CN102708386A - Optical/SAR (synthetic aperture radar) heterogeneous image matching method - Google Patents

Abstract

The invention provides an optical/SAR (synthetic aperture radar) heterogeneous image matching method and aims to provide a heterogeneous image matching method which is characterized in that the matching and searching real-time performance is strong and the SAR real-time image is free from two-valued or multi-valued processing. According to the technical scheme of the invention, the optical/SAR heterogeneous image matching method comprises the following steps: selecting a variety of surface-feature areas from a historically-accumulated millimeter-wave SAR image to count grayscales, and constructing a millimeter-wave SAR surface-feature grayscale characteristic database; selecting a plurality of interested effective matching areas from an optical remote-sensing image, classifying surface features and segmenting the areas; filling different surface-feature segmentation areas with different grayscales, and acquiring a multi-valued reference graph of the areas; and directly carrying out grayscale normalization cross-correlation matching by using the millimeter-wave SAR real-time graph and the multi-valued reference graph of the areas, wherein the multi-valued reference graph of the areas is made by the optical remote-sensing image; pre-processing the correlation matching of the multi-valued reference graph of the areas, and processing the correlation matching of the reference graph and the real-time graph in real time. According to the optical/SAR (synthetic aperture radar) heterogeneous image matching method provided by the invention, the problems of inaccurate real-time image segmentation and mismatching caused by binarization can be solved.

Description

Optics/SAR foreign peoples's image matching method

Technical field

The present invention relates to a kind of optics/SAR foreign peoples's image matching method, particularly relate to remote optical sensing reference map and SAR radar foreign peoples's image matching method of figure in real time based on the many-valued reference map in zone.

Background technology

The mode that Missile Terminal Guidance target seeker information is obtained has synthetic-aperture radar (SAR, Synthetic Aperture Radar), infrared, optical sensor etc.The real-time figure that the target seeker utilization is obtained in real time carries out related operation with the reference map of storage in advance; Realize scene matching aided navigation; Thereby obtain the physical location of target in scheming in real time by the priori position reckoning of target in reference map; Thereby, revise flight error with this, pinpointing according to the positional information that real-time figure obtains target.Advantages such as SAR imaging terminal guidance is round-the-clock with it, round-the-clock, antijamming capability are strong have become the research focus of precise guidance.

The U.S. has used battleax cruise missile BGM-109 to attack Iraq's strategic objective in the Gulf War in 1992 first, and this indicates the arriving in precision guided weapon epoch.The battleax cruise missile adopts the about 30m of circular proable error of (for example BCM-109A) under inertia and the topographic contour coupling combined guidance mode, and precision is to several meters during as if employing scape picture coupling terminal guidance (for example BGM-109C).Meanwhile, the USSR (Union of Soviet Socialist Republics) has has also researched and developed same weapon, for example SS-NX-21 and AS-X-15 cruise missile.

The Pan Xing II guided missile of U.S.'s development is a kind of intermediate range ballistic missile, and effective range reaches 2500 kilometers, and attack precision CEP is 30 meters.The Pan Xing II Missile Terminal Guidance course of work is following.Real aperture radar antenna from body apart from ground 10km about beginning enclose conical scanning over the ground with per second two, realize imaging on a surface target.The radar beam orientation is to wide 2.2 °, and is radially wide 22 °, on differing heights, carries out the imaging Matching Location three to four times.

The scene matching aided navigation of scheming in real time based on millimeter-wave SAR is one of target search key technique in identification in the SAR imaging terminal guidance.Because the shortage of China's High Resolution SAR remote sensing image resource, make that utilizing the High Resolution SAR image to carry out similar image as reference map matees that often the combat support difficulty is big, thereby we adopt usually optical remote sensing image to prepare reference map.Because optical sensor is different with the imaging mechanism of SAR radar, gamma characteristic is often opposite sometimes, thereby need just can mate after the denominator of two types of images that different sensors obtains of extraction.The present invention is exactly according to the millimeter-wave SAR imaging characteristics, with optical remote sensing image be prepared into the approaching reference map of SAR image after could with SAR in real time figure mate.

The big figure in real time of little reference map and big two kinds of schemes of the little real-time figure of reference map are arranged in existing optics/SAR foreign peoples's image matching method.The former operand is little but preparation requires very high to reference map; Reference map according to optical imagery preparation requires with figure is closely similar in real time, though through radar imagery mechanism realize optics to the emulation of SAR but since high being difficult to of the combat support difficulty of peopleware, device resource in reality is fought, fast and effeciently implement; The latter matees reliability and accuracy height but operand is bigger, need be optimized search and matching algorithm to satisfy real-time treatment requirement.

The preparation method of existing optical reference figure, often with reference map with mate after figure is prepared into bianry image simultaneously in real time, and scheme in real time onlinely to cut apart often inaccurately, sometimes even the situation of mismatch occurs, so binaryzation is worthless.

Summary of the invention

The weak point that the objective of the invention is above-mentioned prior art; Provide a kind of easy to use; Be easy to grasp; Can prepare rapidly and accurately in optical reference figure, real-time, the real-time matching process of match search need to SAR in real time figure carry out two-value or many-valuedization pre-service, practicality is optics/SAR foreign peoples's image matching method fast reliably, to solve the problem of object matching identification in the millimeter-wave SAR guidance.

To achieve these goals, the matching process of a kind of optics/SAR foreign peoples's image comprises the steps:

1) utilizes Photoshop software, in the millimeter-wave SAR image of history accumulation, choose all kinds of ground object area and carry out gray-scale statistical, make up millimeter-wave SAR atural object gray feature database;

2) in optical remote sensing image, choose a plurality of effective matching areas interested, and each effective matching area is carried out terrain classification and Region Segmentation;

3) according to the gray-scale statistical characteristic of atural object different atural object cut zone are filled different gray scales; Utilize all kinds of atural object gray-scale statistical empirical values in the millimeter-wave SAR gray feature database; Segmentation result image to each effective matching area carries out the filling of many-valued zone, obtains regional many-valued reference map;

4) the regional many-valued reference map and the real-time figure of millimeter-wave SAR that utilize optical remote sensing image to be prepared into directly carry out gray scale normalization simple crosscorrelation coupling; Earlier the many-valued reference map in zone is carried out the relevant matches pre-service; Carrying out reference map then handles with the relevant matches of scheming in real time in real time; And earlier said reference figure and real-time figure are carried out pyramid decomposition, on real-time figure of lowest hierarchical level and reference map basis, carry out interlacing at a distance from the row search matched.

Above-mentioned relevant matches pre-service is: the many-valued reference map in zone is carried out pyramid decomposition; Calculate the part that to calculate in advance with scheming in real time to have nothing to do in the lowest hierarchical level relevant matches; Comprise that reference map goes average square data, and with the quadratic sum data of all benchmark subgraphs of real-time figure with size.

Above-mentioned relevant matches is handled in real time: earlier real-time figure is carried out pyramid decomposition, calculate the part that can calculate in advance before the lowest hierarchical level relevant matches traversal search; Carry out interlacing at a distance from being listed as search and going average normalized crosscorrelation coupling in lowest hierarchical level then, the top in the correlation surface is thick matched position; Turn back to the upper level reference map at last step by step and scheme to carry out the essence coupling in real time, highest master reference figure is the Optimum Matching position with the smart matched position of real-time figure.

The present invention has following beneficial effect than prior art:

Regional many-valued reference map and millimeter-wave SAR radar image that the present invention utilizes optical remote sensing image to be prepared into directly carry out gray scale normalization simple crosscorrelation coupling; Need not scheme in real time to carry out two-value or many-valuedization pre-service in the matching process in real time, reduce real-time figure is cut apart the inaccurate mismatch problems of bringing with binaryzation SAR.

The instrument that the present invention adopts business software Photoshop ripe on the market to carry carries out the reference map preparation, and is easy to use, is easy to grasp, and can prepare reference map rapidly and accurately.

The present invention has adopted pyramid decomposition and interlacing at a distance from the search strategy that is listed as, and earlier reference map is carried out pyramid decomposition with real-time figure, on real-time figure of lowest hierarchical level and reference map basis, carries out interlacing at a distance from the row search matched.This search strategy has improved match search speed greatly.

The present invention schemes to have adopted by thick to smart matching strategy to many-valued optical reference figure in zone and SAR in real time; Earlier in the basic enterprising line search coupling of the little figure of lowest hierarchical level; Find the thick matched position of lowest hierarchical level; Progressively near the thick matched position of the big figure of high-level, carry out the essence coupling then, the smart matching result of the big figure of highest level is final matching results.Thisly reduced the matching operation time widely by thick matching strategy and do not lost matching precision again to essence.

The optimization computing strategy of quick coupling has been adopted in the real-time processing that many-valued optical reference figure in zone and SAR scheme to mate in real time; Except above-mentioned search strategy can improve the matching treatment real-time greatly; Can precalculated operand in the time of also will mating at every turn as far as possible in advance, pre-service is placed on before the real-time processing like reference map, that has calculated reference map earlier removes average square and each benchmark subgraph quadratic sum two-dimensional array; Before the pre-service of figure in real time is placed on the search matched circulation; Calculate earlier go average square and each level of good figure in real time and scheme quadratic sum in real time, so each search matched round-robin computation amount, thus shortened each coupling computing time greatly.

The present invention fully excavates the denominator of optical imagery and millimeter-wave SAR image; Utilize regional many-valued reference map and SAR to scheme to carry out scene matching aided navigation in real time; Realize the coupling of optics/SAR foreign peoples's image; Have high reliability and the characteristics of handling in real time, make target seeker possess the function of identification and localizing objects from the millimeter-wave SAR image.Solved at High Resolution SAR imaging terminal guidance target seeker scape and adopted the coupling of optical reference figure to guide problem owing to lack High Resolution SAR image reference map in as matching technique, its real-time processing method has very high engineering practical value.

The present invention adopts the matching process of big reference map and little real-time figure; The high reliability and the pinpoint accuracy of relevant matches have been ensured; In order to reduce the difficulty of optical reference figure preparation, improve preparation efficiency simultaneously, and don't reduce reliability and matching precision; Creatively proposed the notion of regional many-valued reference map, selected characteristic a plurality of effective matching area obvious, that be easy to prepare prepares reference map; The present invention proposes the method for many-valued reference map, prepare many-valued reference map, thereby directly carry out relevant matches with real-time figure according to optical reference figure terrain classification split image; The present invention is also consuming time huge and be difficult to realize the characteristics handled in real time to existing big reference map and little real-time figure traversal search; Pyramid decomposition and the interlacing search strategy at a distance from row has been proposed; Simultaneously the correlation matching algorithm computing method are optimized, thereby have improved the coupling real-time greatly.The invention solves owing to lacking the coupling guidance problem that High Resolution SAR image reference map adopts optical reference figure.

Description of drawings

In order more to be expressly understood the present invention, will simultaneously with reference to accompanying drawing, the present invention be described through embodiment of the present invention at present, wherein:

Fig. 1 is based on the scene matching aided navigation treatment step block diagram of regional many-valued optical reference figure and the real-time figure of millimeter-wave SAR.

Fig. 2 is the practical implementation process flow diagram of Fig. 1.

Fig. 3 is certain airport optical remote sensing image figure.

Fig. 4 is that certain airport millimeter-wave SAR is schemed in real time.

Fig. 5 is the regional many-valued reference map for preparing according to Fig. 3.

Fig. 6 carries out the matching effect figure after normalized crosscorrelation matees to Fig. 4 and Fig. 5.

Fig. 7 is that step S1 makes up millimeter-wave SAR image atural object gray feature database flowchart among Fig. 1.

Fig. 8 is the process flow diagram that the quick multizone of step S2 is cut apart among Fig. 1.

Fig. 9 is the process flow diagram that the many-valued gray scale in step S3 zone is filled among Fig. 1.

Figure 10 is the pretreated process flow diagram of step S4 relevant matches among Fig. 1.

Figure 11 is the process flow diagram that step S5 relevant matches is handled in real time among Fig. 1.

Embodiment

For a better understanding of the present invention, at first the algorithm principle and the optimized calculation method of removing average normalized crosscorrelation matching algorithm are introduced.Alleged reference map is regional many-valued reference map among the present invention, and alleged real-time figure is that millimeter-wave SAR is schemed in real time.

1) algorithm principle: when real-time figure travels through in the possible range in reference map; Keep figure in real time and on each possible position of reference map, remove average normalized crosscorrelation metric; The set of all cross correlation values forms correlation surface, and the maximal value note in the relevant peaks of correlation surface is done the top.Usually pairing position, top is the matched position of real-time figure in reference map.

Known reference figure X is of a size of M * N, schemes Y in real time and is of a size of m * n, goes the related coefficient computing formula of average normalized crosscorrelation matching algorithm to be:

$r_k=\frac{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}(x_{\mathrm{ij}}-{\stackrel{\‾}{x}}_k)(y_{\mathrm{ij}}-\stackrel{\‾}{y})}{\sqrt{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{(x_{\mathrm{ij}}-{\stackrel{\‾}{x}}_k)}^2}\sqrt{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{(y_{\mathrm{ij}}-\stackrel{\‾}{y})}^2}}---\left(1\right)$

In the formula:

x _IjBe k reference map subgraph X _kIn the gray-scale value of capable, the j row pixel of i;

Be k reference map subgraph X _kThe average of all pixels;

y _IjThe gray-scale value of, j row pixel capable for i among the real-time figure Y;

is the average of scheming all pixels of Y in real time.

The position of the corresponding subgraph of related coefficient maximum value on reference map is best match position.

2) search strategy optimization: the simple crosscorrelation matching algorithm is actually real-time figure traversal search correct match point on reference diagram.Can carry out pyramid decomposition with real-time figure to reference map, downscaled images is carried out multistage coupling and is calculated step by step, generally divides two or three layer just can.In the lowest hierarchical level image, at a distance from the row search, calculate correlation surface earlier, obtain tower decomposition bottom and search for thick matched position, return the real-time figure and the reference map of the tower decomposition of upper level then, in thick coupling neighborhood of a point, carry out the essence coupling through interlacing.

3) matching algorithm optimization: calculate correlation coefficient r _kValue can be converted into and calculate related coefficient square

$r_k^2=\frac{{\left(\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}(x_{\mathrm{ij}}-{\stackrel{\‾}{x}}_k)(y_{\mathrm{ij}}-\stackrel{\‾}{y})\right)}^2}{\left(\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{(x_{\mathrm{ij}}-{\stackrel{\‾}{x}}_k)}^2\right)\left(\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{(y_{\mathrm{ij}}-\stackrel{\‾}{y})}^2\right)}---\left(2\right)$

Wherein:

pixel quadratic sum for removing average benchmark subgraph; Calculate through reference map, can before cyclic search, calculate.At first calculate many-valued reference map amplitude square array, calculate the quadratic sum of first benchmark subgraph then, in the match search process, current subgraph quadratic sum can just can by the value of last subgraph quadratic sum increase and minimizing delegation one row pixel quadratic sum again;

is the plain quadratic sum of realtime graphic; Calculate through real-time figure; It and each reference map subgraph are all constant when mating, and can before cyclic search, calculate;

is the benchmark subgraph and the cross correlation value of figure in real time, when each cyclic search, calculate subgraph and the cross correlation value of figure in real time.

It is thus clear that; The calculating of matching algorithm; The value that needing to calculate

and

real-time coupling in advance only needs to calculate

gets final product, and has significantly reduced the time of real-time calculating.

For a better understanding of the present invention, introduce method here, mainly realize according to following flow process based on the scene matching aided navigation of many-valued optical reference figure in zone and the real-time figure of millimeter-wave SAR:

Consult Fig. 1.The preparation of optical reference figure is not that all atural object scenes among the optical reference figure are prepared; But the tangible a plurality of effective matching areas of selected characteristic prepare; Like airfield runway and lawn, near harbour seashore and the harbour thereof, highway and grade separation thereof, vegetation along the road; Lake, river and bridge thereof or the like can prepare regional many-valued reference map rapidly and accurately.

The preparation of optical reference figure is not simple two-value reference map preparation, but effective matching area is carried out terrain classification and the preparation of many-valued reference map, according to the gray-scale statistical characteristic of atural object different atural object cut zone is filled corresponding gray scale.

The matching process of optics/SAR foreign peoples's image comprises that mainly structure millimeter-wave SAR atural object gray feature database S1, quick multizone are cut apart S2, regional many-valued gray scale is filled S3, relevant matches pre-service S4,5 five steps of the real-time treatment S of relevant matches.

Step S1 utilizes the manual work in the millimeter-wave SAR image of history accumulation of Photoshop software to choose all kinds of ground object area and carries out gray-scale statistical, makes up all kinds of atural object gray feature databases;

Step S2, in optical remote sensing image, choose a plurality of effective matching areas interested and to each effective matching area carry out terrain classification with cut apart;

Step S3 utilizes all kinds of atural object gray-scale statistical empirical values, the segmentation result image of each effective matching area is carried out many-valued zone fill, and obtains regional many-valued reference map;

Step S4, the regional many-valued reference map that optical remote sensing image is prepared into carries out the relevant matches pre-service, comprises regional many-valued reference map pyramid decomposition, calculates reference map and goes the average square number to reach the quadratic sum data of all benchmark subgraphs according to this;

Step S5 carries out normalized crosscorrelation coupling and handles in real time, comprises real-time figure pyramid decomposition, and calculating real-time figure goes the average quadratic sum, and interlacing is mated to smart at a distance from the row search with by thick.

Consult Fig. 2.According to the inventive method, the treatment scheme of the matching process of optics/SAR foreign peoples's image is following:

1) makes up millimeter-wave SAR atural object gray feature database: utilize the manual work in the millimeter-wave SAR image of history accumulation of Photoshop software to choose all kinds of ground object area and carry out gray-scale statistical, make up atural object gray feature database;

2) multizone is cut apart fast: a plurality of effective matching areas are chosen in manual work in optical remote sensing image; In each effective matching area, carry out Region Segmentation then, comprise and manually cut apart and cut apart automatically dual mode, if effectively matching area is that even matter zone of gray scale and intensity contrast are obvious; Can directly carry out automatic threshold cuts apart; If effectively the matching area gray feature is complicated, can utilize Photoshop software manually to delineate the zone boundary, carry out manual work and cut apart;

3) regional many-valued filling: choose in the millimeter-wave SAR gray feature database corresponding atural object gray-scale statistical empirical value and the segmentation result image of each effective matching area is carried out many-valued zone fill; To the whole zero fillings of invalid matching area, obtain regional many-valued reference map;

4) relevant matches pre-service: the many-valued reference map in zone is carried out pyramid decomposition; Calculate the part that to calculate in advance with scheming in real time to have nothing to do in the lowest hierarchical level relevant matches; Comprise that reference map goes average square data, and with the quadratic sum data of all benchmark subgraphs of real-time figure with size.

5) relevant matches is handled in real time: relevant matches is handled in real time: earlier real-time figure is carried out pyramid decomposition, calculate the part that can calculate in advance before the lowest hierarchical level relevant matches traversal search; Carry out interlacing at a distance from being listed as search and going average normalized crosscorrelation coupling in lowest hierarchical level then, the top in the correlation surface is thick matched position; Turn back to the upper level reference map at last step by step and scheme to carry out the essence coupling in real time, highest master reference figure is the Optimum Matching position with the smart matched position of real-time figure.

In order to understand the present invention better, below Fig. 8, Fig. 9, Figure 10, Figure 11 totally four width of cloth images method based on optics/SAR foreign peoples's image coupling of the many-valued reference map in zone is described.

Consult Fig. 3, this figure is near the original optical remote sensing image waiting to strike target, and is used to prepare regional many-valued reference map.

Consult Fig. 4, near the millimeter-wave SAR radar image that this figure target seeker obtains in real time waiting to strike target, just millimeter-wave SAR is schemed in real time.

Consult Fig. 5, this figure is the regional many-valued reference map that utilizes reference map preparation method of the present invention to produce according to Fig. 8, can directly carry out the normalized crosscorrelation matching operation with the real-time figure of millimeter-wave SAR.

Consult Fig. 6, this figure utilizes method of the present invention that the many-valued reference map in zone shown in Figure 5 and the real-time figure of millimeter-wave SAR shown in Figure 4 are directly carried out the matching effect figure after the normalized crosscorrelation matching operation.

Consult Fig. 7.Near SAR remote sensing image step S11 utilizes PhotoShop software to open to wait to strike target; Can be multiple forms such as bmp, tiff, gif, jpg; It also can be self-defining raw picture format; Data for self-defined raw form also will be removed file header, and the line number and the columns of definition image could correctly open file.

Step S12 utilizes choosing of Photoshop software frame or polygon enclosed region to choose the zone of representative invariant feature, like waters, airfield runway, lawn, forest or the like to different types of ground objects.

Step S13 opens the average statistical that the histogram extended view can show institute's favored area in real time.

Step S14 is input to the average statistical of dissimilar atural objects in the database and gets final product.

Consult Fig. 8.Step S21 utilizes PhotoShop software to open and waits to strike target near optical remote sensing image; Can be multiple forms such as bmp, tiff, gif, jpg; It also can be self-defining raw picture format; Data for self-defined raw form also will be removed file header, and the line number and the columns of definition image could correctly open file.

Step S22; Choose significantly a plurality of effective matching areas of characters of ground object, effectively matching area is generally chosen the strong even matter zone of contrast, like airfield runway and near lawn thereof; Highway and viaduct thereof, roadside vegetation; Port and pier and waters thereof, river and riverbank, vacant lot or the like around large-scale typical buildings reaches.

Step S23 carries out Threshold Segmentation, the Threshold Segmentation function that can adopt Photoshop to carry to the even effective matching area of matter; According to grey level histogram; Threshold value is chosen in two positions, the lowest point between the peak value, can the real time inspection segmentation effect, revise segmentation result; The deletion isolated noise point makes even matter gray scale cut zone continuous.

Step S24 for the complicated effective matching area of surface feature background, need prepare many-valued reference map according to the gray scale of different atural objects, can utilize the Photoshop pen tool manually to delineate the atural object border, fills corresponding grey scale then and gets final product.

Consult Fig. 9.Step S31 chooses the corresponding grey scale value according to different types of ground objects from atural object gray feature database, the segmentation result to a plurality of effective matching areas carries out the gray scale filling then.

Step S32 is to inactive area; Other zones of promptly not mating, characters of ground object is not obvious often, perhaps the type of ground objects complicacy zone that is difficult to prepare; If being used as the effective coverage, this type zone is prepared into reference map; Tend to reduce the degree of confidence of coupling with real-time figure coupling, thus all zero fillings of inactive area, though participation relevant matches computing not contributing to the related coefficient of matching operation.

Step S33; Add for a plurality of often figure of regional many-valued reference map of Photoshop preparation are range upon range of; Need carry out the figure layer and be merged into single gray-scale map layer, the output result saves as 8 gray level image files of a raw data layout, supplies follow-up relevant matches to handle and uses.

Consult Figure 10.The many-valued reference map gray average in step S41 zoning deducts average with the many-valued reference map in zone then and obtains the average reference map, and the purpose of doing like this is in order to eliminate the influence that reference map brightness DC component is brought to relevant matches.

Step S42 carries out pyramid decomposition step by step to removing the average reference map; The pyramid decomposition number of plies is selected to take into account the real-time requirement and don't is influenced matching effect; Generally select 2～3 grades; The length and width of each grade reference map all are reduced into 1/2 of upper level length and width, and the pixel in each grade reference map is near the mean value of 4 pixels upper level correspondence position.

Step S43 calculates the pixel square two-dimensional array of lowest hierarchical level reference map, is used for relevant matches and handles use in real time.In order to practice thrift the time that the relevant matches cyclic search is handled in real time, can precalculated data calculate earlier as far as possible.

Step S44 calculates the quadratic sum two-dimensional array of interlacing each benchmark subgraph when the row traversal search; Array size is the ranks scope of interlacing at a distance from the row traversal search; Value in the array is the quadratic sum of this searching position place benchmark subgraph, and the square number set of calculated of being calculated by step S43 obtains.

What step S45 calculated smart coupling reference maps at different levels goes average square data, and calculates near the benchmark subgraph quadratic sum value of 7 * 7 search neighborhood smart matched position, supplies the subsequent fine coupling to calculate and uses.

Consult Figure 11.Step S51 carries out smothing filtering to real-time figure, generally adopts 5 * 5 mean value smoothings, weakens coherent speckle noise, improves the relevant matches accuracy.

Step S52 calculates the average of level and smooth filtered real-time figure, then with smothing filtering in real time figure deduct average and obtain average and scheme in real time.

Step S53 schemes to carry out pyramid decomposition step by step in real time to smothing filtering; Specific operation process and step S2 are similar; The length and width that each grade schemed in real time all are reduced into upper level and scheme 1/2 of length and width in real time, and the pixel during each grade schemed in real time is the mean value that upper level is schemed near 4 pixels of correspondence position in real time.

Step S54 calculates lowest hierarchical level and removes the average pixel square two-dimensional array of figure in real time, obtains average after adding up and schemes quadratic sum in real time, is used for relevant matches and handles use in real time.In order to practice thrift the time that the relevant matches cyclic search is handled in real time, can precalculated data outside circulation, calculate earlier as far as possible.

Step S55 calculates the smart figure quadratic sums in real time of mating at different levels, uses when supplying smart the coupling.

Step S56 carries out interlacing at a distance from the thick coupling of row search to the lowest hierarchical level reference map with real-time figure, this step be relevant matches most crucial also be coupling maximum step consuming time in real time.

Step S57 finds the maximal value of thick coupling correlation surface to be the Optimum Matching position.

Step S58 finds corresponding matched position in the last layer level image according to smart matched position, near these position 7 * 7 neighborhoods, carries out 49 normalized crosscorrelation matching operations;

Step S59 finds the maximum value position of 49 related coefficients to be the smart matched position of this level.

Step S60 is after the essence coupling of all levels is accomplished, and last smart matched position is the Optimum Matching position of final output.

Claims (8)

1. the matching process of optics/SAR foreign peoples's image is characterized in that comprising the steps:

1) utilizes Photoshop software, in the millimeter-wave SAR image of history accumulation, choose all kinds of ground object area and carry out gray-scale statistical, make up millimeter-wave SAR atural object gray feature database;

2) in optical remote sensing image, choose a plurality of effective matching areas interested, and each effective matching area is carried out terrain classification and Region Segmentation;

3) according to the gray-scale statistical characteristic of atural object different atural object cut zone are filled different gray scales; Utilize all kinds of atural object gray-scale statistical empirical values in the millimeter-wave SAR gray feature database; Segmentation result image to each effective matching area carries out the filling of many-valued zone, obtains regional many-valued reference map;

4) the regional many-valued reference map and the real-time figure of millimeter-wave SAR that utilize optical remote sensing image to be prepared into directly carry out gray scale normalization simple crosscorrelation coupling; Earlier the many-valued reference map in zone is carried out the relevant matches pre-service, carry out reference map then and handle in real time with the relevant matches of scheming in real time; And earlier said reference figure and real-time figure are carried out pyramid decomposition, on real-time figure of lowest hierarchical level and reference map basis, carry out interlacing at a distance from the row search matched.

2. the matching process of optics as claimed in claim 1/SAR foreign peoples's image; It is characterized in that: described relevant matches pre-service is that the many-valued reference map in zone is carried out pyramid decomposition; Calculate the part that to calculate in advance with scheming in real time to have nothing to do in the lowest hierarchical level relevant matches; Comprise that reference map goes average square data, and with the quadratic sum data of all benchmark subgraphs of real-time figure with size.

3. the matching process of optics as claimed in claim 1/SAR foreign peoples's image is characterized in that: it is earlier real-time figure to be carried out pyramid decomposition that described relevant matches is handled in real time, calculates the part that can calculate in advance before the lowest hierarchical level relevant matches traversal search; Carry out interlacing at a distance from being listed as search and going average normalized crosscorrelation coupling in lowest hierarchical level then, the top in the correlation surface is thick matched position; Turn back to the upper level reference map at last step by step and scheme to carry out the essence coupling in real time, highest master reference figure is the Optimum Matching position with the smart matched position of real-time figure.

4. the matching process of optics as claimed in claim 1/SAR foreign peoples's image is characterized in that: utilize two-value that optical remote sensing image is prepared into or many-valued reference map and millimeter-wave SAR radar image directly to carry out gray scale normalization simple crosscorrelation coupling.

5. the matching process of optics as claimed in claim 1/SAR foreign peoples's image; It is characterized in that: regional many-valued optical reference figure and SAR scheme to mate employing in real time by thick coupling to essence; Earlier in the basic enterprising line search coupling of the little figure of lowest hierarchical level; Find the thick matched position of lowest hierarchical level, progressively near the thick matched position of the big figure of high-level, carry out the essence coupling then, the smart matching result of the big figure of highest level is final matching results.

6. the matching process of optics as claimed in claim 3/SAR foreign peoples's image; It is characterized in that: the pyramid decomposition number of plies is selected 2～3 grades; The length and width of each grade reference map all are reduced into 1/2 of upper level length and width, and the pixel in each grade reference map is near the mean value of 4 pixels upper level correspondence position.

7. the matching process of optics as claimed in claim 1/SAR foreign peoples's image; It is characterized in that: calculate interlacing when the row traversal search; The quadratic sum two-dimensional array of each benchmark subgraph; Array size is the ranks scope of interlacing at a distance from the row traversal search, and the value in the array is the quadratic sum of this searching position place benchmark subgraph.

8. the matching process of optics as claimed in claim 1/SAR foreign peoples's image; It is characterized in that: that calculates smart coupling reference maps at different levels goes average square data; And calculate near the benchmark subgraph quadratic sum value of 7 * 7 search neighborhood smart matched position, supply the subsequent fine coupling to calculate and use.

Images