Prior art
In the simulation camera work of present existing digital photogrammetry technology and use film, for the ease of sting out the method for ground control point position on image, pre-cloth mark and the sign that uses are the same.Indicate at the field operation reference mark that object space is laid in advance as shown in Fig. 1 .1,1.2,1.3 before the photographic imagery, and the application engineering measuring technique, the coordinate of measuring sign geometric center (as: right-angled intersection point) is as the reference mark coordinate.Behind the filmed image, film and photograph are found out the position at reference mark on image with the method for acupuncture points.Digital picture is pointed out with mouse, and the geometric center of digital picture picture dot is represented the correspondence position at reference mark.Give the reference mark coordinate figure that field operation is measured to acupuncture points and picture dot geometric center, finish image same place coupling, use for mapping work.
Adopting pre-cloth mark technology is to find the object space reference mark in order to be easy to from image, and concrete grammar is the artificial feature dot mark to be set and to measure its location parameter at object space before the imaging.Generally do not use as the reference mark with natural thing.Carry in the imaging sensor remote sensing of the earth process at satellite, aviation platform, be meant before photography is implemented and waiting the process of area laying of photographing in advance the sky sign.Indicated in " photogrammetric measurement field operation standard " and laid the requirement that is used for aerophotogrammetric surface mark, shape and size are shown in Fig. 1 .3.In the close-range photogrammetry process, be meant and lay various artificial targets in advance as the reference mark or as location point to be measured, at present in the planar cloth bidding will that requires use " photogrammetric standard " in as (a) among Fig. 1 .1 (b) shown in (c), Fig. 1 .2.
Along with the development of photogrammetric technology, digital camera progressively replaces traditional jelly-roll type camera.The continuous gray scale textural difference in space on pixel structure that disperses on the digital image and the film image, when making the pixel center, sign place identified on the digitized video as the object space reference mark, there is a side-play amount at random in the same place of the side of elephant and object space, and the pixel center can't overlap fully with actual object space reference mark coordinate.Producing above-mentioned reasons of error is: the reference mark yardstick of the physical size of each pixel of digital picture and object space is not the same order of magnitude, promptly under the condition of eye-observation, the object space reference mark belongs to the data of millimeter magnitude, but the pixel of digital picture is several centimetres when airborne remote sensing arrives tens of centimeter scale, during satellite remote sensing being tens of centimetres and arriving the number metrical scale, is 1 centimetre to several centimetres during close-range photogrammetry.After refineing to the Pixel-level degree, which position that this pixel is dropped at the reference mark actually is a random number, it may just drop on the positive center of pixel, also might drop on any position of departing from the pixel central point, this offset vector value has influence on the matching precision of ground control point point position and image same place pixel position, and then influences plotting accuracy.
Summary of the invention
Above shortcoming in view of prior art, the purpose of this invention is to provide a kind of technology of eliminating digital photogrammetry object space reference mark and image deviations, the geometric element of the sign laid and CCD, cmos imaging unit is rationally mated, the error when locate on digitized video at actual object space reference mark when eliminating digital photogrammetry effectively.The objective of the invention is to realize by following means.
A kind of technology of eliminating digital photogrammetry object space reference mark and image deviations, be used for CCD, CMOS is the digital photogrammetry of imaging target surface, comprise that on-the-spot sign is provided with requirement, making indicates used parameter, digitized video and three major technique steps of actual object space reference mark location parameter error correction calculating that imaging device obtains, wherein
When (1) laying sign, keep the straight line of sign parallel with the straight line of imaging target surface pixel.
(2) the planar sign that uses of pre-cloth mark has following feature: planar sign be shaped as rectangle, form by center with around two zones, district with same geometric center and different optical reflection feature; Satisfy between the width 2A of described center and the focal distance f of image formation optical device, image-forming range H and imaging target surface pixel width a: A=a*H/f, satisfy between the length 2B of center and imaging target surface pixel length b: B=b*H/f.
When (3) calculating the error of object space reference mark when on digitized video, locating, based on the pixel value of place, reference mark pixel and peripheral pixel thereof on the digitized video that obtains, the deviation of calculation control point pixel central point and actual object space sign central point.Choosing on the digitized video that obtains with place, reference mark pixel is the 5*5 window at center, amounts to 25 pixels, and generally speaking, central point pixel and outmost turns pixel are non-mixed pixel, and remaining is a mixed pixel.Resolve the area ratio of mixed pixel center and marginarium by the pixel value of non-mixed pixel, and the area ratio is converted into the length ratio, with the ground resolution and the deviation of length of single pixel than calculation control point pixel central point and actual ground sign central point; If having 4 non-mixed pixels of 2*2 in the window is its special cases, 4 pixel levels are the center, reference mark with vertical marginal intersection point.Equate if occupy the pixel value of its 4 adjacent up and down gray scale picture dots of non-mixed pixel at center, show that then the non-mixed pixel geometric center at the center of occuping overlaps with the object space reference mark.
Adopt technology of the present invention, can find out the deviation of pixel off-center point effectively.
In order to satisfy the mathematical model of calculation deviation, need be manufactured with the distinctive mark of determining reflectivity, the size of the electromagnetic wave bands that the size of sign and material utilize along with photography, photographic scale, imaging target surface pixel and geometric element change and change.It is different that the optical signature of the used planar sign of the present invention central area and circle zone has relative big light reflection difference.Be convenient on image, discern.The sign that is used for the visible light wave range imaging, its difference is presented as inverse (black-and-white two color) each other during the single band imaging, is complementary colors (as red and blue) during colour imaging.The sign that is used for the imaging of reflection absorption ftir spectroscopy district, its difference are presented as that two zones adopt infrared strong reflection material and infrared strong absorbing material respectively.The position, reference mark is the sign geometric center.The size of sign depends on the resolution of image, and the center is corresponding with 4 picture element matrixs of 2*2, and so no matter how the camera imaging element is cut apart, and has non-mixed pixel (the pixel P (i, j)) of a complete representative center sign all the time.Pre-designed sign size, relation during laying between the pixel geometry element of satisfied sign and image, adopt the inferior pixel cutting techniques of deriving again based on reflected energy, just can accurately eliminate the side-play amount between the pixel center, place, reference mark and working control point coordinate on the image (for example GPS measured value of open-air pre-cloth sign geometric center point), both positions are overlapped.
Embodiment
Planar sign be shaped as square or rectangle.When the image-forming component of digital camera or remote sensor was shaped as square, the shape of sign was got square; When the image-forming component face of digital camera or remote sensor was shaped as rectangle, the shape of sign was got rectangle.Planar sign is formed by the center with around 2 zones, district.Geometric center with square or rectangle is the position at reference mark.Sign center and come the manufactured size of calculation flag with the size of image-forming range (or the relative height between remote sensor and surface mark) and camera (or remote sensor) focal length, imaging target surface unit (CCD, CMOS) around the physical dimension in district.
Setting up a rectangular coordinate system, is true origin with the object space reference mark, cross initial point and sign mutually perpendicular two respectively the limit be X-axis and Y-axis.Image-forming range or terrain clearance are H, and camera (or remote sensor optical system) focal length is f, and imaging target surface (CCD, cmos cell etc.) pixel dimension is a, b.As shown in Figure 2, bidding will center point coordinate is (0,0), indicates that then central location is in (A, A;-B is B) in 4 straight line institute regions.Be connected continuously with the center outer boundary around district's inner boundary.Be not less than (3A, 3A around district's outer boundary;-3B, 3B) border of 4 straight line institute regions is (in square, A=B).A=a*H/f,B=b*H/f。
Fig. 3 is square sign, and wherein the sign part is a black, and the length of side is 2A, and background parts is a white, and the length of side is 6A; Fig. 4 is the rectangle sign, and wherein sign is a black partly, and long is 2A, and wide is 2B, and background parts is a white, and long is 6A, and wide is 6B.
No matter planar sign is to make at the scene or indoor prefabricated, the reflecting properties As soon as possible Promising Policy youth of index plane material uncle face feature avoids generating the index plane with strong directional reflection from technology, when condition is permitted, measures its spectral reflectivity.
According to time, water-fast, anti-degeneration, the anti-damage requirement that the sign needs keep, the material and the technology of selecting to make sign.The center of monumented point and around the district constitute by one matter respectively.Spectral reflectance energy that obtains from the monumented point center or emitted energy and all proportional with the image picture elements value around distinguishing the spectral reflectance energy that obtains.The pixel that drops on fully on the sign center is non-mixed pixel with the pixel that drops on fully around in the district, and the energy that obtains (pixel value) is maximum (little) value and minimum (greatly) value of the target reflection energy around sign cloth is set up an office.
Sign cloth set direction is arranged according to the corresponding relation of imaging target surface and the image-generating unit length of side and sign.When imaging target surface unit (CCD, CMOS) size x=y, be masked as the square of proportional amplification.When laying sign, wherein parallel with one side of imaging target surface on one side.When imaging target surface unit (CCD, CMOS) size x ≠ y, be masked as the rectangle of proportional amplification.The long limit of sign was parallel with the long limit of imaging target surface unit when sign was laid.The requirement of imaging (or three-dimensional imaging) is wanted to satisfy in the position of sign.As shown in Figure 5.
Adopt cloth mark of the present invention, the digitized video that is obtained is in the center of sign with around distinguishing the pixel that a pure individual reflection material is at least respectively arranged.As shown in Figure 6, black dotted lines grid background is a surface mark, and wherein slanted bar line fill area is atural object around the sign.No frame grid is the true field of actual image picture element correspondence, and black solid line surrounds part and is the sign viewing area.If when the center shows four identical pixels, show that then the central point of sign is positioned at the inner edge joint of 4 pixels, the reference mark is the grid central point.As: numbering 1 pixel is P (i-1, j) pixel among the figure; Numbering 2 pixels are P (i+1, j) pixel among the figure; Numbering 3 pixels are P (i, j-1) pixel among the figure; Numbering 4 pixels are P (i, j+1) pixel among the figure.Most of actual imaging situations are that the CCD border is not overlapping fully with the sign border, diagram shows in 36 corresponding with sign in black solid line grid picture dots, has only a complete black squares (P (i, j) pixel part), be close to the mixed pixel that white that pixel around it is black part and background mixes.Because black proportion difference in the mixed pixel make mixed pixel show as the grey pixel, and the gray-scale value of each grey pixel is not quite similar.But the pixel in the complete black pixel left and right sides is made up of highly identical black rectangle subarea and white rectangle subarea respectively, and the black and white area ratio on the gray scale pixel of the left and right sides is just in time opposite, the area that is white in the left pixel equates that with the area of black in the right pixel area of black equates with the area of white in the right pixel in the left pixel.The composition of both sides pixel is identical with the pixel of the left and right sides about complete black pixel.
With reference to figure (6) and figure (2), and ater pixel P (i, pixel value j) is g (i, j), the grey pixel P on the ater pixel left side (i, gray-scale value g (i j-1), j-1), grey pixel P (i, gray-scale value g (i j+1) on ater pixel the right, j+1), the grey pixel P of ater pixel top (i-1, gray-scale value g (i-1 j), j), the grey pixel P of ater pixel bottom (i+1, and gray-scale value g j) (i+1, j).
Set up rectangular coordinate system with black pixel left side grey pixel, the mutually perpendicular left side of pixel and base are respectively the longitudinal axis and transverse axis, and the lower left corner is initial point; I, j represent the ranks number of pixel on the image respectively; The system gain factor that radiation energy on the camera CCD unit is converted to electric signal, generation pixel value is k; In this cell coordinate system, black and white separatrix coordinate is x; The pixel cross-directional length is L, and vertical-direction length is L ', x
0=0.
g(i,j-1)=k×E
i(i,j-1)
=k×E
i(x,y)×{ρ
b(x,y)×(x-x
0)×L′+ρ
w(x,y)×[L-(x-x
0)]×L′}
(4-1)
g(i,j+1)=k×E
i(i,j+1)
=k×E
i(x,y)×{ρ
w(x,y)×(x-x
0)×L′+ρ
b(x,y)×[L-(x-x
0)]×L′}
(4-2)
Following formula (4-1) (4-2) in, E
i(i j-1) is the reflected energy of left pixel; E
i(i j+1) is the reflected energy of right pixel; E
i(x is that camera exposure is incident upon the emittance on the CCD unit in the time y); ρ
w(x y) is the spectral reflectivity in pure white zone; ρ
b(x y) is the spectral reflectivity in ater zone.
Utilize formula (4-1) and (4-2) two formulas can draw,
From formula (4-3) as seen, the difference of left and right grey pixel gray-scale value is converted into difference in length, thereby can obtain the value of x.When about two grey pixel gray scales equate, promptly g (i, j-1)=g (i, j+1), and x
0=0 o'clock,
At this moment, the reference mark does not have skew in the horizontal direction.
In the following formula (4-3), g
wGray-scale value for the pure white pixel; g
bGray-scale value for the ater pixel.
Upper and lower sides pixel gray scale processing method is identical with left and right sides pixel gray scale processing method.4 pixels do not participate in handling on the diagonal line.
In like manner, can obtain the value of y by grey pixel gray-scale value about calculating.
When two grey pixel gray scales up and down equate, promptly g (i-1, j)=g (i+1, in the time of j), y
0=0,
At this moment, the reference mark does not have skew in vertical direction.
X and y with under the cell coordinate system convert actual geographic coordinate system upper/lower positions side-play amount to, and the reference mark is revised.
If when showing 4 ater pixels on the image shown in Figure 7, (0, the 0) point in reference mark and Fig. 2 coordinate system overlaps, 4 pixels adhere to 4 quadrants separately, and the pixel of selection first quartile is as place, reference mark pixel, and the X component of reference mark coordinate figure is added
The Y component adds
Be geographic coordinate with coordinate conversion again.
Whole subsequent treatment can be expressed by the workflow of Fig. 8.