after training, to obtain the eyes of spectators
Image, and the eyes image based on spectators the viewing environment image position and occupy ratio and the eyes image of spectators
Size itself determine the sagging amplitude of spectators' eyes.
Can also include in the Intelligent TV machine monitoring platform:
SD storage card is connect with the recognition decision equipment, for default complexity threshold to be stored in advance, is also used to store
The default trained quantity of the recognition decision equipment output.
In the Intelligent TV machine monitoring platform:
The lighting source is when the real-time light luminance transfinites, for the viewing environment image that equipment is taken on site
Data acquisition provides floor light light:It is provided based on the real-time light luminance degree of transfiniting corresponding, varying strength
Floor light light.
And in the Intelligent TV machine monitoring platform:
The model executes equipment and also connect with the display screen of television set, the sagging amplitude hair of spectators' eyes for will determine
It is sent on the display screen of television set to carry out real-time display.
Using Intelligent TV machine monitoring platform of the invention, it is limited for TV set intelligent direction in the prior art
Technical problem obtains the eyes image of spectators by way of image recognition, and the eyes image based on spectators is in the position of image
Itself size for setting and occupying ratio and the eyes image of spectators determines the sagging amplitude of spectators' eyes, while by determining spectators
The sagging amplitude of eyes is sent on the display screen of television set to carry out real-time display, to solve above-mentioned technical problem.
It is understood that although the present invention has been disclosed in the preferred embodiments as above, above-described embodiment not to
Limit the present invention.For any person skilled in the art, without departing from the scope of the technical proposal of the invention,
Many possible changes and modifications all are made to technical solution of the present invention using the technology contents of the disclosure above, or are revised as
With the equivalent embodiment of variation.Therefore, anything that does not depart from the technical scheme of the invention are right according to the technical essence of the invention
Any simple modifications, equivalents, and modifications made for any of the above embodiments still fall within the range of technical solution of the present invention protection
It is interior.