CN105335944A - Onboard staggered TDI (Transport Driver Interface) infrared image recovering method and system based on geometric correction - Google Patents

Abstract

The invention provides an onboard staggered TDI (Transport Driver Interface) infrared image recovering method and system based on geometric correction. The onboard staggered TDI infrared image recovering method comprises the following steps: carrying out odd and even row separation on an original image; determining a coverage range AOP (Aspect-Oriented Programming) of the original image on an object space projection plane according to strip information of the original image; determining a projection coverage region AOI (Automated Optical Inspection) of a correction image on the object space projection plane; determining a sampling interval of the correction image according to space resolution of the original image, and dividing the AOI by using the sampling interval to form equal-interval grids; primarily determining a strip range of each grid point and determining the range of track data according to the strip range; searching an optimal scanning row of the current grid point in each strip to obtain the track data of the current grid point in each strip; and acquiring a plurality of most adjacent sampling points on an odd and even independent image corresponding to the grid point of the correction image, so as to sample the gray level again. With the adoption of the technical scheme provided by the invention, an infrared image can be strictly recovered based on the geometric correction.

Description

Based on airborne staggered TDI infrared image restored method and the system of geometric correction

Technical field

The present invention relates to remote sensing image technical field, especially relate to a kind of recovery technique scheme of the airborne staggered TDI infrared image based on geometric correction.

Background technology

During infrared imaging, aviation time delay integration (Timedelayandintegration, TDI) sensor can obtain the image higher than common line array CCD (Charge-coupleddevice) signal to noise ratio (S/N ratio), is a kind of important means obtaining high resolving power infrared image.Each image-generating unit of TDI is being rearranged perpendicular to TDI linear array direction by N number of photosensitive unit, and N is the exponent number of TDI.When sensor imaging, the signal of each photosensitive unit transmits cumulative along the opposite direction of direction of scanning successively, obtains final signal value.Compared to the ordinary lines array ccd sensor of ad eundem, the signal to noise ratio (S/N ratio) of the image that TDI obtains can improve N ^1/2doubly.Along with the increase of TDI array length, by restriction physically, the staggered pattern of general employing parity rows.This design effectively can reduce the loss of information between neighbor while expanding sensor areas imaging.But due to the increase of integral time and the interval between odd field and even field, this makes sensor become large to vibration sensing degree.In TDI infrared sensor imaging process, except the vibration of sensor platform (aircraft), also there is lens barrel in sensor electronic cooling system, servomotor, the vibration of platform and sweeping imaging process and to rotate around the axis the vibration etc. caused.Although sensor can consider these factors when Machine Design, and partial vibration can be weakened to a certain extent, or some remainder errors can be there are.These vibration factors make odd-numbered line array and even number line array be in different positions and attitude when obtaining same terrestrial object information, thus cause the changing of the relative positions between staggered TDI image parity rows and image geometry distortion.Except vibration factor, topographic relief and CCD alignment error also can cause the changing of the relative positions between parity rows.Need to restore the image of the changing of the relative positions for this reason.

Image restoration carries out the correction of atmospheric effect, geometry correction and the correction to the scan line leakage, dislocation etc. that cause due to Equipment to remote sensing images data, degraded image is reconstructed into close to or completely without the process of the original ideal image of degenerating.Scholar both domestic and external has done a lot of relevant research to staggered tdi sensor image restoration.RaiterS etc. (2003) propose several method based on Block-matching, consider that sensor vibration and atural object move the parity rows changing of the relative positions problem caused simultaneously.What YitahakyY etc. (2003) estimated odd-numbered line image and even number line image respectively falls clear function, then with the parity rows image after block matching method alignment deblurring.HochmanG etc. (2004) optical flow method estimates the displacement between parity rows image, then uses convex set projection (Projectionontoconvexsets, POCS) method deblurring.AvrinA etc. (2006) build local filter by least-squares algorithm, thus realize the alignment of parity rows image.HaikO etc. (2006) think that the displacement of each row pixel of parity rows image can describe by transversal displacement, length travel and the anglec of rotation three parameters, estimate this three parameters, thus realize parity rows image alignment by least square method.Above method is based in the same culture point hypothesis that gray scale is consistent on odd-numbered line image with even number line image, and if parity rows image greyscale there are differences above method by inapplicable.The domestic scholar of having has done relevant research for this reason.Shi Junxia etc. (2012) are based on the method for detecting vibration of quick CCD, in vibration detection technology, the gray projection algorithm that make use of a kind of improvement calculates Vibration Parameter, proposes the method recovered line by line degraded image, and recovers blurred picture.Chen Wangli etc. (2014) are by adopting based on the relevant (Phase-onlycorrelation of phase place, POC) displacement estimation method estimates the sub-pixel precision displacement field between odd-numbered line image and even number line image, point-to-point gray correction is done according to Displacement Estimation result dual numbers row image, and the even number line image after align odd-numbered line image and gray correction.Finally adopt BTV (Bilateraltotalvariation) regularization super resolution ratio reconstruction method counter fall clear, obtain correct image.Above for staggered tdi sensor image restoration method mainly for be due to vibration and atural object move the image degradation phenomenon caused.But but cannot restore because topographic relief and each CCD of parity rows visit the degeneration of the local image quality that first alignment error causes.And also need to remove bar interband overlapping phenomenon and odd even and visit for the airborne aviation TDI raw video that interlocks the pixel aliasing that first overlap sampling causes.

Summary of the invention

For existing restored method be all research cause image degradation for vibration, the invention provides a kind of recovery technique scheme of the airborne staggered TDI infrared image based on geometric correction.

Technical solution of the present invention provides a kind of airborne staggered TDI infrared image restored method based on geometric correction, corrects for the TDI infrared image that interlocks to arbitrary airborne aviation, generates and corrects image, comprise the following steps,

Step 1, interlocks TDI infrared image for raw video with airborne aviation, carries out parity rows separation to raw video, obtains very going independent image and the independent image of even row;

Step 2, definition object space projecting plane, according to the coverage AOP of stripe information determination raw video on object space projecting plane of raw video;

Step 3, determines to correct the projection overlay area AOI of image on object space projecting plane;

Step 4, determines the sampling interval of correcting image according to the spatial resolution of raw video, utilizes sampling interval to be divided into graticule mesh at equal intervals to AOI; Determine that first grid points is current mesh point;

Step 5, tentatively determine current mesh point band, each grid points can only be in two bands at the most simultaneously, if be in two bands simultaneously, if first band residing for strip1 representative, second band residing for strip2 representative, if be only in a band, if strip2=strip1

Step 6, residing for step 5 gained current mesh point, band determines the scope of track data, and the optimum scanning of search current mesh point in each band is capable, obtains the track data of each band residing for current mesh point; Comprise following sub-step,

Step 6.1, according to current mesh point band strip1 ~ strip2, if current search band strip is strip1;

Step 6.2, according to the scan line scope of strip by capable for initial sweep assignment to variable lastline, determine lastline respective scanned behavior current scan line;

Step 6.3, tries to achieve the image space point coordinate (x, y) of certain object space point corresponding according to the elements of exterior orientation of current scan line;

Step 6.4, judges in image space point coordinate (x, y), whether y is less than default limit value,

Then make lastline=lastline+y/scale if not, wherein scale is scale factor; Return step 6.3;

If then enter step 6.5;

Step 6.5, judges whether lastline respective scanned row is within the scope of the scan line of strip,

If it is the lastline tried to achieve is that optimum scanning is capable, add corresponding targeted message, try to achieve the picpointed coordinate astrip of current mesh point raw video corresponding on strip, then judge whether current strip is more than or equal to strip2, enter step 7, otherwise upgrading strip is next reel number, then returns step 6.2;

If not, then make n=n-1, then judge whether that strip is more than or equal to strip2, be, illustrate that current mesh point is in outside image, enter step 8, otherwise renewal strip is next reel number, then returns step 6.2;

Step 7, judges whether that n is greater than 1,

The picpointed coordinate of raw video corresponding on strip1 and strip2 for current mesh point is designated as astrip1 and astrip2 respectively, judges whether astrip1 and astrip2 is in parity rows pixel overlay region,

Utilize 16 pixels the most contiguous with current mesh point to carry out resampling; Described 16 pixels the most contiguous with current mesh point, be for this grid points in the region, band overlay region of raw video and odd even pixel overlay region, the target mesh respectively searched in strip1, strip2 and in the independent image of strange row, the occasionally independent image of row, gets summit totally 16 points of these four target mesh;

Otherwise utilize 8 pixels the most contiguous with current mesh point to carry out resampling; Described 8 pixels the most contiguous with current mesh point, be this grid points in the region, band overlay region of raw video, the target mesh respectively searched in strip1 and strip2, gets summit totally 8 points of these two target mesh;

Otherwise the picpointed coordinate of raw video corresponding on strip1 for current mesh point is designated as astrip, judges whether astrip is in parity rows pixel overlay region,

Utilize 8 pixels the most contiguous with current mesh point to carry out resampling; Described 8 pixels the most contiguous with current mesh point, be this grid points in odd even pixel overlay region, at the independent image of strange row and each corresponding target mesh point of the independent image of even row, get summit totally 8 points of these two target mesh;

Otherwise utilize 4 pixels the most contiguous with current mesh point to carry out resampling; Described 4 pixels the most contiguous with current mesh point, be this grid points in the non-overlapped area of the band of raw video, only in a band, search a target mesh, get summit totally 4 points of this target mesh;

Step 8, judges whether image rectification completes, and then upgrading current mesh point is if not next grid points, forwards step 6 to, if complete, and end loop, Output rusults.

And, in step 5, tentatively determine that current mesh point band implementation is as follows,

First projection overlay area AOI is evenly divided, judge band residing for the upper left corner, zonule and lower right corner grid points, be all same band if residing, judge that band number residing for this region is 1, if two grid points residing different band respectively, then continue to be divided into region, band overlay region and the non-overlapped area of band by evenly dividing gained region;

According to the band divided residing for gained region, and current mesh point region, obtain the band residing for current mesh point.

And in step 7, utilize the most contiguous pixel, after carrying out gray resample by inverse distance weighted method, assignment is to correcting image.

The invention provides a kind of airborne staggered TDI infrared image recovery system based on geometric correction, correct for the TDI infrared image that interlocks to arbitrary airborne aviation, generate and correct image, comprise with lower module,

First module, for interlocking TDI infrared image for raw video with airborne aviation, carries out parity rows separation to raw video, obtains very going independent image and the independent image of even row;

Second module, for defining object space projecting plane, according to the coverage AOP of stripe information determination raw video on object space projecting plane of raw video;

3rd module, for determining to correct the projection overlay area AOI of image on object space projecting plane;

Four module, for determining the sampling interval of correcting image according to the spatial resolution of raw video, utilizes sampling interval to be divided into graticule mesh at equal intervals to AOI; Determine that first grid points is current mesh point;

5th module, for tentatively determining current mesh point band, each grid points can only be in two bands at the most simultaneously, if be in two bands simultaneously, if first band residing for strip1 representative, second band residing for strip2 representative, if be only in a band, if strip2=strip1

6th module, determines the scope of track data for band residing for the 5th module gained current mesh point, and the optimum scanning of search current mesh point in each band is capable, obtains the track data of each band residing for current mesh point; Comprise following submodule,

First submodule, for according to current mesh point band strip1 ~ strip2, if current search band strip is strip1;

Second submodule, for according to the scan line scope of strip by capable for initial sweep assignment to variable lastline, determine lastline respective scanned behavior current scan line;

3rd submodule, for trying to achieve the image space point coordinate (x, y) of certain object space point corresponding according to the elements of exterior orientation of current scan line;

4th submodule, for judging in image space point coordinate (x, y), whether y is less than default limit value,

Then make lastline=lastline+y/scale if not, wherein scale is scale factor; Order the 3rd submodule work;

If then order the 5th submodule work;

5th submodule, for judging whether lastline respective scanned row is within the scope of the scan line of strip,

If it is the lastline tried to achieve is that optimum scanning is capable, add corresponding targeted message, try to achieve the picpointed coordinate astrip of current mesh point raw video corresponding on strip, then judge whether current strip is more than or equal to strip2, order the 7th module work, otherwise upgrading strip is next reel number, then orders the second submodule work;

If not, then make n=n-1, then judge whether that strip is more than or equal to strip2, be, illustrate that current mesh point is in outside image, order the 8th module work, otherwise renewal strip is next reel number, then orders the second submodule work;

7th module, for judging whether that n is greater than 1,

The picpointed coordinate of raw video corresponding on strip1 and strip2 for current mesh point is designated as astrip1 and astrip2 respectively, judges whether astrip1 and astrip2 is in parity rows pixel overlay region,

Utilize 16 pixels the most contiguous with current mesh point to carry out resampling; Described 16 pixels the most contiguous with current mesh point, be for this grid points in the region, band overlay region of raw video and odd even pixel overlay region, the target mesh respectively searched in strip1, strip2 and in the independent image of strange row, the occasionally independent image of row, gets summit totally 16 points of these four target mesh;

Otherwise utilize 8 pixels the most contiguous with current mesh point to carry out resampling; Described 8 pixels the most contiguous with current mesh point, be this grid points in the region, band overlay region of raw video, the target mesh respectively searched in strip1 and strip2, gets summit totally 8 points of these two target mesh;

Otherwise the picpointed coordinate of raw video corresponding on strip1 for current mesh point is designated as astrip, judges whether astrip is in parity rows pixel overlay region,

Utilize 8 pixels the most contiguous with current mesh point to carry out resampling; Described 8 pixels the most contiguous with current mesh point, be this grid points in odd even pixel overlay region, at the independent image of strange row and each corresponding target mesh point of the independent image of even row, get summit totally 8 points of these two target mesh;

Otherwise utilize 4 pixels the most contiguous with current mesh point to carry out resampling; Described 4 pixels the most contiguous with current mesh point, be this grid points in the non-overlapped area of the band of raw video, only in a band, search a target mesh, get summit totally 4 points of this target mesh;

8th module, for judging whether image rectification completes, then upgrading current mesh point is if not next grid points, order the 6th module work, if complete, end loop, Output rusults.

And, in the 5th module, tentatively determine that current mesh point band implementation is as follows,

First projection overlay area AOI is evenly divided, judge band residing for the upper left corner, zonule and lower right corner grid points, be all same band if residing, judge that band number residing for this region is 1, if these two grid points are residing different band respectively, then continue to be divided into region, band overlay region and the non-overlapped area of band by evenly dividing gained region;

According to the band divided residing for gained region, and current mesh point region, obtain the band residing for current mesh point.

And in the 7th module, utilize the most contiguous pixel, after carrying out gray resample by inverse distance weighted method, assignment is to correcting image.

The present invention proposes a kind of recovery technique scheme of the airborne staggered TDI infrared image based on geometric correction, comprises and carries out raw video separation, obtain parity rows independence image; Determine the covering rectangular area AOI correcting image, the sampling interval of correcting image is determined according to the resolution of parity rows independence image, according to sampling interval, graticule mesh is at equal intervals divided into AOI, judge band number scope residing for grid points successively, obtain and be used for gray resample with the most contiguous several sampled points on the odd even independence image corrected corresponding to image grid points, realize removing the square flute profile phenomenon on the overlapping phenomenon of bar interband and raw video.Technical solution of the present invention can make full use of rigorous geometry model and the targeted message of raw video, effective removal parity rows TDICCD is staggered the overlapping phenomenon of square flute profile dislocation that imaging causes and bar interband, and the installation error of CCD can be taken subtly into account, be a kind of infrared image restored method based on geometric correction strictly.

Accompanying drawing explanation

Fig. 1 is the schematic diagram of the embodiment of the present invention based on optimum scanning line search.

Fig. 2 is embodiment of the present invention process flow diagram.

Specific implementation method

Understand for the ease of those of ordinary skill in the art and implement the present invention, below in conjunction with drawings and Examples, the present invention is described in further detail, should be appreciated that exemplifying embodiment described herein is only for instruction and explanation of the present invention, is not intended to limit the present invention.

Embodiment is corrected arbitrary the airborne aviation infrared raw video of TDI that interlocks, generate the correction image eliminating Bowtie effect and staggered effect, correct for pixel aliasing and square flute profile phenomenon between geometric warping, bar interband overlapping phenomenon, parity rows, described raw video adopts the complete machine sweep type imaging of long detector array, be spliced by several scanning strips, the line number of each scanning strip equals the number visiting unit.Such as, the infrared image of a Large visual angle infrared scanner adopts 1024 to visit first sweeping imaging, and the size of a scape raw video is 10240 × 1354, is spliced by about 10 scanning strips, and each scanning arranges with 1024, often row 1354 pixels.Raw video the 1st arranges to the 1024th behavior the 1st scanning strip, and the 1025th row is to the 2048th behavior the 2nd scanning strip, and the rest may be inferred.

See Fig. 2, the method that embodiment provides comprises the following steps, and during concrete enforcement, those skilled in the art can adopt software mode to realize automatically running:

Step.1 raw video is that airborne aviation interlocks TDI infrared image, carries out parity rows be separated and obtain very go independent image and even capable independent image raw video;

The airborne aviation that aviation time delay integration sensor obtains interlock TDI infrared image be parity rows merge after image, there is Bowtie phenomenon and staggered phenomenon in image part subregion.Bowtie phenomenon is bow-tie effect again, is the data overlap phenomenon occurred in a kind of remote sensing images.It is when detector transversal scanning exceedes certain angle, because detector is to the joint effect of the factors such as the shake in the curvature of the visual field geometrical property of earth observation, earth surface, topographic relief and detector motion, between adjacent scan bands ground coverage, there is overlapping phenomenon.In TDI infrared sensor imaging process, except the motion of sensor platform (aircraft), also exist and vibrate frequently, these vibrations comprise the vibration of sensor electronic cooling system, servomotor, platform, and in sweeping imaging process, lens barrel rotates around the axis the vibration etc. caused.Although sensor can consider these factors when Machine Design, and partial vibration can be weakened to a certain extent, can not eliminate completely.These vibration factors make odd-numbered line array and even number line array be in different positions and attitude when obtaining same terrestrial object information, thus cause the staggered phenomenon of image.From airborne aviation interlocks TDI infrared image, sort out parity rows image assist in removing staggered phenomenon.The present invention interlocks TDI infrared image for raw video with airborne aviation.

Step.2 defines object space projecting plane, according to the coverage AOP of the stripe information determination raw video on object space projecting plane (areaofprojection) of raw video;

Under the tangent plan coordinate system be defined in corresponding to imaging region by object space projecting plane, elevation gets photographing region dispersed elevation H.Can set up according to the scanning imagery geometric model of raw video and height value H and correct the mapping relations of image to raw video, calculate the head and the tail pixel of first row on raw video and the projection coordinate of head and the tail pixel on object space projecting plane of last row respectively.Using the view field of calculated four subpoints composition as its coverage AOP on object space projecting plane.

Step.3 determines to correct projection overlay area AOI (areaofprojection) of image on object space projecting plane;

Along heading, determine the minimum gabarit rectangle AOI of AOP, namely correct the projection overlay area of image on object space projecting plane.Because the region of four subpoint compositions may be irregular quadrilateral, therefore maximin in horizontal ordinate and ordinate can be chosen combine, obtain four new subpoints, the view field that four subpoints form is its coverage AOI on object space projecting plane.

Step.4 determines the sampling interval of correcting image according to the spatial resolution of raw video, utilizes sampling interval to be divided into graticule mesh at equal intervals to AOI; Determine that first grid points is current mesh point;

What the sampling interval of correcting image adopted is ground sampling interval, and ground sampling interval is limited to the spatial resolution of scanner.Therefore can determine to correct the sampling interval of image according to the interlock spatial resolution of TDI infrared image of airborne aviation.

Step.5 tentatively determines current mesh point band: judge whether current mesh point is in band overlay region, and determines that its institute is across band;

Determine current mesh point band, according to the coordinate of grid points and the stripe information of raw video, the present invention proposes to divide AOI region further, determines the method for the band residing for each zoning.Defining method is as follows:

First AOI region is evenly divided, during concrete enforcement, those skilled in the art can preset the size of even zoning, and in embodiment, the size of each even division gained zonule is 256 × 256 pixels, and each zonule can comprise multiple net point.Each grid points can only be in two bands at the most simultaneously, if be only in a band, is designated as strip1, if be in two bands simultaneously, if first band residing for strip1 representative, and second band residing for strip2 representative.Judge band residing for the upper left corner, zonule and lower right corner grid points, be all same band if residing, judge that residing for this region, band number is 1, wherein grid points is all in strip1.If two grid points residing different band respectively, then continue to be divided into region, band overlay region and the non-overlapped area of band by evenly dividing gained region.If band is strip1 ~ strip2 residing for band overlapping region, wherein grid points is in strip1 and strip2 all simultaneously, and residing for the non-overlapped district of band, band is strip1, and wherein grid points is all in strip1.According to the band residing for the known each zoning of division result, and zoning, current mesh point place, the band residing for current grid points.

Step.6 band residing for Step.5 gained current mesh point determines the scope of track data, and the optimum scanning of search current mesh point in each band is capable, obtains the track data of each band residing for current mesh point.

Elements of exterior orientation corresponding to object space point is unknown, according to the feature of sweeping, needs the scan line of trying to achieve its correspondence.Search sweep row has a lot of method, first focal plane coordinate this basis for estimation consistent should be visited with imaging CCD in theory based on the picpointed coordinate value calculated, present invention further proposes the method for the linear array sweep type image coordinate back projection calculating based on the constraint of TDI image space, detailed process is as follows:

Step 6.1, according to Step.5 gained current mesh point band strip1 ~ strip2, if current search band strip is strip1, residing band number is n.Because each grid points can only be in two bands at the most simultaneously, n=1 or 2, determines according to the result of step 5.Step 5 is searched for known, and each grid points can only be in two bands at the most simultaneously, if be only in a band, is designated as strip1, i.e. n=1, is simple flow, makes strip2=strip1.If be in two bands simultaneously, if first band residing for strip1 representative, second band, i.e. n=2 residing for strip2 representative.

Step 6.2, according to the scan line scope of strip by capable for initial sweep assignment to variable lastline, determine lastline respective scanned behavior current scan line.During concrete enforcement, can carry out label according to the scan line scope of strip to wherein scan line, the line number that such as initial sweep is capable is 1, by by line number assignment to variable lastline, current scan line can be determined easily.

Step 6.3, brings the elements of exterior orientation of current scan line (lastline respective scanned row) into formula and tries to achieve image space point coordinate (x, y).

Can obtain elements of exterior orientation corresponding to object space point according to current scan line, if Current Scan behavior i-th scan line, the coordinate having certain corresponding object space point is (X _i, Y _i, Z _i), according to formula (1), the image space point coordinate that this object space point is corresponding can be calculated;

$\left[\begin{array}{c}{X}_i\\ Y_i\\ Z_i\end{array}\right]=\left[\begin{array}{c}{X}_{si}\\ Y_{si}\\ Z_{si}\end{array}\right]+\mathrm{\λ}\·\left(R\left[\begin{array}{c}x\\ y\\ -f\end{array}\right]\right)$

Be deformed into,

$\mathrm{\λ}\left[\begin{array}{c}x\\ y\\ -f\end{array}\right]=\left\{\left[\begin{array}{c}{X}_i-X_{si}\\ Y_i-Y_{si}\\ Z_i-Z_{si}\end{array}\right]\right\}\·\left(\begin{array}{c}U\\ V\\ W\end{array}\right)$

Wherein $\left(\begin{array}{c}U\\ V\\ W\end{array}\right)$ For R ^-1

$\left\{\begin{array}{c}\mathrm{\λ}=-W/f\\ x=U/\mathrm{\λ}\\ y=V/\mathrm{\λ}\end{array}\right.---\left(1\right)$

In formula, (X _si, Y _si, Z _si) be i-th scan line projection centre coordinate, R is the rotation matrix of the angle element formation that scan line is corresponding, and U, V, W are R ^-1corresponding three vectors, (x, y) is object space point (X _i, Y _i, Z _i) corresponding picpointed coordinate, f is sensor main distance, and λ is projection coefficient.

Step 6.4, judges that in image space point coordinate (x, y), whether y is less than default limit value, then makes lastline=lastline+y/scale if not, returns step 6.3, if then enter step 6.5.

During concrete enforcement, those skilled in the art can the corresponding limit value of sets itself, and such as get 0.1, unit is pixel.

According to the ultimate principle of linear array sweep type sensor imaging, that adopt during sweeping is TDICCD, and sweeping speed should match with image motion velocity.The capable scan line number differed of picture point and its optimum scanning can be estimated based on this constraint.If TDICCD is spaced apart scale at the TDICCD in sweeping direction, can be described as scale factor.

By current scan line, the image space coordinate y value of trying to achieve, can be considered as current scan line and the capable distance of optimum scanning, according to y and scale factor can be similar to try to achieve this scan line and optimum scanning capable between the number of scanning lines;

lastline＝lastline+y/scale(2)

Step 6.5, judges whether lastline respective scanned row is within the scope of the scan line of strip,

If it is the lastline tried to achieve is that optimum scanning is capable, add targeted message, try to achieve the picpointed coordinate astrip of current mesh point at the upper capable corresponding raw video of optimum scanning of strip (i.e. strip1), then judge whether current strip is more than or equal to strip2, enter Step.7, otherwise upgrading strip is next reel number, then returns step 6.2;

If not, then make n=n-1, then judge whether that strip is more than or equal to strip2, be, illustrate that current mesh point is in outside image, enter Step.8, otherwise renewal strip is next reel number, then returns step 6.2.Special, the band that Step.5 judges is rough judgement, if only searched in a band, optimum scanning is capable illustrates that this grid points is only positioned on a band.N=0, illustrates that current mesh point is in outside image.

Upgrading strip is next reel number, and namely upgrading strip is strip2, within the scope of the scan line of strip2, continue search.If it is capable all to search optimum scanning respectively at strip1 and strip2, the picpointed coordinate astrip of the capable corresponding raw video of current mesh point optimum scanning on strip1 and strip2 is designated as astrip1 and astrip2 respectively.

According to upper type, optimum scanning can be searched out capable by a small amount of iteration, as shown in Figure 1, if A is desired place millet cake, bringing capable the tried to achieve picture point of preliminary sweep into is a1, focal plane (i.e. imaging plane) the coordinate difference dy of unit is visited with imaging CCD, according to scale factor estimation differ from the number of scanning lines, bring collinearity equation into and try to achieve picture point a2, circulation like this is until the picpointed coordinate of trying to achieve is with to visit the focal plane coordinate of unit consistent with imaging CCD, as three search 0,1,2 in Fig. 1, obtain the corresponding y of a3 and be less than limit value, it is capable namely to try to achieve optimum scanning.

In the step 6.5 of embodiment, add targeted message, utilize formula (3) to try to achieve the picpointed coordinate corrected corresponding to image current mesh point, implementation is as follows:

Because the optimum scanning of optimum scanning behavior not containing picture point error of trying to achieve is capable, therefore targeted message need be utilized try to achieve more accurate picpointed coordinate.Calibration is the parameters accurately determining building sensor rigorous geometry model, is mainly divided into interior orientation error parameter and the large class of outer orientation error parameter two.The picture point that the distortion that interior orientation error parameter mainly describes image geometry shape and position causes departs from the error in point measurement of its correct image space, except elements of interior orientation (principal point and phase owner distance) knots modification, also has the optical distortion of camera shooting object lens, the deformation of CCD itself, displacement and rotation etc.Due to the impact of these factors, image picture element produces extruding, stretching, extension, distortion or skew relative to the physical location on ground.Outer orientation error parameter mainly comprises image elements of exterior orientation (position and attitude) error characterising parameter and multisensor relative position relation parameter.In Airborne GPS/IMU system, there is the eccentric vector of GPS and IMU, IMU Collimation axis eccentricity angle and GPS and IMU drift error.Can calculate according to (3)

$R_{ac}^{-1}{R}_{IMU}^{-1}\left(\begin{array}{c}X-X_{GPS}\\ Y-Y_{GPS}\\ Z-X_{GPS}\end{array}\right)+\left(\begin{array}{c}{\mathrm{\ΔX}}_{ac}\\ {\mathrm{\ΔY}}_{ac}\\ {\mathrm{\ΔX}}_{ac}\end{array}\right)=R_{Scan}\left({\mathrm{\θ}}_r,{\mathrm{\θ}}_s,{\mathrm{\θ}}_p\right)(\left(\begin{array}{c}{\mathrm{\ΔX}}_c\\ {\mathrm{\ΔY}}_c\\ {\mathrm{\ΔZ}}_c\end{array}\right)+{\mathrm{\λR}}_C\left(\begin{array}{c}tg\left({\mathrm{\ψ}}_x\right)\\ tg\left({\mathrm{\ψ}}_y\right)\\ 1\end{array}\right))$

$R_{Scan}^{-1}\left({\mathrm{\θ}}_r,{\mathrm{\θ}}_s,{\mathrm{\θ}}_p\right)\left(R_{ac}^{-1}{R}_{IMU}^{-1}\left(\begin{array}{c}X-X_{GPS}\\ Y-Y_{GPS}\\ Z-X_{GPS}\end{array}\right)+\left(\begin{array}{c}{\mathrm{\ΔX}}_{ac}\\ {\mathrm{\ΔY}}_{ac}\\ {\mathrm{\ΔX}}_{ac}\end{array}\right)\right)=\left(\begin{array}{c}{\mathrm{\ΔX}}_c\\ {\mathrm{\ΔY}}_c\\ {\mathrm{\ΔZ}}_c\end{array}\right)+{\mathrm{\λR}}_C\left(\begin{array}{c}tg\left({\mathrm{\ψ}}_x\right)\\ tg\left({\mathrm{\ψ}}_y\right)\\ 1\end{array}\right)$

$\mathrm{\λ}\left(\begin{array}{c}tg\left({\mathrm{\ψ}}_x\right)\\ tg\left({\mathrm{\ψ}}_y\right)\\ 1\end{array}\right)=R_C^{-1}\left(R_{Scan}^{-1}\right({\mathrm{\θ}}_r,{\mathrm{\θ}}_s,{\mathrm{\θ}}_p\left)\right(R_{ac}^{-1}{R}_{IMU}^{-1}\left(\begin{array}{c}X-X_{GPS}\\ Y-Y_{GPS}\\ Z-X_{GPS}\end{array}\right)+\left(\begin{array}{c}{\mathrm{\ΔX}}_{ac}\\ {\mathrm{\ΔY}}_{ac}\\ {\mathrm{\ΔX}}_{ac}\end{array}\right))-\left(\begin{array}{c}{\mathrm{\ΔX}}_c\\ {\mathrm{\ΔY}}_c\\ {\mathrm{\ΔZ}}_c\end{array}\right))\left(\begin{array}{c}U\\ V\\ W\end{array}\right)$

:

$\left\{\begin{array}{c}tg\left({\mathrm{\ψ}}_X\right)=U/W\\ tg\left({\mathrm{\ψ}}_Y\right)=V/W\end{array}\right.$

Be converted to the picpointed coordinate on focal plane:

$\left\{\begin{array}{c}x=f\·tg\left({\mathrm{\ψ}}_X\right)=f\·U/W\\ y=f\·tg\left({\mathrm{\ψ}}_Y\right)=f\·V/W\end{array}\right.---\left(3\right)$

(X, Y, Z) is object space point coordinate;

(X _gPS, Y _gPS, Z _gPS) be GPS phase center coordinate through Correction of Errors;

R _iMUfor the IMU rotation matrix corrected through drift;

Rac is the rotation matrix of base coordinate system and IMU coordinate system;

(Δ Xac, Δ Yac, Δ Zac) is under base coordinate system, the side-play amount of scanning platform rotation center and gps antenna phase center;

R _scan(θ _r, θ _s, θ _p) for scanning the rotation matrix produced, by roll angle θ _r, pitching angle theta _p, the non-perpendicularity θ of pitch axis and the axis of rolling _scommon decision;

(Δ Xc, Δ Yc, Δ Zc) is initial point (i.e. projection centre) skew under scan coordinate system of camera coordinates system;

(x, y) be by step 6 determine the capable corresponding picpointed coordinate of optimum scanning, f is sensor main distance, (ψ _x, ψ _y) be the sensing angle of (x, y).

R _cfor the rotation matrix between camera coordinates system and scan coordinate system;

λ is projection coefficient.

Step.7 judges whether that n is greater than 1,

Be judge astrip1, whether astrip2 is in parity rows pixel overlay region, is, utilizes 16 pixels the most contiguous with current mesh point to carry out resampling, otherwise utilizes 8 pixels the most contiguous with current mesh point to carry out resampling;

Otherwise judge whether astrip is in parity rows pixel overlay region, is, utilize 8 pixels the most contiguous with current mesh point to carry out resampling, otherwise utilize 4 pixels the most contiguous with current mesh point to carry out resampling.

During concrete enforcement, according to arrangement and the corresponding picpointed coordinate of detector pixel, judge the parity rows residing for grid points;

Concrete determination methods is, the photosensitive region of unit and the arrangement of pixel is visited according to parity rows TDICCD, the coordinate visiting unit because parity rows TDICCD is each is all known, therefore corresponding spy unit can be found according to picpointed coordinate, judge that the picture point of required grid points is in strange row CCD photosensitive area and is still in even row CCD photosensitive area or is in the first overlapping photosensitive area of parity rows spy.

According to the picpointed coordinate of tried to achieve grid points information and correspondence thereof, several sampled points that grid points is the most contiguous on residing each band can be tried to achieve;

Utilize and the most contiguous 4 pixels of current mesh point, be after carrying out gray resample by inverse distance weighted method assignment to correcting image.Described 4 pixels the most contiguous with this grid points, be this grid points in the non-overlapped area of the band of raw video, and not in odd even pixel overlay region, only in a band, search a target mesh, get summit totally 4 points of this target mesh.

Utilize and the most contiguous 8 pixels of current mesh point, be after carrying out gray resample by inverse distance weighted method assignment to correcting image.Described 8 pixels the most contiguous with this grid points, be this grid points in the region, band overlay region of raw video, but not in odd even pixel overlay region, the target mesh respectively searched in two bands, gets summit totally 8 points of these two target mesh; Or this grid points is in odd even pixel overlay region, but in the non-overlapped area of band, at the independent image of strange row and each corresponding target mesh point of the independent image of even row, get summit totally 8 points of these two target mesh.

Utilize and the most contiguous 16 pixels of current mesh point, be after carrying out gray resample by inverse distance weighted method assignment to correcting image.Described 16 pixels the most contiguous with this grid points, that this grid points is in the region, band overlay region of raw video and odd even pixel overlay region, in strip1 and strip2, the target mesh respectively searched in the independent image of strange row and the occasionally independent image of row, gets summit totally 16 points of these four target mesh.

Step.8, judges whether image rectification completes, and then upgrading current mesh point is if not next grid points, forwards Step.6 to and continues same process, if complete the process to possessive case site, end loop, Output rusults to next grid points.

During concrete enforcement, computer software technology can be adopted to realize the automatic operation of above flow process, modular mode also can be adopted to provide corresponding system.The embodiment of the present invention provides a kind of airborne staggered TDI infrared image recovery system based on geometric correction, corrects for the TDI infrared image that interlocks to arbitrary airborne aviation, generates and corrects image, comprise with lower module,

First module, for interlocking TDI infrared image for raw video with airborne aviation, carries out parity rows separation to raw video, obtains very going independent image and the independent image of even row;

Second module, for defining object space projecting plane, according to the coverage AOP of stripe information determination raw video on object space projecting plane of raw video;

3rd module, for determining to correct the projection overlay area AOI of image on object space projecting plane;

Four module, for determining the sampling interval of correcting image according to the spatial resolution of raw video, utilizes sampling interval to be divided into graticule mesh at equal intervals to AOI; Determine that first grid points is current mesh point;

5th module, for tentatively determining current mesh point band, each grid points can only be in two bands at the most simultaneously, if be in two bands simultaneously, if first band residing for strip1 representative, second band residing for strip2 representative, if be only in a band, if strip2=strip1

6th module, determines the scope of track data for band residing for the 5th module gained current mesh point, and the optimum scanning of search current mesh point in each band is capable, obtains the track data of each band residing for current mesh point; Comprise following submodule,

First submodule, for according to current mesh point band strip1 ~ strip2, if current search band strip is strip1;

Second submodule, for according to the scan line scope of strip by capable for initial sweep assignment to variable lastline, determine lastline respective scanned behavior current scan line;

3rd submodule, for trying to achieve the image space point coordinate (x, y) of certain object space point corresponding according to the elements of exterior orientation of current scan line;

4th submodule, for judging in image space point coordinate (x, y), whether y is less than default limit value,

Then make lastline=lastline+y/scale if not, wherein scale is scale factor; Order the 3rd submodule work;

If then order the 5th submodule work;

5th submodule, for judging whether lastline respective scanned row is within the scope of the scan line of strip,

If it is the lastline tried to achieve is that optimum scanning is capable, add corresponding targeted message, try to achieve the picpointed coordinate astrip of current mesh point raw video corresponding on strip, then judge whether current strip is more than or equal to strip2, order the 7th module work, otherwise upgrading strip is next reel number, then orders the second submodule work;

If not, then make n=n-1, then judge whether that strip is more than or equal to strip2, be, illustrate that current mesh point is in outside image, order the 8th module work, otherwise renewal strip is next reel number, then orders the second submodule work;

7th module, for judging whether that n is greater than 1,

The picpointed coordinate of raw video corresponding on strip1 and strip2 for current mesh point is designated as astrip1 and astrip2 respectively, judges whether astrip1 and astrip2 is in parity rows pixel overlay region,

Utilize 16 pixels the most contiguous with current mesh point to carry out resampling; Described 16 pixels the most contiguous with current mesh point, be for this grid points in the region, band overlay region of raw video and odd even pixel overlay region, the target mesh respectively searched in strip1, strip2 and in the independent image of strange row, the occasionally independent image of row, gets summit totally 16 points of these four target mesh;

Otherwise utilize 8 pixels the most contiguous with current mesh point to carry out resampling; Described 8 pixels the most contiguous with current mesh point, be this grid points in the region, band overlay region of raw video, the target mesh respectively searched in strip1 and strip2, gets summit totally 8 points of these two target mesh;

Otherwise the picpointed coordinate of raw video corresponding on strip1 for current mesh point is designated as astrip, judges whether astrip is in parity rows pixel overlay region,

Utilize 8 pixels the most contiguous with current mesh point to carry out resampling; Described 8 pixels the most contiguous with current mesh point, be this grid points in odd even pixel overlay region, at the independent image of strange row and each corresponding target mesh point of the independent image of even row, get summit totally 8 points of these two target mesh;

Otherwise utilize 4 pixels the most contiguous with current mesh point to carry out resampling; Described 4 pixels the most contiguous with current mesh point, be this grid points in the non-overlapped area of the band of raw video, only in a band, search a target mesh, get summit totally 4 points of this target mesh;

8th module, for judging whether image rectification completes, then upgrading current mesh point is if not next grid points, order the 6th module work, if complete, end loop, Output rusults.

Each module realizes can see corresponding steps, and it will not go into details in the present invention.

Specific embodiment described herein is only to the explanation for example of the present invention's spirit.Those skilled in the art can make various amendment or supplement or adopt similar mode to substitute to described specific embodiment, but can't depart from spirit of the present invention or surmount the scope that appended claims defines.

Claims (6)

1. based on an airborne staggered TDI infrared image restored method for geometric correction, correct for the TDI infrared image that interlocks to arbitrary airborne aviation, generate and correct image, it is characterized in that: comprise the following steps,

Step 1, interlocks TDI infrared image for raw video with airborne aviation, carries out parity rows separation to raw video, obtains very going independent image and the independent image of even row;

Step 2, definition object space projecting plane, according to the coverage AOP of stripe information determination raw video on object space projecting plane of raw video;

Step 3, determines to correct the projection overlay area AOI of image on object space projecting plane;

Step 4, determines the sampling interval of correcting image according to the spatial resolution of raw video, utilizes sampling interval to be divided into graticule mesh at equal intervals to AOI; Determine that first grid points is current mesh point;

Step 5, tentatively determine current mesh point band, each grid points can only be in two bands at the most simultaneously, if be in two bands simultaneously, if first band residing for strip1 representative, second band residing for strip2 representative, if be only in a band, if strip2=strip1

Step 6, residing for step 5 gained current mesh point, band determines the scope of track data, and the optimum scanning of search current mesh point in each band is capable, obtains the track data of each band residing for current mesh point; Comprise following sub-step,

Step 6.1, according to current mesh point band strip1 ~ strip2, if current search band strip is strip1;

Step 6.2, according to the scan line scope of strip by capable for initial sweep assignment to variable lastline, determine lastline respective scanned behavior current scan line;

Step 6.3, tries to achieve the image space point coordinate (x, y) of certain object space point corresponding according to the elements of exterior orientation of current scan line;

Step 6.4, judges in image space point coordinate (x, y), whether y is less than default limit value,

Then make lastline=lastline+y/scale if not, wherein scale is scale factor; Return step 6.3;

If then enter step 6.5;

Step 6.5, judges whether lastline respective scanned row is within the scope of the scan line of strip,

If it is the lastline tried to achieve is that optimum scanning is capable, add corresponding targeted message, try to achieve the picpointed coordinate astrip of current mesh point raw video corresponding on strip, then judge whether current strip is more than or equal to strip2, enter step 7, otherwise upgrading strip is next reel number, then returns step 6.2;

If not, then make n=n-1, then judge whether that strip is more than or equal to strip2, be, illustrate that current mesh point is in outside image, enter step 8, otherwise renewal strip is next reel number, then returns step 6.2;

Step 7, judges whether that n is greater than 1,

The picpointed coordinate of raw video corresponding on strip1 and strip2 for current mesh point is designated as astrip1 and astrip2 respectively, judges whether astrip1 and astrip2 is in parity rows pixel overlay region,

Utilize 16 pixels the most contiguous with current mesh point to carry out resampling; Described 16 pixels the most contiguous with current mesh point, be for this grid points in the region, band overlay region of raw video and odd even pixel overlay region, the target mesh respectively searched in strip1, strip2 and in the independent image of strange row, the occasionally independent image of row, gets summit totally 16 points of these four target mesh;

Otherwise utilize 8 pixels the most contiguous with current mesh point to carry out resampling; Described 8 pixels the most contiguous with current mesh point, be this grid points in the region, band overlay region of raw video, the target mesh respectively searched in strip1 and strip2, gets summit totally 8 points of these two target mesh;

Otherwise the picpointed coordinate of raw video corresponding on strip1 for current mesh point is designated as astrip, judges whether astrip is in parity rows pixel overlay region,

Utilize 8 pixels the most contiguous with current mesh point to carry out resampling; Described 8 pixels the most contiguous with current mesh point, be this grid points in odd even pixel overlay region, at the independent image of strange row and each corresponding target mesh point of the independent image of even row, get summit totally 8 points of these two target mesh;

Otherwise utilize 4 pixels the most contiguous with current mesh point to carry out resampling; Described 4 pixels the most contiguous with current mesh point, be this grid points in the non-overlapped area of the band of raw video, only in a band, search a target mesh, get summit totally 4 points of this target mesh;

Step 8, judges whether image rectification completes, and then upgrading current mesh point is if not next grid points, forwards step 6 to, if complete, and end loop, Output rusults.

2. according to claim 1 based on the airborne staggered TDI infrared image restored method of geometric correction, it is characterized in that: in step 5, tentatively determine that current mesh point band implementation is as follows,

First projection overlay area AOI is evenly divided, judge band residing for the upper left corner, zonule and lower right corner grid points, be all same band if residing, judge that band number residing for this region is 1, if these two grid points are residing different band respectively, then continue to be divided into region, band overlay region and the non-overlapped area of band by evenly dividing gained region;

According to the band divided residing for gained region, and current mesh point region, obtain the band residing for current mesh point.

3. according to claim 1 or 2 based on the airborne staggered TDI infrared image restored method of geometric correction, it is characterized in that: in step 7, utilize the most contiguous pixel, after carrying out gray resample by inverse distance weighted method, assignment is to correcting image.

4. based on an airborne staggered TDI infrared image recovery system for geometric correction, correct for the TDI infrared image that interlocks to arbitrary airborne aviation, generate and correct image, it is characterized in that: comprise with lower module,

First module, for interlocking TDI infrared image for raw video with airborne aviation, carries out parity rows separation to raw video, obtains very going independent image and the independent image of even row;

Second module, for defining object space projecting plane, according to the coverage AOP of stripe information determination raw video on object space projecting plane of raw video;

3rd module, for determining to correct the projection overlay area AOI of image on object space projecting plane;

Four module, for determining the sampling interval of correcting image according to the spatial resolution of raw video, utilizes sampling interval to be divided into graticule mesh at equal intervals to AOI; Determine that first grid points is current mesh point;

5th module, for tentatively determining current mesh point band, each grid points can only be in two bands at the most simultaneously, if be in two bands simultaneously, if first band residing for strip1 representative, second band residing for strip2 representative, if be only in a band, if strip2=strip1

6th module, determines the scope of track data for band residing for the 5th module gained current mesh point, and the optimum scanning of search current mesh point in each band is capable, obtains the track data of each band residing for current mesh point; Comprise following submodule,

First submodule, for according to current mesh point band strip1 ~ strip2, if current search band strip is strip1;

Second submodule, for according to the scan line scope of strip by capable for initial sweep assignment to variable lastline, determine lastline respective scanned behavior current scan line;

3rd submodule, for trying to achieve the image space point coordinate (x, y) of certain object space point corresponding according to the elements of exterior orientation of current scan line;

4th submodule, for judging in image space point coordinate (x, y), whether y is less than default limit value,

Then make lastline=lastline+y/scale if not, wherein scale is scale factor; Order the 3rd submodule work;

If then order the 5th submodule work;

5th submodule, for judging whether lastline respective scanned row is within the scope of the scan line of strip,

If it is the lastline tried to achieve is that optimum scanning is capable, add corresponding targeted message, try to achieve the picpointed coordinate astrip of current mesh point raw video corresponding on strip, then judge whether current strip is more than or equal to strip2, order the 7th module work, otherwise upgrading strip is next reel number, then orders the second submodule work;

If not, then make n=n-1, then judge whether that strip is more than or equal to strip2, be, illustrate that current mesh point is in outside image, order the 8th module work, otherwise renewal strip is next reel number, then orders the second submodule work;

7th module, for judging whether that n is greater than 1,

The picpointed coordinate of raw video corresponding on strip1 and strip2 for current mesh point is designated as astrip1 and astrip2 respectively, judges whether astrip1 and astrip2 is in parity rows pixel overlay region,

Utilize 16 pixels the most contiguous with current mesh point to carry out resampling; Described 16 pixels the most contiguous with current mesh point, be for this grid points in the region, band overlay region of raw video and odd even pixel overlay region, the target mesh respectively searched in strip1, strip2 and in the independent image of strange row, the occasionally independent image of row, gets summit totally 16 points of these four target mesh;

Otherwise utilize 8 pixels the most contiguous with current mesh point to carry out resampling; Described 8 pixels the most contiguous with current mesh point, be this grid points in the region, band overlay region of raw video, the target mesh respectively searched in strip1 and strip2, gets summit totally 8 points of these two target mesh;

Otherwise the picpointed coordinate of raw video corresponding on strip1 for current mesh point is designated as astrip, judges whether astrip is in parity rows pixel overlay region,

Utilize 8 pixels the most contiguous with current mesh point to carry out resampling; Described 8 pixels the most contiguous with current mesh point, be this grid points in odd even pixel overlay region, at the independent image of strange row and each corresponding target mesh point of the independent image of even row, get summit totally 8 points of these two target mesh;

Otherwise utilize 4 pixels the most contiguous with current mesh point to carry out resampling; Described 4 pixels the most contiguous with current mesh point, be this grid points in the non-overlapped area of the band of raw video, only in a band, search a target mesh, get summit totally 4 points of this target mesh;

8th module, for judging whether image rectification completes, then upgrading current mesh point is if not next grid points, order the 6th module work, if complete, end loop, Output rusults.

5. according to claim 4 based on the airborne staggered TDI infrared image recovery system of geometric correction, it is characterized in that: in the 5th module, tentatively determine that current mesh point band implementation is as follows,

First projection overlay area AOI is evenly divided, judge band residing for the upper left corner, zonule and lower right corner grid points, be all same band if residing, judge that band number residing for this region is 1, if these two grid points are residing different band respectively, then continue to be divided into region, band overlay region and the non-overlapped area of band by evenly dividing gained region;

According to the band divided residing for gained region, and current mesh point region, obtain the band residing for current mesh point.

6. according to claim 4 or 5 based on the airborne staggered TDI infrared image recovery system of geometric correction, it is characterized in that: in the 7th module, utilize the most contiguous pixel, after carrying out gray resample by inverse distance weighted method, assignment is to correcting image.