CN110415208A - A kind of adaptive targets detection method and its device, equipment, storage medium - Google Patents
A kind of adaptive targets detection method and its device, equipment, storage medium Download PDFInfo
- Publication number
- CN110415208A CN110415208A CN201910498094.2A CN201910498094A CN110415208A CN 110415208 A CN110415208 A CN 110415208A CN 201910498094 A CN201910498094 A CN 201910498094A CN 110415208 A CN110415208 A CN 110415208A
- Authority
- CN
- China
- Prior art keywords
- image
- target
- interest
- contrast
- filtering processing
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- 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
- G06T2207/10048—Infrared image
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20004—Adaptive image processing
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20101—Interactive definition of point of interest, landmark or seed
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
Abstract
The invention discloses a kind of adaptive targets detection method and its device, equipment, storage medium, this method includes obtaining original image；Differential filtering processing is carried out to the original image, obtains filtering processing image；Cluster merging processing is carried out to the filtering processing image, obtains several interesting image regions；Calculate separately the contrast of each described image area-of-interest and the original image；And target image is obtained according to the contrast.The present invention is filtered original image using difference filter, eliminate most of background area, it is then based on Cluster merging method and obtains several interesting image regions ROIS, realize effective segmentation to original image, the contrast of interesting image regions ROIS and original image is finally combined to carry out the further exclusion of false target, false alarm rate is reduced in simple background or the target detection of complex background to realize, and improves detection effect.
Description
Technology neighborhood
The invention belongs to target detection technique neighborhoods, and in particular to a kind of adaptive targets detection method and its device are set
Standby, storage medium.
Background technique
With the development of target detection technique, detection, tracking and the identification of infrared small target mainly from infrared reconnaissance with
Tracking system, how detecting from the infrared image of acquisition and tracking target just becomes the most important thing.Therefore, infrared small target
Detection be always infrared acquisition neighborhood hot subject, study the detection method of infrared small target have to Anti-TBM it is far-reaching
Meaning.
Existing infrared small target detection method mainly includes two steps: carrying out background inhibition processing, to back to image
Image after scape inhibits is split processing.Wherein, background inhibition is carried out to image and handles the enhancing realized to target, it is common
Background suppression method include airspace filter method, morphologic filtering method and based on background forecast method, these background suppression methods
The separation for realizing low-and high-frequency signal in image achievees the effect that prominent target；Image after inhibiting to background is split place
Reason realizes effective segmentation of image, obtains target image by effective segmented image, common dividing method includes thresholding
Segmentation and compartmentalization segmentation, typical thresholding segmentation have gray level threshold segmentation method, and segmentation effect depends on the choosing of gray threshold
It takes, compartmentalization segmentation has region-growing method and split degree method, and segmentation effect depends on the selection and similarity criterion of seed point
Selection.
Above-mentioned existing infrared small target detection method is under simple scenario, because image effective information is more, background interference
Information is few, so detection effect is preferable, but under complex background, because being influenced by background interference information, leads to detection effect
It is deteriorated.
Summary of the invention
In order to solve the above-mentioned problems in the prior art, the present invention provides a kind of adaptive targets detection method and
Its device, equipment, storage medium.
The embodiment of the invention provides a kind of adaptive targets detection methods, this method comprises:
Obtain original image；
Differential filtering processing is carried out to the original image, obtains filtering processing image；
Cluster merging processing is carried out to the filtering processing image, obtains several interesting image regions；
Calculate separately the contrast of each described image area-of-interest and the original image；And
Target image is obtained according to the contrast.
In one embodiment of the invention, differential filtering processing is carried out to the original image, obtains filtering processing figure
Picture, comprising:
Differential filtering processing is carried out to the original image using Difference of Gaussian filter, obtains the filtering processing figure
Picture.
In one embodiment of the invention, Cluster merging processing is carried out to the filtering processing image, obtains several figures
As area-of-interest, comprising:
Several target pixel points are obtained from the filtering processing image using the first preset threshold；
Cluster merging processing is carried out to several target pixel points, obtains several interesting image regions.
In one embodiment of the invention, Cluster merging processing is carried out to several target pixel points, obtained described
Several interesting image regions, comprising:
According to distance threshold, gray threshold, neighborhood sample number threshold value, using Density Clustering method to several target pictures
Vegetarian refreshments carries out Cluster merging processing, obtains several interesting image regions.
In one embodiment of the invention, each described image area-of-interest and the original image are calculated separately
Contrast, comprising:
The adjacent several original picture blocks of each described image area-of-interest are obtained respectively；
Calculate separately first the second gray value of sum of the grayscale values of each described image area-of-interest；
Calculate separately the third gray value of each of the adjacent original picture block of each described image area-of-interest；
According to third gray value described in first gray value, second sum of the grayscale values, each figure is calculated separately
As the contrast between area-of-interest several original picture blocks adjacent with described image area-of-interest.
In one embodiment of the invention, the size of described image area-of-interest and described image area-of-interest
The size of each adjacent original picture block is identical, wherein the size of described image area-of-interest is to include the figure
As the minimum circumscribed rectangle of all target pixel points in area-of-interest.
In one embodiment of the invention, target image is obtained according to the contrast, comprising:
The target image is obtained according to the contrast and the second preset threshold.
Another embodiment of the present invention provides a kind of adaptive targets detection device, described device includes:
Data acquisition module, for obtaining the original image；
First data processing module obtains the filtering processing for carrying out differential filtering processing to the original image
Image；
Second data processing module obtains described several for carrying out Cluster merging processing to the filtering processing image
Interesting image regions；
Third data processing module, for calculating separately the institute of each described image area-of-interest Yu the original image
State contrast；
Data determining module, for obtaining the target image according to the contrast.
Further embodiment of the present invention provides a kind of adaptive targets detection electronic equipment, which includes processing
Device, communication interface, memory and communication bus, wherein the processor, the communication interface, the memory pass through described
Communication bus completes mutual communication；
The memory, for storing computer program；
The processor when for executing the computer program stored on the memory, realizes any of the above-described institute
The method stated.
Another embodiment of the invention provides a kind of computer readable storage medium, in the computer readable storage medium
It is stored with computer program, the computer program realizes any of the above-described method when being executed by processor.
Compared with prior art, beneficial effects of the present invention:
The present invention is filtered original image using difference filter, eliminates most of background area, then
Several interesting image regions ROIS are obtained based on Cluster merging method, effective segmentation to original image is realized, finally combines
The contrast of interesting image regions ROIS and original image carry out false target further exclusion, thus realize regardless of
In simple background or the target detection of complex background, false alarm rate can be reduced, detection effect is improved.
The present invention is described in further details below with reference to accompanying drawings and embodiments.
Detailed description of the invention
Fig. 1 is a kind of flow diagram of adaptive targets detection method provided in an embodiment of the present invention；
Fig. 2 is the structural representation of the eight neighborhood block in a kind of adaptive targets detection method provided in an embodiment of the present invention
Figure；
Fig. 3 is a kind of structural schematic diagram of adaptive targets detection device provided in an embodiment of the present invention；
Fig. 4 is a kind of structural schematic diagram of adaptive targets detection electronic equipment provided in an embodiment of the present invention；
Fig. 5 is a kind of structural schematic diagram of computer readable storage medium provided in an embodiment of the present invention.
Specific embodiment
Further detailed description is done to the present invention combined with specific embodiments below, but embodiments of the present invention are not limited to
This.
Embodiment one
Existing infrared small target detection method mainly includes two steps: background inhibits and image segmentation.Wherein, traditional
Background inhibit by extract image in middle low-and high-frequency signal, separate background or noise region with target image, reach prominent
The effect of target image out, common background suppression method include traditional high-pass filtering or low-pass filtering method, based on gray scale
Morphologic method, the method based on frequency domain filtering and the method based on background forecast, such method are all by given pixel
Pixel near point carries out realizing background estimating after linear transformation or nonlinear transformation, can be converted in specific implementation
Convolution operation based on Filtering Template；Traditional image segmentation is that the image after inhibiting to background carries out effective image segmentation, into
And target image is obtained, common image partition method includes the dividing method based on threshold value and the dividing method based on region,
Wherein, optimum gradation threshold value, more commonly used threshold value are solved based on the dividing method of threshold value is the suitable criterion function of selection
Dividing method includes fixed threshold method, fuzzy binary images, maximum variance between clusters, and the dividing method based on region is by image
It is divided into different regions according to similarity criterion, more commonly used region segmentation method includes seed mediated growth method, regional split conjunction
And method, dividing ridge method.
Detection effect of the above-mentioned traditional infrared small target detection method under simple scenario is preferable, has very high real-time
Property, and under complex background, background suppression technology can also come out contextual information extraction while extracting target image, back
The effect that scape inhibits is poor, and then more false-alarm point is generated after image dividing processing.In addition, above-mentioned traditional infrared small target
Detection method can only detect the target under fixed size, lose validity for multiscale target, and multiscale target is detected
It generally requires and establishes scale space, but will cause the target detection time in this way and be multiplied, be unfavorable for the real-time detection of target.
Based on above-mentioned problem, referring to Figure 1, Fig. 1 is a kind of adaptive targets inspection provided in an embodiment of the present invention
The flow diagram of survey method.The embodiment of the invention provides a kind of adaptive targets detection method, this method includes following step
It is rapid:
Step 1 obtains original image.
In the present embodiment, original image includes infrared small target image data, background area data and noise data.
Step 2 carries out differential filtering processing to original image, obtains filtering processing image.
Specifically, traditional background suppression method essence is background forecast, the pixel around given pixel point is utilized
Transformed value replaces the gray value of central pixel point, more or less there is deviation, even if there are faint noise and background information,
Testing result can all have more false-alarm point, to bring biggish interference to target detection.Therefore, the present embodiment is using poor
Filter-divider carries out differential filtering processing to original image, can exclude most background area or noise information.Wherein,
Difference filter and original image carry out convolution, and the target of infrared small target in original image is captured by Filtering Template central area
Pixel is more advantageous to the extraction to original image edge destination pixel, makes it under more effective target pixel points
Target detection is carried out, the effect of target detection is improved.
Preferably, difference filter is Difference of Gaussian filter, differential low-pass filter, differential bandpass filter.
Step 3 carries out Cluster merging processing to filtering processing image, obtains several interesting image regions.
Specifically, traditional image partition method is mostly the operation based on pixel scale, image pixel by pixel is clicked through
Row processing, and due to being limited by Filtering Template size, target detection, nothing can only be carried out with the target detection frame of fixed size
Method carries out multiscale target detection effectively in real time, and there may be false target, the alert problem of false, traditional more rulers for testing result
Degree detection method usually requires repeatedly to be filtered or established scale pyramid, such multi-dimension testing method is time-consuming serious, figure
As the poor robustness of segmentation, there are a large amount of false-alarm points for testing result.Therefore, it present embodiments provides a kind of based on clustering method
Image partition method, using Cluster merging method to Difference of Gaussian filter filter filtering processing image in it is several
Target pixel points carry out Cluster merging processing, and Cluster merging result corresponds to several interesting image regions of the present embodiment
(Region of Interest, abbreviation ROIS), these interesting image regions ROIS, which is remained in original image, largely to be had
The target pixel points of effect further eliminate background area and noise information, reduce false target, false police, improve target inspection
The effect of survey.Meanwhile shape, the size of each interesting image regions ROIS of the present embodiment may be different, according to each image
Region of interest ROI S size can constantly adjust the size of its corresponding target detection frame, realize the dimension self-adaption inspection of target
It surveys.The present embodiment is the target scale self-adapting detecting method based on block rank, compared with the detection method based on pixel scale,
Significantly reduce the time of target detection.
Preferably, Cluster merging method includes mixed Gaussian clustering method, Density Clustering method, based on neural network model
Clustering method, level agglomerate clustering method.
Step 4, the contrast for calculating separately each interesting image regions and original image.
Specifically, the present embodiment exists not in view of target pixel points and ambient background pixel or noise pixel point
Continuity, therefore the contrast function comprising interesting image regions and original image information is designed, pass through the contrast function meter
Each interesting image regions ROIS and the contrast around it between original image are calculated, the value of contrast is smaller, correspondence image
The contrast of region of interest ROI S is lower, shows that interesting image regions ROIS and peripheral region have certain continuity,
There are similar areas around i.e., then interesting image regions ROIS is likely to be background area or noise region.
Step 5 obtains target image according to contrast.
Specifically, the present embodiment according to the contrast function in step 4, obtain each interesting image regions ROIS with
The contrast of original image retains the big interesting image regions ROIS of contrast, the small interesting image area of removal contrast
Domain ROIS, to obtain target image.The present embodiment utilizes target pixel points and ambient background pixel or noise pixel point
Discontinuity, exclude false edge region, reduce false alarm rate, improve detection effect.Wherein, contrast it is big with it is small, with reality
Empirical value setting in the scene of border is related, and the specific value of the empirical value can be also finely adjusted according to actual scene.
In conclusion the present embodiment is filtered original image using difference filter, most of back is eliminated
Scene area or noise information are then based on Cluster merging method and obtain several interesting image regions ROIS, realize to original graph
Effective segmentation of picture finally combines the contrast of interesting image regions ROIS and original image to carry out the further of false target
It excludes, to realize in simple background or Target under Complicated Background detection, reduces false alarm rate, improve detection effect.
Embodiment two
On the basis of the above embodiment 1, the present embodiment uses the differential filtering processing of step 2 in embodiment one
It is Difference of Gaussian filter, the Cluster merging of step 3 in embodiment one is handled using improved Density Clustering method, so
Comparison of design degree function afterwards is realized in conjunction with the contrast of interesting image regions ROIS and original image to the adaptive of target
Detection, specific implementation process include:
Step 1 obtains original image.
Step 2 carries out differential filtering processing to original image using Difference of Gaussian filter, obtains filtering processing image.
Specifically, the present embodiment is in view of the gray scale in usual infrared small target region is generally than ambient background or noise
Gray scale is big, and the intensity profile in infrared small target region has an apparent peak value, and the infrared small mesh in entire original image
There is biggish contrast in mark region and the region of surrounding all directions, therefore, the present embodiment proposes to filter using difference of Gaussian
Wave device effectively extracts the target pixel points of infrared small target in original image, at the same eliminate some background areas or
It is noise information, to effectively enhance the target pixel points of infrared small target.
The present embodiment has sought the difference of two different Gaussian functions by the deduction to Gaussian function, obtains one
Differential filtering processor of the Difference of Gaussian filter that center is positive, surrounding is negative as the present embodiment, difference of Gaussian filtering
Implement body design is as follows:
Wherein, x and y is the target pixel points in original image, and σ 1 and σ 2 are respectively Gaussian function G1(x, y) and Gauss
Function G2The scale parameter of (x, y), and make 1 < σ of σ 2.
Preferably, 1 value of σ is 3*3 matrix, and 2 value of σ is 5*5 matrix.
The present embodiment carries out process of convolution as Filtering Template, and with original image by Difference of Gaussian filter, by filtering
Wave die plate central area captures the infrared small target in original image, and the filtering processing image filtered is a differential form
Region contrast figure, the bigger pixel of region contrast is more likely to be target pixel points, to realize to major part
The inhibition of background or noise, the enhancing to target pixel points.Wherein, the present embodiment is because the three-dimensional of Difference of Gaussian filter is special
Property be more in line with the three-dimensional character of infrared small target image, so carrying out difference filter to original image by Difference of Gaussian filter
Wave processing, can exclude most background pixel point and noise pixel point, i.e. reduction false-alarm point, to more effectively capture
The target pixel points of infrared small target, the present embodiment three-dimensional include position coordinates and the pixel of the pixel in x-axis, y-axis
The gray value of point.
Step 3 carries out Cluster merging processing to filtering processing image using improved Density Clustering method, obtains several figures
As area-of-interest.
The present embodiment step 3 pair is filtered image and carries out Cluster merging processing using improved Density Clustering method, obtains
To several interesting image regions, specific implementation may include steps of 3.1 and step 3.2, wherein
Step 3.1 obtains several target pixel points from filtering processing image using the first preset threshold.
Specifically, the present embodiment passes through a fixed threshold before carrying out Cluster merging to filtering processing image first
Value, such as the first preset threshold select several target pixel points from filtering processing image, and then to several target pictures selected
Vegetarian refreshments carries out subsequent processing, not only can reduce subsequent calculation amount, improves image processing speed, while by filtering
The pixel and the first preset threshold for handling image judge, select several target pixel points, can quickly exclude one in this way
Divide background pixel point and noise pixel point, reduces false alarm rate.Wherein, several target pixel points are suspected target pixel, this is doubted
It may be target pixel points like target pixel points, may be background pixel point or noise pixel point；The tool of first preset threshold
Body value needs are finely adjusted under different scenes.
Preferably, the value of the first preset threshold is 100.
Step 3.2 carries out Cluster merging processing to several target pixel points, obtains several interesting image regions.
Specifically, the present embodiment improves traditional Density Clustering method in order to improve the effect of target detection, according to
Distance threshold, gray threshold, neighborhood sample number threshold value, gather several target pixel points using improved Density Clustering method
Class merging treatment obtains several interesting image regions ROIS.Specifically, the present embodiment by distance threshold and gray threshold simultaneously
It is as the constraint condition of Density Clustering method, then specific to input for improved Density Clustering method are as follows: above-mentioned pre- through first
If the object pixel point set D={ d of threshold value screening1, d2..., dmAnd the parameter that is related to of improved Density Clustering method, it should
Parameter includes distance threshold α, gray threshold β, neighborhood sample threshold MinP, wherein d1, d2..., dmIt is target pixel points, away from
Distance metric mode from threshold alpha uses Euclidean distance, and the gray scale metric form of gray threshold β is using gray scale difference, this reality
It applies example and comprehensive measurement is carried out using Euclidean distance and gray scale difference；Output are as follows: interesting image regions ROIS set B={ B1,
B2..., Bk, the number of the interesting image regions ROIS of the present embodiment output is k, and k is uncertain data, specifically by improved
The target pixel points of Density Clustering method input and the parameter designing of this method determine, pass through improved Density Clustering method
The corresponding several interesting image regions ROIS detected of cluster result.
Preferably, the value of distance threshold α is 5, and the value of gray threshold β is 30, and neighborhood sample number threshold value MinP's takes
Value is 3.
Improved Density Clustering method provided in this embodiment is compared with the Density Clustering method before improvement, in conjunction with gray scale threshold
Value carries out self-adaption cluster, can preferably adapt to target pixel points, amalgamation result is more rationally more accurate, while ensure that cluster
Speed；Improved Density Clustering method provided in this embodiment does not need to determine seed picture compared with traditional seed mediated growth method
The selection of vegetarian refreshments and the number of seed, it is completely adaptive according to distance threshold α, gray threshold β and neighborhood sample number threshold value MinP
Carry out Cluster merging with answering, therefore, the interesting image regions ROIS that the present embodiment is obtained by improved Density Clustering method
It is not only restricted to target image shape for arbitrary shape, and does not need the number for knowing target pixel points in advance, with adaptive
Target detection frame obtains the exact scale of target while effectively segmentation, realize the dimension self-adaption of target, ensure that image
The real-time of detection, while a part of background pixel point and noise pixel point can be excluded.Wherein, for each adaptive mesh
Detection block is marked, the present embodiment preferably takes a minimum including all target pixel points in interesting image regions ROIS
Boundary rectangle.
It should be noted that the adaptive target detection frame of the present embodiment is not limited to above-mentioned take method.
Step 4, the contrast for calculating separately each interesting image regions and original image.
The present embodiment step 4 calculates separately the specific implementation packet of the contrast of each interesting image regions and original image
Include following steps 4.1, step 4.2, step 4.3 and step 4.4, wherein
Step 4.1 obtains the adjacent several original picture blocks of each interesting image regions respectively.
Specifically, the present embodiment first navigates to an interesting image regions ROIS, the figure first in original image
Picture region of interest ROI S is interesting image regions ROIS obtained in an above-mentioned steps 3, according to interesting image regions
All target pixel points in ROIS obtain a minimum circumscribed rectangle as target image block, and target image block corresponds to image
Specific location and scale of the region of interest ROI S in original image, then obtain around target image block adjacent thereto
Several original picture blocks, the size of each original picture block are equal to the size of target image block.For example, referring to Fig. 2, Fig. 2 is
The structural schematic diagram of eight neighborhood block in a kind of adaptive targets detection method provided in an embodiment of the present invention, as shown in Fig. 2, this
Implementation example figure is taken adjacent with target image block as the corresponding target image block of region of interest ROI S is as shown in 0 position in Fig. 2
8 original picture blocks, respectively such as in Fig. 21,2 ..., shown in 8 positions.
Step 4.2, first the second gray value of sum of the grayscale values for calculating separately each interesting image regions.
Specifically, calculating separately each interesting image regions ROIS by gray count method and corresponding to target image
The gray average of block, the gray average are the first gray value, and the first gray value is denoted as m0, while calculating each interesting image
Maximum gradation value in the corresponding target image block of region ROIS, the maximum gradation value are the second gray value, the second gray value
It is denoted as Lmax。
Step 4.3, the third gray value for calculating separately the adjacent each original picture block of each interesting image regions.
Specifically, calculating separately corresponding 8 original picture blocks of each target image block by gray count method
Gray average, the gray average are third gray value, and third gray value is denoted as m respectively1..., m8.Wherein, target image block with
Interesting image regions ROIS is corresponding.
Step 4.4, according to the first gray value, the second sum of the grayscale values third gray value, calculate separately each interesting image
Contrast between several original picture blocks adjacent with interesting image regions of region.
Specifically, the first gray value being calculated based on step 4.2, the second sum of the grayscale values step 4.3 are calculated
Third gray value, the present embodiment constructs the corresponding target image block of interesting image regions ROIS and 8 are adjacent thereto
The contrast function C of original picture block, specifically more as follows than degree function C design:
In the present embodiment, each interesting image regions ROIS and original image can be calculated by formula (2)
The contrast value of contrast, contrast function C is smaller, and the contrast of corresponding interesting image regions ROIS is lower, shows this
There are certain continuity in interesting image regions ROIS and peripheral region, i.e., there are similar areas for surrounding, then the interesting image
Region ROIS is likely to be background area or noise region.
Step 5 obtains target image according to contrast.
The present embodiment step 5 be specifically according to the contrast and the second preset threshold that contrast function C is obtained in step 4 come
Determine whether interesting image regions ROIS is object region.
Specifically, the second preset threshold is rule of thumb arranged in the present embodiment, pass through above-mentioned contrast function C formula (2)
The contrast of interesting image regions ROIS and original image is obtained, which is compared with the second preset threshold, it is right
It is less than or equal to the second preset threshold than degree, shows that interesting image regions ROIS is object region, contrast is greater than the
Two preset thresholds show that interesting image regions ROIS is background area or noise region, thus exclude background area or
It is these false targets of noise region, further excludes false edge region, lower false alarm rate, improves detection effect, obtain most
True object region eventually.Wherein, the second preset threshold specific value needs are finely adjusted according to actual scene.
Preferably, the value of the second preset threshold is 1.1.
By contrast function C formula (2), each interesting image regions ROIS degree of comparing is calculated, meter is passed through
Obtained contrast confirms whether each interesting image regions ROIS is object-image region compared with the second preset threshold
Domain excludes non-object image region, retains object region, obtains final target image by these object regions.
In conclusion adaptive targets detection method provided in this embodiment, improves traditional infrared small target deteection side
The detection effect of method reduces false alarm rate, while can accurately realize the adaptive of target detection scale in real time, improves mesh
The robustness for marking detection, makes it preferably be applied to actual scene.
Embodiment three
On the basis of above-described embodiment two, Fig. 3 is referred to, Fig. 3 is a kind of adaptive mesh provided in an embodiment of the present invention
Mark the structural schematic diagram of detection device.A kind of adaptive targets detection device is present embodiments provided, which includes:
Data acquisition module, for obtaining original image.
First data processing module obtains filtering processing image for carrying out differential filtering processing to original image.
Specifically, the present embodiment carries out differential filtering processing to original image using Difference of Gaussian filter, filtered
Wave handles image.
It is emerging to obtain several image senses for carrying out Cluster merging processing to filtering processing image for second data processing module
Interesting region；
Specifically, the present embodiment obtains several object pixels from filtering processing image first with the first preset threshold
Point, it is poly- using improved density then to several target pixel points according to distance threshold, gray threshold, neighborhood sample number threshold value
Class method carries out Cluster merging processing to several target pixel points, obtains several interesting image regions.
Third data processing module, for calculating separately the contrast of each interesting image regions and original image；
Specifically, the present embodiment obtains the adjacent several original picture blocks of each interesting image regions respectively first,
Then first the second gray value of sum of the grayscale values of each interesting image regions is calculated separately, and calculates separately each image sense
The third gray value of the adjacent each original picture block in interest region, according to the first gray value, the second sum of the grayscale values third gray scale
Value, calculates separately the comparison between each interesting image regions several original picture blocks adjacent with interesting image regions
Degree.
Data determining module, for obtaining target image according to contrast.
Specifically, when the present embodiment obtains target image according to contrast, in combination with the second preset threshold, according to right
Target image is obtained than degree and the second preset threshold.
A kind of adaptive targets detection device provided in this embodiment can execute above method embodiment, realize former
Reason is similar with technical effect, and details are not described herein.
Example IV
On the basis of above-described embodiment three, Fig. 4 is referred to, Fig. 4 is a kind of adaptive mesh provided in an embodiment of the present invention
Mark detection electronic devices structure schematic diagram.Present embodiments provide a kind of adaptive targets detection electronic equipment, the electronic equipment
Including processor, communication interface, memory and communication bus, wherein processor, communication interface, memory pass through communication bus
Complete mutual communication；
Memory, for storing computer program；
Processor, when for executing the computer program stored on memory, which is executed by processor
When perform the steps of
Step 1 obtains original image.
Step 2 carries out differential filtering processing to original image, obtains filtering processing image.
Specifically, the present embodiment carries out differential filtering processing to original image using Difference of Gaussian filter, filtered
Wave handles image.
Step 3 carries out Cluster merging processing to filtering processing image, obtains several interesting image regions.
Specifically, the present embodiment obtains several object pixels from filtering processing image first with the first preset threshold
Point, it is poly- using improved density then to several target pixel points according to distance threshold, gray threshold, neighborhood sample number threshold value
Class method carries out Cluster merging processing to several target pixel points, obtains several interesting image regions.
Step 4, the contrast for calculating separately each interesting image regions and original image.
Specifically, the present embodiment obtains the adjacent several original picture blocks of each interesting image regions respectively first,
Then first the second gray value of sum of the grayscale values of each interesting image regions is calculated separately, and calculates separately each image sense
The third gray value of the adjacent each original picture block in interest region, according to the first gray value, the second sum of the grayscale values third gray scale
Value, calculates separately the comparison between each interesting image regions several original picture blocks adjacent with interesting image regions
Degree.
Step 5 obtains target image according to contrast.
Specifically, when the present embodiment obtains target image according to contrast, in combination with the second preset threshold, according to right
Target image is obtained than degree and the second preset threshold.
A kind of adaptive targets provided in this embodiment detect electronic equipment, can execute above method embodiment and above-mentioned
Installation practice, it is similar that the realization principle and technical effect are similar, and details are not described herein.
Embodiment five
On the basis of above-described embodiment four, Fig. 5 is referred to, Fig. 5 is that a kind of computer provided in an embodiment of the present invention can
Read the structural schematic diagram of storage medium.A kind of computer readable storage medium provided in this embodiment, is stored thereon with computer
Program, above-mentioned computer program perform the steps of when being executed by processor
Step 1 obtains original image.
Step 2 carries out differential filtering processing to original image, obtains filtering processing image.
Specifically, the present embodiment carries out differential filtering processing to original image using Difference of Gaussian filter, filtered
Wave handles image.
Step 3 carries out Cluster merging processing to filtering processing image, obtains several interesting image regions.
Specifically, the present embodiment obtains several object pixels from filtering processing image first with the first preset threshold
Point, it is poly- using improved density then to several target pixel points according to distance threshold, gray threshold, neighborhood sample number threshold value
Class method carries out Cluster merging processing to several target pixel points, obtains several interesting image regions.
Step 4, the contrast for calculating separately each interesting image regions and original image.
Specifically, the present embodiment obtains the adjacent several original picture blocks of each interesting image regions respectively first,
Then first the second gray value of sum of the grayscale values of each interesting image regions is calculated separately, and calculates separately each image sense
The third gray value of the adjacent each original picture block in interest region, according to the first gray value, the second sum of the grayscale values third gray scale
Value, calculates separately the comparison between each interesting image regions several original picture blocks adjacent with interesting image regions
Degree.
Step 5 obtains target image according to contrast.
Specifically, when the present embodiment obtains target image according to contrast, in combination with the second preset threshold, according to right
Target image is obtained than degree and the second preset threshold.
A kind of computer readable storage medium provided in this embodiment, can execute above method embodiment, above-mentioned apparatus
Embodiment and above-mentioned electronic equipment embodiment, it is similar that the realization principle and technical effect are similar, and details are not described herein.
The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and it cannot be said that
Specific implementation of the invention is only limited to these instructions.For the those of ordinary skill of the technology neighborhood belonging to the present invention, In
Under the premise of not departing from present inventive concept, a number of simple deductions or replacements can also be made, all shall be regarded as belonging to of the invention
Protection scope.
Claims (10)
1. a kind of adaptive targets detection method characterized by comprising
Obtain original image；
Differential filtering processing is carried out to the original image, obtains filtering processing image；
Cluster merging processing is carried out to the filtering processing image, obtains several interesting image regions；
Calculate separately the contrast of each described image area-of-interest and the original image；And
Target image is obtained according to the contrast.
2. being obtained the method according to claim 1, wherein carrying out differential filtering processing to the original image
Image is filtered, comprising:
Differential filtering processing is carried out to the original image using Difference of Gaussian filter, obtains the filtering processing image.
3. the method according to claim 1, wherein to the filtering processing image carry out Cluster merging processing,
Obtain several interesting image regions, comprising:
Several target pixel points are obtained from the filtering processing image using the first preset threshold；
Cluster merging processing is carried out to several target pixel points, obtains several interesting image regions.
4. according to the method described in claim 3, it is characterized in that, being carried out at Cluster merging to several target pixel points
Reason, obtains several interesting image regions, comprising:
According to distance threshold, gray threshold, neighborhood sample number threshold value, using Density Clustering method to several target pixel points
Cluster merging processing is carried out, several interesting image regions are obtained.
5. the method according to claim 1, wherein calculate separately each described image area-of-interest with it is described
The contrast of original image, comprising:
The adjacent several original picture blocks of each described image area-of-interest are obtained respectively；
Calculate separately first the second gray value of sum of the grayscale values of each described image area-of-interest；
Calculate separately the third gray value of each of the adjacent original picture block of each described image area-of-interest；
According to third gray value described in first gray value, second sum of the grayscale values, each described image sense is calculated separately
The contrast between several original picture blocks adjacent with described image area-of-interest of interest region.
6. according to the method described in claim 5, it is characterized in that, the size of described image area-of-interest and described image sense
The size of the adjacent original picture block in each of interest region is identical, wherein the size of described image area-of-interest is
Minimum circumscribed rectangle comprising all target pixel points in described image area-of-interest.
7. the method according to claim 1, wherein obtaining target image according to the contrast, comprising:
The target image is obtained according to the contrast and the second preset threshold.
8. a kind of adaptive targets detection device, which is characterized in that described device includes:
Data acquisition module, for obtaining the original image；
First data processing module obtains the filtering processing image for carrying out differential filtering processing to the original image；
Second data processing module obtains several images for carrying out Cluster merging processing to the filtering processing image
Area-of-interest；
Third data processing module, for calculating separately the described right of each described image area-of-interest and the original image
Degree of ratio；
Data determining module, for obtaining the target image according to the contrast.
9. a kind of electronic equipment of adaptive targets detection, which is characterized in that the electronic equipment includes that processor, communication connect
Mouth, memory and communication bus, wherein the processor, the communication interface, the memory are complete by the communication bus
At mutual communication；
The memory, for storing computer program；
The processor when for executing the computer program stored on the memory, realizes claim 1~7 times
Method described in one.
10. a kind of computer readable storage medium, which is characterized in that be stored with computer in the computer readable storage medium
Program realizes claim 1~7 any method when the computer program is executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910498094.2A CN110415208B (en) | 2019-06-10 | 2019-06-10 | Self-adaptive target detection method and device, equipment and storage medium thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910498094.2A CN110415208B (en) | 2019-06-10 | 2019-06-10 | Self-adaptive target detection method and device, equipment and storage medium thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110415208A true CN110415208A (en) | 2019-11-05 |
| CN110415208B CN110415208B (en) | 2023-10-17 |
Family
ID=68358917
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910498094.2A Active CN110415208B (en) | 2019-06-10 | 2019-06-10 | Self-adaptive target detection method and device, equipment and storage medium thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110415208B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110838123A (en) * | 2019-11-06 | 2020-02-25 | 南京止善智能科技研究院有限公司 | Segmentation method for illumination highlight area of indoor design effect image |
| CN110910421A (en) * | 2019-11-11 | 2020-03-24 | 西北工业大学 | Weak and small moving object detection method based on block characterization and variable neighborhood clustering |
| CN111047624A (en) * | 2019-12-27 | 2020-04-21 | 成都英飞睿技术有限公司 | Image dim target detection method, device, equipment and storage medium |
| CN111476971A (en) * | 2020-04-09 | 2020-07-31 | 英大智能电气有限公司 | Transmission line closely mountain fire monitoring devices |
| CN111985555A (en) * | 2020-08-19 | 2020-11-24 | 中国科学院上海微系统与信息技术研究所 | Millimeter wave three-dimensional holographic image denoising method |
| CN112837335A (en) * | 2021-01-27 | 2021-05-25 | 上海航天控制技术研究所 | Medium-long wave infrared composite anti-interference method |
| WO2021146952A1 (en) * | 2020-01-21 | 2021-07-29 | 深圳市大疆创新科技有限公司 | Following method and device, movable platform, and storage medium |
| CN113810720A (en) * | 2021-08-09 | 2021-12-17 | 北京博雅慧视智能技术研究院有限公司 | Image processing method, device, equipment and medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103198480A (en) * | 2013-04-02 | 2013-07-10 | 西安电子科技大学 | Remote sensing image change detection method based on area and Kmeans clustering |
| CN103996198A (en) * | 2014-06-04 | 2014-08-20 | 天津工业大学 | Method for detecting region of interest in complicated natural environment |
| US20160035106A1 (en) * | 2014-07-30 | 2016-02-04 | Olympus Corporation | Image processing apparatus, image processing method and medium storing image processing program |
| CN105374029A (en) * | 2015-10-12 | 2016-03-02 | 国家电网公司 | Segmenting method and system of transformer substation equipment infrared image interest areas |
| CN105513080A (en) * | 2015-12-21 | 2016-04-20 | 南京邮电大学 | Infrared image target salience evaluating method |
| CN107256560A (en) * | 2017-05-16 | 2017-10-17 | 北京环境特性研究所 | A kind of method for detecting infrared puniness target and its system |
| CN107392885A (en) * | 2017-06-08 | 2017-11-24 | 江苏科技大学 | A kind of method for detecting infrared puniness target of view-based access control model contrast mechanism |
| CN107424146A (en) * | 2017-06-28 | 2017-12-01 | 北京理工大学 | A kind of infrared polarization method for objectively evaluating image quality and system |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104123561B (en) * | 2014-07-10 | 2018-04-13 | 中国矿业大学 | Fuzzy C-mean algorithm remote sensing image automatic classification method based on spatial attraction model |
| CN105654453B (en) * | 2014-11-10 | 2018-09-28 | 华东师范大学 | A kind of FCM image partition methods of robustness |
| CN105261004B (en) * | 2015-09-10 | 2018-03-06 | 西安电子科技大学 | Fuzzy C-mean algorithm image partition method based on average drifting and neighborhood information |
-
2019
- 2019-06-10 CN CN201910498094.2A patent/CN110415208B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103198480A (en) * | 2013-04-02 | 2013-07-10 | 西安电子科技大学 | Remote sensing image change detection method based on area and Kmeans clustering |
| CN103996198A (en) * | 2014-06-04 | 2014-08-20 | 天津工业大学 | Method for detecting region of interest in complicated natural environment |
| US20160035106A1 (en) * | 2014-07-30 | 2016-02-04 | Olympus Corporation | Image processing apparatus, image processing method and medium storing image processing program |
| CN105374029A (en) * | 2015-10-12 | 2016-03-02 | 国家电网公司 | Segmenting method and system of transformer substation equipment infrared image interest areas |
| CN105513080A (en) * | 2015-12-21 | 2016-04-20 | 南京邮电大学 | Infrared image target salience evaluating method |
| CN107256560A (en) * | 2017-05-16 | 2017-10-17 | 北京环境特性研究所 | A kind of method for detecting infrared puniness target and its system |
| CN107392885A (en) * | 2017-06-08 | 2017-11-24 | 江苏科技大学 | A kind of method for detecting infrared puniness target of view-based access control model contrast mechanism |
| CN107424146A (en) * | 2017-06-28 | 2017-12-01 | 北京理工大学 | A kind of infrared polarization method for objectively evaluating image quality and system |
Non-Patent Citations (1)
| Title |
|---|
| J.H. HAN等: "A Robust Infrared Small Target Detection Algorithm Based on Human Visual System", 《IEEE GEOSCIENCE AND REMOTE SENSING LETTERS》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110838123A (en) * | 2019-11-06 | 2020-02-25 | 南京止善智能科技研究院有限公司 | Segmentation method for illumination highlight area of indoor design effect image |
| CN110838123B (en) * | 2019-11-06 | 2022-02-11 | 南京止善智能科技研究院有限公司 | Segmentation method for illumination highlight area of indoor design effect image |
| CN110910421A (en) * | 2019-11-11 | 2020-03-24 | 西北工业大学 | Weak and small moving object detection method based on block characterization and variable neighborhood clustering |
| CN111047624A (en) * | 2019-12-27 | 2020-04-21 | 成都英飞睿技术有限公司 | Image dim target detection method, device, equipment and storage medium |
| WO2021146952A1 (en) * | 2020-01-21 | 2021-07-29 | 深圳市大疆创新科技有限公司 | Following method and device, movable platform, and storage medium |
| CN111476971A (en) * | 2020-04-09 | 2020-07-31 | 英大智能电气有限公司 | Transmission line closely mountain fire monitoring devices |
| CN111985555A (en) * | 2020-08-19 | 2020-11-24 | 中国科学院上海微系统与信息技术研究所 | Millimeter wave three-dimensional holographic image denoising method |
| CN112837335A (en) * | 2021-01-27 | 2021-05-25 | 上海航天控制技术研究所 | Medium-long wave infrared composite anti-interference method |
| CN112837335B (en) * | 2021-01-27 | 2023-05-09 | 上海航天控制技术研究所 | Medium-long wave infrared composite anti-interference method |
| CN113810720A (en) * | 2021-08-09 | 2021-12-17 | 北京博雅慧视智能技术研究院有限公司 | Image processing method, device, equipment and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110415208B (en) | 2023-10-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110415208A (en) | A kind of adaptive targets detection method and its device, equipment, storage medium | |
| CN107833220B (en) | Fabric defect detection method based on deep convolutional neural network and visual saliency | |
| CN109685060B (en) | Image processing method and device | |
| CN109086724B (en) | Accelerated human face detection method and storage medium | |
| EP2085928B1 (en) | Detection of blobs in images | |
| CN112598713A (en) | Offshore submarine fish detection and tracking statistical method based on deep learning | |
| CN109685045A (en) | A kind of Moving Targets Based on Video Streams tracking and system | |
| Wazalwar et al. | A design flow for robust license plate localization and recognition in complex scenes | |
| CN110135312B (en) | Rapid small target detection method based on hierarchical LCM | |
| CN112364865B (en) | Method for detecting small moving target in complex scene | |
| CN111369570B (en) | Multi-target detection tracking method for video image | |
| CN113780110A (en) | Method and device for detecting weak and small targets in image sequence in real time | |
| CN107862262A (en) | A kind of quick visible images Ship Detection suitable for high altitude surveillance | |
| CN111222511B (en) | Infrared unmanned aerial vehicle target detection method and system | |
| CN115171218A (en) | Material sample feeding abnormal behavior recognition system based on image recognition technology | |
| CN117115117B (en) | Pathological image recognition method based on small sample, electronic equipment and storage medium | |
| CN106778822B (en) | Image straight line detection method based on funnel transformation | |
| CN111539980B (en) | Multi-target tracking method based on visible light | |
| CN112085683B (en) | Depth map credibility detection method in saliency detection | |
| CN110276260B (en) | Commodity detection method based on depth camera | |
| CN111027560B (en) | Text detection method and related device | |
| CN110472472B (en) | Airport detection method and device based on SAR remote sensing image | |
| Huang et al. | Invasion detection on transmission lines using saliency computation | |
| CN113284135B (en) | SAR ship detection method based on global and local context information | |
| Singh et al. | Multi-level threshold based edge detector using logical operations |
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 |