CN109002777A - A kind of infrared small target detection method towards complex scene - Google Patents
A kind of infrared small target detection method towards complex scene Download PDFInfo
- Publication number
- CN109002777A CN109002777A CN201810694988.4A CN201810694988A CN109002777A CN 109002777 A CN109002777 A CN 109002777A CN 201810694988 A CN201810694988 A CN 201810694988A CN 109002777 A CN109002777 A CN 109002777A
- Authority
- CN
- China
- Prior art keywords
- target
- image
- background
- fractional order
- detection
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The invention belongs to unmanned plane detection technique field, specifically a kind of infrared small target detection method towards complex scene.The present invention, which passes through, inhibits a large amount of backgrounds using a kind of improved morphologic filtering, isolates background.Propose a kind of conspicuousness detection algorithm of combination fractional order differential, using all directions to fractional order differential operator enhance target gray, ambient noise can slightly be inhibited simultaneously, then conspicuousness target detection is carried out using spectrum residual error, extract target, this is direct more preferable come the effect detected using spectrum residual error method than single, and obtained target conspicuousness is stronger, achievees the purpose that carry out the infrared small object in the visual field high detection rate and the detection of low false alarm rate in several scenes.
Description
Technical field
The invention belongs to unmanned plane detection technique field, specifically a kind of infrared small target inspection towards complex scene
Survey method.
Background technique
Currently, most of Method of Target Detection in Infrared is only capable of having good detection effect for sky background, majority is calculated
Method cannot be applicable in air-ground background, and the verification and measurement ratio under complex background is lower and false alarm rate is higher, for complicated air-ground background
Or relatively difficult problem.
Summary of the invention
The purpose of the present invention, aiming at the above problem, by utilizing a kind of improved morphologic filtering, in conjunction with fractional order
Differential operator and a kind of conspicuousness detection method solve the purpose of the infrared small target high-precision detection under Various Complex background.
The technical solution of the present invention is as follows:
A kind of infrared small target detection method towards complex scene, which comprises the following steps:
S1, background inhibition is carried out by morphologic filtering:
Grayscale morphology filtering algorithm is exactly that the one kind for expanding to come in binary morphology filtering is directed to gray-value image
Filtering processing algorithm.Since infrared image is substantially the gray level image of a secondary more Pixel-levels.Therefore gray scale can directly be utilized
Smaller using gray scale morphology processing image calculation amount in Morphology Algorithm processing, most cases can reach satisfied filtering effect
Fruit.
1. gray value dilation operation is defined as follows:
The gray scale that central point grey scale pixel value can be made to replace with given structural element corresponding region by dilation operation is maximum
Value.
2. a kind of dual operations that corrosion is expansion.Gray value erosion operation is defined as follows:
The same gray scale that central point grey scale pixel value can be made to replace with given structural element region by erosion operation is minimum
Value.
Wherein F is original image, and B is the structural element of corresponding operation.
Soft morphology operations formula are as follows:
Wherein, F is original image, and g (x, y) is that coordinate is the gray value of the pixel of (x, y) in G, and h (x, y) is background
Image H coordinate after inhibition is the gray value of the pixel of (x, y)；
S2, to background inhibit after image, carry out fractional order targets improvement:
Set image I (x, y) in the fractional order differential of X, Y-axis both direction be under certain condition it is separable, simultaneously
Due to the gray scale Gaussian distribution feature of Weak target, eight directions of image are handled using fractional order differential operator, it will
The duration [a, t] of picture signal I (x, y) presses unit gap h=1 equal part, obtains the fractional order differential operator edge under G-L is defined
X and the numerical value calculation expression of Y direction are as follows:
S3, conspicuousness Objective extraction:
Using spectrum residual error method, image I (x) is mapped to frequency domain from spatial domain using the Fourier transformation of following formula, is mentioned
Its amplitude spectrum p (f) is taken, keeps phase spectrum constant:
A (f)=| F [I (x)] | (5)
L (f)=log [A (f)] (7)
Q (f)=L (f) * n (f) (8)
D (f)=L (f)-Q (f) (10)
S (I (x))=G (x, y) * | F-1[exp{D(f)+iP(f)}]|2 (11)
Log transformation is carried out to its amplitude spectrum, obtains log frequency spectrum L (f)；It is obtained with the mean operator convolution of a n × n
Image Q (f) after to smooth background；
The amplitude spectrum D (f) that original amplitude spectrum and smoothed out amplitude spectrum subtraction obtain, is the area being smoothed in frequency domain
Domain, that is, salient region recycle a Gaussian filter to realize better in conjunction with back to spatial domain is mapped after its phase spectrum
Display effect obtains the boundary in interesting target region, obtains final Saliency maps as S (I (x)).
The solution of the present invention detects low latitude distance small target in visual field using infrared image detection technique.It is logical
Cross morphologic filtering, isolate after background using after fractional order differential operator enhancing image with a kind of conspicuousness detection method knot
It closes, can more highlight target than the detection of single conspicuousness, extract complete Weak target region, achieve the purpose that effectively to detect.
Beneficial effects of the present invention are to adapt in the small target deteection of more scenes, will be had extensively in unmanned plane early warning field
General application prospect.
Detailed description of the invention
Fig. 1 is the improved morphology top-hat Operator structure element of the present invention；
Fig. 2 is fractional order image enhancement operator of the present invention.
Specific embodiment
With reference to the accompanying drawings and examples, the technical schemes of the invention are described in detail.
Detection method of the invention was specifically divided into for three stages: a kind of improved top-hat background suppression method, fractional order
Targets improvement and spectrum residual error method conspicuousness Objective extraction.The invention firstly uses improved soft Mathematical morphology filter wave operators to be carried on the back
Scape inhibits, and can remove the interference of major part background area, since target and background clutter can be weakened after this operation,
Targets improvement is carried out using fractional order differential, target gray value can be enhanced, while slightly increasing background clutter, in combination with one
Kind conspicuousness detection method extracts target.
Improved top-hat background suppression method:
Mathematical morphology is molecular by one group of morphologic algebraic operation, its basic operation includes 4: expansion, corruption
Erosion, opening operation and closed operation.Top-hat operator be it is one such, in classical morphology operations, carry out expansion and erosion operation
Structural element be consistent, although above-mentioned top-hat operator is overall good in the relatively high Scene Representation of some signal-to-noise ratio
It is good, as when signal noise ratio (snr) of image is higher under sky background.Complex scene is simultaneously not suitable for.
Shown in soft morphology operations such as formula (12), F is original image, and g (x, y) is the pixel that coordinate is (x, y) in G
Gray value, h (x, y) be background inhibit after image H coordinate be (x, y) pixel gray value.
For improved structural element as shown in Figure 1, structural element size and size for corrosion remain unchanged, change is swollen
Swollen structural element keeps its hollow, inhibits very well when can contain vertical bar and treetop in complex scene.Improved top-
Hat operation has better background inhibitory effect and scene adaptability.
Fractional order targets improvement:
Fractional order differential operator can be while enhancing signal high frequency components, and the low frequency in nonlinear stick signal divides
Amount.Image is 2D signal, it is assumed that image I (x, y) is that can divide under certain condition in the fractional order differential of X, Y-axis both direction
From, simultaneously because eight directions of image can be utilized fractional order differential in this way by the gray scale Gaussian distribution feature of Weak target
Operator is handled, and the duration [a, t] of picture signal I (x, y) is pressed unit gap h=1 equal part, G-L available in this way
Fractional order differential operator under definition is along shown in X and the numerical value calculation expression such as formula (3) (4) of Y direction.
What Fig. 2 was indicated is an all directions to image enhancement mask operator E.Template size is m*m.The enhancing operator energy base
Rest of pixels point in the synchronous enhancing target pixel region of central target pixel, enhances the specific filter factor such as formula in operator E
(14) shown in.
Enhanced image is significantly enhanced on target gray, although background pixel point gray value is also increased
By force, but generally for, the grey value difference of background and target pixel points is increasing.
Conspicuousness target detection:
Conspicuousness detection method reflects larger significant difference just for colors some in image, area and surrounding
Target is usually all some biggish targets of area accounting, is not suitable for the direct detection of the Weak target in complex scene.Its
In fact for the infrared small target in gray level image, we similarly can consider that Small object is to be different from showing for peripheral region
Work property exists, as long as pretreatment and background suppression method are proper, so that it may which the Small object in the image after in pretreatment is regarded
A kind of conspicuousness background presence.
The basic principle of spectrum residual error method is by image I (x) using the inner Fourier transformation of formula (5)~formula (11) from sky
Between domain mapping to frequency domain, extract its amplitude spectrum p (f), keep phase spectrum constant；A log transformation then is carried out to its amplitude spectrum,
Obtain log frequency spectrum L (f).Image Q (f) after obtaining smooth background with the mean operator convolution of a n × n.Original amplitude spectrum with
The amplitude spectrum D (f) that smoothed out amplitude spectrum subtraction obtains is the region being smoothed in frequency domain, that is, salient region, knot
Back to spatial domain is mapped after closing its phase spectrum, convenient for intuitive display.Due to the obscurity boundary of target in infrared image, a height is utilized
This filter reaches better display effect, to obtain the boundary in interesting target region, obtains final Saliency maps as S
(I(x))。
The present invention, which passes through, inhibits a large amount of backgrounds using a kind of improved morphologic filtering, isolates background.Propose one kind
In conjunction with the conspicuousness detection algorithm of fractional order differential, using an all directions to fractional order differential operator enhance target ash
Degree, while can slightly inhibit ambient noise, conspicuousness target detection is then carried out using spectrum residual error, extracts target, this is than single
Direct using composing, residual error method is more preferable come the effect detected, and obtained target conspicuousness is stronger, reaches the equal energy in several scenes
The purpose of high detection rate and the detection of low false alarm rate is carried out to the infrared small object in the visual field.
Claims (1)
1. a kind of infrared small target detection method towards complex scene, which comprises the following steps:
S1, background inhibition is carried out by gray scale morphology filtering algorithm:
Gray value dilation operation is defined as follows:
Central point grey scale pixel value is set to replace with the gray scale maximum value of given structural element corresponding region by dilation operation；
Gray value erosion operation is defined as follows:
Central point grey scale pixel value is set to replace with the minimum gray value in given structural element region by erosion operation；
Wherein F is original image, and B is the structural element of corresponding operation；
Soft morphology operations formula are as follows:
Wherein, g (x, y) is that coordinate is the gray value of the pixel of (x, y) in G, and h (x, y) is the image H coordinate after background inhibits
For the gray value of the pixel of (x, y)；
It can inhibit most of background area by the step, extract target area and noise range comprising a small amount of background clutter；
S2, to background inhibit after image, carry out fractional order targets improvement:
It is separable under certain condition that image I (x, y), which is set, in the fractional order differential of X, Y-axis both direction, simultaneously because
The gray scale Gaussian distribution feature of Weak target is handled eight directions of image using fractional order differential operator, by image
The duration [a, t] of signal I (x, y) presses unit gap h=1 equal part, obtains the fractional order differential operator under G-L is defined along X and Y
The numerical value calculation expression of axis direction is as follows:
Wherein:
The template related parameter values of horizontal direction and vertical direction can be obtained according to above formula, the image after targets improvement is in target ash
It is significantly enhanced on degree；
S3, conspicuousness Objective extraction:
After targets improvement, although target becomes more apparent upon, at this time high-frequency noise also enhance to a certain extent, therefore using aobvious
Work property extracting method further extracts target；
Using spectrum residual error method, image I (x) is mapped to frequency domain from spatial domain using the Fourier transformation of following formula, extracts it
Amplitude spectrum p (f) keeps phase spectrum constant:
A (f)=| F [I (x)] |
L (f)=log [A (f)]
Q (f)=L (f) * n (f)
D (f)=L (f)-Q (f)
S (I (x))=G (x, y) * | F-1[exp{D(f)+iP(f)}]|2
Log transformation is carried out to its amplitude spectrum, obtains log frequency spectrum L (f)；It is put down with the mean operator convolution of a n × n
Image Q (f) after sliding background；
The amplitude spectrum D (f) that original amplitude spectrum and smoothed out amplitude spectrum subtraction obtain, is the region being smoothed in frequency domain,
It is exactly salient region, in conjunction with back to spatial domain is mapped after its phase spectrum, a Gaussian filter is recycled to realize preferably display
Effect obtains the boundary in interesting target region, obtains final Saliency maps as S (I (x)).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810694988.4A CN109002777B (en) | 2018-06-29 | 2018-06-29 | Infrared small target detection method for complex scene |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810694988.4A CN109002777B (en) | 2018-06-29 | 2018-06-29 | Infrared small target detection method for complex scene |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109002777A true CN109002777A (en) | 2018-12-14 |
| CN109002777B CN109002777B (en) | 2021-03-30 |
Family
ID=64600726
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810694988.4A Active CN109002777B (en) | 2018-06-29 | 2018-06-29 | Infrared small target detection method for complex scene |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109002777B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110490155A (en) * | 2019-08-23 | 2019-11-22 | 电子科技大学 | A kind of no-fly airspace unmanned plane detection method |
| CN110599509A (en) * | 2019-08-02 | 2019-12-20 | 西安理工大学 | Edge detection method based on eight-direction fractional order differential operator |
| CN110765948A (en) * | 2019-10-24 | 2020-02-07 | 长沙品先信息技术有限公司 | Target detection and identification method and system based on unmanned aerial vehicle |
| WO2021138994A1 (en) * | 2020-01-10 | 2021-07-15 | 大连理工大学 | Infrared small target detection method based on depth map under complex scene |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102034239A (en) * | 2010-12-07 | 2011-04-27 | 北京理工大学 | Local gray abrupt change-based infrared small target detection method |
| CN102096824A (en) * | 2011-02-18 | 2011-06-15 | 复旦大学 | Multi-spectral image ship detection method based on selective visual attention mechanism |
| CN103226695A (en) * | 2013-04-02 | 2013-07-31 | 北京林业大学 | Multi-scale intelligent Chinese red pine identification model based on selective vision attention mechanism |
| US20130243277A1 (en) * | 2012-03-17 | 2013-09-19 | Sony Corporation | Flourescent dot counting in digital pathology images |
| CN105930811A (en) * | 2016-04-26 | 2016-09-07 | 济南梦田商贸有限责任公司 | Palm texture feature detection method based on image processing |
| CN106845457A (en) * | 2017-03-02 | 2017-06-13 | 西安电子科技大学 | Method for detecting infrared puniness target based on spectrum residual error with fuzzy clustering |
| CN107194896A (en) * | 2017-06-05 | 2017-09-22 | 华中科技大学 | A kind of background suppression method and system based on neighbour structure |
| CN107358585A (en) * | 2017-06-30 | 2017-11-17 | 西安理工大学 | Misty Image Enhancement Method based on fractional order differential and dark primary priori |
| CN107516315A (en) * | 2017-08-25 | 2017-12-26 | 华南理工大学 | A kind of development machine based on machine vision is slagged tap monitoring method |
-
2018
- 2018-06-29 CN CN201810694988.4A patent/CN109002777B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102034239A (en) * | 2010-12-07 | 2011-04-27 | 北京理工大学 | Local gray abrupt change-based infrared small target detection method |
| CN102096824A (en) * | 2011-02-18 | 2011-06-15 | 复旦大学 | Multi-spectral image ship detection method based on selective visual attention mechanism |
| US20130243277A1 (en) * | 2012-03-17 | 2013-09-19 | Sony Corporation | Flourescent dot counting in digital pathology images |
| CN103226695A (en) * | 2013-04-02 | 2013-07-31 | 北京林业大学 | Multi-scale intelligent Chinese red pine identification model based on selective vision attention mechanism |
| CN105930811A (en) * | 2016-04-26 | 2016-09-07 | 济南梦田商贸有限责任公司 | Palm texture feature detection method based on image processing |
| CN106845457A (en) * | 2017-03-02 | 2017-06-13 | 西安电子科技大学 | Method for detecting infrared puniness target based on spectrum residual error with fuzzy clustering |
| CN107194896A (en) * | 2017-06-05 | 2017-09-22 | 华中科技大学 | A kind of background suppression method and system based on neighbour structure |
| CN107358585A (en) * | 2017-06-30 | 2017-11-17 | 西安理工大学 | Misty Image Enhancement Method based on fractional order differential and dark primary priori |
| CN107516315A (en) * | 2017-08-25 | 2017-12-26 | 华南理工大学 | A kind of development machine based on machine vision is slagged tap monitoring method |
Non-Patent Citations (3)
| Title |
|---|
| ROMAN JCM ET AL: "《Top-Hat Transform for Enhancement of Aerial Thermal Images》", 《2017 30TH SIBGRAPI》 * |
| 吴林春: "《基于分数阶微积分的路面裂缝检测研究》", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 * |
| 张智丰等: "《加权融合特征耦合Top-Hat变换的红外目标检测算法》", 《计算机工程与设计》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110599509A (en) * | 2019-08-02 | 2019-12-20 | 西安理工大学 | Edge detection method based on eight-direction fractional order differential operator |
| CN110490155A (en) * | 2019-08-23 | 2019-11-22 | 电子科技大学 | A kind of no-fly airspace unmanned plane detection method |
| CN110490155B (en) * | 2019-08-23 | 2022-05-17 | 电子科技大学 | Method for detecting unmanned aerial vehicle in no-fly airspace |
| CN110765948A (en) * | 2019-10-24 | 2020-02-07 | 长沙品先信息技术有限公司 | Target detection and identification method and system based on unmanned aerial vehicle |
| WO2021138994A1 (en) * | 2020-01-10 | 2021-07-15 | 大连理工大学 | Infrared small target detection method based on depth map under complex scene |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109002777B (en) | 2021-03-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109002777A (en) | A kind of infrared small target detection method towards complex scene | |
| CN106709426B (en) | Ship target detection method based on infrared remote sensing image | |
| CN103761731A (en) | Small infrared aerial target detection method based on non-downsampling contourlet transformation | |
| CN109460764B (en) | Satellite video ship monitoring method combining brightness characteristics and improved interframe difference method | |
| CN107392095A (en) | A kind of small IR targets detection algorithm based on mask image | |
| CN110400294B (en) | Infrared target detection system and detection method | |
| Rani et al. | Detection of edges using mathematical morphological operators | |
| Lian et al. | A novel method on moving-objects detection based on background subtraction and three frames differencing | |
| CN105184804A (en) | Sea surface small target detection method based on airborne infrared camera aerially-photographed image | |
| CN106327464A (en) | Edge detection method | |
| CN110728668A (en) | Airspace high-pass filter for maintaining small target form | |
| CN108230365A (en) | SAR image change detection based on multi-source differential image content mergence | |
| Jiang et al. | Automatic space debris extraction channel based on large field of view photoelectric detection system | |
| KR101426864B1 (en) | Region-based edge enhancement method for short-distance thermal target tracking | |
| CN115035350B (en) | Edge detection enhancement-based method for detecting small objects on air-ground and ground background | |
| Li et al. | An infrared dim and small target detection method based on fractional differential | |
| Liu et al. | Temporal noise suppression for small target detection in infrared image sequences | |
| Mai et al. | An improved method for Vietnam License Plate location | |
| Qi-chang et al. | A method of vehicle license plate de-noising and location in low light level | |
| Bae et al. | Recursive multi-SEs NWTH method for small target detection in infrared images | |
| Si-ming et al. | Moving shadow detection based on Susan algorithm | |
| Qin et al. | A background extraction and shadow removal algorithm based on clustering for ViBe | |
| Sun et al. | Infrared small targets detection based on MRF model | |
| Zhang et al. | Infrared small target detection using modified order morphology and weighted local entropy | |
| Wu et al. | Moving-object Detection Based on Shadow Removal and Prospect Reconstruction |
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 |