CN109727279A - An automatic registration method of vector data and remote sensing images - Google Patents

An automatic registration method of vector data and remote sensing images Download PDF

Info

Publication number
CN109727279A
CN109727279A CN201910083298.XA CN201910083298A CN109727279A CN 109727279 A CN109727279 A CN 109727279A CN 201910083298 A CN201910083298 A CN 201910083298A CN 109727279 A CN109727279 A CN 109727279A
Authority
CN
China
Prior art keywords
matrix
remote sensing
sensing image
edge
vector data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201910083298.XA
Other languages
Chinese (zh)
Other versions
CN109727279B (en
Inventor
张文涵
李安波
李安营
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing Normal University
Original Assignee
Nanjing Normal University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nanjing Normal University filed Critical Nanjing Normal University
Publication of CN109727279A publication Critical patent/CN109727279A/en
Application granted granted Critical
Publication of CN109727279B publication Critical patent/CN109727279B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Landscapes

  • Image Analysis (AREA)

Abstract

Open remotely-sensed data has been done certain DecryptDecryption by the requirement of safety policy and has been handled, when user attempts own vector data and stacked remotely-sensed data, it may appear that the unmatched situation of the two.For this problem, the invention discloses the autoegistration methods of a kind of vector data and remote sensing image.Specific steps of the invention are as follows: 1) be based on vector data, extract three value matrix A using scan-line algorithm；2) gray matrix G is extracted based on remote sensing image, and obtains the matrix of edge E of gray matrix G using Canny operator；3) it is traversed on three value matrix A using matrix of edge E, generates normalized edge registration factor matrix F'Side；4) it is traversed on three value matrix A using gray matrix G, generates normalized gray scale registration factor matrix F'Ash；5) based on the characteristic for matching quasi-divisor, the autoregistration of vector data and remote sensing image is carried out；6) remote sensing image data in vector element prescribed limit is extracted.

Description

A kind of autoegistration method of vector data and remote sensing image
Technical field
The invention belongs to the fields GIS, and in particular to a kind of to utilize remote sensing image pixel characteristic and vector data element characteristic Consistency carry out autoregistration method.
Background technique
Currently, had it is all it is multi-platform provide open remotely-sensed data service, be numerous space navigations and retrieval apply Good data background is provided to support.Especially in grant land use right, environmental assessment, roading etc. is in application, usually need Remote sensing image within the scope of certain vector element is individually extracted and is analyzed.However, since related open data are pressed Certain DecryptDecryption processing has been done in the requirement of safety policy, when user attempts own vector data and stacked remotely-sensed data, meeting There is the unmatched situation of the two.Therefore, it is badly in need of a kind of efficient, automatic method, for own vector data and open remote sensing Geometrical deviation between image data is corrected.
Since vector data is different from the property of remote sensing image, and there is centainly rotationally-varying, dimensional variation and position Variation.Currently, mainstream way is that the transforming function transformation function for selecting certain ground control point to obtain between the two by hand is calibrated.It is this Method is time-consuming and laborious to the more demanding of operation precision, and there are map datums to download, and naked eyes calibrate inaccurate, scaling The problems such as degree is inconsistent is not appropriate for the conventional means as processing magnanimity geographic information data.Also one can not be used as During rigorous scientific method is applied to high-precision, large batch of data processing and updates.
In recent years, for the autoregistration problem of remote sensing image and vector data, Chinese and overseas scholars propose many methods. Joachim Hohle (2008) passes through orthography and map vector using old orthography and existing map vector as reference The image template for obtaining road junction carries out matching on new image and generates new control point, realizes that remote sensing image is automatically outer Orientation.Experiments have shown that can satisfy the requirement that orthography and map vector update substantially；Heiner Hild etc. (2001) with The registration of SPOT image and map vector is research object, and approximation between the two is determined using the initial control point manually chosen Transformation relation extracts control point in Polygonal Boundary and carries out aerial triangulation, realizes the registration of image and map；Zhang Xiaodong The autoregistration of remote sensing image and vector data is carried out based on area feature polygon feature Deng (2006)；Liu Zhiqing (2012) makes With Canny operator and Edge track, screening extracts suitable remote sensing image linear feature and carries out similarity measure, realizes remote sensing Image is registrated with the automation of vector data.
The processing step of the above method generally comprises: character selection and abstraction, vector data pretreatment, characteristic matching and Realize arrow grid registration.The problem of these research methods are just for certain concrete type and carry out, most of method still needs More man-machine interactively is wanted, there are no the mature system being fully automated occur, it is difficult to meet efficient extensive image data Process demand.
Summary of the invention
Present invention is generally directed to shape distortion and angular distortion between own vector data and open remote sensing image data are more micro- Weak, this larger feature of positional shift is matched automatically using remote sensing image pixel characteristic and the consistency of vector element feature Standard, to support the automatic cutting of remote sensing image data and extraction within the scope of certain element or element collection.It is registrated between the two extracting It is accurate, quick by calculating the relative deviation between open remote sensing image data and the own vector data of user on the basis of the factor Realize that the remote sensing image within the scope of certain element or element collection automatically extracts in ground.
The present invention provides the autoegistration methods of a kind of vector data and remote sensing image, comprising the following steps:
Step 1 is based on vector data, extracts three value matrix A using scan-line algorithm；
Step 2 extracts gray matrix G based on remote sensing image, and obtains the edge square of gray matrix G using Canny operator Battle array E；
Step 3 is traversed on three value matrix A using matrix of edge E, generates normalized edge registration factor matrix F'Side；
Step 4 is traversed on three value matrix A using gray matrix G, generates normalized gray scale registration factor matrix F'Ash；
Step 5, based on match quasi-divisor characteristic, carry out vector data and remote sensing image autoregistration；
Remote sensing image data in step 6, extraction vector element prescribed limit.
Step 1 specifically includes:
1.1, the minimum outsourcing rectangle for extracting vector data, is converted into binaryzation grid matrix Y=according to formula (1) Y (i, j) | i=0 ..., m-1；J=0 ..., n-1 }, wherein m is the height of vector data outsourcing rectangle, and n is vector data The width of outsourcing rectangle；
1.2, create Boolean type mark matrix F=f (i, j) | i=0 ..., m-1；J=0 ..., n-1 } judge whether it is Vector element edge；
1.3, to every horizontal scanning line Lt=(t, j) | j=0 ..., n-1 } (t ∈ [0, m-1]), it is right using formula (2) Indicate that matrix F carries out assignment；
1.4, according to mark matrix F, the boundary and inside of vector element are identified respectively；Creation matrix A=a (i, j) | i= 0,...,m-1；J=0 ..., n-1 }, initial default value 0 carries out three value matrixs after assignment obtains assignment according to formula (3) A, wherein vector data marginal point is identified as 2, and internal point is identified as 1, and background dot is identified as 0；
Step 2 specifically includes:
2.1, gray matrix is extracted based on open remote sensing image: reading remote sensing image data to Matrix C=c (i, j) | i= 0,...,p-1；J=0 ..., q-1 } in, according to formula (4) be processed into gray matrix G=g (i, j) | i=0 ..., p- 1；J=0 ..., q-1 }；
G (i, j)=0.299*cr(i,j)+0.587*cg(i,j)+0.114*cb(i,j) (4)
Wherein, p is remote sensing image data height, and q is remote sensing image data width, and meets condition (p > m) and (q > n)；cr(i,j)、cg(i,j)、cb(i, j) respectively indicates R, G, B value of remote sensing image C at point (i, j)；
2.2, smooth operation is carried out to gray matrix G using Gaussian filter；
A) user's intended size (2k+1) * (2k+1) and variances sigma are calculated using formula (5)2Lower Gaussian convolution core G'={ g' (x, y) | x=-k ..., k；Y=-k ..., k }；
B) convolution kernel G' and remote sensing image gray matrix G are subjected to convolution, obtain it is smooth after image array S=s (i, j) | I=0 ..., p-1；J=0 ..., q-1 }；
2.3, gradient magnitude and direction are calculated；Using formula (6), (7), (8) calculate smoothing matrix S gradient amplitude with Direction；
θ (i, j)=arctan (Px(i,j)/Py(i,j)) (8)
Wherein Px, PyRespectively gradient operator of the image on x, the direction y, Px(i, j), Py(i, j) is ladder at point (i, j) The product of operator and smoothing matrix is spent, arctan indicates tangent function, and M (i, j) is amplitude of the smoothing matrix S at (i, j), θ (i, j) is direction of the smoothing matrix S at (i, j)；
2.4, non-maxima suppression: the gradient square according to formula (9), after obtaining non-maxima suppression is carried out to gradient magnitude Battle array Grad=grad (i, j) | i=0 ..., p-1；J=0 ..., q-1 }；
Wherein MBeforeIndicate the gradient magnitude of the previous point of point (i, j) along gradient direction, MAfterwardsIt indicates along gradient direction The gradient magnitude of the latter point of point (i, j)；
2.5, according to dual-threshold voltage carry out edge detection with connect: set up high threshold δIt is highWith Low threshold δIt is low, meet condition (10) point (i, j) can be determined as marginal point, by these point be attached, obtain final edge image matrix E=e (i, J) | i=0 ..., p-1；J=0 ..., q-1 }；
G (i, j) > δIt is highOr (g (i, j) > δIt is highAnd g (i.j) > δIt is lowAnd flag (i, j)=true) (10)
Step 3 specifically includes:
3.1, the point for setting the remote sensing image upper left corner is obtained as origin using edge image matrix E and three value matrix A User has vector data by oneself in different location, and the point in the upper left corner and the point of remote sensing image origin relative displacement deviation are to set D={ (dx,dy)|0≤dx≤p-m-1；0≤dy≤ q-n-1 }, according to any deviation (d in formula (11) set of computations Dx,dy) Corresponding matrix of consequence
3.2, according to formula (12), the sum of the element in the corresponding matrix of consequence T of deviation f is calculated；
3.3, in set D each element corresponding element and composition matrix of consequence FSide={ fSide(x, y) | x=0 ..., p-m- 1；Y=0 ..., q-n-1 }, as edge registration factor matrix；
3.4, the edge registration factor matrix F that will be obtainedSide, it is normalized according to formula (13), obtains normalization edge It is registrated factor matrix F'Side={ f'Side(dx,dy)|dx=0 ..., p-m-1；dy=0 ..., q-n-1 }；
Wherein, fSide maxIndicate the maximum value of edge registration factor matrix, fSide minIndicate the minimum of edge registration factor matrix Value.
Step 4 specifically includes:
4.1, using gray level image matrix G and three value matrix A, to any deviation { (d in set Dx,dy), according to formula (14) calculate its corresponding matrix of consequence T'=t'(i, j) | i=0 ..., m-1；J=0 ..., n-1 }；
4.2, the variance f' of nonzero value in the matrix of consequence corresponding to it is calculated according to formula (15), (16)；
Wherein R is the number of nonzero value in matrix of consequence T'；
4.3, the matrix of consequence F that the corresponding variance of each element is constituted in set DAsh={ fAsh(x, y) | x=0 ..., p-m-1； Y=0 ..., q-n-1 } it is gray scale registration factor matrix；
4.4, obtained gray scale is registrated factor matrix FAsh, it is normalized according to formula (17), obtains Normalized Grey Level It is registrated factor matrix F'Ash={ f'Ash(dx,dy)|dx=0 ..., p-m-1；dy=0 ..., q-n-1 }；
Wherein, fGrey maxIndicate the maximum value of gray scale registration factor matrix, fGrey minIndicate the minimum of gray scale registration factor matrix Value.
Step 5 specifically includes:
5.1, by registration property it is found that the edge registration factor of certain differential location is bigger and gray scale is smaller with quasi-divisor When, vector data is more matched with remote sensing image at the differential location, it then follows this principle uses the comprehensive registration of formula (18) creation Factor matrix FSentence={ fSentence(i, j) | i=0 ..., p-m-1；J=0 ..., q-n-1 }；
5.2, discriminant value matrix F is traversedSentence, obtain the corresponding deviation (i of minimum value0,j0), as vector data and remote sensing The optimal deviation of Image registration.
Step 6 specifically includes:
According to formula (19) generate matrix of consequence R=r (x, y) | x=0 ..., m-1；Y=0 ..., n-1 }；The matrix Corresponding image is the remote sensing image image within the scope of vector data element or element collection, and rest part is black background；
The utility model has the advantages that compared with prior art, the present invention the invention proposes a kind of comprehensive utilization remote sensing gray level images to obtain To gray scale registration factor matrix and the edge registration factor matrix that obtains in conjunction with Canny operator, utilize the characteristic of the two to carry out The solution and optimization of remote sensing image and own vector data deviation, and finally obtain the remote sensing image within the scope of certain vector element Method.This method mainly has the following characteristics that
1) it is larger to take full advantage of positional shift between own vector data and remote sensing image, shape, angular distortion can neglect Slightly this feature；
2) comprehensive utilization Remote Sensing Image Edge image in the consistency of vector element and particular range atural object it is related Property, realize the autoregistration of vector element and remote sensing image.
Detailed description of the invention
Fig. 1 is the flow chart of the method for the present invention；
Fig. 2 is the deviation schematic diagram of remote sensing image and vector data；
Fig. 3 is the remote sensing image data that embodiment uses；
Fig. 4 is the vector data that embodiment uses；
Fig. 5 is the initial relative position schematic diagram of remote sensing image data and vector data；
Fig. 6 is three value matrix schematic diagram of vector data；
Fig. 7 is remote sensing image gray level image；
Fig. 8 is Remote Sensing Image Edge image；
Fig. 9 be deviation be (0,0) when three value matrixs dot product partial schematic diagram corresponding with edge image；
Figure 10 is edge registration factor matrix schematic diagram；
Figure 11 be deviation be (0,0) when three value matrixs dot product partial schematic diagram corresponding with gray level image；
Figure 12 is that gray scale is registrated factor matrix schematic diagram；
Figure 13 is comprehensive registration factor matrix schematic diagram；
Figure 14 is relative position schematic diagram after being registrated of remote sensing image and vector data；
Figure 15 is the result map for extracting remote sensing image within the scope of vector element.
Specific embodiment
The present invention is further explained with reference to the accompanying drawings and examples.
The present embodiment selects day map-Jiangsu publication Jiangsu Province's remote sensing image base map in 2016 as remote sensing image data (Fig. 3)；User has a part (Fig. 4) that vector data is Wuxi City park element resource by oneself.Both relative position started As shown in Figure 5.Wherein, the map sheet size of remote sensing map is 708*571 pixel, and the map sheet size that user has vector data by oneself is 612*512 pixel.
Step 1, the extraction for carrying out three value matrix of vector data；
1.1, the minimum outsourcing rectangle for extracting vector data, is converted into binaryzation grid matrix Y=according to formula (1) Y (i, j) | i=0 ..., m-1；J=0 ..., n-1 }, as shown in figure 4, in the present embodiment, m=708, n=571；
1.2, create Boolean mark matrix F=f (i, j) | i=0 ..., m-1；J=0 ..., n-1 } judge whether it is Vector element edge；
1.3, to every horizontal scanning line Lt=(t, j) | j=0 ..., n-1 } (t ∈ [0, m-1]), it is right using formula (2) Indicate that matrix F carries out assignment；
1.4, according to mark matrix F, the boundary and inside of vector element are identified respectively.Creation matrix A=a (i, j) | i= 0,...,m-1；J=0 ..., n-1 }, initial value 0 carries out assignment according to formula (3) and obtains three value matrix A, wherein vector number 2 are identified as according to marginal point, internal point is identified as 1, and background dot is identified as 0.In the present embodiment, obtained three value matrixs such as Fig. 6 institute Show, needed for display, matrix is respectively worth to pixel value of 255/2 times of the amplification as image.
Step 2 extracts gray matrix and matrix of edge based on remote sensing image；
2.1, gray matrix is extracted based on open remote sensing image, read remote sensing image data to Matrix C=c (i, j) | i= 0,...,p-1；J=0 ..., q-1 } in, according to formula (4) be processed into gray matrix G=g (i, j) | i=0 ..., p- 1；J=0 ..., q-1 }, as shown in fig. 7, in the present embodiment, p=612, q=512；
2.2, smooth operation is carried out to gray matrix G using Gaussian filter；
A) intended size (2k+1) * (2k+1) and variances sigma are calculated using formula (5)2Lower Gaussian convolution core G'=g'(x, Y) | x=-k ..., k；Y=-k ..., k }, k=1, σ are taken in the present embodiment2=1；
B) convolution kernel G' and remote sensing image gray matrix G are subjected to convolution, obtain it is smooth after image array S=s (i, j) | I=0 ..., p-1；J=0 ..., q-1 }；
2.3, gradient magnitude and direction are calculated, using formula (6), (7), (8) calculate the amplitude of the gradient of smoothing matrix S with Direction；
2.4, non-maxima suppression is carried out to gradient magnitude.Gradient square according to formula (9), after obtaining non-maxima suppression Battle array Grad=grad (i, j) | i=0 ..., p-1；J=0 ..., q-1 }；
2.5, according to dual-threshold voltage carry out edge detection with connect.Set up high threshold δIt is high=150 and Low threshold δIt is low=50.It is full The point (i, j) of sufficient condition (10) can be determined as marginal point.These points are attached, final edge image matrix E is obtained =e (i, j) | i=0 ..., p-1；J=0 ..., q-1 }, as shown in Figure 8.
Step 3 generates normalized edge registration factor matrix；
3.1, the point for setting the remote sensing image upper left corner is obtained as origin using edge image matrix E and three value matrix A User has vector data by oneself in different location, and the point of the deviation of the point and remote sensing image origin relative displacement in the upper left corner is to collection Close D={ (dx,dy)|0≤dx≤p-m-1；0≤dy≤q-n-1}.Deviation (0,0) corresponding result square is calculated according to formula (11) Battle array T=t (i, j) | i=0 ..., m-1；J=0 ..., n-1 }, partial element is as shown in Figure 9；
3.2, according to formula (12), its corresponding matrix element and f=550 are calculated；
3.3, the corresponding matrix of consequence F of set DSide={ fSide(x, y) | x=0 ..., p-m-1；Y=0 ..., q-n-1 } i.e. For edge registration factor matrix.
3.4, the edge registration factor matrix F that will be obtainedSideIt is normalized according to formula (13), obtains normalization edge and match Quasi-divisor matrix F 'Side={ f'Side(dx,dy)|dx=0 ..., p-m-1；dy=0 ..., q-n-1 }, as shown in Figure 10.
Step 4 generates normalized gray scale registration factor matrix；
4.1, using gray level image matrix G and three value matrix A, to deviation (0,0), it is corresponding that its is calculated according to formula (14) Matrix of consequence T'=t'(i, j) | i=0 ..., m-1；J=0 ..., n-1 }, partial element is as shown in figure 11；
4.2, the variance f'=1555 of nonzero value in the matrix of consequence corresponding to it is calculated according to formula (15), (16)；
4.3, the corresponding variance matrix F of set DAsh={ fAsh(x, y) | x=0 ..., p-m-1；Y=0 ..., q-n-1 } i.e. Factor matrix is registrated for gray scale.
4.4, obtained gray scale is registrated factor matrix FAshIt is normalized according to formula (17), obtains Normalized Grey Level and match Quasi-divisor matrix F 'Ash={ f'Ash(dx,dy)|dx=0 ..., p-m-1；dy=0 ..., q-n-1 }, as shown in figure 12.
Step 5, the characteristic based on registration factor matrix, carry out the autoregistration of vector data and remote sensing image；
5.1, by registration property it is found that the edge registration factor of certain differential location is bigger and gray scale is smaller with quasi-divisor When, vector data is more matched with remote sensing image at the differential location.This principle is followed, creates comprehensive registration using formula (18) Factor matrix FSentence={ fSentence(dx,dy)|dx=0 ..., p-m-1；dy=0 ..., q-n-1 }, as shown in figure 13；
5.2, discriminant value matrix F is traversedSentence, obtain the corresponding deviation of minimum value (65,47).This be vector data with it is distant Feel the optimal deviation of Image registration, registration result is as shown in figure 14.
Remote sensing image data in step 6, extraction vector element prescribed limit.
According to formula (19) generate result images matrix R=r (x, y) | x=0 ..., m-1；Y=0 ..., n-1 }.It should The corresponding image of matrix (Figure 15) is the corresponding remote sensing image image of vector data element, and rest part is black background.
As can be seen from the above embodiments, this method can be more automatic and accurately extracts the remote sensing shadow in certain vector scope As data.Compared with existing method for registering, this method has vector data and open remote sensing image data by oneself mainly for user Between autoregistration, more convenient quick in the way of the registration of pixel characteristic, high degree of automation can satisfy high-volume and wants The registration process of element needs.
In the present embodiment, the edge registration factor is selected to be registrated with gray scale with quasi-divisor in the equal mode of weight, base Originally the needs that remote sensing image image is extracted within the scope of this are met.Different types of vector data, the weight with quasi-divisor are set It sets different.If the vector data indicate be the bulks homogeneity such as forest, lake natural feature on a map, should allow gray scale registration because The specific gravity of sub- Zhan Geng great；And if its indicate be residential quarter etc. not homogeneity but with obvious boundary man-made features when, edge With quasi-divisor should Zhan Geng great specific gravity.

Claims (7)

1.一种矢量数据与遥感影像的自动配准方法，其特征在于：包括以下步骤：1. an automatic registration method of vector data and remote sensing image, is characterized in that: comprise the following steps: 步骤1、基于矢量数据，利用扫描线算法提取三值矩阵A；Step 1. Based on the vector data, use the scan line algorithm to extract the ternary matrix A; 步骤2、基于遥感影像提取灰度矩阵G，并利用Canny算子得到灰度矩阵G的边缘矩阵E；Step 2. Extract the grayscale matrix G based on the remote sensing image, and use the Canny operator to obtain the edge matrix E of the grayscale matrix G; 步骤3、使用边缘矩阵E在三值矩阵A上进行遍历，生成归一化的边缘配准因子矩阵F'边；Step 3, use the edge matrix E to traverse the three-valued matrix A to generate the normalized edge registration factor matrix F'side; 步骤4、使用灰度矩阵G在三值矩阵A上进行遍历，生成归一化的灰度配准因子矩阵F'灰；Step 4. Use the grayscale matrix G to traverse the three-valued matrix A to generate a normalized grayscale registration factor matrix F'gray; 步骤5、基于配准因子的特性，进行矢量数据与遥感影像的自动配准；Step 5, based on the characteristics of the registration factor, perform automatic registration of the vector data and the remote sensing image; 步骤6、提取矢量要素规定范围内的遥感影像数据。Step 6: Extract remote sensing image data within a specified range of vector elements. 2.根据权利要求1所述的一种矢量数据与遥感影像的自动配准方法，其特征在于：所述步骤1具体包括：2. the automatic registration method of a kind of vector data and remote sensing image according to claim 1, is characterized in that: described step 1 specifically comprises: 1.1、提取矢量数据的最小外包矩形，根据公式(1)将其转换为二值化栅格矩阵Y＝{y(i,j)|i＝0,...,m-1；j＝0,...,n-1}，其中m为矢量数据外包矩形的高度，n为矢量数据外包矩形的宽度；1.1. Extract the minimum outer rectangle of the vector data, and convert it into a binarized grid matrix according to formula (1) Y={y(i,j)|i=0,...,m-1; j=0 ,...,n-1}, where m is the height of the outer rectangle of the vector data, and n is the width of the outer rectangle of the vector data; 1.2、创建布尔型标志矩阵F＝{f(i,j)|i＝0,...,m-1；j＝0,...,n-1}判断是否为矢量要素边缘；1.2. Create a Boolean flag matrix F={f(i,j)|i=0,...,m-1; j=0,...,n-1} to judge whether it is a vector element edge; 1.3、对每条行扫描线Lt＝{(t,j)|j＝0,...,n-1}(t∈[0,m-1])，利用公式(2)对标志矩阵F进行赋值；1.3. For each row scan line L t ={(t,j)|j=0,...,n-1}(t∈[0,m-1]), use formula (2) to F for assignment; 1.4、根据标志矩阵F，分别标识矢量要素的边界与内部；创建矩阵A＝{a(i,j)|i＝0,...,m-1；j＝0,...,n-1}，初始默认值为0，根据公式(3)进行赋值得到赋值后的三值矩阵A，其中矢量数据边缘点标识为2，内部点标识为1，背景点标识为0；1.4. Identify the boundaries and interiors of vector elements according to the marker matrix F; create a matrix A={a(i,j)|i=0,...,m-1; j=0,...,n- 1}, the initial default value is 0, and the assignment is performed according to formula (3) to obtain a three-valued matrix A after assignment, wherein the edge point identifier of the vector data is 2, the internal point identifier is 1, and the background point identifier is 0; 3.根据权利要求2所述的一种矢量数据与遥感影像的自动配准方法，其特征在于：所述步骤2具体包括：3. the automatic registration method of a kind of vector data and remote sensing image according to claim 2, is characterized in that: described step 2 specifically comprises: 2.1、基于公开遥感影像提取灰度矩阵：读取遥感影像数据到矩阵C＝{c(i,j)|i＝0,...,p-1；j＝0,...,q-1}中，根据公式(4)将其处理为灰度矩阵G＝{g(i,j)|i＝0,...,p-1；j＝0,...,q-1}；2.1. Extract grayscale matrix based on public remote sensing images: read remote sensing image data to matrix C={c(i,j)|i=0,...,p-1; j=0,...,q- 1}, it is processed as a grayscale matrix G={g(i,j)|i=0,...,p-1; j=0,...,q-1} according to formula (4). ; g(i,j)＝0.299*cr(i,j)+0.587*cg(i,j)+0.114*cb(i,j) (4)g(i,j)=0.299*c r (i,j)+0.587*c g (i,j)+0.114*c b (i,j) (4) 其中，p为遥感影像数据高度，q为遥感影像数据宽度，且满足条件(p＞m)and(q＞n)；cr(i,j)、cg(i,j)、cb(i,j)分别表示点(i,j)处遥感影像C的R、G、B值；Among them, p is the height of remote sensing image data, q is the width of remote sensing image data, and satisfy the conditions (p>m) and (q>n); cr (i,j), c g ( i, j), c b ( i, j) represent the R, G, and B values of the remote sensing image C at point (i, j), respectively; 2.2、使用高斯滤波器对灰度矩阵G进行平滑操作；2.2. Use a Gaussian filter to smooth the grayscale matrix G; a)利用公式(5)计算用户给定尺寸(2k+1)*(2k+1)与方差σ2下高斯卷积核G'＝{g'(x,y)|x＝-k,...,k；y＝-k,...,k}；a) Use formula (5) to calculate the Gaussian convolution kernel G'={ g '(x,y)|x=-k, . ..,k; y=-k,...,k}; b)将卷积核G'与遥感影像灰度矩阵G进行卷积，得到平滑后图像矩阵S＝{s(i,j)|i＝0,...,p-1；j＝0,...,q-1}；b) Convolve the convolution kernel G' with the grayscale matrix G of the remote sensing image to obtain the smoothed image matrix S={s(i,j)|i=0,...,p-1; j=0, ...,q-1}; 2.3、计算梯度幅值和方向；利用公式(6)、(7)、(8)计算平滑矩阵S的梯度的幅值与方向；2.3. Calculate the gradient magnitude and direction; use formulas (6), (7), (8) to calculate the magnitude and direction of the gradient of the smoothing matrix S; θ(i,j)＝arctan(Px(i,j)/Py(i,j)) (8)θ(i,j)=arctan( Px (i,j)/ Py (i,j)) (8) 其中Px，Py分别为图像在x，y方向上的梯度算子，Px(i,j)，Py(i,j)为点(i,j)处梯度算子与平滑矩阵的乘积，arctan表示正切函数，M(i,j)为平滑矩阵S在(i,j)处的幅值，θ(i,j)为平滑矩阵S在(i,j)处的方向；Among them, P x and P y are the gradient operators of the image in the x and y directions, respectively, and P x (i,j) and P y (i, j) are the gradient operators at the point (i, j) and the smoothing matrix. Product, arctan represents the tangent function, M(i,j) is the amplitude of the smoothing matrix S at (i,j), θ(i,j) is the direction of the smoothing matrix S at (i,j); 2.4、对梯度幅值进行非极大值抑制：根据公式(9)，得到非极大值抑制后的梯度矩阵Grad＝{grad(i,j)|i＝0,...,p-1；j＝0,...,q-1}；2.4. Non-maximum suppression of gradient amplitude: According to formula (9), the gradient matrix after non-maximum suppression is obtained Grad={grad(i,j)|i=0,...,p-1 ;j=0,...,q-1}; 其中M前表示沿梯度方向上点(i,j)的前一个点的梯度幅值，M后表示沿梯度方向上点(i,j)的后一个点的梯度幅值；Wherein front M represents the gradient magnitude of the previous point along the gradient direction (i, j), and rear M represents the gradient magnitude of the next point along the gradient direction (i, j); 2.5、依据双阈值法进行边缘检测与连接：设立高阈值δ高和低阈值δ低，满足条件(10)的点(i,j)即可判定为边缘点，将这些点进行连接，得到最终的边缘图像矩阵E＝{e(i,j)|i＝0,...,p-1；j＝0,...,q-1}；2.5. Edge detection and connection according to the double-threshold method: establish a high threshold δ high and a low threshold δ low , and the point (i, j) that satisfies the condition (10) can be determined as an edge point, and these points are connected to obtain the final The edge image matrix E={e(i,j)|i=0,...,p-1; j=0,...,q-1}; g(i,j)＞δ高or(g(i,j)＜δ高and g(i.j)＞δ低and flag(i,j)＝true) (10)g(i,j)> δhigh or(g(i,j)< δhigh and g(ij)> δlow and flag(i,j)=true) (10) 4.根据权利要求3所述的一种矢量数据与遥感影像的自动配准方法，其特征在于：所述步骤3具体包括：4. the automatic registration method of a kind of vector data and remote sensing image according to claim 3, is characterized in that: described step 3 specifically comprises: 3.1、设定遥感影像左上角的点作为原点，利用边缘图像矩阵E及三值矩阵A,得到用户自有矢量数据在不同位置时，其左上角的点与遥感影像原点相对位移偏差的点对集合D＝{(dx,dy)|0≤dx≤p-m-1；0≤dy≤q-n-1}，根据公式(11)计算集合D中的任一偏差(dx,dy)对应的结果矩阵Tdxdy＝{t(i,j)|i＝0,...,m-1；j＝0,...,n-1}；3.1. Set the point in the upper left corner of the remote sensing image as the origin, and use the edge image matrix E and the ternary matrix A to obtain the point pair of the relative displacement deviation between the upper left corner of the user's own vector data and the origin of the remote sensing image when the user's own vector data is in different positions. Set D={(d x ,d y )|0≤d x ≤pm-1; 0≤d y ≤qn-1}, according to formula (11), calculate any deviation (d x ,d y in set D) ) corresponding result matrix T dxdy ={t(i,j)|i=0,...,m-1; j=0,...,n-1}; 3.2、根据公式(12)，计算该偏差对应的结果矩阵T内的要素之和f；3.2. According to formula (12), calculate the sum f of the elements in the result matrix T corresponding to the deviation; 3.3、集合D中各要素对应的要素和构成的结果矩阵F边＝{f边(x,y)|x＝0,...,p-m-1；y＝0,...,q-n-1}，即为边缘配准因子矩阵；3.3. The elements corresponding to each element in the set D and the resulting matrix F edge = {f edge (x, y) | x = 0, ..., pm-1; y = 0, ..., qn-1 } is the edge registration factor matrix; 3.4、将得到的边缘配准因子矩阵F边，根据公式(13)进行归一化，得到归一化边缘配准因子矩阵F'边＝{f'边(dx,dy)|dx＝0,...,p-m-1；dy＝0,...,q-n-1}；3.4. Normalize the obtained edge registration factor matrix F edge according to formula (13) to obtain the normalized edge registration factor matrix F' edge = {f' edge (d x , dy )|d x =0,...,pm-1; dy = 0,...,qn-1}; 其中，f边max表示边缘配准因子矩阵的最大值，f边min表示边缘配准因子矩阵的最小值。Among them, the f- side max represents the maximum value of the edge registration factor matrix, and the f- side min represents the minimum value of the edge registration factor matrix. 5.根据权利要求4所述的一种矢量数据与遥感影像的自动配准方法，其特征在于：所述步骤4具体包括：5. the automatic registration method of a kind of vector data and remote sensing image according to claim 4, is characterized in that: described step 4 specifically comprises: 4.1、利用灰度图像矩阵G及三值矩阵A,对集合D中的任一偏差{(dx,dy)}，根据公式(14)计算其对应的结果矩阵T'＝{t'(i,j)|i＝0,...,m-1；j＝0,...,n-1}；4.1. Using the grayscale image matrix G and the three-valued matrix A, for any deviation {(d x , dy )} in the set D, calculate the corresponding result matrix T'={t'( i,j)|i=0,...,m-1; j=0,...,n-1}; 4.2、根据公式(15)、(16)计算其所对应的结果矩阵中非零值的方差f'；4.2. Calculate the variance f' of non-zero values in the corresponding result matrix according to formulas (15) and (16); 其中R为结果矩阵T'中非零值的个数；where R is the number of non-zero values in the result matrix T'; 4.3、集合D中各要素对应的方差构成的结果矩阵F灰＝{f灰(x,y)|x＝0,...,p-m-1；y＝0,...,q-n-1}即为灰度配准因子矩阵；4.3. The resulting matrix F gray = {f gray (x, y)|x = 0,..., pm-1; y = 0,..., qn-1} composed of the variance corresponding to each element in the set D is the grayscale registration factor matrix; 4.4、将得到的灰度配准因子矩阵F灰，根据公式(17)进行归一化，得到归一化灰度配准因子矩阵F'灰＝{f'灰(dx,dy)|dx＝0,...,p-m-1；dy＝0,...,q-n-1}；4.4. Normalize the obtained grayscale registration factor matrix F gray according to formula (17) to obtain a normalized grayscale registration factor matrix F' gray = {f' gray (d x , dy )| d x =0,...,pm-1; dy =0,...,qn-1}; 其中，f灰max表示灰度配准因子矩阵的最大值，f灰min表示灰度配准因子矩阵的最小值。Among them, f gray max represents the maximum value of the gray registration factor matrix, and f gray min represents the minimum value of the gray registration factor matrix. 6.根据权利要求5所述的一种矢量数据与遥感影像的自动配准方法，其特征在于：所述步骤5具体包括：6. the automatic registration method of a kind of vector data and remote sensing image according to claim 5, is characterized in that: described step 5 specifically comprises: 5.1、由配准因子特性可知，某偏差位置的边缘配准因子越大而灰度配准因子越小时，该偏差位置处矢量数据与遥感影像越匹配，遵循这一原则，使用公式(18)创建综合配准因子矩阵F判＝{f判(i,j)|i＝0,...,p-m-1；j＝0,...,q-n-1}；5.1. From the characteristics of the registration factor, it can be known that the larger the edge registration factor of a certain deviation position and the smaller the gray registration factor, the more matched the vector data at the deviation position with the remote sensing image. Following this principle, formula (18) is used. Create a comprehensive registration factor matrix F = {fjudgment ( i,j)|i=0,...,pm-1; j=0,...,qn-1}; 5.2、遍历判别值矩阵F判，得到最小值对应的偏差值(i0,j0)，即为矢量数据与遥感影像配准的最优偏差值。5.2. Traverse the discriminant value matrix F , and obtain the deviation value (i 0 , j 0 ) corresponding to the minimum value, which is the optimal deviation value of the registration of the vector data and the remote sensing image. 7.根据权利要求6所述的一种矢量数据与遥感影像的自动配准方法，其特征在于：所述步骤6具体包括：7. the automatic registration method of a kind of vector data and remote sensing image according to claim 6, is characterized in that: described step 6 specifically comprises: 根据公式(19)生成结果矩阵R＝{r(x,y)|x＝0,...,m-1；y＝0,...,n-1}；该矩阵对应的图像为矢量数据要素或要素集范围内的遥感影像图像，其余部分为黑色背景；Generate the result matrix R={r(x,y)|x=0,...,m-1; y=0,...,n-1} according to formula (19); the image corresponding to this matrix is a vector Remote sensing imagery images within the range of data elements or feature sets, with the rest on a black background;
CN201910083298.XA 2018-06-04 2019-01-28 Automatic registration method of vector data and remote sensing image Active CN109727279B (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN2018105622875 2018-06-04
CN201810562287 2018-06-04

Publications (2)

Publication Number Publication Date
CN109727279A true CN109727279A (en) 2019-05-07
CN109727279B CN109727279B (en) 2022-07-29

Family

ID=66301207

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910083298.XA Active CN109727279B (en) 2018-06-04 2019-01-28 Automatic registration method of vector data and remote sensing image

Country Status (1)

Country Link
CN (1) CN109727279B (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110378316A (en) * 2019-07-29 2019-10-25 苏州中科天启遥感科技有限公司 A kind of method and system for extracting remote sensing image Objects recognition sample
CN111696121A (en) * 2020-06-05 2020-09-22 中国人民解放军火箭军工程设计研究院 Planar water area extraction method and system
CN112950680A (en) * 2021-02-20 2021-06-11 哈尔滨学院 Satellite remote sensing image registration method
CN114897659A (en) * 2022-05-09 2022-08-12 南京师范大学 A zero-watermark generation algorithm for vector geographic data and a method for detecting zero-watermark information
CN115830087A (en) * 2022-12-09 2023-03-21 陕西航天技术应用研究院有限公司 Batch rapid registration technology for translational motion continuous frame image sets

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040052409A1 (en) * 2002-09-17 2004-03-18 Ravi Bansal Integrated image registration for cardiac magnetic resonance perfusion data
CN101957991A (en) * 2010-09-17 2011-01-26 中国科学院上海技术物理研究所 Remote sensing image registration method
CN102842137A (en) * 2012-08-14 2012-12-26 中山大学 Automatic registration method for multi-temporal empty spectrum remote sensing image based on space comprehensive mutual information
CN104167003A (en) * 2014-08-29 2014-11-26 福州大学 Method for fast registering remote-sensing image
CN105654423A (en) * 2015-12-28 2016-06-08 西安电子科技大学 Area-based remote sensing image registration method
CN107301661A (en) * 2017-07-10 2017-10-27 中国科学院遥感与数字地球研究所 High-resolution remote sensing image method for registering based on edge point feature

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040052409A1 (en) * 2002-09-17 2004-03-18 Ravi Bansal Integrated image registration for cardiac magnetic resonance perfusion data
CN101957991A (en) * 2010-09-17 2011-01-26 中国科学院上海技术物理研究所 Remote sensing image registration method
CN102842137A (en) * 2012-08-14 2012-12-26 中山大学 Automatic registration method for multi-temporal empty spectrum remote sensing image based on space comprehensive mutual information
CN104167003A (en) * 2014-08-29 2014-11-26 福州大学 Method for fast registering remote-sensing image
CN105654423A (en) * 2015-12-28 2016-06-08 西安电子科技大学 Area-based remote sensing image registration method
CN107301661A (en) * 2017-07-10 2017-10-27 中国科学院遥感与数字地球研究所 High-resolution remote sensing image method for registering based on edge point feature

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110378316A (en) * 2019-07-29 2019-10-25 苏州中科天启遥感科技有限公司 A kind of method and system for extracting remote sensing image Objects recognition sample
CN110378316B (en) * 2019-07-29 2023-06-27 苏州中科天启遥感科技有限公司 Method and system for extracting ground object identification sample of remote sensing image
CN111696121A (en) * 2020-06-05 2020-09-22 中国人民解放军火箭军工程设计研究院 Planar water area extraction method and system
CN112950680A (en) * 2021-02-20 2021-06-11 哈尔滨学院 Satellite remote sensing image registration method
CN112950680B (en) * 2021-02-20 2022-07-05 哈尔滨学院 Satellite remote sensing image registration method
CN114897659A (en) * 2022-05-09 2022-08-12 南京师范大学 A zero-watermark generation algorithm for vector geographic data and a method for detecting zero-watermark information
CN114897659B (en) * 2022-05-09 2023-12-29 南京师范大学 A zero-watermark generation method for vector geographic data and a zero-watermark information detection method
CN115830087A (en) * 2022-12-09 2023-03-21 陕西航天技术应用研究院有限公司 Batch rapid registration technology for translational motion continuous frame image sets
CN115830087B (en) * 2022-12-09 2024-02-20 陕西航天技术应用研究院有限公司 A fast batch registration method for continuous frame image sets of translational motion

Also Published As

Publication number Publication date
CN109727279B (en) 2022-07-29

Similar Documents

Publication Publication Date Title
CN109727279A (en) An automatic registration method of vector data and remote sensing images
CN111968172B (en) Method and system for measuring volume of stock ground material
CN107563438B (en) A Fast and Robust Multimodal Remote Sensing Image Matching Method and System
CN110727747B (en) Paper map rapid vectorization method and system based on longitude and latitude recognition
CN110263717B (en) A land-use category determination method incorporating street view imagery
CN107463918B (en) Lane line extraction method based on fusion of laser point cloud and image data
Fan et al. Registration of optical and SAR satellite images by exploring the spatial relationship of the improved SIFT
CN106340044B (en) Camera external parameter automatic calibration method and calibration device
CN103383773B (en) The remote sensing satellite image of a kind of dynamic extraction Image Control Point is the most just penetrating framework and the method for correction
Huang et al. Feature-based image registration using the shape context
CN102855628B (en) Automatic matching method for multisource multi-temporal high-resolution satellite remote sensing image
CN107688777B (en) A method for extracting urban green space from collaborative multi-source remote sensing images
CN112184785B (en) Multi-mode remote sensing image registration method based on MCD measurement and VTM
CN106920235A (en) Star-loaded optical remote sensing image automatic correction method based on the matching of vector base map
CN110070545B (en) A Method for Automatically Extracting Urban Built-up Areas from Urban Texture Feature Density
CN108230375A (en) Visible images and SAR image registration method based on structural similarity fast robust
CN103136525A (en) High-precision positioning method for special-shaped extended target by utilizing generalized Hough transformation
CN103761722A (en) A Method for Accurate Stitching of Fixed-wing UAV Patrol Images on Power Transmission Lines
JP2012073845A (en) Computer system and method for alignment of image and graph
JP6188052B2 (en) Information system and server
CN101246595A (en) Multi-view point cloud data combination method in optical 3D scanning system
Gonçalves et al. CHAIR: Automatic image registration based on correlation and Hough transform
CN110852176A (en) High-resolution three-number SAR image road detection method based on Mask-RCNN
CN106097319A (en) A kind of black and white gridiron pattern Angular Point Extracting Method for camera calibration
CN119516386A (en) A method and system for fine division of urban neighborhood land based on deep learning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant