CN108734175A - A kind of extracting method of characteristics of image, device and electronic equipment - Google Patents
A kind of extracting method of characteristics of image, device and electronic equipment Download PDFInfo
- Publication number
- CN108734175A CN108734175A CN201810404934.XA CN201810404934A CN108734175A CN 108734175 A CN108734175 A CN 108734175A CN 201810404934 A CN201810404934 A CN 201810404934A CN 108734175 A CN108734175 A CN 108734175A
- Authority
- CN
- China
- Prior art keywords
- image
- region
- pending
- frequency spectrum
- pending image
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
Abstract
An embodiment of the present invention provides a kind of extracting method of characteristics of image, device and electronic equipment, the method includes:Obtain pending image;Pending image is divided into multiple images region;According to predetermined manner, pending image-region is determined from multiple images region, wherein the quantity of pending image-region is less than the quantity of image-region;Feature extraction is carried out to pending image-region, obtains characteristics of image.Since the quantity of pending image-region is less than the quantity of image-region, it is therefore not necessary to carry out image characteristics extraction to all parts in pending image, the time needed for extraction characteristics of image can be shortened, improve image characteristics extraction efficiency.
Description
Technical field
The present invention relates to technical field of image processing, more particularly to a kind of extracting method of characteristics of image, device and electricity
Sub- equipment.
Background technology
In image processing field, the extraction of characteristics of image is a kind of to be widely used general image procossing mode.For example,
Image similarity detection, object category identification, visual odometry etc., the extraction of characteristics of image has critical role.
By taking visual odometry as an example, during determining image capture device motion state by visual odometry, one
As can be divided into image characteristics extraction, Image Feature Matching, determine that image capture device motion state is several based on matching result
Process.And the most of the time of entire calculating process is occupied the time required to the extraction of characteristics of image, directly determine whole process institute
The time needed.
And in the extracting mode of conventional images feature, for example, ORB (Oriented FAST and Rotated
BRIEF, swift nature extraction), SIFT (Scale-invariant feature transform, Scale invariant features transform)
It is that feature extraction, the number of the characteristics of image extracted are carried out to all parts of image in equal image characteristics extractions mode
Amount is very more, and calculation amount is very big, and the extraction process used time of characteristics of image is very long, less efficient.
Invention content
The embodiment of the present invention is designed to provide a kind of extracting method of characteristics of image, device and electronic equipment, with contracting
Time needed for short extraction characteristics of image improves image characteristics extraction efficiency.Specific technical solution is as follows:
In a first aspect, an embodiment of the present invention provides a kind of extracting method of characteristics of image, the method includes:
Obtain pending image;
The pending image is divided into multiple images region;
According to predetermined manner, pending image-region is determined from described multiple images region, wherein the pending figure
As the quantity in region is less than the quantity in described image region;
Feature extraction is carried out to the pending image-region, obtains characteristics of image.
Optionally, described according to predetermined manner, from described multiple images region the step of determining pending image-region,
Including:
Described multiple images region is converted into frequency domain, obtains the corresponding frequency spectrum of each image-region;
Judge whether each frequency spectrum high frequency components meet preset condition, wherein the high fdrequency component is more than for frequency values
The component of preset value;
High fdrequency component is met into the image-region corresponding to the frequency spectrum of the preset condition and is determined as pending image-region.
Optionally, described to judge the step of whether each frequency spectrum high frequency components meet preset condition, including:
Calculate the gross energy of each frequency spectrum high frequency components;
Judge whether the gross energy of each frequency spectrum high frequency components is not less than preset energy value;
If so, determining that the frequency spectrum high frequency components meet preset condition;
If not, determining that the frequency spectrum high frequency components are unsatisfactory for preset condition.
Optionally, the described the step of pending image is divided into multiple images region, including:
The pending image is divided into the multiple images region that size is (N*N), wherein N=2n, n is positive integer;
It is described that described multiple images region is converted into frequency domain, the step of obtaining each image-region corresponding frequency spectrum,
Including:
Using dct algorithm or Wavelet Transformation Algorithm, described multiple images region is converted into frequency domain, is obtained
To the corresponding frequency spectrum of each image-region.
Optionally, the method for determination of the preset value, including:
The pending image is inputted into target network model, obtains the corresponding preset value of the pending image, wherein
The target network model is the default corresponding to the corresponding frequency spectrum high frequency components of image for determining of training completion in advance
The network model of value.
Optionally, described according to predetermined manner, from described multiple images region the step of determining pending image-region,
Including:
Image detection is carried out to described multiple images region, determines whether each image-region includes object;
Image-region including object is determined as pending image-region.
Optionally, the method further includes:
Obtained characteristics of image is matched with the characteristics of image of target image, obtains matching result;
Based on the matching result, determine that the image capture device for acquiring the pending image and the target image exists
Acquire the operating status in the time interval of the pending image and the target image.
Second aspect, an embodiment of the present invention provides a kind of extraction element of characteristics of image, described device includes:
Image collection module, for obtaining pending image;
Image division module, for the pending image to be divided into multiple images region;
Area determination module, for according to predetermined manner, pending image-region to be determined from described multiple images region,
Wherein, the quantity of the pending image-region is less than the quantity in described image region;
Characteristic extracting module obtains characteristics of image for carrying out feature extraction to the pending image-region.
Optionally, the area determination module includes:
Image conversion unit obtains each image-region and corresponds to for described multiple images region to be converted to frequency domain
Frequency spectrum;
Frequency spectrum judging unit, for judging whether each frequency spectrum high frequency components meet preset condition, wherein the high frequency
Component is the component that frequency values are more than the preset value determined by preset value determining module;
First area determination unit, for high fdrequency component to be met to the image-region corresponding to the frequency spectrum of the preset condition
It is determined as pending image-region.
Optionally, the frequency spectrum judging unit includes:
Gross energy computation subunit, the gross energy for calculating each frequency spectrum high frequency components;
Gross energy judgment sub-unit, for judging whether the gross energy of each frequency spectrum high frequency components is not less than preset energy
Value;
First determination subelement, for when the gross energy of frequency spectrum high frequency components is not less than preset energy value, determining should
Frequency spectrum high frequency components meet preset condition;
The and determination subelement, for when the gross energy of frequency spectrum high frequency components is less than preset energy value, determining the frequency
Spectrum high frequency components are unsatisfactory for preset condition.
Optionally, described image division module includes:
Image division unit, for the pending image to be divided into the multiple images region that size is (N*N),
In, N=2n, n is positive integer;
Described image converting unit includes:
Image conversion subunit, for using dct algorithm or Wavelet Transformation Algorithm, by described multiple images
Region is converted to frequency domain, obtains the corresponding frequency spectrum of each image-region.
Optionally, the preset value determining module includes:
Preset value determination unit obtains the pending figure for the pending image to be inputted target network model
As corresponding preset value, wherein the target network model is training completion in advance for determining in the corresponding frequency spectrum of image
The network model of preset value corresponding to high fdrequency component.
Optionally, the area determination module includes:
Whether image detecting element determines each image-region for carrying out image detection to described multiple images region
Including object;
Second area determination unit, for will include that the image-region of object is determined as pending image-region.
Optionally, described device further includes:
Characteristic matching module obtains for matching obtained characteristics of image with the characteristics of image of target image
With result;
Determining module is moved, for being based on the matching result, determines and acquires the pending image and the target figure
Operating status of the image capture device of picture in the time interval for acquiring the pending image and the target image.
The third aspect, the embodiment of the present invention additionally provide a kind of electronic equipment, which is characterized in that including processor, storage
Device and communication bus, wherein processor, memory complete mutual communication by communication bus;
Memory, for storing computer program;
Processor when for executing the program stored on memory, realizes the extracting method of any of the above-described characteristics of image
Step.
Fourth aspect, the embodiment of the present invention additionally provide a kind of computer readable storage medium, which is characterized in that the meter
Computer program is stored in calculation machine readable storage medium storing program for executing, the computer program realizes any of the above-described figure when being executed by processor
As the extracting method step of feature.
In the scheme that the embodiment of the present invention is provided, pending image is obtained first, pending image is divided into multiple
Image-region determines pending image-region then according to predetermined manner from multiple images region, pending image-region
Quantity is less than the quantity of image-region, in turn, carries out feature extraction to pending image-region, obtains characteristics of image.Due to waiting for
The quantity for handling image-region is less than the quantity of image-region, it is therefore not necessary to carry out figure to all parts in pending image
As feature extraction, the time needed for extraction characteristics of image can be shortened, improve image characteristics extraction efficiency.
Description of the drawings
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
Obtain other attached drawings according to these attached drawings.
A kind of flow chart of the extracting method for characteristics of image that Fig. 1 is provided by the embodiment of the present invention;
Fig. 2 is the particular flow sheet of step S103 in embodiment illustrated in fig. 1;
Fig. 3 is the particular flow sheet of step S202 in embodiment illustrated in fig. 2;
A kind of structural schematic diagram of the extraction element for characteristics of image that Fig. 4 is provided by the embodiment of the present invention;
The structural schematic diagram for a kind of electronic equipment that Fig. 5 is provided by the embodiment of the present invention.
Specific implementation mode
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation describes, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
In order to shorten the time needed for extraction characteristics of image, the extraction efficiency of characteristics of image is improved, the embodiment of the present invention carries
A kind of extracting method of characteristics of image, device, electronic equipment and computer readable storage medium are supplied.
A kind of extracting method of characteristics of image is provided for the embodiments of the invention first below to be introduced.
A kind of extracting method for characteristics of image that the embodiment of the present invention is provided can be applied to arbitrarily can to image into
Row processing, extracts the electronic equipment of characteristics of image, hereinafter referred to as electronic equipment, for example, can be robot, computer, processor
Equal electronic equipments, are not specifically limited herein.
As shown in Figure 1, a kind of extracting method of characteristics of image, the method includes:
S101 obtains pending image;
The pending image is divided into multiple images region by S102;
S103 determines pending image-region according to predetermined manner from described multiple images region;
Wherein, the quantity of the pending image-region is less than the quantity in described image region.
S104 carries out feature extraction to the pending image-region, obtains characteristics of image.
As it can be seen that in the scheme that the embodiment of the present invention is provided, electronic equipment can obtain pending image first, will wait locating
Reason image is divided into multiple images region, and then according to predetermined manner, pending image-region is determined from multiple images region,
The quantity of pending image-region is less than the quantity of image-region, in turn, carries out feature extraction to pending image-region, obtains
Characteristics of image.Since the quantity of pending image-region is less than the quantity of image-region, it is therefore not necessary to in pending image
All parts carry out image characteristics extraction, can shorten the time needed for extraction characteristics of image, improve image characteristics extraction efficiency.
In above-mentioned steps S101, electronic equipment can obtain pending image, it is to be understood that pending image is
To need to carry out the image of image characteristics extraction.Can be robot during determining motion state by visual odometry,
The image of the image capture device acquisition of robot installation, or the image etc. obtained in image-detection process, herein not
It is specifically limited.
After obtaining above-mentioned pending image, pending image can be divided into multiple images region by electronic equipment, specifically
Pending image can be divided into the identical or of different sizes multiple images region of size.Size for image-region and
The quantity embodiment of the present invention is not specifically limited herein.For example, identical 4 image-regions of size can be divided into, it can also
4 image-regions of different sizes are divided into, identical 6 regions of size, etc. can also be divided into.
In turn, in above-mentioned steps S103, electronic equipment can be according to predetermined manner, from the multiple images area that division obtains
Pending image-region is determined in domain, wherein the quantity of pending image-region is less than the quantity of image-region.For example, image
Region shares 8, then electronic equipment can determine the processing figure less than 8 according to predetermined manner from 8 image-regions
As region, for example, 7,6,5 etc..
In order to which the characteristics of image ensured includes valuable characteristics of image more as possible, so waiting locating in above-mentioned determination
When managing image-region, the more image-region of the valuable characteristics of image for including can be retained as possible, that is, by including
The more image-region of valuable characteristics of image is determined as pending image-region, and by including characteristics of image it is less or
It is not that especially valuable image-region is given up.In this way, can not only improve the efficiency of image characteristics extraction, while can also protect
The characteristics of image that card later use obtains carries out the accuracy rate of the processes such as characteristic matching.Wherein, " valuable characteristics of image " can
To be interpreted as representative characteristics of image, the characteristics of image that can be utilized in subsequent characteristics matching process.For example, for
For one piece of blackboard in image, the characteristics of image of the edge of blackboard may act as valuable characteristics of image, and blackboard
Middle part is the region of a piece of pure color, does not also just have the characteristics of image that can be utilized during characteristic matching, therefore this part figure
As feature is not just valuable characteristics of image yet.
So, it is assumed that pending image is divided into 8 image-regions, the picture material that one of image-region includes
For pure color metope, it is evident that, which does not include valuable characteristics of image, then the image-region can not yet
As pending image-region.
In turn, electronic equipment can execute above-mentioned steps S104, i.e., carrying out feature to determining pending image-region carries
It takes, obtains characteristics of image.Wherein, image processing field times may be used in the mode that feature extraction is carried out to pending image-region
The image characteristics extraction mode of meaning, for example, can be determined according to the difference of gray value between certain pixel and surrounding pixel point should
Whether pixel can be used as the modes such as characteristics of image, be not specifically limited and illustrate herein.
As a kind of embodiment of the embodiment of the present invention, as shown in Fig. 2, above-mentioned according to predetermined manner, from the multiple
In image-region the step of determining pending image-region, may include:
Described multiple images region is converted to frequency domain by S201, obtains the corresponding frequency spectrum of each image-region;
In order to give up the image-region not including especially valuable characteristics of image, electronic equipment can be by multiple images
Region is changed to frequency domain from transform of spatial domain, also can be obtained by the corresponding frequency spectrum of each image-region.Wherein, by image-region from
The mode that transform of spatial domain is changed to frequency domain may be used image processing field image be arbitrarily changed to frequency domain from transform of spatial domain
Mode is not specifically limited herein.
S202, judges whether each frequency spectrum high frequency components meet preset condition;
It is understood that each image-region is substantially a matrix, after an image-region is converted to frequency domain,
Obtained frequency spectrum is a matrix identical with its ranks number, and the element in matrix is spectrum component, and the numerical value of element is
The range value of spectrum component.
In order to facilitate calculating, the frequency spectrum that each image-region can be converted to frequency domain normalizes to [0,2 π], will frequency
The range value of spectral component normalizes to [0,1], then the position of obtained frequency spectrum spectral components can use [0,2 π] to indicate,
For example, it is assumed that frequency spectrum is the matrix of a 16*16, then, ranks are the matrix lower right corner in the spectrum component of [1.5 π, 2 π]
16 elements.
Spectrum component can be divided into high fdrequency component and low frequency component, wherein high fdrequency component is that frequency values are more than preset value
Component, and low frequency component can be then the component that frequency values are not more than preset value.For frequency spectrum after normalized, high frequency
Component is the component that place ranks are more than preset value, which can correspond to according to valuable feature in most of images
Frequency spectrum in frequency values setting.For example, by statistics, it is assumed that in 1000 images are converted to the frequency spectrum that frequency domain obtains,
The frequency values of the corresponding spectrum component of more valuable feature are all higher than numerical value A, then can be using numerical value A as above-mentioned pre-
If value.
After obtaining the corresponding frequency spectrum of each image-region, whether electronic equipment may determine that each frequency spectrum high frequency components
Meet preset condition, whether includes valuable characteristics of image with the corresponding image-region of each frequency spectrum of determination.
In one embodiment, electronic equipment may determine that each high fdrequency component in each frequency spectrum range value whether
More than predetermined threshold value, if it is greater than predetermined threshold value, it is determined that the high fdrequency component is effective high fdrequency component, in turn, it can be determined that every
Whether effective high fdrequency component in a frequency spectrum is more than preset quantity, if it is greater than preset quantity, it is determined that the frequency spectrum medium-high frequency point
Amount meets preset condition, is otherwise unsatisfactory for preset condition.
For example, pending image is divided into 6 image-regions, it is respectively converted into frequency domain, obtains corresponding 6 frequency spectrums
A-F, it is assumed that predetermined threshold value 0.5, preset quantity 500, then, for frequency spectrum B, the range value of high fdrequency component is more than 0.5
It is 600, it is evident that 600 are more than 500, then can determines that the high fdrequency component in frequency spectrum B meets preset condition.
In another implementation, electronic equipment may determine that the range value of all high fdrequency components in each frequency spectrum
Whether adduction is more than default adduction threshold value, if it is greater than default adduction threshold value, it is determined that the frequency spectrum high frequency components meet default
Otherwise condition is unsatisfactory for preset condition.
For example, pending image is divided into 16 image-regions, it is respectively converted into frequency domain, obtains corresponding 6 frequencies
Compose A-P, it is assumed that default adduction threshold value is 5.5, then, for frequency spectrum M, high fdrequency component shares 300, the adduction of range value
It is 5.0, it is evident that 5.0 are less than 5.5, then can determines that the high fdrequency component in frequency spectrum M is unsatisfactory for preset condition.
It can certainly determine whether each frequency spectrum high frequency components meet preset condition using other reasonable manners,
This is not specifically limited.
High fdrequency component is met the image-region corresponding to the frequency spectrum of the preset condition and is determined as pending image by S203
Region.
In turn, after whether each frequency spectrum high frequency components of determination meet preset condition, due to corresponding frequency spectrum medium-high frequency
Component meets preset condition, illustrates that the image-region includes more valuable characteristics of image, then electronic equipment can
High fdrequency component is met into the image-region corresponding to the frequency spectrum of above-mentioned preset condition and is determined as pending image-region, for follow-up
Extract characteristics of image.
As it can be seen that in the present embodiment, multiple images region can be converted to frequency domain by electronic equipment, judgement obtains each
Whether the corresponding frequency spectrum high frequency components of image-region meet preset condition, and high fdrequency component is then met the preset condition
Image-region corresponding to frequency spectrum is determined as pending image-region.Can will include comparing by the screening to high fdrequency component
The image-region of valuable characteristics of image is determined as pending image-region, not only can reduce image characteristics extraction
The calculation amount of journey improves the extraction efficiency of characteristics of image, while can also ensure to extract as far as possible valuable in pending image
The characteristics of image of value ensures the accuracy rate of image characteristics extraction.
As a kind of embodiment of the embodiment of the present invention, as shown in figure 3, each frequency spectrum high frequency components of above-mentioned judgement are
No the step of meeting preset condition, may include:
S301 calculates the gross energy of each frequency spectrum high frequency components;
Since the gross energy of frequency spectrum high frequency components reflects high fdrequency component proportion shared in frequency spectrum, electronics
Equipment can calculate the gross energy of each frequency spectrum high frequency components.
It is understood that for the frequency spectrum being converted to by image, the gross energy of high fdrequency component can use high frequency
The quadratic sum of the range value of component indicates.The value that the summed square of the range value of namely all high fdrequency components obtains.
For example, for frequency spectrum a, if it includes 10 high fdrequency components, corresponding range value be respectively a1,
A2, a3, a4, a5, a6, a7, a8, a9 and a10, then the gross energy of frequency spectrum a high frequency components is (a1)2+(a2)2+(a3)2+
(a4)2+(a5)2+(a6)2+(a7)2+(a8)2+(a9)2+(a10)2Value.
S302, judges whether the gross energy of each frequency spectrum high frequency components is not less than preset energy value, if it is, executing
Step S303;If not, thening follow the steps S304;
S303 determines that the frequency spectrum high frequency components meet preset condition;
S304 determines that the frequency spectrum high frequency components are unsatisfactory for preset condition.
Next, electronic equipment may determine that whether the gross energy of each frequency spectrum high frequency components is not less than preset energy
Value, if the gross energy of frequency spectrum high frequency components is not less than preset energy value, illustrate the gross energies of the frequency spectrum high frequency components compared with
Height, corresponding image-region will likely include ones which more valuable characteristics of image, then the height in the frequency spectrum can be determined
Frequency component meets preset condition, subsequently then can the corresponding image-region of the frequency spectrum be determined as pending image-region, is used for
Extract characteristics of image.
If the gross energy of frequency spectrum high frequency components is less than preset energy value, illustrate the gross energy of the frequency spectrum high frequency components
Relatively low, corresponding image-region is likely to not include more valuable characteristics of image, then can determine in the frequency spectrum
High fdrequency component is unsatisfactory for preset condition, subsequently then the corresponding image-region of the frequency spectrum is not determined as pending image-region, no
For extracting characteristics of image.
For example, the frequency spectrum after being normalized isIn, it is assumed that default energy
Magnitude is 0.05, and high fdrequency component is spectrum component of the ranks between [π, 2 π]That is these spectrum components
The quadratic sum of range value is 1.8246, it is clear that is more than 0.05, then can determine that the high fdrequency component satisfaction in the frequency spectrum is default
Condition.
As it can be seen that in the present embodiment, electronic equipment can be by the gross energy of each frequency spectrum high frequency components of calculating, in turn
Judge whether the gross energy of each frequency spectrum high frequency components not less than the mode of preset energy value determines that frequency spectrum high frequency components are
It is no to meet preset condition, rapidly and accurately it can determine whether frequency spectrum high frequency components meet preset condition, and then quick and precisely
Ground determines whether for the corresponding image-region of frequency spectrum to be determined as the pending image-region for extracting characteristics of image.
Frequency domain is converted to for above by by above-mentioned multiple images region, obtains the corresponding frequency spectrum of each image-region
Mode for, it is above-mentioned that the pending image is divided into multiple images as a kind of embodiment of the embodiment of the present invention
The step of region may include:
The pending image is divided into the multiple images region that size is (N*N).
Multiple images region is subsequently converted into frequency domain for convenience, it, can be with when being divided to pending image
Pending image is divided into the multiple images region that size is (N*N), wherein N=2n, n is positive integer.That is, can
Pending image is divided multiple square areas, index that the length of side of each square area is 2.
Electronic equipment can select suitable size to draw pending area according to the resolution ratio of pending image
Point, for example, it is the image-regions such as 4*4,8*8,16*16 that can be divided into size.
It should be noted that the size in multiple images region may be the same or different, as long as each image-region is
The length of side is 2nSquare area.For example, the size for dividing some image-regions in obtained image-region can be 8*
8, the size of other image-regions can be 16*16.
Correspondingly, above-mentioned be converted to frequency domain by described multiple images region, the corresponding frequency spectrum of each image-region is obtained
The step of, may include:
Using dct algorithm or Wavelet Transformation Algorithm, described multiple images region is converted into frequency domain, is obtained
To the corresponding frequency spectrum of each image-region.
After pending image to be divided into the multiple images region that size is (N*N), electronic equipment can use discrete
Multiple images region is converted to frequency domain, it is corresponding to obtain each image-region by cosine transform algorithm or Wavelet Transformation Algorithm
Frequency spectrum.Wherein, symmetry transformation algorithm can be dct algorithm (Discrete Cosine Transform, DCT),
Wavelet Transformation Algorithm (wavelet) etc., is not specifically limited herein.
Image area transitions are in the way of frequency domain obtains frequency spectrum, specifically to come by explanation by taking dct algorithm as an example
It says, frequency spectrum can be obtained by following formula:
Wherein, f (x, y) is the coordinate value of the pixel in the image-region of N*N, and F (u, v) is then corresponding spectrum component
Range value.Each image-region can be converted to frequency domain with upper type by using, and obtain corresponding frequency spectrum.
As it can be seen that in the present embodiment, pending image can be divided into the multiple images that size is (N*N) by electronic equipment
Multiple images region is converted to frequency domain by dct algorithm, Wavelet Transformation Algorithm etc., obtained by region in turn
The corresponding frequency spectrum of each image-region.The corresponding frequency spectrum of each image-region can be rapidly and accurately obtained, it is convenient follow-up determining
Pending image-region.
The preset value being more accurately suitble in order to obtain, it is above-mentioned default as a kind of embodiment of the embodiment of the present invention
The method of determination of value may include:
The pending image is inputted into target network model, obtains the corresponding preset value of the pending image.
Wherein, target network model can be training completion in advance for determining the corresponding frequency spectrum high frequency components of image
The network model of corresponding preset value.That is, for the frequency spectrum that frequency domain obtains, target network mould is converted the image into
Which component in frequency spectrum be determined for and may be considered high fdrequency component for type.Target network model can be convolutional neural networks
Model even depth learning network model, is not specifically limited herein.
Due to the characteristics of image that for the pending image acquired under different application scene, pending image includes
Whether valuable, evaluation criteria may have difference.For example, if pending image is used for vehicle detection, it is pending
The characteristics of image of vehicle in image is valuable characteristics of image;If pending image is determined for movable equipment
Position, then the characteristics of image with significant object in pending image is valuable characteristics of image.So in order to
Above-mentioned preset value is more accurately set, can training objective network model in advance, for determining the corresponding frequency spectrum medium-high frequency of image
Preset value corresponding to component.
It,, can be in training objective network model in order to adapt to various application scenarios in the first realization method
Structure initial network model in advance acquires multiple images sample under various application scenarios, then divides each image pattern
For multiple images area sample, and each image-region sample is converted into frequency domain, obtains each image-region sample and correspond to frequency spectrum
Sample.According to the valuable characteristics of image that each image-region sample includes, the frequency spectrum point in corresponding spectral samples is marked
It measures, and determines the preset value of the corresponding high fdrequency component of each image-region sample according to these spectrum components.First way, can be with
Using these spectrum components frequency values average value as preset value;These spectrum components may be used in the second way
As preset value, this is all reasonable minimum value in frequency values.
In turn, can by image-region sample and its initial network model that build in advance of corresponding preset value input into
Row training, in the training process, initial network model can gradually establish figure by learning the characteristics of image of image-region sample
As the correspondence of area sample and preset value, and then obtain target network model.
In another implementation, in order to more targeted to various application scenarios, for each application scenarios
It can determine more accurate preset value, each application scenarios can be trained to obtain a target network model.So just
Multiple images sample can be acquired under each application scenarios, each image pattern is divided into multiple images area sample, so
The target network model under the application scenarios is obtained by image-region sample training afterwards.It waits locating in this way, obtaining in electronic equipment
After managing image, suitable target network model can be selected to determine above-mentioned preset value according to its application scenarios.Due to every
The training method of the corresponding target network model of a application scenarios in the first above-mentioned realization method to target network model
Training method is identical, therefore details are not described herein.
It should be noted that due to whether dividing an image into multiple images region and have no effect on the image that image includes
The essence of feature therefore, can not also be to image pattern in training objective network model in order to reduce the cumbersome degree of realization
It is divided, directly the corresponding preset value of image pattern is inputted in the initial network model built in advance and is trained,
Have no effect on the accuracy that target network model determines preset value.
As it can be seen that in the present embodiment, pending image can be inputted target network model and is detected by electronic equipment, be obtained
To the corresponding preset value of pending image, since target network model can be a large amount of previously according to what is obtained in practical application scene
Image pattern trains to obtain, it may be determined that more accurate and suitable preset value, the pending image-region that can make are most
It may include valuable characteristics of image, improve the accuracy of image characteristics extraction.
It is above-mentioned according to predetermined manner as a kind of embodiment of the embodiment of the present invention, from described multiple images region
The step of determining pending image-region may include:
Image detection is carried out to described multiple images region, determines whether each image-region includes object;It will include object
The image-region of body is determined as pending image-region.
May not include object due in pending image, there is some regions, for example, a piece of sky, a sidewalls,
The pure color part of one piece of blackboard does not include valuable characteristics of image in such region, then in order to from pending image
Middle to screen such region, electronic equipment can carry out image detection to the multiple images region that division obtains, and determine
Whether each image-region includes object.The arbitrary image detection side of technical field of image processing may be used in specific detection mode
Formula is not specifically limited and illustrates as long as can detect whether image-region includes object herein.
In turn, after the testing result for obtaining each image-region, electronic equipment then can be by the image including object
Region is determined as pending image-region, need not then be further continued for handling without the image-region including object.
As it can be seen that in the present embodiment, electronic equipment can carry out image detection to above-mentioned multiple images region, determine each
Whether image-region includes object, and then the image-region including object is determined as pending image-region.By this way
The region for not including valuable characteristics of image in pending image can be removed, the calculating of subsequent image feature extraction is reduced
Amount, improves the efficiency of image characteristics extraction.
As a kind of embodiment of the embodiment of the present invention, the extracting method of above-mentioned characteristics of image can be used for utilizing vision
Odometer estimates in the scene of movable equipment movement that in this case, the above method can also include:
Obtained characteristics of image is matched with the characteristics of image of target image, obtains matching result;Based on described
With as a result, determining that the image capture device for acquiring the pending image and the target image is acquiring the pending image
And the operating status in the time interval of the target image.
In the scene using visual odometry estimation movable equipment movement, Image Acquisition is installed on movable equipment
Equipment, the image capture device acquire an image at regular intervals, using two images, can estimate to acquire this two
In the time interval of image, the motion state of image capture device.Since image capture device is installed on movable equipment, then
Motion state of the motion state i.e. movable equipment of image capture device in the time interval.
Above-mentioned pending image can be as one in two images for estimating movable equipment motion state, mesh
Logo image is another.It is understood that pending image and target image are the image for being installed on movable equipment
Collecting device acquired image.In this way, the image that electronic equipment after obtaining target image, can extract target image is special
Sign, in order to improve the extraction efficiency of characteristics of image, above-mentioned image can also be used by extracting the mode of the characteristics of image of target image
The figure area method of feature.
In turn, after extraction obtains the characteristics of image of pending image, electronic equipment can be by the figure of pending image
As feature is matched with the characteristics of image of target image, matching result is obtained.In turn, it is based on the matching result, also
Determine operating status of the image capture device in the time interval for acquiring pending image and target image.Wherein, due to figure
Determine that the mode of the operating status of image capture device can be led with image procossing as characteristic matching and based on the matching result
The relevant way in domain, is not specifically limited and illustrates herein.For example, description that characteristics of image may be used determines two images
Whether feature matches;Epipolar geometry algorithm, PnP (perspective n point locations) derivation algorithm, ICP algorithm (Iterative may be used
Closest Point, iteration closest approach algorithm) etc. determine image capture device operating status.
As it can be seen that in the present embodiment, in the application scenarios using visual odometry estimation movable equipment movement, being based on
The extracting method of above-mentioned characteristics of image, the characteristics of image for the target image that electronic equipment is extracted by obtained characteristics of image and in advance
It is matched, obtains matching result, and then be based on matching result, determine the Image Acquisition for acquiring pending image and target image
Operating status of the equipment in the time interval for acquiring pending image and target image.Also it is assured that installing the image adopts
Collect the motion state of the movable equipment of equipment, it is possible to reduce the calculation amount of visual odometry, raising are estimated using visual odometry
Count the efficiency of movable equipment movement.
Corresponding to above method embodiment, the embodiment of the present invention additionally provides a kind of extraction element of characteristics of image.Below
A kind of extraction element of characteristics of image is provided for the embodiments of the invention to be introduced.
As shown in figure 4, a kind of extraction element of characteristics of image, described device include:
Image collection module 410, for obtaining pending image;
Image division module 420, for the pending image to be divided into multiple images region;
Area determination module 430, for according to predetermined manner, pending image district to be determined from described multiple images region
Domain, wherein the quantity of the pending image-region is less than the quantity in described image region;
Characteristic extracting module 440 obtains characteristics of image for carrying out feature extraction to the pending image-region.
As it can be seen that in the scheme that the embodiment of the present invention is provided, electronic equipment can obtain pending image first, will wait locating
Reason image is divided into multiple images region, and then according to predetermined manner, pending image-region is determined from multiple images region,
The quantity of pending image-region is less than the quantity of image-region, in turn, carries out feature extraction to pending image-region, obtains
Characteristics of image.Since the quantity of pending image-region is less than the quantity of image-region, it is therefore not necessary to in pending image
All parts carry out image characteristics extraction, can shorten the time needed for extraction characteristics of image, improve image characteristics extraction efficiency.
As a kind of embodiment of the embodiment of the present invention, above-mentioned zone determining module 430 may include:
Image conversion unit (is not shown) in Fig. 4, for described multiple images region to be converted to frequency domain, obtains each
The corresponding frequency spectrum of image-region;
Frequency spectrum judging unit (is not shown) in Fig. 4, for judging whether each frequency spectrum high frequency components meet preset condition,
Wherein, the high fdrequency component is point that frequency values are more than the preset value determined by preset value determining module (being not shown in Fig. 4)
Amount;
First area determination unit (is not shown) in Fig. 4, the frequency spectrum institute for high fdrequency component to be met to the preset condition
Corresponding image-region is determined as pending image-region.
As a kind of embodiment of the embodiment of the present invention, above-mentioned frequency spectrum judging unit may include:
Gross energy computation subunit (is not shown) in Fig. 4, the gross energy for calculating each frequency spectrum high frequency components;
Gross energy judgment sub-unit (is not shown) in Fig. 4, for judge each frequency spectrum high frequency components gross energy whether
Not less than preset energy value;
First determination subelement (is not shown) in Fig. 4, for being not less than default energy in the gross energy of frequency spectrum high frequency components
When magnitude, determine that the frequency spectrum high frequency components meet preset condition;
The and determination subelement (being not shown in Fig. 4), be less than preset energy for the gross energy in frequency spectrum high frequency components
When value, determine that the frequency spectrum high frequency components are unsatisfactory for preset condition.
As a kind of embodiment of the embodiment of the present invention, above-mentioned image division module 420 may include:
Image division unit (is not shown) in Fig. 4, is the multiple of (N*N) for the pending image to be divided into size
Image-region, wherein N=2n, n is positive integer;
Described image converting unit includes:
Image conversion subunit (is not shown) in Fig. 4, will for using dct algorithm or Wavelet Transformation Algorithm
Described multiple images region is converted to frequency domain, obtains the corresponding frequency spectrum of each image-region.
As a kind of embodiment of the embodiment of the present invention, above-mentioned preset value determining module may include:
Preset value determination unit (is not shown) in Fig. 4, for the pending image to be inputted target network model, obtains
The corresponding preset value of the pending image, wherein the target network model be training in advance complete for determining image
The network model of preset value corresponding to corresponding frequency spectrum high frequency components.
As a kind of embodiment of the embodiment of the present invention, above-mentioned zone determining module 430 may include:
Image detecting element (is not shown) in Fig. 4, for carrying out image detection to described multiple images region, determines each
Whether image-region includes object;
Second area determination unit (is not shown) in Fig. 4, for will include that the image-region of object is determined as pending figure
As region.
As a kind of embodiment of the embodiment of the present invention, above-mentioned apparatus can also include:
Characteristic matching module (is not shown) in Fig. 4, for by the characteristics of image of obtained characteristics of image and target image into
Row matching, obtains matching result;
Determining module (being not shown in Fig. 4) is moved, for being based on the matching result, determines and acquires the pending image
And the image capture device of the target image is in the time interval for acquiring the pending image and the target image
Operating status.
The embodiment of the present invention additionally provides a kind of electronic equipment, as shown in figure 5, including processor 501, communication interface 502,
Memory 503 and communication bus 504, wherein processor 501, communication interface 502, memory 503 are complete by communication bus 504
At mutual communication,
Memory 503, for storing computer program;
Processor 501 when for executing the program stored on memory 503, realizes following steps:
Obtain pending image;
The pending image is divided into multiple images region;
According to predetermined manner, pending image-region is determined from described multiple images region, wherein the pending figure
As the quantity in region is less than the quantity in described image region;
Feature extraction is carried out to the pending image-region, obtains characteristics of image.
As it can be seen that in the scheme that the embodiment of the present invention is provided, electronic equipment can obtain pending image first, will wait locating
Reason image is divided into multiple images region, and then according to predetermined manner, pending image-region is determined from multiple images region,
The quantity of pending image-region is less than the quantity of image-region, in turn, carries out feature extraction to pending image-region, obtains
Characteristics of image.Since the quantity of pending image-region is less than the quantity of image-region, it is therefore not necessary to in pending image
All parts carry out image characteristics extraction, can shorten the time needed for extraction characteristics of image, improve image characteristics extraction efficiency.
The communication bus that above-mentioned electronic equipment is mentioned can be Peripheral Component Interconnect standard (Peripheral Component
Interconnect, PCI) bus or expanding the industrial standard structure (Extended Industry Standard
Architecture, EISA) bus etc..The communication bus can be divided into address bus, data/address bus, controlling bus etc..For just
It is only indicated with a thick line in expression, figure, it is not intended that an only bus or a type of bus.
Communication interface is for the communication between above-mentioned electronic equipment and other equipment.
Memory may include random access memory (Random Access Memory, RAM), can also include non-easy
The property lost memory (Non-Volatile Memory, NVM), for example, at least a magnetic disk storage.Optionally, memory may be used also
To be at least one storage device for being located remotely from aforementioned processor.
Above-mentioned processor can be general processor, including central processing unit (Central Processing Unit,
CPU), network processing unit (Network Processor, NP) etc.;It can also be digital signal processor (Digital Signal
Processing, DSP), it is application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), existing
It is field programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic device, discrete
Door or transistor logic, discrete hardware components.
Wherein, above-mentioned according to predetermined manner, it the step of pending image-region is determined from described multiple images region, can
To include:
Described multiple images region is converted into frequency domain, obtains the corresponding frequency spectrum of each image-region;
Judge whether each frequency spectrum high frequency components meet preset condition, wherein the high fdrequency component is more than for frequency values
The component of preset value;
High fdrequency component is met into the image-region corresponding to the frequency spectrum of the preset condition and is determined as pending image-region.
Wherein, above-mentioned that the step of whether each frequency spectrum high frequency components meet preset condition judged, may include:
Calculate the gross energy of each frequency spectrum high frequency components;
Judge whether the gross energy of each frequency spectrum high frequency components is not less than preset energy value;
If so, determining that the frequency spectrum high frequency components meet preset condition;
If not, determining that the frequency spectrum high frequency components are unsatisfactory for preset condition.
Wherein, the above-mentioned the step of pending image is divided into multiple images region, may include:
The pending image is divided into the multiple images region that size is (N*N), wherein N=2n, n is positive integer;
It is described that described multiple images region is converted into frequency domain, the step of obtaining each image-region corresponding frequency spectrum,
Including:
Using dct algorithm or Wavelet Transformation Algorithm, described multiple images region is converted into frequency domain, is obtained
To the corresponding frequency spectrum of each image-region.
Wherein, the method for determination of above-mentioned preset value may include:
The pending image is inputted into target network model, obtains the corresponding preset value of the pending image, wherein
The target network model is the default corresponding to the corresponding frequency spectrum high frequency components of image for determining of training completion in advance
The network model of value.
Wherein, above-mentioned according to predetermined manner, it the step of pending image-region is determined from described multiple images region, can
To include:
Image detection is carried out to described multiple images region, determines whether each image-region includes object;
Image-region including object is determined as pending image-region.
Wherein, the above method can also include:
Obtained characteristics of image is matched with the characteristics of image of target image, obtains matching result;
Based on the matching result, determine that the image capture device for acquiring the pending image and the target image exists
Acquire the operating status in the time interval of the pending image and the target image.
The embodiment of the present invention additionally provides a kind of computer readable storage medium, the computer readable storage medium memory
Computer program is contained, the computer program realizes following steps when being executed by processor:
Obtain pending image;
The pending image is divided into multiple images region;
According to predetermined manner, pending image-region is determined from described multiple images region, wherein the pending figure
As the quantity in region is less than the quantity in described image region;
Feature extraction is carried out to the pending image-region, obtains characteristics of image.
As it can be seen that in the scheme that the embodiment of the present invention is provided, when computer program is executed by processor, can obtain first
Pending image is divided into multiple images region by pending image, then according to predetermined manner, from multiple images region really
Fixed pending image-region, the quantity of pending image-region is less than the quantity of image-region, in turn, to pending image-region
Feature extraction is carried out, characteristics of image is obtained.Since the quantity of pending image-region is less than the quantity of image-region, nothing
Image characteristics extraction need to be carried out to all parts in pending image, the time needed for extraction characteristics of image can be shortened, carried
Hi-vision feature extraction efficiency.
Wherein, above-mentioned according to predetermined manner, it the step of pending image-region is determined from described multiple images region, can
To include:
Described multiple images region is converted into frequency domain, obtains the corresponding frequency spectrum of each image-region;
Judge whether each frequency spectrum high frequency components meet preset condition, wherein the high fdrequency component is more than for frequency values
The component of preset value;
High fdrequency component is met into the image-region corresponding to the frequency spectrum of the preset condition and is determined as pending image-region.
Wherein, above-mentioned that the step of whether each frequency spectrum high frequency components meet preset condition judged, may include:
Calculate the gross energy of each frequency spectrum high frequency components;
Judge whether the gross energy of each frequency spectrum high frequency components is not less than preset energy value;
If so, determining that the frequency spectrum high frequency components meet preset condition;
If not, determining that the frequency spectrum high frequency components are unsatisfactory for preset condition.
Wherein, the above-mentioned the step of pending image is divided into multiple images region, may include:
The pending image is divided into the multiple images region that size is (N*N), wherein N=2n, n is positive integer;
It is described that described multiple images region is converted into frequency domain, the step of obtaining each image-region corresponding frequency spectrum,
Including:
Using dct algorithm or Wavelet Transformation Algorithm, described multiple images region is converted into frequency domain, is obtained
To the corresponding frequency spectrum of each image-region.
Wherein, the method for determination of above-mentioned preset value may include:
The pending image is inputted into target network model, obtains the corresponding preset value of the pending image, wherein
The target network model is the default corresponding to the corresponding frequency spectrum high frequency components of image for determining of training completion in advance
The network model of value.
Wherein, above-mentioned according to predetermined manner, it the step of pending image-region is determined from described multiple images region, can
To include:
Image detection is carried out to described multiple images region, determines whether each image-region includes object;
Image-region including object is determined as pending image-region.
Wherein, the above method can also include:
Obtained characteristics of image is matched with the characteristics of image of target image, obtains matching result;
Based on the matching result, determine that the image capture device for acquiring the pending image and the target image exists
Acquire the operating status in the time interval of the pending image and the target image.
It should be noted that for above-mentioned apparatus, equipment and computer readable storage medium embodiment, due to its base
Originally it is similar to embodiment of the method, so description is fairly simple, the relevent part can refer to the partial explaination of embodiments of method.
Need further exist for explanation, herein, relational terms such as first and second and the like be used merely to by
One entity or operation are distinguished with another entity or operation, without necessarily requiring or implying these entities or operation
Between there are any actual relationship or orders.Moreover, the terms "include", "comprise" or its any other variant meaning
Covering non-exclusive inclusion, so that the process, method, article or equipment including a series of elements includes not only that
A little elements, but also include other elements that are not explicitly listed, or further include for this process, method, article or
The intrinsic element of equipment.In the absence of more restrictions, the element limited by sentence "including a ...", is not arranged
Except there is also other identical elements in the process, method, article or apparatus that includes the element.
Each embodiment in this specification is all made of relevant mode and describes, identical similar portion between each embodiment
Point just to refer each other, and each embodiment focuses on the differences from other embodiments.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the scope of the present invention.It is all
Any modification, equivalent replacement, improvement and so within the spirit and principles in the present invention, are all contained in protection scope of the present invention
It is interior.
Claims (10)
1. a kind of extracting method of characteristics of image, which is characterized in that the method includes:
Obtain pending image;
The pending image is divided into multiple images region;
According to predetermined manner, pending image-region is determined from described multiple images region, wherein the pending image district
The quantity in domain is less than the quantity in described image region;
Feature extraction is carried out to the pending image-region, obtains characteristics of image.
2. the method as described in claim 1, which is characterized in that it is described according to predetermined manner, from described multiple images region
The step of determining pending image-region, including:
Described multiple images region is converted into frequency domain, obtains the corresponding frequency spectrum of each image-region;
Judge whether each frequency spectrum high frequency components meet preset condition, wherein the high fdrequency component is frequency values more than default
The component of value;
High fdrequency component is met into the image-region corresponding to the frequency spectrum of the preset condition and is determined as pending image-region.
3. method as claimed in claim 2, which is characterized in that described to judge whether each frequency spectrum high frequency components meet default
The step of condition, including:
Calculate the gross energy of each frequency spectrum high frequency components;
Judge whether the gross energy of each frequency spectrum high frequency components is not less than preset energy value;
If so, determining that the frequency spectrum high frequency components meet preset condition;
If not, determining that the frequency spectrum high frequency components are unsatisfactory for preset condition.
4. method as claimed in claim 2 or claim 3, which is characterized in that described that the pending image is divided into multiple images
The step of region, including:
The pending image is divided into the multiple images region that size is (N*N), wherein N=2n, n is positive integer;
It is described that described multiple images region is converted into frequency domain, the step of obtaining each image-region corresponding frequency spectrum, including:
Using dct algorithm or Wavelet Transformation Algorithm, described multiple images region is converted into frequency domain, is obtained every
The corresponding frequency spectrum of a image-region.
5. method as claimed in claim 2 or claim 3, which is characterized in that the method for determination of the preset value, including:
The pending image is inputted into target network model, obtains the corresponding preset value of the pending image, wherein described
Target network model is the preset value being used to determine corresponding to the corresponding frequency spectrum high frequency components of image that training in advance is completed
Network model.
6. the method as described in claim 1, which is characterized in that it is described according to predetermined manner, from described multiple images region
The step of determining pending image-region, including:
Image detection is carried out to described multiple images region, determines whether each image-region includes object;
Image-region including object is determined as pending image-region.
7. the method as described in claim 1, which is characterized in that the method further includes:
Obtained characteristics of image is matched with the characteristics of image of target image, obtains matching result;
Based on the matching result, determine that the image capture device for acquiring the pending image and the target image is acquiring
Operating status in the time interval of the pending image and the target image.
8. a kind of extraction element of characteristics of image, which is characterized in that described device includes:
Image collection module, for obtaining pending image;
Image division module, for the pending image to be divided into multiple images region;
Area determination module, for according to predetermined manner, pending image-region to be determined from described multiple images region,
In, the quantity of the pending image-region is less than the quantity in described image region;
Characteristic extracting module obtains characteristics of image for carrying out feature extraction to the pending image-region.
9. a kind of electronic equipment, which is characterized in that including processor, memory and communication bus, wherein processor, memory
Mutual communication is completed by communication bus;
Memory, for storing computer program;
Processor when for executing the program stored on memory, realizes any method and steps of claim 1-7.
10. a kind of computer readable storage medium, which is characterized in that be stored with computer in the computer readable storage medium
Program realizes claim 1-7 any method and steps when the computer program is executed by processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810404934.XA CN108734175A (en) | 2018-04-28 | 2018-04-28 | A kind of extracting method of characteristics of image, device and electronic equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810404934.XA CN108734175A (en) | 2018-04-28 | 2018-04-28 | A kind of extracting method of characteristics of image, device and electronic equipment |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108734175A true CN108734175A (en) | 2018-11-02 |
Family
ID=63939952
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810404934.XA Pending CN108734175A (en) | 2018-04-28 | 2018-04-28 | A kind of extracting method of characteristics of image, device and electronic equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108734175A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110544234A (en) * | 2019-07-30 | 2019-12-06 | 北京达佳互联信息技术有限公司 | Image noise detection method, image noise detection device, electronic equipment and storage medium |
CN111563416A (en) * | 2020-04-08 | 2020-08-21 | 安徽舒州农业科技有限责任公司 | Automatic steering method and system based on rice transplanter |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102957898A (en) * | 2011-08-15 | 2013-03-06 | 索尼公司 | Image processing device, image processing method and program |
CN103049446A (en) * | 2011-10-13 | 2013-04-17 | 中国移动通信集团公司 | Image retrieving method and device |
CN104079800A (en) * | 2014-06-24 | 2014-10-01 | 上海波汇通信科技有限公司 | Shaking preventing method for video image in video surveillance |
US9275302B1 (en) * | 2012-08-24 | 2016-03-01 | Amazon Technologies, Inc. | Object detection and identification |
CN106931962A (en) * | 2017-03-29 | 2017-07-07 | 武汉大学 | A kind of real-time binocular visual positioning method based on GPU SIFT |
CN107105150A (en) * | 2016-02-23 | 2017-08-29 | 中兴通讯股份有限公司 | A kind of method, photographic method and its corresponding intrument of selection photo to be output |
-
2018
- 2018-04-28 CN CN201810404934.XA patent/CN108734175A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102957898A (en) * | 2011-08-15 | 2013-03-06 | 索尼公司 | Image processing device, image processing method and program |
CN103049446A (en) * | 2011-10-13 | 2013-04-17 | 中国移动通信集团公司 | Image retrieving method and device |
US9275302B1 (en) * | 2012-08-24 | 2016-03-01 | Amazon Technologies, Inc. | Object detection and identification |
CN104079800A (en) * | 2014-06-24 | 2014-10-01 | 上海波汇通信科技有限公司 | Shaking preventing method for video image in video surveillance |
CN107105150A (en) * | 2016-02-23 | 2017-08-29 | 中兴通讯股份有限公司 | A kind of method, photographic method and its corresponding intrument of selection photo to be output |
CN106931962A (en) * | 2017-03-29 | 2017-07-07 | 武汉大学 | A kind of real-time binocular visual positioning method based on GPU SIFT |
Non-Patent Citations (2)
Title |
---|
刘丽等: ""图像纹理特征提取方法综述"", 《中国图象图形学报A》 * |
刘传才: "《图像理解与计算机视觉》", 30 September 2002, 厦门大学出版社 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110544234A (en) * | 2019-07-30 | 2019-12-06 | 北京达佳互联信息技术有限公司 | Image noise detection method, image noise detection device, electronic equipment and storage medium |
CN111563416A (en) * | 2020-04-08 | 2020-08-21 | 安徽舒州农业科技有限责任公司 | Automatic steering method and system based on rice transplanter |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109657716B (en) | Vehicle appearance damage identification method based on deep learning | |
CN109029381B (en) | Tunnel crack detection method and system and terminal equipment | |
US10430951B2 (en) | Method and device for straight line detection and image processing | |
US10410084B2 (en) | Devices, systems, and methods for anomaly detection | |
CN111179419B (en) | Three-dimensional key point prediction and deep learning model training method, device and equipment | |
CN110493595B (en) | Camera detection method and device, storage medium and electronic device | |
EP2154630A1 (en) | Image identification method and imaging apparatus | |
CN110910445B (en) | Object size detection method, device, detection equipment and storage medium | |
CN113724259B (en) | Well lid abnormity detection method and device and application thereof | |
CN104243815A (en) | Focusing method and electronic equipment | |
CN104951765A (en) | Remote sensing image target division method based on shape priori information and vision contrast ratio | |
CN106331848B (en) | The recognition methods of panoramic video and equipment play video method and equipment | |
CN110349216A (en) | Container method for detecting position and device | |
CN105825228B (en) | Image-recognizing method and device | |
CN108734175A (en) | A kind of extracting method of characteristics of image, device and electronic equipment | |
CN113283439B (en) | Intelligent counting method, device and system based on image recognition | |
JP2012123631A (en) | Attention area detection method, attention area detection device, and program | |
CN108447084B (en) | Stereo matching compensation method based on ORB characteristics | |
CN104268550A (en) | Feature extraction method and device | |
CN104573698A (en) | Image texture main direction recognition method oriented to imaging measurement | |
CN110135422B (en) | Dense target detection method and device | |
CN110738229B (en) | Fine-grained image classification method and device and electronic equipment | |
CN116363583A (en) | Human body identification method, device, equipment and medium for top view angle | |
CN110222704A (en) | A kind of Weakly supervised object detection method and device | |
CN113286086B (en) | Camera use control method and device, electronic equipment 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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181102 |