CN105654068B - A kind of target detection background estimating method based on fractal theory - Google Patents
A kind of target detection background estimating method based on fractal theory Download PDFInfo
- Publication number
- CN105654068B CN105654068B CN201610073478.6A CN201610073478A CN105654068B CN 105654068 B CN105654068 B CN 105654068B CN 201610073478 A CN201610073478 A CN 201610073478A CN 105654068 B CN105654068 B CN 105654068B
- Authority
- CN
- China
- Prior art keywords
- pixel
- background
- point
- value
- fractal
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
Abstract
The target detection background estimating method based on fractal theory that the present invention relates to a kind of, comprising the following steps: obtain the image sequence comprising target to be detected；Obtain brightness value array of each coordinate points pixel in whole image sequence；Calculate the fractal characteristics value of each coordinate points pixel；Classification obtains background, prospect and interference pixel；Respective background estimation is carried out point by point for type of pixel to calculate；Initial value is replaced using obtained background estimating value, obtains final background image data.The present invention can adapt at random and there may be the use environment of uncertain noises factor, strong real-times, high reliablity.
Description
Technical field
The present invention relates to Real-time Monitor Technique fields, estimate more particularly to a kind of target detection background based on fractal theory
Meter method.
Background technique
Currently, machine vision, real time monitoring and intelligent safety and defence system identify key technology carrying out moving object detection
In be directed to a key problem, i.e., to the accurate estimation of background scene image.System whether can at the scene in the case of
To accurately and reliably background image, the accuracy and stability of detection of the System Back-end to moving target are directly affected, especially
Be it is random and there may be in the case of the use environment of uncertain noises factor carry out target detection when, include to what is currently obtained
Real-time, accuracy and the robustness of the background image estimation of the image sequence of target to be detected are very important.Therefore, exist
It is random and when there may be carrying out target detection under the use environment of uncertain noises factor, can obtain in real time accuracy it is high,
The algorithm of the background image of strong robustness becomes machine vision, key technology in real time monitoring and intelligent security guard area research
One of.
When carrying out the background estimating of target detection, existing will adapt to random and make there may be uncertain noises factor
With demands such as environment, strong real-times, and accuracy wants high, and it is strong to need to meet availability in actual use, robustness
Good wait requires.
In order to realize target detection background estimating, can be used mixed Gauss model background estimating, mean value background estimating or
Person's optical flow method background estimating etc..Mixed Gauss model background estimating is that the pixel value in image is regarded as some Gaussian Profiles
Comprehensive function, the i.e. mixture of prospect Gaussian Profile and background Gaussian Profile.Certain point pixel value of image meets prospect Gauss minute
When cloth, it is considered as the point and belongs to foreground target；When meeting background Gaussian Profile, it is considered as the point and belongs to background, and carry out background
It updates, this is computationally intensive, and it is impossible to meet real-times, and more sensitive to noise；And mean value background estimating is that frame accumulation is made even
The process of mean value, needs a large amount of original data source that can just obtain accurate background data, and real-time and noise immunity are poor；Light
Stream method background estimating is computationally intensive, high to frame per second requirement, and to light noise-sensitive.
Summary of the invention
The target detection background estimating method based on fractal theory that technical problem to be solved by the invention is to provide a kind of,
It can adapt to random and there may be the use environment of uncertain noises factor, strong real-times, high reliablity, for subsequent target
Detection provides effective background data foundation, improves the accuracy in detection of target.
The technical solution adopted by the present invention to solve the technical problems is: providing a kind of target detection based on fractal theory
Background estimating method, comprising the following steps:
(1) image sequence comprising target to be detected is obtained；
(2) brightness value array of each coordinate points pixel in whole image sequence is obtained；
(3) fractal characteristics value of each coordinate points pixel is calculated；
(4) classification obtains background, prospect and interference pixel；
(5) respective background estimation is carried out point by point for type of pixel to calculate；
(6) initial value is replaced using obtained background estimating value, obtains final background image data.
It further include being filtered pretreatment removal to every frame image of the image sequence of acquisition frame by frame to do in the step (1)
The step of disturbing.
The step (2) specifically: each of traversal image pixel includes the figure of target to be detected using acquisition
As sequence data, extracts every frame image in image sequence and correspond to the pixel brightness value of same coordinate position point to get the coordinate is arrived
Brightness value array of the point pixel in whole image sequence.
The step (3) specifically: fractal characteristics value calculating is carried out to the brightness array of each coordinate pixel, is divided
Shape characteristic value has obtained fractal characteristics value array after traversing all pixels point.
The step (4) specifically: according to the regularity of brightness Distribution value, by all pixels point according to its fractal characteristics value
Classification is divided into foreground pixel point, background pixel point and interference pixel.
The step (5) specifically: point-by-point background estimating is carried out to all pixels point, is selected according to the type of the pixel
Different background estimating methods carries out background estimating calculating, obtains the estimated value of each pixel；Wherein, it for background dot, uses
The corresponding brightness value array of the point carries out background pixel value estimation；For foreground point, using the corresponding brightness value array of the point and
Fractal characteristics value carries out background pixel value estimation；For noise spot, background pixel is carried out using the corresponding brightness value array of the point
Value estimation.
Further include the steps that carrying out local processing to distortion point in the step (6).
Beneficial effect
Due to the adoption of the above technical solution, compared with prior art, the present invention having the following advantages that and actively imitating
Fruit: the present invention is realized at random and there may be under the use environment of uncertain noises factor, and estimation obtains current background number
According to strong real-time, high reliablity provide effective background data foundation for subsequent target detection, improve the inspection of target
Survey accuracy.The present invention does not need to carry out a large amount of data acquisition, strong interference immunity, speed it is fast, it can be achieved that system real-time,
It is particularly suitable in machine vision, real time monitoring and the more demanding environment of this real-time validity of intelligent safety and defence system.
Detailed description of the invention
Fig. 1 is flow chart of the invention；
Fig. 2 is the flow chart that the image sequence comprising target to be detected is obtained in the present invention；
Fig. 3 is the schematic diagram of the fractal characteristics value calculated in the present invention；
Fig. 4 is that classification obtains background, prospect and the flow chart for interfering pixel in the present invention；
Fig. 5 is the foreground and background schematic diagram obtained in the specific embodiment of the invention.
Specific embodiment
Present invention will be further explained below with reference to specific examples.It should be understood that these embodiments are merely to illustrate the present invention
Rather than it limits the scope of the invention.In addition, it should also be understood that, after reading the content taught by the present invention, those skilled in the art
Member can make various changes or modifications the present invention, and such equivalent forms equally fall within the application the appended claims and limited
Range.
Fig. 1 is a kind of flow chart of the target detection background estimating algorithm based on fractal theory of the present invention, as shown in Figure 1,
The method, including the following steps:
Step 100: obtaining the image sequence comprising target to be detected；
Step 200: obtaining brightness value array of each coordinate points pixel in whole image sequence；
Step 300: calculating the fractal characteristics value of each coordinate points pixel；
Step 400: classification obtains background, prospect and interference pixel；
Step 500: carrying out respective background estimation point by point for type of pixel and calculate；
Step 600: obtaining background image data.
Fig. 2 is the structural schematic diagram that the image sequence comprising target to be detected is obtained in the present invention, illustrates that acquisition includes
The image sequence data and Filter Preprocessing Method of target to be detected.
Step 101: the acquisition for obtaining most original includes the image sequence N { 1,2 ..., n } of target to be detected, n >=9.
Step 102: pretreatment, removal interference noise, especially electronics are filtered to the original image sequence N of acquisition
Interference noise point.Filter Preprocessing Method generally takes M*M median filtering, M=3,5 or other numerical value.
Step 103: obtaining image sequence data after obtaining filter preprocessing.
In step 200, each of image pixel is traversed, includes the image sequence of target to be detected using acquisition
Data extract every frame image in image sequence and correspond to the pixel brightness value of same coordinate position point to get the coordinate points pixel is arrived
Brightness value array V [1,2 ..., n] in whole image sequence, wherein n is the number of image frames that image sequence includes.N is general
Need to only meet can normal use more than or equal to 9 frames.
In step 300, its fractal characteristics value calculating is carried out to the brightness array V of each coordinate pixel, it is right obtains this
The fractal characteristics value for the data sequence for answering the pixel intensity of coordinate points to form has obtained fractal characteristic after traversing all pixels point
It is worth array F [x] [y], wherein x, y are the transverse and longitudinal resolution ratio of image.Wherein fractal characteristics value mainly selects pixel brightness array
Short Time Fractal Numbers, Short Time Fractal Numbers be capable of it is significantly more efficient reflection signal variation characteristic calculation method it is as follows: assuming that
Fractal characteristics value array F [i] [j] is the fractal characteristics value of any coordinate pixel, then F [i] [j]=1+log2 D(s)/D2
(2s), wherein
D (s) and D2It is s that (2s), which is respectively indicated with width, and the square net of 2s covers data sequence grid number.Obtain D (s) and D2
Fractal dimension will be calculated in log logarithmic coordinates after (2s), as fractal characteristics value.
Fig. 3 is a kind of fractal characteristics value signal that the target detection background estimating algorithm based on fractal theory calculates of the present invention
Figure, wherein comprising a certain frame in original sequence, and obtain the fractal characteristics value schematic diagram of all the points.
Fig. 4 is that classification obtains background, prospect in a kind of target detection background estimating algorithm based on fractal theory of the present invention
And the structural schematic diagram of interference pixel, illustrate that classification obtains prospect, background and the method for interfering pixel.
Step 401: to arbitrary coordinate pixel, the fractal characteristics value of calculating before having obtained.
Step 402: using obtained fractal characteristics value, judging that the pixel extremely meets fractal characteristic, accorded with when being judged as
When closing fractal characteristic, illustrate that the point is background dot；Otherwise the point is non-background dot.
After obtaining the fractal characteristics value of all the points, pixel class discrimination is carried out.In the acquisition process of whole image sequence
In, its rule and characteristic is distributed in brightness value, these determine the distribution of its fractal characteristics value, if the coordinate pixel
Point is not comprising target point and noise spot, then it meets fractal characteristic, has self-similarity, therefore fractal dimension meets a point shape spy
Range is levied, is background dot；And the coordinate points do not include target point and noise spot, then do not meet a point shape self-similarity characteristics, therefore
It is background dot or noise spot that fractal dimension, which does not meet fractal characteristic range,.And background dot and the fractal characteristics value of noise spot pixel divide
Cloth classifies all pixels point according to its fractal characteristics value in different range, according to this feature, point of ordinary circumstance background dot
Shape characteristic value is distributed near [1.8,2.2], and foreground point, near [1,1.2], noise spot is 2.2 or more.Therefore it sets suitable
The threshold value of conjunction can get respectively background, prospect and interference pixel.
Step 403: further classification judgement being carried out to non-background dot, when the fractal characteristics value changed the time meets foreground point spy
When value indicative range, judge that it, for foreground point, is otherwise interference pixel.
In step 500, the background value of the point is estimated, for background dot, uses the corresponding brightness value number of the point
Group V [1,2 ..., n] carries out following background pixel value estimation: D [i] [j]=MEAN (V), whereinFor foreground point, using the corresponding brightness value array of the point and fractal characteristics value carry out with
Lower background value estimation: D [i] [j]=MEAN (V) ± F [i] [j] * AVE [i] [j], whereinFor noise spot, using the corresponding brightness value array V of the point [1,
2 ..., n] carry out following background brightness value estimation: D [i] [j]=MID (V), wherein MID (V)=mid (V [1,2 ..., N]).
In step 600, after carrying out point-by-point background estimating calculating to all pixels point, the estimation of each pixel is obtained
Value replaces initial value using this estimated value, and carries out local processing to distortion point, obtains final background image data.
Fig. 5 is that a kind of foreground and background of the acquisition of the target detection background estimating algorithm based on fractal theory of the present invention is shown
It is intended to, wherein the background image data comprising estimating to obtain and the foreground image data schematic diagram obtained by difference.
A kind of target detection background estimating algorithm based on fractal theory of the invention, realize it is random and there may be
Under the use environment of uncertain noises factor, estimation obtains current background data, and strong real-time, high reliablity are subsequent mesh
Mark detection provides effective background data foundation, improves the accuracy in detection of target.It does not need to carry out a large amount of data to adopt
Collection, strong interference immunity, speed are fastly, it can be achieved that the real-time of system, is particularly suitable in machine vision, real time monitoring and intelligent security guard
The more demanding environment of this real-time validity of system.
Claims (7)
1. a kind of target detection background estimating method based on fractal theory, which comprises the following steps:
(1) image sequence comprising target to be detected is obtained；
(2) brightness value array of each coordinate points pixel in whole image sequence is obtained；
(3) fractal characteristics value of each coordinate points pixel is calculated；
(4) classification obtains background, prospect and interference pixel；
(5) respective background estimation is carried out point by point for type of pixel to calculate；Specifically: point-by-point background is carried out to all pixels point and is estimated
Meter selects different background estimating methods to carry out background estimating calculating, obtains estimating for each pixel according to the type of the pixel
Evaluation；Wherein, for background dot, background pixel value estimation is carried out using the corresponding brightness value array of the point；For foreground point, make
Background pixel value estimation is carried out with the corresponding brightness value array of the point and fractal characteristics value；For noise spot, corresponded to using the point
Brightness value array carry out background pixel value estimation；
(6) initial value is replaced using obtained background estimating value, obtains final background image data.
2. the target detection background estimating method according to claim 1 based on fractal theory, which is characterized in that the step
Suddenly further include the steps that being filtered pretreatment removal noise spot to every frame image of the image sequence of acquisition frame by frame in (1).
3. the target detection background estimating method according to claim 1 based on fractal theory, which is characterized in that the step
Suddenly (2) specifically: each of traversal image pixel is mentioned using the image sequence data comprising target to be detected is obtained
The pixel brightness value that every frame image corresponds to same coordinate position point in image sequence is taken entirely to scheme to get to the coordinate points pixel
As the brightness value array in sequence.
4. the target detection background estimating method according to claim 1 based on fractal theory, which is characterized in that the step
Suddenly (3) specifically: fractal characteristics value calculating is carried out to the brightness array of each coordinate pixel, fractal characteristics value is obtained, traverses
After all pixels point, fractal characteristics value array has been obtained.
5. the target detection background estimating method according to claim 4 based on fractal theory, which is characterized in that described point
The Short Time Fractal Numbers of shape characteristic value selection pixel brightness array.
6. the target detection background estimating method according to claim 1 based on fractal theory, which is characterized in that the step
Suddenly (4) specifically: according to the regularity of brightness Distribution value, all pixels point is classified according to its fractal characteristics value, is divided into prospect
Pixel, background pixel point and interference pixel.
7. the target detection background estimating method according to claim 1 based on fractal theory, which is characterized in that the step
Suddenly further include the steps that carrying out local processing to distortion point in (6).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610073478.6A CN105654068B (en) | 2016-02-02 | 2016-02-02 | A kind of target detection background estimating method based on fractal theory |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610073478.6A CN105654068B (en) | 2016-02-02 | 2016-02-02 | A kind of target detection background estimating method based on fractal theory |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105654068A CN105654068A (en) | 2016-06-08 |
| CN105654068B true CN105654068B (en) | 2019-03-26 |
Family
ID=56489089
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610073478.6A Active CN105654068B (en) | 2016-02-02 | 2016-02-02 | A kind of target detection background estimating method based on fractal theory |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105654068B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109753960B (en) * | 2018-12-24 | 2022-03-11 | 西安理工大学 | Fractal theory-based isolated forest underwater artificial target detection method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101604383A (en) * | 2009-07-24 | 2009-12-16 | 哈尔滨工业大学 | A kind of method for detecting targets at sea based on infrared image |
| CN101630369A (en) * | 2009-07-30 | 2010-01-20 | 南京航空航天大学 | Pedestrian detection method based on wavelet fractal characteristic |
| CN101930610A (en) * | 2009-06-26 | 2010-12-29 | 思创影像科技股份有限公司 | Method for detecting moving object by using adaptable background model |
-
2016
- 2016-02-02 CN CN201610073478.6A patent/CN105654068B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101930610A (en) * | 2009-06-26 | 2010-12-29 | 思创影像科技股份有限公司 | Method for detecting moving object by using adaptable background model |
| CN101604383A (en) * | 2009-07-24 | 2009-12-16 | 哈尔滨工业大学 | A kind of method for detecting targets at sea based on infrared image |
| CN101630369A (en) * | 2009-07-30 | 2010-01-20 | 南京航空航天大学 | Pedestrian detection method based on wavelet fractal characteristic |
Non-Patent Citations (1)
| Title |
|---|
| 基于分形分析和光流估计的运动目标检测技术研究;马龙;《中国博士学位论文全文数据库》;20140315(第03期);第2章 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105654068A (en) | 2016-06-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103077384B (en) | A kind of method and system of vehicle-logo location identification | |
| CN106856002B (en) | Unmanned aerial vehicle shooting image quality evaluation method | |
| CN106530281B (en) | Unmanned plane image fuzzy Judgment method and system based on edge feature | |
| CN104361582B (en) | Method of detecting flood disaster changes through object-level high-resolution SAR (synthetic aperture radar) images | |
| CN109816644A (en) | A kind of bearing defect automatic checkout system based on multi-angle light source image | |
| Zhang et al. | Infrared small target detection based on spatial-temporal enhancement using quaternion discrete cosine transform | |
| CN107392885A (en) | A kind of method for detecting infrared puniness target of view-based access control model contrast mechanism | |
| CN109086724B (en) | Accelerated human face detection method and storage medium | |
| CN110874592B (en) | Forest fire smoke image detection method based on total bounded variation | |
| CN106127253B (en) | A kind of method for detecting infrared puniness target using sample characteristics learning classification | |
| CN108985169A (en) | Across the door operation detection method in shop based on deep learning target detection and dynamic background modeling | |
| CN104834915B (en) | A kind of small infrared target detection method under complicated skies background | |
| CN104766079B (en) | A kind of remote method for detecting infrared puniness target | |
| CN106770087A (en) | Greasy dirt remote sensing module, system and method | |
| CN109087363B (en) | HSV color space-based sewage discharge detection method | |
| CN106686377A (en) | Algorithm for determining video key area based on deep neural network | |
| CN109993797A (en) | Door and window method for detecting position and device | |
| CN108765406A (en) | A kind of snow mountain detection method based on infrared remote sensing image | |
| Lian et al. | A novel method on moving-objects detection based on background subtraction and three frames differencing | |
| CN110321855A (en) | A kind of greasy weather detection prior-warning device | |
| CN108038856B (en) | Infrared small target detection method based on improved multi-scale fractal enhancement | |
| CN103456123B (en) | A kind of video smoke detection method based on flowing with diffusion characteristic | |
| CN106408563B (en) | A kind of snow noise detection method based on the coefficient of variation | |
| CN105654068B (en) | A kind of target detection background estimating method based on fractal theory | |
| Kumar et al. | Comparative analysis for edge detection techniques |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |