CN108389205A - A kind of rail monitoring foreign bodies method and device based on space base platform image - Google Patents
A kind of rail monitoring foreign bodies method and device based on space base platform image Download PDFInfo
- Publication number
- CN108389205A CN108389205A CN201810225219.XA CN201810225219A CN108389205A CN 108389205 A CN108389205 A CN 108389205A CN 201810225219 A CN201810225219 A CN 201810225219A CN 108389205 A CN108389205 A CN 108389205A
- Authority
- CN
- China
- Prior art keywords
- feature
- rail
- picture
- foreign matter
- pending picture
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/0008—Industrial image inspection checking presence/absence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
Abstract
The present invention provides a kind of rail monitoring foreign bodies method and device based on space base platform image, and wherein method includes：Pending picture is obtained, pending picture is the rail picture of low latitude unmanned plane shooting；Obtain the effective gradient information of pending picture；Coded treatment is carried out to effective gradient information according to the type of effective gradient information and obtains fisrt feature；Second feature is obtained according to the tone of pending picture, saturation degree, transparency hsv color model；According to the fisrt feature and the second feature, judge that rail whether there is foreign matter in the pending picture.A kind of rail monitoring foreign bodies method and device based on space base platform image provided by the invention, judges rail with the presence or absence of foreign matter in conjunction with the effective gradient information and colouring information of picture, improves the monitoring efficiency to rail foreign matter.
Description
Technical field
The present invention relates to aviation surveillance technology more particularly to a kind of rail monitoring foreign bodies methods based on space base platform image
And device.
Background technology
Rail traffic is the main path of the passenger traffic shipping of country, has critically important strategic importance.Wherein, railway transportation
Play the effect being even more important.Since China human mortality is numerous, the railway traffic of big flow needs more rigorous security inspection
To ensure travelling safety.
In the prior art, the extensive use with low latitude unmanned plane in inspection monitoring field more and more uses low latitude
The space base platform of unmanned plane is used for Along Railway inspection with the cost that uses manpower and material resources sparingly.Rail maintenance personnel passes through space base platform
The rail picture passed back of low latitude unmanned plane the situation of railway is monitored, when finding that the when of having foreign matter on rail is showed in time
It safeguards field.
Using the prior art, to ensure that safe railway operation, the low latitude unmanned plane inspection needs of space base platform are more accurate
Timely feedback Along Railway information, and the monitoring side that only low latitude unmanned plane picture is observed by rail maintenance personnel
Formula causes relatively low to the monitoring efficiency of rail foreign matter and needs to expend a large amount of manpower and materials.
Invention content
The present invention provides a kind of rail monitoring foreign bodies method and device based on space base platform image, improves different to rail
The monitoring efficiency of object.
The present invention provides a kind of rail monitoring foreign bodies method based on space base platform image, including：
Pending picture is obtained, the pending picture is the rail picture of low latitude unmanned plane shooting；
Obtain the effective gradient information of the pending picture；
Coded treatment is carried out to the effective gradient information according to the type of the effective gradient information and obtains fisrt feature；
Second feature is obtained according to the tone of the pending picture, saturation degree, transparency hsv color model；
According to the third feature obtained after the fisrt feature and second feature fusion, the pending picture is judged
Middle rail whether there is foreign matter.
In an embodiment of the present invention, a kind of rail monitoring foreign bodies method based on space base platform image as described above,
The effective gradient information for obtaining the pending picture, including：
Establish the integral image of the pending picture；
The integral image is established into scale space by tank filters；
Position the characteristic point of the scale space；
Son is described by construction feature point, obtains the effective gradient information of the pending picture.
In an embodiment of the present invention, a kind of rail monitoring foreign bodies method based on space base platform image as described above,
Coded treatment is carried out to the effective gradient information according to the type of the effective gradient information and obtains fisrt feature, including：
The effective gradient information is clustered by clustering algorithm, using cluster centre as basic code word；
Coded treatment is carried out to the effective gradient information using bag of words according to the basic code word, obtains fixed volume
The fisrt feature of code format.
In an embodiment of the present invention, a kind of rail monitoring foreign bodies method based on space base platform image as described above,
It is described that second feature is obtained according to the tone of the pending picture, saturation degree, the transparency hsv color aspect of model, including：
Obtain the hsv color model data of the pending picture；
The hsv color model data is counted according to color histogram to obtain the hsv color aspect of model, is made
For the second feature.
In an embodiment of the present invention, a kind of rail monitoring foreign bodies method based on space base platform image as described above,
It is described to judge that rail whether there is foreign matter in the pending picture according to the fisrt feature and the second feature, including：
Fusion Features are carried out to the fisrt feature and the second feature and obtain third feature；
By grader by the third feature judge the rail in the pending picture whether there is foreign matter, described point
Class device includes：There are the third feature of the rail picture of foreign matter and there is no the third of the rail picture of foreign matter spies
Sign.
In an embodiment of the present invention, a kind of rail monitoring foreign bodies method based on space base platform image as described above,
Before the pending picture of acquisition, further include：
Obtaining N, there are the institutes of the third feature of the rail picture of foreign matter and the M rail pictures that foreign matter is not present
State third feature, wherein N and M is positive integer；
The N are opened there are the third feature of the rail picture of foreign matter and M the rail picture that foreign matter is not present
The third feature is stored in the grader.
In an embodiment of the present invention, a kind of rail monitoring foreign bodies method based on space base platform image as described above,
The grader is support vector machines.
In an embodiment of the present invention, a kind of rail monitoring foreign bodies method based on space base platform image as described above,
It is described that rail is judged in the pending picture with the presence or absence of after foreign matter according to the fisrt feature and the second feature,
Further include：
If there are foreign matters for the rail for judging in the pending picture, the foreign matter is obtained by the third feature
Attribute.
In an embodiment of the present invention, a kind of rail monitoring foreign bodies method based on space base platform image as described above,
The attribute of the foreign matter includes below one or more：Type, size, shape and the color of foreign matter.
The present invention provides a kind of rail monitoring foreign bodies device based on space base platform image, including：Acquisition module, it is described to obtain
Modulus block is the rail picture of low latitude unmanned plane shooting for obtaining pending picture, the pending picture；
Characteristic extracting module, the processing module are used to obtain the effective gradient information of the pending picture；
The characteristic extracting module is additionally operable to, according to the type of the effective gradient information to the effective gradient information into
Row coded treatment obtains fisrt feature；
The characteristic extracting module is additionally operable to, according to the tone of the pending picture, saturation degree, transparency hsv color
Model obtains second feature；
Sort module, the sort module are used for according to the obtained after the fisrt feature and second feature fusion
Three features judge that rail whether there is foreign matter in the pending picture.
The present invention provides a kind of rail monitoring foreign bodies method and device based on space base platform image, and wherein method includes：
Pending picture is obtained, pending picture is the rail picture of low latitude unmanned plane shooting；Obtain the effective gradient of pending picture
Information；Coded treatment is carried out to effective gradient information according to the type of effective gradient information and obtains fisrt feature；According to pending
Tone, saturation degree, the transparency hsv color model of picture obtain second feature；It is waited for by fisrt feature and second feature judgement
It handles rail in picture and whether there is foreign matter.A kind of rail monitoring foreign bodies method based on space base platform image provided by the invention
And device, rail is judged with the presence or absence of foreign matter in conjunction with the effective gradient information and colouring information of picture, is improved to iron
The monitoring efficiency of rail foreign matter.
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 without having to pay creative labor, may be used also for those of ordinary skill in the art
With obtain other attached drawings according to these attached drawings.
Fig. 1 is a kind of flow signal of rail monitoring foreign bodies embodiment of the method one based on space base platform image of the present invention
Figure；
Fig. 2 is a kind of structural representation of the rail monitoring foreign bodies device embodiment one based on space base platform image of the present invention
Figure.
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.
Term " first ", " second ", " third " in description and claims of this specification and above-mentioned attached drawing, "
The (if present)s such as four " are for distinguishing similar object, without being used to describe specific sequence or precedence.It should manage
The data that solution uses in this way can be interchanged in the appropriate case, so that the embodiment of the present invention described herein for example can be to remove
Sequence other than those of illustrating or describe herein is implemented.In addition, term " comprising " and " having " and theirs is any
Deformation, it is intended that cover it is non-exclusive include, for example, containing the process of series of steps or unit, method, system, production
Product or equipment those of are not necessarily limited to clearly to list step or unit, but may include not listing clearly or for this
The intrinsic other steps of processes, method, product or equipment or unit a bit.
Technical scheme of the present invention is described in detail with specifically embodiment below.These specific implementations below
Example can be combined with each other, and same or analogous concept or process may be repeated no more in some embodiments.
Fig. 1 is a kind of flow signal of rail monitoring foreign bodies embodiment of the method one based on space base platform image of the present invention
Figure.As shown in Figure 1, a kind of foreign matter detecting method based on space base platform image provided in this embodiment includes：
S101：Pending picture is obtained, pending picture is the rail picture of low latitude unmanned plane shooting.
Specifically, the present embodiment executive agent can be the low latitude unmanned plane of space base platform, and the space base platform is for leading to
At least one unmanned plane is crossed to be monitored rail foreign matter；The present embodiment executive agent can also be can by internet and its
His mode obtains terminal device (Terminal), the user equipment (User Equipment) of the rail picture of space base platform shooting
And any one in server apparatus etc. or a variety of combinations.Wherein, terminal device can be desktop computer
(computer), laptop (notebook), tablet computer (PAD) etc..User equipment can be smart mobile phone
(smart phone), smartwatch (smart watch), intelligent glasses etc..It should be understood that the example above is only to illustrate, no
Specific restriction should be constituted.
In this step, low latitude unmanned plane is continuous to monitored rail when the rail segment for needing to monitor carries out inspection
It is shot, obtains continuous rail picture.It, can be direct by low latitude unmanned plane after low latitude unmanned plane takes rail picture
As pending picture；Alternatively, low latitude unmanned plane stores rail picture by internet upload server, then by end
The image processing devices such as end, user equipment obtain the rail picture stored in server as pending picture.
S102：Obtain the effective gradient information of pending picture.
Specifically, the effective gradient information of pending picture obtained in S101 is sought, wherein it is alternatively possible to passing through ladder
Degree feature extractor extracts the effective gradient information in pending picture.
Optionally, S102 extracts pending picture effective gradient and may comprise steps of：
S1021：Establish the integral image of pending picture.
Specifically, the integral image established in this step is to carry out the image that integral and calculating obtains, product to pending picture
The every bit of partial image be expressed as original image from origin to the pixel of the rectangular area of the point and, the foundation of integral image is why
It can accelerate calculating speed, be any rectangle in original image after carrying out integral image traversal to entire image because of us
The sum of the pixel in region can be completed by plus and minus calculation, and unrelated with the area of rectangle, and rectangle is bigger, the calculating of saving
Time is more.
S1022：Integral image is established into scale space by tank filters.
Specifically, scale space is established by tank filters in this step, height is replaced come approximate using tank filters
This kernel function so that convolution mask is made of simple rectangle.The introducing of integral image solves rectangular area and quickly calculates
The problem of, the approximate maximum of tank filters improves calculating speed.In order to ensure that images match has scale invariability, need
Image is layered, the scale space of image is established, then finds characteristic point on the image of different scale.The present embodiment
The foundation in mesoscale space is to maintain that pending image size is constant, by changing the size of tank filters come to pending figure
The integral image that picture is calculated is filtered, to form the scale space of image.
S1023：Position the characteristic point of scale space.
Specifically, by the scale space established in S102, using quick on each tomographic image of scale space
Hessian matrixes carry out the extreme point of detection image.For any point (x, y) in space, the scale in corresponding scale space is
σ, then shown in Hessian matrixes are defined as follows：
Wherein Lxx(xσ)、Lxy(xσ)、Lyy(x σ) be point on image respectively with Gauss second order derviation numberConvolution as a result, wherein g be Gaussian function.
It, can be to scale space into row interpolation, thus meanwhile in order in the settling position and scale-value for obtaining characteristic point
The positional value of characteristic point and the scale-value of characteristic point are obtained.
S1024：Son is described by construction feature point, obtains the effective gradient information of pending picture.
Specifically, in this step, the principal direction of characteristic point is sought first, can ensure the rotational invariance of algorithm in this way,
Then feature neighborhood of a point is rotated into principal direction, characteristic point is described.In order to make the matching of image that there is invariable rotary
Property, introduce the concept of principal direction.The calculating of principal direction is centered on characteristic point, and it is that (s is special to 6s to take radius around characteristic point
A scale-value where sign point) border circular areas, calculate Haar wavelet transform response of the pixel on x, the directions y in neighborhood.It is right
The response being calculated is assigned certain weight coefficient by distance, then carries out statistics with histogram to the response after weighting.
Statistics obtains a new vector since x-axis, to the Haar wavelet transform response addition calculation within the scope of 60 degree of border circular areas.Often
Vector is calculated in the same way every 5 degree, traverses entire border circular areas, can obtain 72 new vectors.We select longest
Principal direction of the direction vector as this feature point.
For the characteristic point detected, centered on characteristic point, the area of 20S*20S sizes within the scope of Selection Center vertex neighborhood
Then the principal direction in region is rotated to the principal direction of characteristic point by domain.In order to preferably utilize the spatial information of image, by 20S*
The region of 20S is divided into 4*4 totally 16 sub-regions, and the pixel value size per sub-regions in this way is 5S*5S.Finally by statistics picture
Characteristic point is described in the Haar wavelet transform response of vegetarian refreshments, obtains the effective gradient information of pending picture.Illustratively,
Assuming that a pictures can obtain P characteristic point by detection, each characteristic point describes one 128 available through the above way
The Feature Descriptor of dimension, then we can indicate effective ladder of pending picture with the one-dimensional vector that a P*128 is tieed up
Spend information.
The effective gradient information computational methods that this example provides are only for example, and unlisted place has with reference to the calculating of this field
Known to the known method for imitating gradient information.It should be noted that this step can also pass through the usual meter of other those skilled in the art
Calculation mode calculates the effective gradient information of pending picture, and the present embodiment is not especially limited this.
S103：Coded treatment is carried out to effective gradient information according to the type of effective gradient information and obtains fisrt feature.
Specifically, in this step, due to the dimension of the different calculated effective gradient information of picture, length not phases
Together, effective gradient information is if desired enable to need to calculate in different pictures as the feature of picture a processing and classification
To different dimensions, length effective gradient information handled so that dimension, the length of the effective gradient information of all pictures
It is identical, it can be compared together and subsequent classification.Optionally, the mode handled effective gradient information can be
First classify to the type of effective gradient information, then effective gradient information is carried out according to classification situation corresponding with type
Coded treatment, the effective gradient information after being encoded is as fisrt feature.Wherein, classification is to determine effective gradient information
Coding mode.
Optionally, a kind of possible realization methods of this step S103 are,
S1031：Effective gradient information is clustered by clustering algorithm, using cluster centre as basic code word；
S1032：Coded treatment is carried out to effective gradient information using bag of words according to basic code word, obtains regular coding
The fisrt feature of format.
Specifically, can be according to K-means clustering methods, it is 1000 to preset cluster classification sum, special by weighing
The distance between sign and feature cluster fisrt feature, obtain cluster centre as basic code word.Then use bag of words mould
Type encodes effective gradient information according to basic code word.The effective gradient information dimension that each pictures obtain is different, leads to
Bag of words cluster is crossed, each characteristic point is belonged into corresponding cluster centre, then counts straight based on cluster centre
Fang Tu obtains the feature vector of 1000 dimensions, as our fisrt feature to be extracted.Next, we are specific to introduce one
Processing example.
Assuming that the picture containing 1w low latitude unmanned plane shootings in sample, wherein the picture containing foreign matter has 7000, no
Picture containing foreign matter has 3000.The gradient information in every pictures can be extracted by Gradient Features extractor, uses several
The feature description subrepresentation of 128 dimensions.So that we can obtainA characteristic point, wherein P are in every pictures
The number of characteristic point.Cluster operation is carried out to features described above point using K-means clustering methods, it is 1000 to specify classification sum,
1000 classifications can be obtained, each classification includes several characteristic points.By calculating 128 Wei Te all in each classification
The mean value of sign description, obtains the cluster centre of each classification, and the one-dimensional vector that can be tieed up with one 128 indicates.So far, we
It can obtain 1000 cluster centres.Then, these cluster centres are as basic dictionary, using the thought of bag of words, to every
Pictures are encoded.Specifically, assuming to contain P characteristic point in picture, counts this P characteristic point and belong to above-mentioned 1000 classes
Not respective frequency, the one-dimensional vector that can finally obtain one 1000 dimension are total to subsequent processing and make as the code word after finally encoding
With.
The clustering algorithm and bag of words method that this example provides are only for example, and unlisted place is known with reference to this field
Known to method.It should be noted that this step can also be by the usual calculation of other those skilled in the art to effective gradient
Information carries out the calculating of unified dimensional, and the present embodiment is simultaneously not especially limited.
S104：Second feature is obtained according to the tone of pending picture, saturation degree, transparency hsv color model.
Specifically, in this step, the color characteristic of pending picture is handled, wherein optionally, this step tool
Body includes：
S1041：Obtain the hsv color model data of pending picture；
S1042：Hsv color model data is counted according to color histogram to obtain the hsv color aspect of model, as
Second feature.
Wherein, hsv color model is the color model towards visual perception, including 3 components H, S, V correspond respectively to coloured silk
Tone, saturation degree and the brightness of chrominance signal, hsv color model can be indicated with an inverted cone.Apart from long axis
Size indicates that saturation degree, long axis indicate brightness, and the angle around long axis indicates color.Due to appreciable colour-difference and Europe it is several in
It obtains apart from proportional.Therefore HSV is more suitable for the perception of people.Color histogram is to indicate the common method of color characteristic, but straight
The data volume of square figure is excessive, therefore generally requires quantization histogram to simplify color characteristic.For this purpose, first having to carry out color threshold
Quantization, quantification manner are as follows：
According to G=HQSQV+SQV+ V structuring one-dimensional feature vectors, wherein QSQVRespectively equal to 3, so G=9H+3S+V.This
Sample color characteristic is just quantized into the integer of a 0-71, and it is one-dimensional characteristic vector to facilitate statistics with histogram, is denoted as second feature.
The computational methods for the hsv color model that this example provides are only for example, calculating of the unlisted place with reference to this field
Known to the known method of hsv color model.It should be noted that this step can also pass through the usual meter of other those skilled in the art
Calculation mode calculates pending picture hsv color model, and the present embodiment is not especially limited this.
S105：According to fisrt feature and second feature, judge that rail whether there is foreign matter in pending picture.
Specifically, special by the first of the pending picture obtained in S103 in a kind of possible realization method of the present embodiment
The second feature of the pending picture obtained in S104 of seeking peace carries out Fusion Features and obtains third feature.Optionally, due to difference
The dimension of the fisrt feature of picture and the dimension of second feature are all identical, therefore can be by by the array of fisrt feature and second
The array of feature obtains the number of the third feature of the array comprising fisrt feature and second feature array simultaneously in a manner of merging
Group.It is merged alternatively, fisrt feature and second feature the modes such as can also be added to, subtract each other or be multiplied, as long as energy can be obtained
The third feature of fisrt feature and second feature, and the dimension of the third feature of all pictures and form all phases are characterized simultaneously
Together, it is not limited thereto.
Specifically, it can judge whether the rail in pending picture is deposited according to third feature by grader in this step
In foreign matter, wherein be stored in grader：There are the third feature of the rail picture of foreign matter and there is no the rail pictures of foreign matter
Third feature.
Such as：The third feature of N number of normal rail picture that foreign matter is not present, collection are stored in the set A of grader
Normal rail picture that N number of third feature in A is N low latitude unmanned planes shootings is closed according to the S101 in above-described embodiment extremely
S105 calculates N number of third feature of gained, and wherein N is positive integer.Rail can be chosen, N low latitude unmanned planes of foreign matter are not present
The rail picture of shooting, and being put into set A by the third feature that those pictures are found out, N number of third feature in set A by
In being free of foreign matter, then all within the scope of same.Meanwhile it being stored with M in the set B of grader there are the iron of foreign matter
It is a there are the third feature of the rail picture of foreign matter that M is stored in the third feature of rail picture, such as the set B of meter grader,
M third feature in set B be M low latitude unmanned planes shootings there are the rail pictures of foreign matter according in above-described embodiment
S101 to S105 calculates M third feature of gained, and wherein M is positive integer, and M and N can be identical or different.Rail can be chosen
There are the rail pictures of the M of foreign matter low latitude unmanned plane shootings, and are put into set B by the third feature that those pictures are found out
In, M third feature in set B is then in the range different from the third feature in set A due to there is foreign matter, and
If foreign matter type is identical, within the scope of the third feature in set B is in same.
In S105, after the third feature that pending picture is got by above-mentioned steps, grader is by pending figure
The third feature of piece is compared according to machine learning algorithm with third feature in set A and set B, and judges pending picture
Third feature be to belong to set A or set B, if the third feature of pending picture is special with the third in set A in the comparison
It levies more similar, then judges that foreign matter is not present in rail in pending picture；Correspondingly, if the third feature of pending picture than
More similar with the third feature in set B compared in, then there are foreign matters for the rail for judging in pending picture.
Optionally, in the above-described embodiments, grader used by step S105 is support vector machines (Support
Vector Machine, referred to as：SVM) linear classifier.
In addition, the present invention, which has been carried out example, also provides a kind of dynamic scene sorter of multiple features fusion, to execute such as
The upper embodiment of the method, technical characteristic and technique effect having the same repeat no more.
Optionally, in the above-described embodiments, after S105, further include：If judging, the rail in the pending picture is deposited
In foreign matter, then the attribute of the foreign matter is obtained by the third feature.
Specifically, when judging the rail in pending picture in S105, there are foreign matters, then can be further advanced by classification
The type of foreign matter is identified in device.Wherein, the attribute of foreign matter can be type, size, shape, color, the mobile speed of foreign matter
The parameter informations such as degree.Such as：There are the pictures that livestock (such as ox, sheep) enters in the rail picture of low latitude unmanned plane acquisition, by this
A little pictures are calculated third feature and are stored in the set C of grader through the above steps；Similar, there are the vehicles
The picture that (such as automobile, motorcycle) enters is stored in after those pictures are calculated third feature in the set D of grader.Then when
After grader receives the third feature of pending picture, according to the third feature in set C and set D to pending picture
Third feature is classified, and show that the type of rail foreign matter in pending picture is livestock or the vehicles.Optionally, this implementation
Foreign matter attribute is judged to classify to the judgement with the presence or absence of foreign matter using different two from above-described embodiment in example
Device or the same grader are realized, if being realized by same grader, above-mentioned set A to set D is deposited in the grader
Storage.Above-mentioned distance is the example of type in foreign matter attribute, and other parameters realization method is identical, repeats no more, it should be noted that different
The picture that the movement speed of object can be shot according to different time, which combines, to be obtained.
To sum up, a kind of rail foreign matter detecting method based on space base platform image provided in this embodiment is waited for by obtaining
Picture is handled, pending picture is the rail picture of low latitude unmanned plane shooting；Obtain the effective gradient information of pending picture；Root
Coded treatment is carried out to effective gradient information according to the type of effective gradient information and obtains fisrt feature；According to the color of pending picture
Tune, saturation degree, transparency hsv color model obtain second feature；Judge pending picture according to fisrt feature and second feature
Middle rail whether there is foreign matter.To by combining the effective gradient information of picture and colouring information to whether there is foreign matter to rail
Judged, improves the monitoring efficiency to rail foreign matter.Meanwhile the one kind provided in an embodiment of the present invention is based on space base
In the rail monitoring foreign bodies embodiment of the method for platform image, the gradient that picture to be sorted is extracted using Gradient Features extractor is special
Sign, then obtains high-level Gradient Features using K-means clustering methods and bag of words according to Gradient Features；Meanwhile it adopting
The color characteristic that picture to be sorted is extracted with color feature extracted device will carry out the monitoring point of rail foreign matter after two kinds of Fusion Features
Class.When picture to be sorted is identified as the picture comprising foreign matter, early warning is carried out.The gradient information in picture is not only allowed for,
Also contemplate the colouring information in picture so that classification is more accurate.
Fig. 2 is a kind of structural representation of the rail monitoring foreign bodies device embodiment one based on space base platform image of the present invention
Figure.As shown in Fig. 2, a kind of rail monitoring foreign bodies device based on unmanned plane image provided in this embodiment includes：Acquisition module
201, characteristic extracting module 202 and sort module 203.Wherein, acquisition module 201 is for obtaining pending picture, pending figure
Piece is the rail picture of low latitude unmanned plane shooting；Characteristic extracting module 202 is used to obtain the effective gradient information of pending picture；
Characteristic extracting module 202 is additionally operable to obtain first to effective gradient information progress coded treatment according to the type of effective gradient information
Feature；Characteristic extracting module 202 is additionally operable to obtain according to the tone of pending picture, saturation degree, transparency hsv color model
Two features；Sort module 203 is used to, according to the fisrt feature and the second feature, judge rail in the pending picture
With the presence or absence of foreign matter.
Device provided in this embodiment is for executing the method provided in embodiment illustrated in fig. 1, realization method and principle
It is identical, it repeats no more.
Optionally, in the above-described embodiments, characteristic extracting module is specifically used for establishing the integral image of pending picture；It will
Integral image establishes scale space by tank filters；Position the characteristic point of scale space；Son is described by construction feature point,
Obtain the effective gradient information of pending picture.
Optionally, in the above-described embodiments, characteristic extracting module is specifically used for through clustering algorithm to effective gradient information
It is clustered, using cluster centre as basic code word；Effective gradient information is compiled using bag of words according to basic code word
Code processing, obtains the fisrt feature of regular coding format.
Optionally, in the above-described embodiments, characteristic extracting module is specifically used for obtaining the hsv color model of pending picture
Data；Hsv color model data is counted according to color histogram to obtain the hsv color aspect of model, as second feature.
Optionally, in the above-described embodiments, sort module is specifically used for melting fisrt feature and second feature progress feature
Conjunction obtains third feature；Judge that the rail in pending picture whether there is foreign matter, grader by third feature by grader
Including：There are the third feature of the rail picture of foreign matter and there is no the third feature of the rail picture of foreign matter.
Optionally, in the above-described embodiments, acquisition module is additionally operable to obtain N that there are the third of the rail picture of foreign matter spies
The M that seeks peace there is no the third feature of the rail picture of foreign matter, wherein N and M is positive integer；By N, there are the rail figures of foreign matter
There is no the third feature of the rail picture of foreign matter to be stored in grader for the third feature and M of piece.
Optionally, in the above-described embodiments, grader is support vector machines.
Optionally, in the above-described embodiments, if sort module is additionally operable to judge the rail in pending picture there are foreign matter,
The attribute of foreign matter is then obtained by third feature.
Optionally, in the above-described embodiments, if sort module is additionally operable to judge the rail in pending picture there are foreign matter,
Information the location of is then obtained when low latitude unmanned plane shoots pending picture, and is sent to alarm server and carries location information
Alarm signal.
Device provided in this embodiment is for executing the method provided in above-described embodiment, realization method and principle phase
Together, it repeats no more.
One embodiment of the invention also provides a kind of computer storage media, is stored thereon with computer program, computer journey
A kind of rail monitoring foreign bodies side based on space base platform image in any of the above-described embodiment is executed when sequence is executed by processor
Method.Wherein, the storage medium in the present embodiment is computer-readable storage medium.
One embodiment of the invention also provides a kind of electronic equipment, including：Processor；And memory, it is handled for storage
The executable instruction of device；Wherein, processor is configured to execute in any of the above-described embodiment via executable instruction is executed
A kind of rail monitoring foreign bodies method based on space base platform image.
One of ordinary skill in the art will appreciate that：Realize that all or part of step of above-mentioned each method embodiment can lead to
The relevant hardware of program instruction is crossed to complete.Program above-mentioned can be stored in a computer read/write memory medium.The journey
When being executed, execution includes the steps that above-mentioned each method embodiment to sequence；And storage medium above-mentioned includes：ROM, RAM, magnetic disc or
The various media that can store program code such as person's CD.
Finally it should be noted that：The above embodiments are only used to illustrate the technical solution of the present invention., rather than its limitations；To the greatest extent
Present invention has been described in detail with reference to the aforementioned embodiments for pipe, it will be understood by those of ordinary skill in the art that：Its according to
So can with technical scheme described in the above embodiments is modified, either to which part or all technical features into
Row equivalent replacement；And these modifications or replacements, various embodiments of the present invention technology that it does not separate the essence of the corresponding technical solution
The range of scheme.
Claims (10)
1. a kind of rail monitoring foreign bodies method based on space base platform image, which is characterized in that including：
Pending picture is obtained, the pending picture is the rail picture of low latitude unmanned plane shooting；
Obtain the effective gradient information of the pending picture；
Coded treatment is carried out to the effective gradient information according to the type of the effective gradient information and obtains fisrt feature；
Second feature is obtained according to the tone of the pending picture, saturation degree, transparency hsv color model；
According to the fisrt feature and the second feature, judge that rail whether there is foreign matter in the pending picture.
2. according to the method described in claim 1, it is characterized in that, the effective gradient letter for obtaining the pending picture
Breath, including：
Establish the integral image of the pending picture；
The integral image is established into scale space by tank filters；
Position the characteristic point of the scale space；
Son is described by construction feature point, obtains the effective gradient information of the pending picture.
3. according to the method described in claim 1, it is characterized in that, the type according to the effective gradient information is to described
Effective gradient information carries out coded treatment and obtains fisrt feature, including：
The effective gradient information is clustered by clustering algorithm, using cluster centre as basic code word；
Coded treatment is carried out to the effective gradient information using bag of words according to the basic code word, obtains regular coding lattice
The fisrt feature of formula.
4. according to the method described in claim 1, it is characterized in that, it is described according to the tone of the pending picture, saturation degree,
The transparency hsv color aspect of model obtains second feature, including：
Obtain the hsv color model data of the pending picture；
The hsv color model data is counted according to color histogram to obtain the hsv color aspect of model, as institute
State second feature.
5. according to the method described in claim 1, it is characterized in that, described according to the fisrt feature and the second feature,
Judge that rail whether there is foreign matter in the pending picture, including：
Fusion Features are carried out to the fisrt feature and the second feature and obtain third feature；
Judge that the rail in the pending picture whether there is foreign matter, the grader by the third feature by grader
Including：There are the third feature of the rail picture of foreign matter and there is no the third feature of the rail picture of foreign matter.
6. according to the method described in claim 5, it is characterized in that, before the pending picture of acquisition, further include：
Obtain described the of N there are the third feature of the rail picture of foreign matter and M there is no the rail picture of foreign matter
Three features, wherein N and M is positive integer；
By the N, there are described in the third feature of the rail picture of foreign matter and the M rail pictures that foreign matter is not present
Third feature is stored in the grader.
7. according to the method described in claim 6, it is characterized in that, the grader is support vector machines.
8. according to claim 1-7 any one of them methods, which is characterized in that described according to the fisrt feature and described
Two features judge that rail further includes with the presence or absence of after foreign matter in the pending picture：
If there are foreign matters for the rail for judging in the pending picture, the category of the foreign matter is obtained by the third feature
Property.
9. according to the method described in claim 8, it is characterized in that, the attribute of the foreign matter includes below one or more：
Type, size, shape and the color of foreign matter.
10. a kind of rail monitoring foreign bodies device based on space base platform image, which is characterized in that including：
Acquisition module, the acquisition module are the shooting of low latitude unmanned plane for obtaining pending picture, the pending picture
Rail picture；
Characteristic extracting module, the processing module are used to obtain the effective gradient information of the pending picture；
The characteristic extracting module is additionally operable to, and is compiled to the effective gradient information according to the type of the effective gradient information
Code processing obtains fisrt feature；
The characteristic extracting module is additionally operable to, according to the tone of the pending picture, saturation degree, transparency hsv color model
Obtain second feature；
Sort module, the sort module are used to, according to the fisrt feature and the second feature, judge the pending figure
Rail whether there is foreign matter in piece.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810225219.XA CN108389205B (en) | 2018-03-19 | 2018-03-19 | Rail foreign matter monitoring method and device based on air-based platform image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810225219.XA CN108389205B (en) | 2018-03-19 | 2018-03-19 | Rail foreign matter monitoring method and device based on air-based platform image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108389205A true CN108389205A (en) | 2018-08-10 |
| CN108389205B CN108389205B (en) | 2022-10-25 |
Family
ID=63067549
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810225219.XA Active CN108389205B (en) | 2018-03-19 | 2018-03-19 | Rail foreign matter monitoring method and device based on air-based platform image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108389205B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112907534A (en) * | 2021-02-18 | 2021-06-04 | 哈尔滨市科佳通用机电股份有限公司 | Fault detection method and device based on door closing part position image |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110255743A1 (en) * | 2010-04-13 | 2011-10-20 | International Business Machines Corporation | Object recognition using haar features and histograms of oriented gradients |
| CN102982350A (en) * | 2012-11-13 | 2013-03-20 | 上海交通大学 | Station caption detection method based on color and gradient histograms |
| CN104680189A (en) * | 2015-03-15 | 2015-06-03 | 西安电子科技大学 | Pornographic image detection method based on improved bag-of-words model |
| CN105448095A (en) * | 2014-06-03 | 2016-03-30 | 杭州海康威视数字技术股份有限公司 | Yellow label car detection method and device |
| CN105989594A (en) * | 2015-02-12 | 2016-10-05 | 阿里巴巴集团控股有限公司 | Image region detection method and device |
-
2018
- 2018-03-19 CN CN201810225219.XA patent/CN108389205B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110255743A1 (en) * | 2010-04-13 | 2011-10-20 | International Business Machines Corporation | Object recognition using haar features and histograms of oriented gradients |
| CN102982350A (en) * | 2012-11-13 | 2013-03-20 | 上海交通大学 | Station caption detection method based on color and gradient histograms |
| CN105448095A (en) * | 2014-06-03 | 2016-03-30 | 杭州海康威视数字技术股份有限公司 | Yellow label car detection method and device |
| CN105989594A (en) * | 2015-02-12 | 2016-10-05 | 阿里巴巴集团控股有限公司 | Image region detection method and device |
| CN104680189A (en) * | 2015-03-15 | 2015-06-03 | 西安电子科技大学 | Pornographic image detection method based on improved bag-of-words model |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112907534A (en) * | 2021-02-18 | 2021-06-04 | 哈尔滨市科佳通用机电股份有限公司 | Fault detection method and device based on door closing part position image |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108389205B (en) | 2022-10-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Bautista et al. | Convolutional neural network for vehicle detection in low resolution traffic videos | |
| Baran et al. | A smart camera for the surveillance of vehicles in intelligent transportation systems | |
| Tang et al. | Compressed-domain ship detection on spaceborne optical image using deep neural network and extreme learning machine | |
| Meethongjan et al. | An intelligent fused approach for face recognition | |
| Balali et al. | Evaluation of multiclass traffic sign detection and classification methods for US roadway asset inventory management | |
| CN109948497A (en) | A kind of object detecting method, device and electronic equipment | |
| CN109447169A (en) | The training method of image processing method and its model, device and electronic system | |
| CN109241985A (en) | A kind of image-recognizing method and device | |
| KR101716646B1 (en) | Method for detecting and recogniting object using local binary patterns and apparatus thereof | |
| CN107273836A (en) | A kind of pedestrian detection recognition methods, device, model and medium | |
| CN107688784A (en) | A kind of character identifying method and storage medium based on further feature and shallow-layer Fusion Features | |
| CN108875542B (en) | Face recognition method, device and system and computer storage medium | |
| CN106778705A (en) | A kind of pedestrian's individuality dividing method and device | |
| US8958641B2 (en) | Moving object detection device | |
| JP2003523587A (en) | Visual attention system | |
| CN104134364B (en) | Real-time traffic sign identification method and system with self-learning capacity | |
| CN109948616A (en) | Image detecting method, device, electronic equipment and computer readable storage medium | |
| CN109871789A (en) | Vehicle checking method under a kind of complex environment based on lightweight neural network | |
| CN106682681A (en) | Recognition algorithm automatic improvement method based on relevance feedback | |
| CN108564111A (en) | A kind of image classification method based on neighborhood rough set feature selecting | |
| CN110084175A (en) | A kind of object detection method, object detecting device and electronic equipment | |
| CN108205649A (en) | Driver drives to take the state identification method and device of phone | |
| CN106203461A (en) | A kind of image processing method and device | |
| WO2021050769A1 (en) | Spatio-temporal interactions for video understanding | |
| CN115238723A (en) | Local vertex detection method and device |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |