CN104143196A - Point object detection method based on multiple-linear time difference scanning and expansion sampling - Google Patents
Point object detection method based on multiple-linear time difference scanning and expansion sampling Download PDFInfo
- Publication number
- CN104143196A CN104143196A CN201410319201.8A CN201410319201A CN104143196A CN 104143196 A CN104143196 A CN 104143196A CN 201410319201 A CN201410319201 A CN 201410319201A CN 104143196 A CN104143196 A CN 104143196A
- Authority
- CN
- China
- Prior art keywords
- pixel
- image
- detector array
- row
- sub
- 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
Landscapes
- Image Analysis (AREA)
Abstract
Provided is a point object detection method based on multiple-linear time difference scanning and expansion sampling. The point object detection method comprises the steps of (1) constructing a double-linear-detector imaging device; (2) performing point object detection through each linear detector in an expansion sampling method; (3) respectively processing Nt groups of image data collected by each the linear detectors to obtain sub-pixel images, performing heterogeneity correction on two sub-pixels processed by the two linear detectors, performing sub pixel matching, and performing calculus of differences to complete background removal; (4) performing threshold filtering on residual images obtained through difference in the step (3), extracting positive and negative point pairs in the residual images through a neighborhood constraint criterion to complete detection and extraction of a moving point object in a scanning process, and forming neighborhood constraint criterion (as specified in the specification) according to imaging time differences of different linear detectors and the target movement speed, wherein delta D represents a distance between a positive area and a negative area, vmin represents the minimum movement speed of the detected object, vmax represents the maximum movement speed of the detected object, and GDS represents ground sampling distances of the linear detectors.
Description
Technical field
The invention belongs to target detection process field, relate to a kind of point target detection method based on multi-thread row moveout scan expansion sampling.
Background technology
Under complex background, the research of small and weak moving target Detection Techniques has important application in civilian, space flight and military affairs.Because image-forming range is far away and the impact of complex background, target becomes point-like in image, lacks enough texture informations and shape information, and target signal to noise ratio in image is very low, and this has increased the difficulty of dim targets detection greatly.At present general optical imagery detection system is general adopts single detector to carry out detection imaging, to obtained image, the method based on filtering of employing, the method based on small echo, based on morphologic method etc., carry out the inhibition of single frames background, on single-frame images, carry out suspected target extraction, the rear mode of utilizing multiframe association is carried out target trajectory matching, thereby realize target detects.This Detecting System is often subject to the impact of complex background, cannot on single-frame images, carry out reliably target extraction, this Detecting System does not make full use of the movable information of target in target leaching process simultaneously, therefore often there is higher false alarm rate, need to adopt complicated algorithm of target detection for improving detection accuracy, reduce data-handling efficiency.
Summary of the invention
Technology of the present invention is dealt with problems and is: overcome the deficiencies in the prior art, propose a kind of point target detection method based on multi-thread row moveout scan expansion sampling, realize moving target quick detection and detect, improve the ability that under complex background environment, moving target is surveyed.
Technical solution of the present invention is: a kind of point target detection method based on multi-thread row moveout scan expansion sampling, and step is as follows:
(1) construct multi-thread row moveout scan sniffer, this device comprises optical system, scanning mechanism and multi-thread row detector; Described scanning mechanism comprises pendulum mirror and drive shaft thereof; Described multi-thread row detector comprises N
dindividual detector array, each detector array is arranged in parallel, and the distance between adjacent two detector array is d
i, wherein i=1,2 ..., N
d-1; Optical system images in focal plane by scene in visual field together with scanning mechanism, and drive shaft drives the rotation of pendulum mirror, and the scene imaging in linear field is with tandem multi-thread row detector on the inswept focal plane of certain speed, and angular scanning rate is ω; Multiple detector array are to successively imaging of same position scene in visual field, any two alignment imaging time intervals
any two detectable movement velocitys of detector array are greater than v
ttarget, wherein
wherein Δ d is distance between any two detector array, and f is optical system focal length, the ground sampled distance that GSD is detector array; Described N
dspan is N
d>=2;
(2) each detector array adopts expansion sample mode to carry out point target detection, and detector array adopts N
tindividual detection array composition, the instantaneous field of view that pixel is corresponding is IFOV, adjacent two detection arrays are arranged in parallel, at the vertical scanning direction 1/N that staggers successively
tindividual pixel, and the detection array S that samples in direction of scanning, in a sampling length is set
tinferior; Described sampling length is the instantaneous field of view that pixel is corresponding; Described N
tbe more than or equal to 2; Described S
tspan is S
t>=2;
(3) N respectively each detector array being gathered
tgroup view data is processed, and obtains subpixel image; Choose two sub-pixels of detector array two width after treatment and carry out carrying out Difference Calculation after Nonuniformity Correction, sub-pix coupling, complete background and eliminate;
(4) difference in step (3) is obtained to residual image and carry out threshold filter, and it is right to adopt neighborhood constraint criterion to extract positive and negative point wherein, completes moving spot targets extraction in single pass process; Forming neighborhood constraint criterion according to the different detector array imaging time differences and target speed is
wherein positive and negative region distance Δ D, the target minimum movement speed v of detection
min, the target maximum movement speed v of detection
max.
The N that each detector array is gathered in described step (3)
tgroup view data is processed, and obtains the mode of subpixel image, and detailed process is as follows:
(3.1) N in each detector array
tthe imaging simultaneously of individual detection array, obtains N
tgroup view data; Proceed to immediately afterwards step (3.2); In this simultaneously, each detector array carries out scanning imagery according to the sampling interval that sampling number is corresponding is set in step (2) in direction of scanning, and each imaging obtains respectively N
tgroup view data, obtains proceeding to immediately step (3.2) after data;
(3.2) respectively by the N of each detector array
tgroup view data is alignd after splicing is processed and is formed a frame detection image;
(3.3) on direction of scanning, complete after default sampling number, the frame detection image that each detector array correspondence is obtained spliced and obtains corresponding subpixel image according to the time.
Before two width subpixel images in described step (3) carry out sub-pix coupling, taking subpixel image corresponding to the detector array of formerly imaging as benchmark, the subpixel image corresponding detector array of rear imaging is moved forward in direction of scanning
oK, L
Δ dround numbers row, detector array array pixel dimension is a × a; Adopt again on this basis a sub-pix method for registering to mate.
Filtering in described step (4) adopts the mode of threshold filter, and detailed process is as follows:
(4.1) the each pixel in the difference image obtaining after step (3) processing is strengthened to processing, obtain image I
sub1;
(4.2) image I
sub1take absolute value, be designated as image I
sub2;
(4.3) by I
sub2adjacent several image column are divided into one group, are divided into N
subnumber of sub images; The columns L of subimage
sub=H/N
sub, H is image I
sub2total columns; Simultaneously to image I
sub1adopt and image I
sub2identical mode is divided into groups, image I
sub1with I
sub2subimage corresponding one by one;
(4.4) from I
sub2the first row of current subimage is designated as current line and starts, and sets a threshold value initial value T, and T gets the pixel value of first pixel, or the value slightly larger than the pixel of first pixel;
(4.5) by the L in current subimage current line
subthe pixel value of individual pixel compares with threshold value initial value T successively, if pixel value exceedes threshold value initial value T, according to formula (a), T is improved to certain amplitude:
T′＝T+0.5
k(V
i-T) (a)
T＝T′
Wherein k is coefficient, adjusts V according to signal noise ratio (snr) of image
irepresent the pixel value of i pixel of current subimage current line;
After last pixel of current line is processed, obtain the final filtering threshold T of current subimage current line r
r;
(4.6) utilize final threshold value T obtained above
rto I
sub1the corresponding row of middle corresponding subimage is carried out threshold filter,, in same a line image, the pixel that is greater than filtering threshold is labeled as to 1; If pixel value, for negative, is labeled as-1 by the pixel that is less than filtering threshold negative value; Remaining pixel is all labeled as 0;
(4.7) threshold value T step (4.5) finally being obtained
rcarry out ratio reduction in formula (b), the threshold value initial value as current subimage next line:
T＝(1-0.5
m)T
r (b)
Wherein m is scale-up factor;
Using current subimage next line as current line, start to carry out from step (4.5), until all row of current subimage are disposed;
(4.8) select next son image as current subimage, repeating step (4.4)～(4.7), until complete image I
sub1the threshold filter processing of all row of all subimages, be designated as image I '
lab;
(4.9) traversal I '
labeach row, if continuous more than 2 pixel is denoted as 1, retains former mark constant, otherwise this pixel be labeled as to 0; If continuous 2 following pixels are marked as-1, retain former mark, otherwise visit pixel is labeled as to 0; Complete I '
labafter all pixel traversals, be designated as image I
lab.
Enhancing processing in described step (4.1), i.e. the pixel value of pixel value * α+neighbours territory pixel of pending pixel pixel value after treatment=pending pixel, α is for strengthening coefficient.
The present invention's advantage is compared with prior art:
1, adopt New Moving Target Detecting System and data processing technique, realize the two-dimensional expansion of Point Target imaging by expansion Sampling techniques, and gap detector obtains the detection image pair with the mistiming while utilizing multi-thread row; By follow-up image sub-pixel coupling and Difference Calculation, realize the inhibition of complex background; Utilize signal Enhancement Method to carry out echo signal enhancing, and then by detection method of small target, realize the extraction of moving target.
2, make full use of the motion feature of moving target, positive and negative marked region according to moving target on two width subpixel image difference images is differentiated extraction to carrying out moving target, can effectively suppress false-alarm, improve target detection accuracy, improve the ability that mobile target in complex background detects.
3, according to the variation of detecting movements of objects speed, can regulate sweep velocity, and in conjunction with the difference of multi-thread row arrangement pitch, choosing arbitrarily the image that two detector array become processes and Difference Calculation, make positive and negative in meeting certain neighborhood constraint condition on corresponding difference image, utilize two alignments that arrangement pitch is little to realize high-speed target detection, utilize two alignments that arrangement pitch is large to realize slower-velocity target detection, can realize thus friction speed moving target and survey.
4, adopt the scanning mechanism mode of carrying out active scan imaging, can realize on a large scale, the target detection of arbitrary motion direction.
5,, when two width subpixel images mate, can directly adopt conventional sub-pix matching process to mate, but matching efficiency and precision are lower; The present invention makes full use of detector array and arranges the imaging mode of feature and setting, taking subpixel image corresponding to the detector array of formerly imaging as benchmark, the subpixel image corresponding detector array of rear imaging is moved forward to L in direction of scanning
Δ doK, carry out two width Rapid Image Registrations, adopt again on this basis a sub-pix method for registering to mate, dwindle two width images match scopes, improve matching efficiency, reduce matching error, improve matching precision.
Brief description of the drawings
Fig. 1,2 is two kinds of mode schematic diagram of the multi-thread row moveout scan of the present invention sniffer;
Fig. 3 is that schematic diagram is processed in detector array two field picture splicing of the present invention;
Fig. 4 is residual image column split schematic diagram of the present invention;
Fig. 5 is neighborhood constraint criterion schematic diagram of the present invention;
Embodiment
Below in conjunction with accompanying drawing and example, the present invention is elaborated.A kind of multi-thread row moveout scan moving target detection method, step is as follows:
(1) construct multi-thread row moveout scan sniffer, this device comprises optical system 1, scanning mechanism 2 and multi-thread row detector 3; Described scanning mechanism comprises pendulum mirror and drive shaft thereof; Optical system images in focal plane by scene in visual field together with scanning mechanism, and drive shaft drives the rotation of pendulum mirror, and the scene imaging in linear field is with tandem multi-thread row detector on the inswept focal plane of certain speed, angular scanning rate ω; Described multi-thread row detector comprises N
dindividual detector array, each detector array is arranged in parallel, and the distance between adjacent two detector array is d
i, wherein i=1,2 ..., N
d-1; Multiple detector array are to successively imaging of same position scene in visual field, any two alignment imaging time intervals
any two detectable movement velocitys of detector array are greater than v
ttarget, wherein
wherein Δ d is distance between any two detector array, and f is optical system focal length, the ground sampled distance that GSD is detector array;
Each detector array adopts expansion sample mode to carry out point target detection, and detector array adopts N
tindividual detection array composition, the instantaneous field of view that pixel is corresponding is IFOV, adjacent two detection arrays are arranged in parallel, at the vertical scanning direction 1/N that staggers successively
tindividual pixel, and the detection array S that samples in direction of scanning, in a sampling length is set
tinferior; Described sampling length is the instantaneous field of view that pixel is corresponding; Described N
tbe more than or equal to 2; Described S
tspan is S
t>=2; Below with N
t=2 describe for example.
What Fig. 1 provided is that front end scans multi-thread column scan sniffer; The incident light of the emittance information that comprises target and background converges to focal plane through optical system 1 after the reflection of pendulum mirror, forms the picture of scenery, and drive shaft drives pendulum mirror according to default angular speed rotation, makes picture inswept each detector array successively of scenery.When the picture of scenery is during with inswept one of them detector array of certain speed, detector is sampled to the picture of scenery.What Fig. 2 provided is the multi-thread column scan sniffer of rear-end scanning.The incident light of the emittance information that comprises target and background converges to pendulum mirror through optical system 1, reflexes to focal plane through pendulum mirror, forms the picture of scenery.Drive shaft drives pendulum mirror according to default angular speed rotation, makes picture inswept each detector array successively of scenery.When the picture of scenery is during with inswept one of them detector array of certain speed, detector is sampled to the picture of scenery.
(2) optical system 1 together with scanning mechanism 2 by visual field in scene image in focal plane, drive shaft drives the rotation of pendulum mirror, scene imaging in linear field is with tandem two-wire row detector on the inswept focal plane of certain speed, two detector array are to successively imaging of same position scene in visual field, imaging time interval
be the N in each detector array
tthe imaging simultaneously of individual detection array, obtains N
tgroup view data; Obtain entering immediately step (3) after view data; In this simultaneously, each detector array still according in step (1), arrange in a sampling length sampling number, carry out scanning imagery in direction of scanning, each imaging obtains respectively N
tgroup view data, obtains proceeding to immediately step (3) after view data equally; For example, S can be set
t>=2, can realize the more than 2 times over-sampling of detector array on direction of scanning;
(3) respectively by the N of two detector array
tgroup view data is alignd after splicing is processed and is formed a frame detection image, and splicing is processed as shown in Figure 3, by N
tgroup pattern intersects splicing mutually.Processed and obtained sub-pixel frame detection image by splicing, realize target is in the stretching of vertical scanning direction.
(4), after default sampling number, the frame detection image that each detector array correspondence is obtained spliced and obtains two width subpixel images according to the time; Default desirable 200～300 row of sampling number, increase default sampling number and can increase scan image details, improve the precision of successive image processing; Reduce the efficiency that default sampling number can improve data processing, therefore default sampling number can regulate according to actual conditions.
(5) adopt multi-thread row moveout scan image to synthesize correcting mode, two width subpixel images are carried out to Nonuniformity Correction, specific as follows:
(5.1) two width subpixel images are intersected to splicing by row, form the new stitching image I of a width
p1;
(5.2) to stitching image I
p1each row image carry out Nonuniformity Correction, all row obtain row to the image I after nonuniformity correction after finishing dealing with
p2;
(5.3) according to the order of the row intersection splicing of two width images in step (5.1), from I
p2middlely extract respectively corresponding row image, rebuild two width and complete the subpixel image of row to Nonuniformity Correction;
(5.4) by complete row to two width subpixel images after nonuniformity correction by same direction half-twist respectively, repeating step (5.1)～(5.3), obtain proofreading and correct rear image I
p3;
(5.5) according to the order of the row intersection splicing of two width images in step (5.1), from I
p3middlely extract respectively corresponding row image, rebuild two width and complete the subpixel image of row to nonuniformity correction;
(5.6) by completing the capable two width subpixel images to nonuniformity correction by 90 ° of the opposite spins of step (5.4) rotation, two width subpixel images of row, column both direction nonuniformity correction have been obtained;
(6) taking subpixel image corresponding to the detector array of formerly imaging as benchmark, the subpixel image corresponding detector array of rear imaging is moved forward in direction of scanning
oK, L
Δ dround numbers row, detector array array pixel dimension is a × a;
(7) adopt sub-pix image registration algorithm to mate to step (6) two width subpixel images after treatment, specific as follows:
(7.11) taking subpixel image corresponding to the detector array of first imaging as reference picture, the subpixel image that the detector array of rear imaging is corresponding is image subject to registration.
(7.12) in reference picture, search for gradient largest block; Gradient largest block is defined as: the gradient image of computing reference two field picture, and pixel the logical value corresponding gradient that is greater than image gradient maximal value 4/5 is labeled as to 1, otherwise is labeled as 0, the image of formation is called logical value image.Select suitably big or small window at logical value image slide, be called gradient largest block containing 1 maximum piece corresponding to window.Determine the position of gradient largest block by the maximal value in comparison moving window.Window size is set up and is adopted 50x50.
(7.13) be to picture to the gradient largest block in (7.12), adopt based on the relevant image matching algorithm of gray scale, carry out reference picture and mate with the sub-pix of image subject to registration.Obtain more intensive grid and supply target image search by reference picture being carried out to interpolation, obtain the registration accuracy of sub-pixel.
(7.14) registration parameter calculating according to (7.13), treats registering images and carries out image conversion, obtains the image after registration.
(8) step (7) two width images after treatment are carried out to Difference Calculation, complete complex background self-adaptation and eliminate, obtain residual image I
sub1;
(9) to the image I obtaining after step (8) processing
sub1in each pixel strengthen processing, i.e. the pixel value of pixel value * α+neighbours territory pixel of pending pixel pixel value after treatment=pending pixel, α generally gets 2～4, obtains image I after processing
sub2;
(10) to image I
sub2carry out filtering, filtering herein can adopt conventional at present several filtering modes, and for example iterative threshold segmentation algorithm, can certainly adopt the Promethean threshold filter mode of the application, and the method detailed process is as follows:
(10.1) image I
sub2take absolute value, be designated as image I
sub3;
(10.2) by I
sub3adjacent several image column (general 6-10 row) are divided into one group, are divided into N
subnumber of sub images, as shown in Figure 4.The columns L of subimage
sub=H/N
sub, H is image I
sub3total columns; By I
sub2according to above-mentioned I
sub3identical mode is divided into groups, and obtains equally N
subnumber of sub images, I
sub2with I
sub3in subimage corresponding one by one;
(10.3) to I
sub3in current subimage the first row (being designated as current line) start, set a threshold value initial value T, T can get the pixel value of first pixel, or the value slightly larger than the pixel of first pixel;
(10.4) by the L in current subimage current line
subthe pixel value of individual pixel compares with threshold value initial value T successively, if pixel value exceedes threshold value initial value T, T is improved to certain amplitude, shown in (a):
T′＝T+0.5
k(V
i-T) (a)
T＝T′
Wherein k is conventionally taken as 4, k value and can adjusts according to signal noise ratio (snr) of image, and signal noise ratio (snr) of image is higher, k value can be got littlely, and signal noise ratio (snr) of image is lower, k value can be got greatly to V
irepresent the pixel value of i pixel of current subimage current line;
The T obtaining after last pixel of current line is processed, is the final filtering threshold T of current subimage current line r
r;
(10.5) utilize final threshold value T obtained above
rto I
sub2the corresponding row of middle corresponding subimage is carried out threshold filter,, in same a line image, the pixel that is greater than filtering threshold is labeled as to 1; If pixel value, for negative, is labeled as-1 by the pixel that is less than filtering threshold negative value; Remaining pixel is all labeled as 0;
(10.6) threshold value T step (10.4) finally being obtained
rreduce in a certain ratio, as the threshold value initial value of current subimage next line, now T computing formula is suc as formula (b):
T＝(1-0.5
m)T
r (b)
Wherein m is conventionally taken as 4, m value and can adjusts according to signal noise ratio (snr) of image, and signal noise ratio (snr) of image is higher, m value can be got greatly, and signal noise ratio (snr) of image is lower, m value can be got little.
Using current subimage next line as current line, start to carry out from step (10.4), until all row of current subimage are disposed;
(10.7) select next son image as current subimage, repeating step (10.3)～(10.6), until complete image I
sub2the threshold filter processing of all row of all subimages, be designated as image I '
lab;
(10.8) traversal I '
labeach row, if continuous more than 2 pixel is denoted as 1, retains former mark constant, otherwise this pixel be labeled as to 0; If continuous 2 following pixels are marked as-1, retain former mark, otherwise visit pixel is labeled as to 0; Complete I '
labafter all pixel traversals, be designated as image I
lab;
(11) as shown in Figure 5, extract I
labin paired positive and negative marked region, complete in single pass process moving target and survey and extract; The neighborhood constraint criterion of in pairs positive negative region is that two positive and negative region distances are Δ D, generally desirable
wherein two detector array trace interval Δ t, the target maximum movement speed v of detection
min, the target maximum movement speed v of detection
max, the ground sampled distance GSD of detector array;
(12) moving target extracting according to step (11) carries out track association analysis through follow-up multiple image, can determine the status information such as direction of motion, movement locus of moving target.
The unspecified part of the present invention belongs to general knowledge as well known to those skilled in the art.
Claims (5)
1. the point target detection method based on multi-thread row moveout scan expansion sampling, is characterized in that step is as follows:
(1) construct multi-thread row moveout scan sniffer, this device comprises optical system, scanning mechanism and multi-thread row detector; Described scanning mechanism comprises pendulum mirror and drive shaft thereof; Described multi-thread row detector comprises N
dindividual detector array, each detector array is arranged in parallel, and the distance between adjacent two detector array is d
i, wherein i=1,2 ..., N
d-1; Optical system images in focal plane by scene in visual field together with scanning mechanism, and drive shaft drives the rotation of pendulum mirror, and the scene imaging in linear field is with tandem multi-thread row detector on the inswept focal plane of certain speed, and angular scanning rate is ω; Multiple detector array are to successively imaging of same position scene in visual field, any two alignment imaging time intervals
any two detectable movement velocitys of detector array are greater than v
ttarget, wherein
wherein Δ d is distance between any two detector array, and f is optical system focal length, the ground sampled distance that GSD is detector array; Described N
dspan is N
d>=2;
(2) each detector array adopts expansion sample mode to carry out point target detection, and detector array adopts N
tindividual detection array composition, the instantaneous field of view that pixel is corresponding is IFOV, adjacent two detection arrays are arranged in parallel, at the vertical scanning direction 1/N that staggers successively
tindividual pixel, and the detection array S that samples in direction of scanning, in a sampling length is set
tinferior; Described sampling length is the instantaneous field of view that pixel is corresponding; Described N
tbe more than or equal to 2; Described S
tspan is S
t>=2;
(3) N respectively each detector array being gathered
tgroup view data is processed, and obtains subpixel image; Choose two sub-pixels of detector array two width after treatment and carry out carrying out Difference Calculation after Nonuniformity Correction, sub-pix coupling, complete background and eliminate;
(4) difference in step (3) is obtained to residual image and carry out threshold filter, and it is right to adopt neighborhood constraint criterion to extract positive and negative point wherein, completes moving spot targets extraction in single pass process; Forming neighborhood constraint criterion according to the different detector array imaging time differences and target speed is
wherein positive and negative region distance Δ D, the target minimum movement speed v of detection
min, the target maximum movement speed v of detection
max.
2. a kind of point target detection method based on multi-thread row moveout scan expansion sampling according to claim 1, is characterized in that: the N that each detector array is gathered in described step (3)
tgroup view data is processed, and obtains the mode of subpixel image, and detailed process is as follows:
(3.1) N in each detector array
tthe imaging simultaneously of individual detection array, obtains N
tgroup view data; Proceed to immediately afterwards step (3.2); In this simultaneously, each detector array carries out scanning imagery according to the sampling interval that sampling number is corresponding is set in step (2) in direction of scanning, and each imaging obtains respectively N
tgroup view data, obtains proceeding to immediately step (3.2) after data;
(3.2) respectively by the N of each detector array
tgroup view data is alignd after splicing is processed and is formed a frame detection image;
(3.3) on direction of scanning, complete after default sampling number, the frame detection image that each detector array correspondence is obtained spliced and obtains corresponding subpixel image according to the time.
3. a kind of point target detection method based on multi-thread row moveout scan expansion sampling according to claim 1, it is characterized in that: before two width subpixel images in described step (3) carry out sub-pix coupling, taking subpixel image corresponding to the detector array of formerly imaging as benchmark, the subpixel image corresponding detector array of rear imaging is moved forward in direction of scanning
oK, L
Δ dround numbers row, detector array array pixel dimension is a × a; Adopt again on this basis a sub-pix method for registering to mate.
4. a kind of point target detection method based on multi-thread row moveout scan expansion sampling according to claim 1, is characterized in that: the filtering in described step (4) adopts the mode of threshold filter, and detailed process is as follows:
(4.1) the each pixel in the difference image obtaining after step (3) processing is strengthened to processing, obtain image I
sub1;
(4.2) image I
sub1take absolute value, be designated as image I
sub2;
(4.3) by I
sub2adjacent several image column are divided into one group, are divided into N
subnumber of sub images; The columns L of subimage
sub=H/N
sub, H is image I
sub2total columns; Simultaneously to image I
sub1adopt and image I
sub2identical mode is divided into groups, image I
sub1with I
sub2subimage corresponding one by one;
(4.4) from I
sub2the first row of current subimage is designated as current line and starts, and sets a threshold value initial value T, and T gets the pixel value of first pixel, or the value slightly larger than the pixel of first pixel;
(4.5) by the L in current subimage current line
subthe pixel value of individual pixel compares with threshold value initial value T successively, if pixel value exceedes threshold value initial value T, according to formula (a), T is improved to certain amplitude:
T′＝T+0.5
k(V
i-T) (a)
T＝T′
Wherein k is coefficient, adjusts V according to signal noise ratio (snr) of image
irepresent the pixel value of i pixel of current subimage current line;
After last pixel of current line is processed, obtain the final filtering threshold T of current subimage current line r
r;
(4.6) utilize final threshold value T obtained above
rto I
sub1the corresponding row of middle corresponding subimage is carried out threshold filter,, in same a line image, the pixel that is greater than filtering threshold is labeled as to 1; If pixel value, for negative, is labeled as-1 by the pixel that is less than filtering threshold negative value; Remaining pixel is all labeled as 0;
(4.7) threshold value T step (4.5) finally being obtained
rcarry out ratio reduction in formula (b), the threshold value initial value as current subimage next line:
T＝(1-0.5
m)T
r (b)
Wherein m is scale-up factor;
Using current subimage next line as current line, start to carry out from step (4.5), until all row of current subimage are disposed;
(4.8) select next son image as current subimage, repeating step (4.4)～(4.7), until complete image I
sub1the threshold filter processing of all row of all subimages, be designated as image I '
lab;
(4.9) traversal I '
labeach row, if continuous more than 2 pixel is denoted as 1, retains former mark constant, otherwise this pixel be labeled as to 0; If continuous 2 following pixels are marked as-1, retain former mark, otherwise visit pixel is labeled as to 0; Complete I '
labafter all pixel traversals, be designated as image I
lab.
5. a kind of point target detection method based on multi-thread row moveout scan expansion sampling according to claim 4, it is characterized in that: the enhancing processing in described step (4.1), be the pixel value of pixel value * α+neighbours territory pixel of pending pixel pixel value after treatment=pending pixel, α is for strengthening coefficient.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410319201.8A CN104143196B (en) | 2014-07-04 | 2014-07-04 | A kind of point target detection method based on many alignment moveout scans extension sampling |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410319201.8A CN104143196B (en) | 2014-07-04 | 2014-07-04 | A kind of point target detection method based on many alignment moveout scans extension sampling |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104143196A true CN104143196A (en) | 2014-11-12 |
| CN104143196B CN104143196B (en) | 2017-05-31 |
Family
ID=51852364
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410319201.8A Active CN104143196B (en) | 2014-07-04 | 2014-07-04 | A kind of point target detection method based on many alignment moveout scans extension sampling |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104143196B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107886538A (en) * | 2017-10-12 | 2018-04-06 | 北京空间机电研究所 | A kind of stellar camera asterism detection method and system |
| CN112729554A (en) * | 2020-12-18 | 2021-04-30 | 中国航空工业集团公司洛阳电光设备研究所 | Self-discrimination variable-frequency sampling multi-sampling method applied to line detector |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101706961A (en) * | 2009-11-10 | 2010-05-12 | 北京航空航天大学 | Image registration method and image registration device |
| CN102663385A (en) * | 2012-03-07 | 2012-09-12 | 北京空间机电研究所 | Detection method for spot target on satellite |
| CN103679766A (en) * | 2012-09-25 | 2014-03-26 | 中国航天科工集团第二研究院二〇七所 | A sub-pixel processing method decreasing boundary back substitution errors |
-
2014
- 2014-07-04 CN CN201410319201.8A patent/CN104143196B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101706961A (en) * | 2009-11-10 | 2010-05-12 | 北京航空航天大学 | Image registration method and image registration device |
| CN102663385A (en) * | 2012-03-07 | 2012-09-12 | 北京空间机电研究所 | Detection method for spot target on satellite |
| CN103679766A (en) * | 2012-09-25 | 2014-03-26 | 中国航天科工集团第二研究院二〇七所 | A sub-pixel processing method decreasing boundary back substitution errors |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107886538A (en) * | 2017-10-12 | 2018-04-06 | 北京空间机电研究所 | A kind of stellar camera asterism detection method and system |
| CN112729554A (en) * | 2020-12-18 | 2021-04-30 | 中国航空工业集团公司洛阳电光设备研究所 | Self-discrimination variable-frequency sampling multi-sampling method applied to line detector |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104143196B (en) | 2017-05-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105245841B (en) | A kind of panoramic video monitoring system based on CUDA | |
| Kim et al. | High-speed drone detection based on yolo-v8 | |
| Yang et al. | Concrete defects inspection and 3D mapping using CityFlyer quadrotor robot | |
| CN110009732B (en) | GMS feature matching-based three-dimensional reconstruction method for complex large-scale scene | |
| CN110569704A (en) | Multi-strategy self-adaptive lane line detection method based on stereoscopic vision | |
| CN107610175A (en) | The monocular vision SLAM algorithms optimized based on semi-direct method and sliding window | |
| CN111899164B (en) | Image splicing method for multi-focal-segment scene | |
| CN111311666A (en) | Monocular vision odometer method integrating edge features and deep learning | |
| CN109544679A (en) | The three-dimensional rebuilding method of inner wall of the pipe | |
| CN109615653A (en) | Percolating water area detecting and recognition methods based on deep learning and visual field projection model | |
| CN106887037B (en) | indoor three-dimensional reconstruction method based on GPU and depth camera | |
| Cvišić et al. | Recalibrating the KITTI dataset camera setup for improved odometry accuracy | |
| CN104156977A (en) | Point target movement velocity detection method based on multiple linear moveout scanning, extending and sampling | |
| Zou et al. | Conditional generative adversarial network-based training image inpainting for laser vision seam tracking | |
| CN107862713B (en) | Camera deflection real-time detection early warning method and module for polling meeting place | |
| CN106504274A (en) | A kind of visual tracking method and system based under infrared camera | |
| CN108596960B (en) | Sub-aperture image alignment method of light field camera | |
| CN106558069A (en) | A kind of method for tracking target and system based under video monitoring | |
| CN117974766B (en) | Multi-target identity judging method of distributed double infrared sensors based on space-time basis | |
| CN104159007A (en) | Point target searching detection device based on multi-linear time lag extension sampling | |
| CN104143196A (en) | Point object detection method based on multiple-linear time difference scanning and expansion sampling | |
| CN104143179A (en) | Method for enhancing moving target through multi-linear-array time difference scanning expansion sampling | |
| Reid et al. | 3D Trajectories from a Single Viewpoint using Shadows. | |
| CN111062968A (en) | Robot path skeleton extraction method based on edge scanning and centerline extraction | |
| CN110533733B (en) | Method for automatically searching target depth based on ghost imaging calculation |
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 |