CN109685837B - Heterologous remote sensing image registration method based on feature structure similarity - Google Patents
Heterologous remote sensing image registration method based on feature structure similarity Download PDFInfo
- Publication number
- CN109685837B CN109685837B CN201811318809.3A CN201811318809A CN109685837B CN 109685837 B CN109685837 B CN 109685837B CN 201811318809 A CN201811318809 A CN 201811318809A CN 109685837 B CN109685837 B CN 109685837B
- Authority
- CN
- China
- Prior art keywords
- points
- shape
- point
- image
- similarity
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 239000011159 matrix material Substances 0.000 claims description 17
- 238000004364 calculation method Methods 0.000 claims description 15
- 238000000605 extraction Methods 0.000 claims description 14
- 238000009826 distribution Methods 0.000 claims description 8
- 238000005070 sampling Methods 0.000 claims description 8
- 230000009466 transformation Effects 0.000 claims description 7
- 230000008569 process Effects 0.000 claims description 6
- 238000011524 similarity measure Methods 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 4
- 238000012952 Resampling Methods 0.000 claims description 3
- 238000003708 edge detection Methods 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims description 3
- 230000003287 optical effect Effects 0.000 abstract description 6
- 238000010586 diagram Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 238000013329 compounding Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10044—Radar image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20021—Dividing image into blocks, subimages or windows
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a heterogeneous remote sensing image registration method based on feature structure similarity, which comprises the steps of firstly uniformly extracting a plurality of characteristic points on a reference image by utilizing a segmented Harris operator and forming a plurality of characteristic point sets; then extracting the image edge by using a canny operator, and calculating a shape similarity descriptor at the feature point according to a shape context algorithm; performing feature point matching according to the shape similarity descriptor to obtain homonymous points, and removing the feature points which are erroneously matched; and finally, establishing a corresponding relation between the reference image and the input image according to the correctly matched characteristic points to match the input image. The invention can effectively improve the image registration precision between the optical image and the SAR image, and is convenient for subsequent use.
Description
Technical Field
The invention relates to a method for registering a heterologous remote sensing image, in particular to a method for registering a heterologous remote sensing image based on the structural similarity of ground objects.
Background
Image registration is an indispensable component for dynamic monitoring, change information extraction, information compounding and other works. The remote sensing images of different sensors, multiple resolutions and multiple phases reflect different characteristics of the ground object, and the remote sensing data need to be matched for obtaining respective advantages. However, due to different imaging mechanisms, gray level differences and geometric differences often exist between images of the heterologous remote sensing images (namely, the optical images and the SAR images), so that characteristic points and homonymous points are difficult to obtain for matching during image matching, and the problem of the registration of the heterologous remote sensing images is always a hot spot and a difficult point of current world research.
The current remote sensing image registration method mainly comprises the following steps: region (gray) based methods and feature based methods. The region-based matching method is a template matching method, and uses a certain similarity measure as a criterion to identify homonymous points between images, and finally, geometric transformation model parameters between the images are estimated according to the homonymous points. Common similarity measures are correlation coefficients, phase correlations, mutual information, etc. The region-based matching method has high calculation efficiency and high registration accuracy, but is very sensitive to nonlinear gray scale differences among images, so that the region-based matching method cannot be well suitable for automatic registration of heterologous remote sensing images. Therefore, the geometric properties between images are less affected by the gray scale differences than the gray scale information.
The feature-based method is that features such as feature points, lines and planes are extracted from a reference image, descriptors of the features are constructed, similarity among the descriptors is used for matching, and finally registration of the images is completed. The method can be well adapted to geometric structure differences among images, but when nonlinear gray scale differences among images are large, stable common features among the images are difficult to detect, the repetition rate of feature point detection is greatly reduced, and finally the image registration effect is poor.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a heterologous remote sensing image registration method based on the structure similarity of ground objects, which can effectively improve the image registration precision between an optical image and an SAR image and is convenient for subsequent use.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows: a heterogeneous remote sensing image registration method based on feature structure similarity comprises the following specific steps:
(1) Feature extraction: extracting characteristic points of the reference image by adopting a segmented Harris operator;
(2) Construction of shape similarity descriptors: constructing a shape similarity descriptor by using a shape context algorithm according to the feature structure similarity;
(3) Feature matching: the method of template matching is adopted, NCC between the constructed shape similarity descriptors is used as similarity measure (SC NCC ) Searching a point with the maximum correlation with the characteristic point in the searching area to be used as a homonymy point of the characteristic point;
(4) And (5) removing the mismatching points: setting a standard error value (such as 1 pixel), calculating residual errors and root mean square errors of all the homonymous points, comparing all the errors with the standard error value, and removing the homonymous points if the errors exceed the standard error value so as to obtain the homonymous points which are correctly matched;
(5) Image registration: and estimating geometric transformation model parameters between images according to the correctly matched homonymous points, geometrically transforming the input images by using a transformation model, resampling the gray values of the transformed images, and finally completing image registration.
Further, the specific process of feature extraction in the step (1) is as follows: the reference image is divided into a plurality of grids which are mutually non-overlapped and regular, the Harris intensity value H of each pixel in each grid is calculated, the H values are arranged in a sequence from big to small, and the first n points with larger H values in each grid are selected as characteristic point sets of each grid, so that the conditions that extracted characteristic points are too dense or rare and the like can be effectively avoided.
Further, as the shape context is a feature description method based on the shape contour, the distribution situation of sampling points on the contour can be well reflected by using the histogram to describe the shape feature in a logarithmic polar coordinate system; based on the shape context principle, the following is concrete:
step1: for a given shape, the contour edge is obtained by an edge detection operator (such as a canny operator), and the contour edge is sampled to obtain a group of discrete point sets p i (i＝1,2,...,N)；
Step2: calculating a shape context; at any point p therein i At p as reference point i And establishing n concentric circles at logarithmic distance intervals in the area with the circle center and the radius R. This region is equally divided in the circumferential direction m to form a target template. Point p i The relative positions of the vectors to other points are reduced to the number of point distributions in each sector on the template. Statistical distribution histogram h of these points i (k) Called point p i The calculation formula of the shape context of (a) is as follows:
h i (k)＝#{q≠p i :(q-p i )∈bin(k)}
wherein k= {1,2,..k }, k=m x n;
the use of logarithmic distance segmentation enables shape context descriptors to be more sensitive to neighboring sampling points than to distant points, thereby enabling local feature enhancement. The shape context at different points of the contour is different, but there tends to be a similar shape context at corresponding points of a similar contour.
For the whole point set p, N points p are respectively used 1 ,p 2 ,...,p n And (3) taking the shape histogram as a reference point, sequentially calculating shape histograms formed by the left N-1 points, finally obtaining N shape histograms, and storing the N shape histograms in a matrix with the size of N (N-1). Thus, for any target, N (N-1) can be usedThe matrix of size represents its shape information, and the matrix of size N x (N-1) is the shape context of the point set P, which characterizes the overall outline shape. The more sampling points, the finer the shape expression, and the more the calculation amount is multiplied.
Step3: similarity calculation between the shape contexts and cost matrix cost;
wherein h is i (k) Point P being the target P i Is a histogram of the shape of (a); h is a i (k) Point Q being target Q i Is a histogram of the shape of (a). And obtaining a cost matrix C between two targets according to a formula, wherein the size of the cost matrix C is N.
Step4: based on the cost matrix C obtained by calculation, carrying out point matching operation by using a Hungary operator to obtain the minimum value by the following formula;
H(π)＝∑C(p i ,q π(i) )
to this end, the method represents the similarity of two objects in a non-vector number. Since the calculation is based on the cost matrix, the larger the result, the less similar the result, and the smaller the result, the more similar the result.
According to the above principle, the specific process of constructing the shape similarity descriptor in the step (2) is as follows:
A. the feature point set extracted from the reference image is recorded as p i (i=1, 2,.. N.) a search area of a certain size is predicted on the input image based on the known geographic coordinate information of the reference image and the input image for homonymous point matching (i.e. the feature points on the reference image are mapped to the input influence by establishing a mapping relationship between the two images by the known ground control points between the reference image and the input image, thereby predicting the size of the search area);
B. extracting feature point set p i Selecting a template window with a set size on a reference image by taking the point as the center (the size of the template window can be set arbitrarily according to the requirement);
C. setting the size of a cell (namely a window unit) and a canny operator extraction threshold (namely a threshold set when the canny operator extracts an image boundary, different boundaries can be extracted by setting different thresholds, namely boundaries with different definition) and extracting the template window boundary by using the canny operator extraction threshold first, and dividing the window into a plurality of cells to form a basic structure of a shape similarity descriptor.
D. The shape context of all cells is collected to form a shape similarity descriptor describing the template window.
The key parameters of the invention are the threshold value set when the image contour is extracted by using the canny operator and the size of the cell, and the optimal matching of the image can be sought by setting the cell with different sizes and the canny operator threshold value.
Further, the residual equation and the root mean square error equation of the homonymous point in the step (4) are respectively:
(dx,dy)＝(x 2 ,y 2 )-(x 1 -y 1 )
compared with the prior art, the method has the advantages that firstly, a plurality of characteristic points are uniformly extracted on a reference image by utilizing the Harris operator of the block; then extracting the image edge by using a canny operator, and calculating a shape similarity descriptor at the feature point according to a shape context algorithm; performing feature point matching according to the shape similarity descriptor to obtain homonymous points, and removing the feature points which are erroneously matched; and finally, establishing a corresponding relation between the reference image and the input image according to the correctly matched characteristic points to match the input image. The invention successfully realizes the image registration between the optical image and the SAR image, and the registration accuracy can reach 0.65 pixel.
Drawings
FIG. 1 is an overall flow chart of the present invention;
FIG. 2 is a flow chart of the shape similarity descriptor construction in the present invention;
FIG. 3 is a diagram of a reference image and an input image according to the present invention;
FIG. 4 is a contour map of the reference image and the input image of FIG. 3 extracted by the canny operator;
FIG. 5 is a feature point extraction diagram of the reference image and the input image of FIG. 3;
FIG. 6 is a shape context of one of the feature points of FIG. 5;
FIG. 7 is a graph of image registration accuracy with different cell sizes and channel operator thresholds;
fig. 8 is a schematic illustration of the registered fig. 3 using the present invention.
In the upper graph: (a) is a reference image; (b) is an input image.
Detailed Description
The present invention will be further described below.
Examples: the optical image data is from Band3 red light wave Band data on the land satellite series Landsat5 TM sensor in the United states, because the experimental area is provided with a plurality of houses, roads and other ground features, and the red light wave Band can just better distinguish the types of the artificial ground features. The spatial resolution was 30m and the image acquisition time was 1 month in 2003. SAR image data come from earth observation satellite series ENVISAT satellite ASAR sensors of European space agency, and the spatial resolution is 20m. The image acquisition time was 3 months 2004.
As shown in the figure, the method for registering the heterologous remote sensing images based on the structural similarity of the ground objects comprises the following specific steps:
(1) Feature extraction: extracting characteristic points of a reference image (namely an optical image) by adopting a segmented Harris operator;
(2) Construction of shape similarity descriptors: constructing a shape similarity descriptor by using a shape context algorithm according to the feature structure similarity;
(3) Feature matching: the method of template matching is adopted, NCC between the constructed shape similarity descriptors is used as similarity measure (SC NCC ) Searching a point with the maximum correlation with the characteristic point in the searching area to be used as a homonymy point of the characteristic point;
(4) And (5) removing the mismatching points: setting a standard error value (such as 1 pixel), calculating residual errors and root mean square errors of all the homonymous points, comparing all the errors with the standard error value, and removing the homonymous points if the errors exceed the standard error value so as to obtain the homonymous points which are correctly matched;
(5) Image registration: estimating geometric transformation model parameters between images according to the correctly matched homonymous points, geometrically transforming the input image (namely SAR image) by utilizing a transformation model, resampling the gray value of the transformed image, and finally completing image registration.
Further, the specific process of feature extraction in the step (1) is as follows: the reference image is divided into a plurality of grids which are mutually non-overlapped and regular, the Harris intensity value H of each pixel in each grid is calculated, the H values are arranged in a sequence from big to small, and the first n points with larger H values in each grid are selected as characteristic point sets of each grid, so that the conditions that extracted characteristic points are too dense or rare and the like can be effectively avoided.
Further, as the shape context is a feature description method based on the shape contour, the distribution situation of sampling points on the contour can be well reflected by using the histogram to describe the shape feature in a logarithmic polar coordinate system; based on the shape context principle, the following is concrete:
step1: for a given shape, the contour edge is obtained by an edge detection operator (such as a canny operator), and the contour edge is sampled to obtain a group of discrete point sets p i (i＝1,2,...,N)；
Step2: calculating a shape context; at any point p therein i At p as reference point i And establishing n concentric circles at logarithmic distance intervals in the area with the circle center and the radius R. This region is equally divided in the circumferential direction m to form a target template. Point p i The relative positions of the vectors to other points are reduced to the number of point distributions in each sector on the template. Statistical distribution histogram h of these points i (k) Called point p i The calculation formula of the shape context of (a) is as follows:
h i (k)＝#{q≠p i :(q-p i )∈bin(k)}
wherein k= {1,2,..k }, k=m x n;
the use of logarithmic distance segmentation enables shape context descriptors to be more sensitive to neighboring sampling points than to distant points, thereby enabling local feature enhancement. The shape context at different points of the contour is different, but there tends to be a similar shape context at corresponding points of a similar contour.
For the whole point set p, N points p are respectively used 1 ,p 2 ,...,p n And (3) taking the shape histogram as a reference point, sequentially calculating shape histograms formed by the N-1 points, and finally obtaining N shape histograms. Stored in a matrix of size N x (N-1). Thus, for any object, its shape information can be represented by an N (N-1) sized matrix, which is the shape context of the point set P, that characterizes the overall outline shape. The more sampling points, the finer the shape expression, and the more the calculation amount is multiplied.
Step3: similarity calculation between the shape contexts and cost matrix cost;
wherein h is i (k) Point P being the target P i Is a histogram of the shape of (a); h is a i (k) Point Q being target Q i Is a histogram of the shape of (a). And obtaining a cost matrix C between two targets according to a formula, wherein the size of the cost matrix C is N.
Step4: based on the cost matrix C obtained by calculation, carrying out point matching operation by using a Hungary operator to obtain the minimum value by the following formula;
H(π)＝∑C(p i ,q π(i) )
to this end, the method represents the similarity of two objects in a non-vector number. Since the calculation is based on the cost matrix, the larger the result, the less similar the result, and the smaller the result, the more similar the result.
According to the above principle, the specific process of constructing the shape similarity descriptor in the step (2) is as follows:
A. the feature point set extracted from the reference image is recorded as p i (i=1, 2,.. N.) a search area of a certain size is predicted on the input image based on the known geographic coordinate information of the reference image and the input image for homonymous point matching (i.e. the feature points on the reference image are mapped to the input influence by establishing a mapping relationship between the two images by the known ground control points between the reference image and the input image, thereby predicting the size of the search area);
B. extracting a point p in the feature point set i Selecting a template window with the size of 100 x 100 on the reference image by taking the point as the center;
C. setting the size of cells and a canny operator extraction threshold (namely, a threshold set when the canny operator extracts the image boundary), extracting different boundaries by setting different thresholds, namely, extracting boundaries with different definition, firstly extracting the boundary of a template window by using the canny operator extraction threshold, and then dividing the window into a plurality of cells to form a basic structure of a shape similarity descriptor, so that the shape context of the central point in each cell is only calculated for improving the calculation efficiency.
D. The shape context of all cells is collected to form a shape similarity descriptor describing the template window.
Therefore, the key parameters of the invention are the threshold value set when the image contour is extracted by using the canny operator and the size of the cell, and the best matching of the image is sought by setting the cell and the canny operator threshold values with different sizes.
Further, the residual equation and the root mean square error equation of the homonymous point in the step (4) are respectively:
(dx,dy)＝(x 2 ,y 2 )-(x 1 -y 1 )
directing attention to fig. 7 and 8, it can be seen that the image registration accuracy under the conditions of different cell sizes and canny operator thresholds in the present invention is that the registration accuracy is the highest, where parameters are set as follows: the cell size is 15×15, the channel operator threshold is 0.2, and the precision value is 0.65 pixels. Table 1 is a comparison of the present invention and the conventional image registration method (NCC) and the geometric similarity measure (HOPC) based image registration method, and it can be seen that: the feature point matching accuracy of the invention is obviously higher than that of the other two algorithms.
Table 1: image registration result comparison
The above embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention, therefore, any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (3)
1. A method for registering heterogeneous remote sensing images based on feature structure similarity is characterized by comprising the following specific steps:
(1) Feature extraction: extracting characteristic points of the reference image by adopting a segmented Harris operator;
(2) Construction of shape similarity descriptors: according to the structural similarity of the ground object, a shape similarity descriptor is constructed by using a shape context algorithm, and the method specifically comprises the following steps:
A. the extracted characteristic point set on the reference image is recorded as k i (i=1, 2,., N) predicting a search area of a certain size on the input image based on the known reference image and the geographic coordinate information of the input image;
B. extracting a point k in the feature point set i Selecting a template window with a set size on a reference image by taking the point as the center;
C. setting the size of cells and a canny operator extraction threshold, extracting edge points of a template window by using the canny operator extraction threshold, dividing the window into a plurality of cells, and forming a basic structure of a shape similarity descriptor; based on the shape context principle, the specific calculation process is as follows:
step1: for a given shape, acquiring a contour edge through a canny edge detection operator, and sampling the contour edge to obtain a group of discrete point sets p i (i＝1,2,...,N)；
Step2: calculating a shape context; at any point p therein i At p as reference point i Establishing n concentric circles at logarithmic distance intervals in a region with a circle center and R as a radius; equally dividing the area along the circumferential direction m to form a target-shaped template; point p i The relative positions of vectors to other points are simplified into the distribution number of points in each sector on the template; statistical distribution histogram h of these points i (k) Called point p i The calculation formula of the shape context of (a) is as follows:
h i (k)＝#{q≠p i :(q-p i )∈bin(k)}
wherein k= {1,2,..k }, k=m x n;
the adoption of logarithmic distance segmentation can make the shape context descriptors more sensitive to adjacent sampling points than distant points, so that local characteristics can be enhanced; the shape context at different points of the contour is different, but there tends to be a similar shape context at corresponding points of a similar contour;
for the whole point set p, N points p are respectively used 1 ,p 2 ,...,p n Sequentially calculating shape histograms formed by the left N-1 points by taking the shape histograms as reference points, finally obtaining N shape histograms, and storing the N shape histograms in a matrix with the size of N (N-1); thus, for any object, its shape information can be represented by a matrix of size N x (N-1), which is the shape context of the point set P, characterizing the overall outline shape; finally, to improve the calculation efficiency, only the shape context of the center point in each cell is calculated;
D. collecting the shape context of all cells to form a shape similarity descriptor describing a template window;
(3) Feature matching: adopting a template matching method, taking NCC between the constructed shape similarity descriptors as similarity measure, and searching a point with the maximum correlation with the feature point in a search area as a homonymy point of the feature point;
(4) And (5) removing the mismatching points: setting a standard error value, calculating residual errors and root mean square errors of all the homonymous points, comparing all the errors with the standard error value, and removing the homonymous points if the errors exceed the standard error value, so as to obtain the homonymous points which are correctly matched;
(5) Image registration: and estimating geometric transformation model parameters between images according to the correctly matched homonymous points, geometrically transforming the input images by using a transformation model, resampling the gray values of the transformed images, and finally completing image registration.
2. The method for registering heterogeneous remote sensing images based on the structural similarity of ground objects according to claim 1, wherein the specific process of feature extraction in the step (1) is as follows: dividing the reference image into a plurality of grids which are mutually non-overlapping and regular, calculating the Harris intensity value H of each pixel in each grid, arranging the H values in a sequence from large to small, and selecting the first n points with larger H values in each grid as a characteristic point set of each grid.
3. The method for registering heterogeneous remote sensing images based on feature structural similarity according to claim 1, wherein the residual formulas and root mean square error formulas of the homonymous points in the step (4) are respectively:
(dx,dy)＝(x 2 ,y 2 )-(x 1 -y 1 )
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811318809.3A CN109685837B (en) | 2018-11-07 | 2018-11-07 | Heterologous remote sensing image registration method based on feature structure similarity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811318809.3A CN109685837B (en) | 2018-11-07 | 2018-11-07 | Heterologous remote sensing image registration method based on feature structure similarity |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109685837A CN109685837A (en) | 2019-04-26 |
| CN109685837B true CN109685837B (en) | 2024-03-22 |
Family
ID=66184563
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811318809.3A Active CN109685837B (en) | 2018-11-07 | 2018-11-07 | Heterologous remote sensing image registration method based on feature structure similarity |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109685837B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110039525A (en) * | 2019-05-25 | 2019-07-23 | 塞伯睿机器人技术（长沙）有限公司 | Robot is used in 6S management |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102200999A (en) * | 2011-04-27 | 2011-09-28 | 华中科技大学 | Method for retrieving similarity shape |
| CN104021556A (en) * | 2014-06-13 | 2014-09-03 | 西南交通大学 | Heterological remote-sensing image registration method based on geometric structure similarity |
-
2018
- 2018-11-07 CN CN201811318809.3A patent/CN109685837B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102200999A (en) * | 2011-04-27 | 2011-09-28 | 华中科技大学 | Method for retrieving similarity shape |
| CN104021556A (en) * | 2014-06-13 | 2014-09-03 | 西南交通大学 | Heterological remote-sensing image registration method based on geometric structure similarity |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109685837A (en) | 2019-04-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111028277B (en) | SAR and optical remote sensing image registration method based on pseudo-twin convolution neural network | |
| CN109029381B (en) | Tunnel crack detection method and system and terminal equipment | |
| CN109409292B (en) | Heterogeneous image matching method based on refined feature optimization extraction | |
| CN107038717B (en) | A method of 3D point cloud registration error is automatically analyzed based on three-dimensional grid | |
| Hou et al. | Deep learning-based subsurface target detection from GPR scans | |
| Tian et al. | Building change detection based on satellite stereo imagery and digital surface models | |
| CN104574347B (en) | Satellite in orbit image geometry positioning accuracy evaluation method based on multi- source Remote Sensing Data data | |
| CN107067405B (en) | Remote sensing image segmentation method based on scale optimization | |
| CN111027446B (en) | Coastline automatic extraction method of high-resolution image | |
| CN103839265A (en) | SAR image registration method based on SIFT and normalized mutual information | |
| CN106886988B (en) | Linear target detection method and system based on unmanned aerial vehicle remote sensing | |
| CN106780503A (en) | Remote sensing images optimum segmentation yardstick based on posterior probability information entropy determines method | |
| CN111369495A (en) | Video-based panoramic image change detection method | |
| CN113947724A (en) | Automatic line icing thickness measuring method based on binocular vision | |
| CN107862319A (en) | A kind of heterologous high score optical image matching error elimination method based on neighborhood ballot | |
| Guo et al. | KD-tree-based euclidean clustering for tomographic SAR point cloud extraction and segmentation | |
| CN109685837B (en) | Heterologous remote sensing image registration method based on feature structure similarity | |
| CN116863349A (en) | Remote sensing image change area determining method and device based on triangular network dense matching | |
| Sidike et al. | Automatic building change detection through adaptive local textural features and sequential background removal | |
| CN117078726A (en) | Different spectrum image registration method based on edge extraction | |
| CN107945218B (en) | Edge large distortion image matching method | |
| Hao et al. | Robust multisource remote sensing image registration method based on scene shape similarity | |
| Wang et al. | Point cloud registration by combining shape and intensity contexts | |
| Wu et al. | An accurate feature point matching algorithm for automatic remote sensing image registration | |
| Ren et al. | SAR image matching method based on improved SIFT for navigation system |
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 |