A kind of real-time joining method of unmanned aerial vehicle remote sensing images and systemTechnical field
The present invention relates to unmanned aerial vehicle remote sensing images processing technology fields, and in particular, to a kind of unmanned aerial vehicle remote sensing images are realWhen joining method and system.
Background technique
Aerial survey of unmanned aerial vehicle mainly passes through unmanned plane field operation acquisition remote sensing image at present, and interior industry processed offline image providesMap Services.This techniqueflow complexity, the separation of interior field operation, response cycle are long, are unable to satisfy the application of rapid response to customer's needScene.In addition, unmanned plane real-time live broadcast video data can only obtain local region information, global situation can not be provided in real time.
In the scene more demanding to real time response speed such as emergency management and rescue, aerophotogrammetric field work is needed to handle remote sensing shadow in real timePicture provides global situation Map Services.
Summary of the invention
The present invention provides the real-time joining methods and system of a kind of unmanned aerial vehicle remote sensing images, can make in unmanned planeDuring industry, real-time positioning and directing, correction and splicing passback image, the universe outdoor scene in unmanned plane region is presented in real timeFigure, meets real-time at figure demand under emergency scene.
For achieving the above object, one aspect of the present invention provides a kind of real-time joining method of unmanned aerial vehicle remote sensing images,The described method includes:
Step 1: the data of taking photo by plane of real-time reception unmanned plane passback, comprising: real-time aerial images and corresponding POS data；
Step 2: being based on POS(Position and Orientation System, positioning and orientation system) data obtain shadowThe geography information of picture；
Step 3: the elements of interior orientation correcting image lens distortion based on unmanned plane camera, based on the foreign side in POS dataBit element geometric correction image is registrated the image after correction in conjunction with the image geography information that step 2 obtains；
Step 4: the image after step 3 registration is sliced；
Step 5: tile that anastomosing and splicing the is sliced and Map Services for being issued as standard are used for browser end.
Preferably, real-time reception unmanned plane data, obtain unmanned plane filmed image when spatial position and three axis appearancesState information, spatial position include the longitude, latitude, elevation of unmanned plane, and three-axis attitude information includes that unmanned plane is sat relative to navigationMark roll angle, the pitch angle, yaw angle of system.
Preferably, step 2 further include: according to the elements of exterior orientation that unmanned plane is shot, based on unmanned plane camera and GPS withIMU(Inertial Measurement Unit) mounting distance error and posture angular variation caused by projection error correctThe geographical location of image.
Preferably, with calculating ground image surface according to camera CCD size, image resolution and opposite unmanned plane flying height on timePlain resolution ratio utilizes the image center point coordinate registration image after pixel resolution and correction.
Preferably, the method also includes tiles after browser dynamic acquisition anastomosing and splicing, and load newest fusion and spellTile after connecing is visualized.
Preferably, the method specifically: service the TMS that the tile after anastomosing and splicing is issued as standard；It utilizesThe newest visual range of the real-time informing browser front end WebSocket load live-action map；Browser end is according to newest visual modelEnclose the newest tile live-action map for loading corresponding level.
Picture microtomy is that the image with geographical coordinate is cut to the technology of small picture according to pyramidal configuration.By nothingThe remote sensing image of man-machine shooting is issued as Map Services, can be accessed by browser, at present there are mainly two types of processing method, oneKind is that all remote sensing images are spliced into a Zhang great Tu, is issued as Map Services after being sliced to big figure；Another kind is to individual shadowAs being issued as Map Services for every image as independent figure layer after progress streaming slice, pass through browser-presented after superposition.
First method needs first splice remote sensing image to be sliced again, is unable to satisfy real-time scene demand.Second methodAlthough can handle in real time, it is only applicable to the scene of less image, as image increases, layer count increases, and works as imageThe load of browser becomes larger when reaching certain amount, and performance decline is unable to satisfy real-time application.The present invention provides it is a kind of nobodyThe dicing method of machine remote sensing image solves the problems, such as real time problems, the boundary tile black and white side, browser of existing microtomyThe performance loading problem at end can be sliced and be handled to the image after correction in unmanned plane operation, in real time by nothingThe remote sensing image of man-machine shooting smoothness in the form of Map Services is presented to the user.Image will be sliced and be merged in this methodThe tile that splicing slice obtains, specifically includes:
Step a: detecting the coordinate system of individual unmanned aerial vehicle remote sensing images, if not Mercator projection coordinate system, then be converted toMercator projection coordinate system, for image to be presented on map；
Step b: maximum zoom the level N, N for determining that image is supported are the integer greater than 1；
Step c: the resolution ratio of image is aligned with the resolution ratio of n-th layer grade；
Step d: the tile that slice obtains n-th layer grade is carried out to image；
Step e: the tile based on n-th layer grade sequentially generates the tile of N-1 level to the 1st level；
Step f: the tile of layer-by-layer the 1st level of anastomosing and splicing to n-th layer grade is identified according to timing and tile；
Preferably, according to the pixel resolution of image and Mercator projection coordinate algorithm, quaternary tree cuts map.It cuts every timeThe initial resolution that figure layer Mercator projection coordinate system is corresponded to after cutting can all become larger, the resolution ratio until being less than image.Image is supportedMaximum zoom level be that maximum cutting times subtract one.
Wherein, when step d and step e generates tile, tile can be stored as any image format.If it is considered that map takesBusiness, then need for tile to be stored as the picture format that HTTP Content-Type is supported and do transparency process to tile.WhenWhen tile is stored as jpg or png format, the tile (i.e. non-boundary tile) not comprising nodata is stored as jpg format, includesThe tile (i.e. boundary tile) of nodata is stored as png format.Map boundary line is uniformly set at Map Services end, can only be requestedTile in bounds.The tile storage format for ignoring request when the tile in bounds is requested, directly returns and actually depositsThe tile of storage can solve the problems, such as to load slow problem caused by boundary tile black and white side and png picture are excessive simultaneously.Wherein,This method is for individual unmanned aerial vehicle remote sensing images after correction, using the tile obtained to n-th layer grade pet chip as tile map goldThe bottom of word tower model；The tile of remaining level all passes through its next layer of tile generation of resampling；It is identified according to timing and tileAnastomosing and splicing tile.
Preferably, the step d is specifically included: determining the size of every tile according to tile data specification and in imagePixel coordinate from left to right, carry out cutting from top to bottom since the image upper left corner, by image cutting be several individualTile, the incomplete part of tile are filled with nodata.
Preferably, tile is converted to matrix in anastomosing and splicing tile by step f.This matrix can obtain tile channelIn identical ranks all pixels point pixel value, these pixel values can be considered as one group of vector v respectively.The picture of nodata pixelPlain value is uniformly considered as scalar 0 in affiliated channel.Therefore it can judge whether new tile includes nodata by vector operation.
Preferably, when step f anastomosing and splicing tile, tile is converted into matrix.It, will for the new tile comprising nodataThe pixel value of the pixel of identical ranks is considered as one group of vector v in image processing channels, and the pixel value of each pixel is each channel mark in vAmount.Vector v seeing image is not all monomial or multinomial as channel number.The tile and new tile that former slice is generated it is corresponding toEvery scalar in amount v does or operation by channel sequence one-to-one correspondence, and replaces the tile that former slice generates.
Wherein, the anastomosing and splicing time-consuming of tile is about down to the 1/6 of mode pixel-by-pixel using above two method.
Preferably, the geometric correction in this method can use traditional geometric correction method, can also be using the present inventionIn improved geometric correction method, in order to small under solving can not to set up control point or emergency application scenarios without control pointDisturbance of data may cause the problem of larger correction error, be splicing in real time the present invention provides improved geometric correction methodFast geometric correction when unmanned aerial vehicle remote sensing images provides support, and the method for the present invention causes for pitch angle, roll angle, yaw angleGeometric deformation, utilize 3-D geometric model fast geometric correct raw video.
Existing no control point geometric correction method is sat IMU coordinate system transformation to ground survey by transformed coordinate systemMark system, by the POS parameter after correction directly as image elements of exterior orientation, passes through collinearity condition equation and indirect method image geometryCorrect unmanned aerial vehicle remote sensing images.The correction course needs the data such as the error of camera and IMU established angle, these data are in emergency fieldIt is difficult to obtain under scape, and may face and calculate error and part matrix caused by multiple matrix products to error sensitive, leadThe problem of causing small data error to cause larger correction error.
The elements of exterior orientation based in POS data carries out geometric correction to image, specifically includes:
Acquire unmanned plane original remote sensing image；Due to the influence of unmanned plane pitch angle, unmanned plane original remote sensing image atTrapezoidal deformation is generated when picture, is corrected trapezoidal deformation caused by pitch angle and is obtained the corresponding first trapezoidal ABEF of raw video；
Using the minimum circumscribed rectangle of the first trapezoidal ABEF as the second rectangle ABCD, i.e., with the upper bottom of the first trapezoidal ABEF withLength of the longer bottom AB as the second rectangle ABCD in bottom, using the height of the first trapezoidal ABEF as the width of the second rectangle ABCD；Due to the influence of unmanned plane roll angle, the second rectangle ABCD generates trapezoidal deformation in imaging, corrects trapezoidal caused by roll angleDeformation obtains the corresponding second trapezoidal A'B'C'D' of the second rectangle ABCD；
Resolve to obtain the first quadrangle A'B'E'F' according to the second trapezoidal A'B'C'D', according to bottom the first trapezoidal ABEF on andRelative position of the two endpoint E and F of shorter bottom EF on the second side rectangle CD, obtains first by perspective transform in bottomPosition of the two vertex E' and F' of quadrangle A'B'E'F' on the C'D' of side；
The is obtained by four vertex that the corresponding Two Dimensional Rotating matrix rotation of unmanned plane yaw angle converts the first quadrangleTwo quadrangles, that is, the image shape after correcting；
According to four vertex of coordinate and second quadrangle of four vertex of the first rectangle under pixel coordinate system in pixelThe corresponding relationship of coordinate under coordinate system resolves the homography conversion matrix of image after obtaining raw video and correction；
According to the pixel value of homography conversion matrix and each pixel of raw video, each picture of image after correction is calculatedThe pixel value of vegetarian refreshments is to get the image to after correcting.
Preferably, the frame rectangle of raw video is the first rectangle.
Preferably, the first trapezoidal height:
；
First trapezoidal upper bottom:
；
First trapezoidal bottom:
；
Wherein, H is flying height of the unmanned plane with respect to ground, w1For the length of the first rectangle, h1For the width of the first rectangle, f is camera lensFocal length, α are pitch angle (unit: radian) of the unmanned plane relative to navigational coordinate system, y0For principal point (photo centre with as planeVertical line and intersection point as plane) ordinate under collimation mark coordinate system, GSD is the ground sampling interval.
Preferably, the second trapezoidal height:
；
Second trapezoidal upper bottom:
；
Second trapezoidal bottom:
；
Wherein, H is flying height of the unmanned plane with respect to ground, w2For the length of the second rectangle, h2For the width of the second rectangle, β is nobodyRoll angle (unit: radian) of the machine relative to navigational coordinate system, x0For principal point, (photo centre is with the vertical line as plane and as flatThe intersection point in face) abscissa under collimation mark coordinate system, f is lens focus, and GSD is the ground sampling interval.
Preferably, four vertex for converting the first quadrangle by the corresponding Two Dimensional Rotating matrix rotation of yaw angle obtainTo the second quadrangle, i.e. image shape after correction, specifically include:
Step A: using the lower-left angular vertex of the first quadrangle as origin o, horizontal direction is positive direction of the x-axis to the right, according to the right sideGimmick then determines positive direction of the y-axis, respectively obtains coordinate of four vertex of the first quadrangle at coordinate system xoy.
Step B: yaw angle (unit: radian) of the γ as unmanned plane relative to navigational coordinate system is set, then Two Dimensional Rotating matrix R
, for calculate point (x1,y1) around origin rotate γ counterclockwise after coordinate(x2,y2) be。
The four obtained vertex step C: step A obtain the second quadrangle after the spin matrix R rotation transformation of step BThe coordinate on four vertex.
Preferably, four of the coordinate and the second quadrangle according to four vertex of the first rectangle under pixel coordinate systemThe corresponding relationship of coordinate of the vertex under pixel coordinate system resolves the homography conversion square of image after obtaining raw video and correctionBattle array, specifically includes:
Step a: since what is finally obtained in camera image is pixel, so needing four vertex of the second quadrangleBe converted to pixel coordinate system (using the top left corner apex of image minimum circumscribed rectangle as origin, horizontal direction be to the right x-axis forward direction,Vertical direction be downwards y-axis forward direction) under coordinate.
Step b: solving homography conversion matrix H, gives homography switching matrix, the homogeneous coordinates respectively (x of raw video and two corresponding points of image after correction1,y1, 1) and (x2,y2, 1), then have:
Homography switching matrix H has 8 unknown numbers, it is known that coordinate of first four, the rectangle vertex under pixel coordinate system withThe corresponding relationship of coordinate of four vertex of the second quadrangle under pixel coordinate system, can solve homography switching matrix H, intoAnd obtain corresponding points of the point on raw video after calibration on image.
Preferably, according to the pixel value of homography conversion matrix and each pixel of raw video, pass through quadratic linear interpolationMethod carries out resampling, the pixel value of each pixel of image after being corrected.
On the other hand, the method in the corresponding present invention, the present invention also provides a kind of unmanned aerial vehicle remote sensing images to splice in real timeSystem, the system comprises:
Communication module, the data of taking photo by plane for the passback of real-time reception unmanned plane, comprising: real-time aerial images and correspondingPOS data；
Geography information obtains module, for obtaining the geography information of image based on POS data positioning and directing image；
Correction module, for the elements of interior orientation correcting image lens distortion based on unmanned plane camera, based in POS dataElements of exterior orientation geometric correction image；
It is sliced module, for being sliced to the image that correction module obtains；
Anastomosing and splicing module, the tile obtained for anastomosing and splicing slice module are simultaneously issued as the Map Services of standard for clearLook at device end use.
One or more technical solution provided by the invention, has at least the following technical effects or advantages:
This method and system can be in unmanned plane operation process, real-time positioning and directing, correction and splicing passback shadowPicture, solve field data acquisition data turn again interior industry processed offline traditional unmanned plane image processing process be not able to satisfy it is live in real timeThe problem of at figure, the universe live-action map in unmanned plane region is presented in real time, is met real-time at figure need under emergency sceneIt asks, such as forest fire, mud-rock flow, the monitoring application of flood calamity emergency.
This method and system, can be while unmanned planes clear by real-time splicing unmanned aerial vehicle remote sensing imagesDevice end real time inspection of looking at take photo by plane during live-action map, in real time on map splicing take photo by plane by region image.Aerial surveyAt the end of the entirely universe outdoor scene taken photo by plane in operating area can be presented.
Due to the various extraneous factors interference such as wind-engaging, air-flow, unmanned plane will appear pitching, rolling and partially in flight courseThe case where boat, causes the image taken photo by plane to generate geometric deformation.These geometric deformations cause image that can not really reflect region of taking photo by planeGround, the shape of object and position need first geometric correction image to splice image again.This method and system use improved image schoolCorrection method solves traditional joining method and causes larger correction to miss dependent on control point and without the small disturbance of data under control pointThe problem of difference, can satisfy the demand of emergency scene；Improved geometric correction method realizes image in this method and systemFast geometric correction, splice in real time for unmanned aerial vehicle remote sensing images and support be provided.
Detailed description of the invention
Attached drawing described herein is used to provide to further understand the embodiment of the present invention, constitutes of the invention onePoint, do not constitute the restriction to the embodiment of the present invention；
Fig. 1 is a kind of flow diagram of the real-time joining method of unmanned aerial vehicle remote sensing images in the present invention；
Fig. 2 is a kind of composition schematic diagram of the real-time splicing system of unmanned aerial vehicle remote sensing images in the present invention；
Fig. 3 is the flow diagram of unmanned aerial vehicle remote sensing images dicing method in the present invention；
Fig. 4 is the flow diagram based on the elements of exterior orientation geometric correction image method in POS data in the present invention；
Fig. 5 is that the image shift process of the unmanned aerial vehicle remote sensing images geometric correction method in the present invention based on POS data showsIt is intended to.
Specific embodiment
To better understand the objects, features and advantages of the present invention, with reference to the accompanying drawing and specific realApplying mode, the present invention is further described in detail.It should be noted that in the case where not conflicting mutually, it is of the inventionFeature in embodiment and embodiment can be combined with each other.
In the following description, numerous specific details are set forth in order to facilitate a full understanding of the present invention, still, the present invention may be used alsoImplemented with being different from the other modes being described herein in range using other, therefore, protection scope of the present invention is not by underThe limitation of specific embodiment disclosed in face.
Referring to FIG. 1, the present invention provides a kind of real-time joining methods of unmanned aerial vehicle remote sensing images, which comprises
(1) data of taking photo by plane of real-time reception unmanned plane passback；The POS system collection DGPS technology and inertia that unmanned plane carries are ledBoat system (INS) technology is in one, spatial position and three-axis attitude information when can obtain unmanned plane filmed image, i.e., nobodyThree line elements (longitude, latitude, elevation) of machine and three angle elements (roll angle, pitch angle, yaw angle).
(2) elements of exterior orientation shot according to unmanned plane considers the mounting distance error and appearance of camera and GPS and IMUProjection error brought by state angular variation, the geographical location of correcting image.
(3) according to the elements of exterior orientation in the elements of interior orientation and POS data of camera, lens distortion school is done to image respectivelyImage deformation when being just imaged with geometric correction.
(4) the image position information for combining (2) to obtain, image after the correction that geographic registration (3) obtains；Root is needed with punctualGround pixel resolution is calculated according to camera CCD size, image resolution and terrain clearance, utilizes pixel resolution and correctionImage center point coordinate geographic registration image afterwards.
(5) image after geographic registration is sliced.
(6) it is classified the tile that layer-by-layer anastomosing and splicing is sliced.
(7) tile after anastomosing and splicing is issued as to the TMS service (Tile Map Service) of standard.
(8) WebSocket technology is utilized, notifies the newest visual range of browser front end real-time loading live-action map.
(9) browser end loads the newest tile live-action map of corresponding level according to newest visual range, and map support is putBig and diminution.
Referring to FIG. 2, the embodiment of the invention provides a kind of real-time splicing system of unmanned aerial vehicle remote sensing images, the system packetIt includes:
Communication module, the data of taking photo by plane for the passback of real-time reception unmanned plane, comprising: real-time aerial images and correspondingPOS data；
Geography information obtains module, for obtaining the geography information of image based on POS data positioning and directing image；
Correction module, for the elements of interior orientation correcting image lens distortion based on unmanned plane camera, based in POS dataElements of exterior orientation geometric correction image；
It is sliced module, for being sliced to the image that correction module obtains；
Anastomosing and splicing module, the tile obtained for anastomosing and splicing slice module are simultaneously issued as the Map Services of standard for clearLook at device end use.
Referring to FIG. 3, unmanned aerial vehicle remote sensing images dicing method under one of embodiment of the present invention real-time scene, comprising:
Step a: the coordinate system of the unmanned aerial vehicle remote sensing images after detecting individual correction, if not Mercator projection coordinateSystem, then be converted to Mercator projection coordinate system, for image to be presented on map.
Step b: the maximum zoom level N that image is supported is determined.According to the pixel resolution of image, Mercator's coordinate is utilizedAlgorithm calculates the maximum zoom level of image support.
Step c: the resolution ratio of image is aligned with the resolution ratio of n-th layer grade.The maximum zoom layer obtained according to step 2The resolution ratio resampling of image is the corresponding resolution ratio of the level, prepared for the slice of n-th layer grade by grade.
Step d: the slicing treatment of n-th layer grade.According to tile data specification determine every tile size (256x256) andPixel coordinate in image from left to right, carries out cutting since the image upper left corner from top to bottom, if being by image cuttingIndividual tile is done, incomplete part is filled with nodata.
Step e: according to the tile of n-th layer grade sequentially generate N-1 level to the 1st level tile.It will be cut in step dBottom of the n-th layer grade tile as pyramid model, above the tile of each level pass through its next layer tile of resamplingIt generates.
Step f: layer-by-layer anastomosing and splicing tile is identified according to timing and tile.When unmanned aerial vehicle remote sensing images are handled in real time,The remote sensing image of unmanned plane will form video stream, have certain degree of overlapping between video stream.Meanwhile coordinate belonging to being slicedAfter system and level determine, every tile just has unique x, y and z coordinate and hierarchical identification, and referred to as tile identifies.For subsequentEvery image, according to step a-e be sliced, then the newly-generated tile of layer-by-layer anastomosing and splicing therewith previous existence at tile.
During tile processing, there are two key difficulties and its solution are as follows:
Black and white side problem comprising nodata tile and png picture problems of too: if tile to be all stored as to jpg latticeFormula will have black surround or white edge to occur comprising nodata tile.Png lattice are stored as by increasing a transparency channel to tileFormula is able to solve the black and white side problem comprising nodata tile.But png picture size is 3-5 times of jpg picture, this can addBig network transport load, the load time for increasing browser end tile.The present invention will not include nodata tile and be stored as jpg latticeFormula is stored as png format comprising nodata tile.Map boundary line is uniformly set at Map Services end, can only request boundary modelEnclose interior tile.The tile storage format for ignoring request when the tile in bounds is requested, watt of actual storage is directly returnedPiece.The method can solve the problems, such as black and white side and png picture problems of too comprising nodata tile simultaneously.
The anastomosing and splicing speed issue of tile: using vector operation replace pixel-by-pixel multilevel iudge tile whether includenodata.Anastomosing and splicing is carried out using the tile that matrix or operation generate the new tile comprising nodata and former slice.This twoTile anastomosing and splicing time-consuming is about down to the 1/6 of mode pixel-by-pixel by kind method.
Using above-mentioned dicing method, can be cut to the real-time of image is completed in the real-time splicing of unmanned aerial vehicle remote sensing imagesPiece, while solving the problems, such as chip property (boundary black and white side and the excessive problem of browser offered load), it is unmanned aerial vehicle remote sensing shadowAs splicing provides strong support in real time.
Referring to FIG. 4, the present invention provides based on the elements of exterior orientation geometric correction image method in POS data, it is describedMethod includes:
Unmanned plane original remote sensing image is acquired, the frame rectangle of unmanned plane raw video is the first rectangle；Due to unmanned planeThe influence of pitch angle, unmanned plane original remote sensing image generate trapezoidal deformation in imaging.Pitch angle is corrected according to the first rectangle to drawThe trapezoidal deformation risen obtains the corresponding first trapezoidal ABEF of raw video；
Using the minimum circumscribed rectangle of the first trapezoidal ABEF as the second rectangle ABCD, i.e., with the upper bottom of the first trapezoidal ABEF withLength of the longer bottom AB as the second rectangle in bottom, using the height of the first trapezoidal ABEF as the width of the second rectangle；Due to nobodyThe influence of machine roll angle, the second rectangle generate trapezoidal deformation in imaging.It is trapezoidal caused by correcting roll angle according to the second rectangleDeformation obtains the second trapezoidal A'B'C'D', the acquisition pattern of the second trapezoidal A'B'C'D' specifically: by resolving the second trapezoidal A'Four vertex of B'C'D', and then obtain the corresponding second trapezoidal A'B'C'D' of the second rectangle ABCD；
The first quadrangle A'B'E'F' is obtained according to the second trapezoidal A'B'C'D', according to bottom on the first trapezoidal ABEF and bottomIn shorter bottom EF relative position of two endpoint E and F on the second side rectangle CD, the one or four side is obtained by perspective transformPosition of the two vertex E' and F' of shape on the C'D' of side；
The two or four side is obtained by four vertex that the corresponding Two Dimensional Rotating matrix rotation of yaw angle converts the first quadrangleShape, that is, the image shape after correcting；
According to four vertex of coordinate and second quadrangle of four vertex of the first rectangle under pixel coordinate system in pixelThe corresponding relationship of coordinate under coordinate system resolves the homography conversion matrix of image after obtaining raw video and correction；
According to the pixel value of homography conversion matrix and each pixel of raw video, each picture of image after correction is calculatedThe pixel value of vegetarian refreshments is to get the image to after correcting.
The unmanned aerial vehicle remote sensing images fast geometric bearing calibration based on POS data in the present invention: it is based on photogrammetric reasonBy (image is trapezoidal because of roll angle or the conformation of pitch angle being formed slopely).(relatively by a line element of POS dataThe flying height in face) and three angle elements (pitch angle, roll angle, yaw angle), it is first corrected according to pitch angle, further according to roll angle schoolJust, finally according to yaw angle rotation correction, pass through the coordinate (x of principal point in correction course0, y0) eliminate boresight misalignments.
Unmanned aerial vehicle remote sensing images fast geometric bearing calibration based on POS data, comprising the following steps:
Step 1, when only considering pitch angle, shape when camera imaging becomes trapezoidal deformation, by resolving trapezoidal upper bottom underBottom length and trapezoidal height obtain the first trapezoidal ABEF(referring to attached drawing 5, and each length is as unit of pixel)；
The height of first trapezoidal ABEF:
；
The upper bottom of first trapezoidal ABEF:
；
The bottom of first trapezoidal ABEF:
；
Wherein, H is flying height (unit: rice) of the unmanned plane with respect to ground, w1For the length (unit: rice) of the first rectangle, h1It isThe width (unit: rice) of one rectangle, f are focal length (unit: rice), and α is pitch angle (unit: radian), y0It is sat for the direction y of principal pointMark, GSD (Ground Sample Distance) are ground sampling interval (ground resolution).
Step 2, using the minimum circumscribed rectangle of the first trapezoidal ABEF as the second rectangle ABCD(referring to attached drawing 5, i.e., with firstLength of the longer bottom AB as the second rectangle ABCD in the upper bottom of trapezoidal ABEF and bottom, using the height of the first trapezoidal ABEF as theThe width of two rectangle ABCD)；The second rectangle ABCD is considered because of trapezoidal deformation caused by roll angle, is passed through and is resolved second trapezoidal four topsPoint obtains the corresponding second trapezoidal A'B'C'D'(of the second rectangle ABCD referring to attached drawing 5, and each length is as unit of pixel)；
The height of second trapezoidal A'B'C'D':
；
The upper bottom of second trapezoidal A'B'C'D':
；
The bottom of second trapezoidal A'B'C'D':
；
Wherein, w2For the length (unit: rice) of the second rectangle ABCD, h2For the width (unit: rice) of the second rectangle ABCD, β is to turn overRoll angle (unit: radian), x0For the abscissa of principal point.
Step 3, the first quadrangle A'B'E'F'(is obtained referring to attached drawing 5 according to the second trapezoidal A'B'C'D'), according to firstRelative position of the two endpoint E and F of shorter bottom EF on the side CD of the second rectangle ABCD in bottom and bottom on trapezoidal ABEF,Position of the two vertex E' and F' of the first quadrangle A'B'E'F' on the C'D' of side is obtained by perspective transform；
Step 4, four tops of the first quadrangle A'B'E'F' are converted by the corresponding Two Dimensional Rotating matrix rotation of yaw anglePoint obtains the second quadrangle (referring to attached drawing 5), that is, the image shape after correcting.
Step 5, four vertex of the coordinate and the second quadrangle according to four vertex of the first rectangle under pixel coordinate systemThe corresponding relationship of coordinate under pixel coordinate system resolves the homography conversion matrix of image after obtaining raw video and correction.
Step 6, according to the pixel value of homography conversion matrix and each pixel of raw video, pass through quadratic linear interpolation methodCarry out resampling, the pixel value of each pixel of image after being corrected.
Above-mentioned unmanned aerial vehicle remote sensing images geometric correction method, suitable for the quick correction process of individual image, treatment processAccelerated using GPU server, the correction time of individual 40,000,000 pixel image controls in 1s or so.
The exterior orientation angle element that the method for the invention uses is the exterior orientation angle element that GPS/INS system obtains, and is used forPosture information of the IMU in navigational coordinate system is described.Since the installation error of camera and IMU are smaller, this method is in emergency demandUnder do not consider.
Assuming that pitch angle is greater than 0 and roll angle is similar less than other situations of 0(, angle is with example angular on the contrary, then shouldTrapezoidal rotation 180 degree after the corresponding correction of angle), the unmanned aerial vehicle remote sensing images fast geometric bearing calibration based on POS dataSchematic diagram is as shown in Figure 5.
Above-mentioned unmanned aerial vehicle remote sensing images real-time processing method and system, the whole process is handled using streamingization, from listThe positioning and directing of image adjustment of image, is sliced, is spliced to and is issued as Map Services, forms a complete image and locates automaticallyManage chain.Adjustment of image, slice and anastomosing and splicing during taking photo by plane are all made of the acceleration of GPU server, so as to will be entireTreatment process controlled within several seconds.
For the processing method of GPS and IMU data exception: GPS and IMU may cause a because of extraneous interference at workOther data exception.In response to this, while the slice of data of spliced big figure slice and every image is stored, on mapOffer takes the function of a little checking original image.Region i.e. biggish for stitching error can directly click and check all of the positionRaw video map.
Using the above method and system, aerial survey of unmanned aerial vehicle operation returns the POS data of image and unmanned plane in real time, can be real-timeThe live-action map for checking unmanned plane region, achieve the effect that " mesh in one's power, figure institute to ".
Although preferred embodiments of the present invention have been described, it is created once a person skilled in the art knows basicProperty concept, then additional changes and modifications may be made to these embodiments.So it includes excellent that the following claims are intended to be interpreted asIt selects embodiment and falls into all change and modification of the scope of the invention.
Obviously, various changes and modifications can be made to the invention without departing from essence of the invention by those skilled in the artMind and range.In this way, if these modifications and changes of the present invention belongs to the range of the claims in the present invention and its equivalent technologiesWithin, then the present invention is also intended to include these modifications and variations.