CN104809724A - Automatic precise registration method for multiband remote sensing images - Google Patents
Automatic precise registration method for multiband remote sensing images Download PDFInfo
- Publication number
- CN104809724A CN104809724A CN201510188808.1A CN201510188808A CN104809724A CN 104809724 A CN104809724 A CN 104809724A CN 201510188808 A CN201510188808 A CN 201510188808A CN 104809724 A CN104809724 A CN 104809724A
- Authority
- CN
- China
- Prior art keywords
- training sample
- matching double
- coordinate
- double points
- matching
- 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.)
- Pending
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses an automatic precise registration method for multiband remote sensing images, including the steps of: screening wave bands from the multiband remote sensing images as training samples, extracting feature points of each training sample and the corresponding standard samples and matching the feature points, respectively performing abnormal match detection on the matching point pair of each training sample, deleting the abnormal matching point pair, making statistics on occurrence numbers of the same matching point pair from the residual matching point pairs, selecting the matching point pairs with occurrence numbers more than a preset threshold as common control point pairs, obtaining a common control point pair set, generating a coordinate conversion model according to the common control point pair set, performing coordinate conversion on all training sample feature points in the other matching point pairs, screening the matching point pairs with small errors according to a preset proportion, at last generating a final registration model according to the common control point pair and the screened matching point pairs, and registering the image to be registered. According to the automatic precise registration method, the established registration model is high in precision, and automatic precise registration of the remote sensing images can be achieved.
Description
Technical field
The invention belongs to Remote Sensing Image Processing Technology field, more specifically say, relate to a kind of automatic precision method for registering of multiband remote sensing image.
Background technology
Along with the fast development of remote sensing technology, satellite remote sensing system can provide the earth observation data of dynamic realtime for people, the continuous progress of remote sensing image obtaining means and raising, also for the remote sensing applications such as the variation monitoring between multi_temporal images provide the data basis of magnanimity.But, because earth rotation, topographic relief, atmospheric refraction and air cloud layer move, add the impact of the image-forming conditions such as sensor track position and attitudes vibration, in the remote sensing image that same atural object obtains in different phase, there will be the situation that locus is inconsistent, the application scenario that sub-pix etc. is higher to accuracy requirement cannot be met.Therefore, by smart registration process, several images are transformed to reference under image coordinate system, thus image is mated on locus be mapped, becoming the important topic of field of remote sensing image processing, is also simultaneously the problem that must solve based on the remote sensing application of multi-temporal data.
The smart method for registering of tradition is mainly through setting up ground control point database, and manually or semi-automatically choose reference mark, but this method operating process is tediously long, precision is not high enough, cannot meet the demand of automatic batch process.Smart method for registering development in recent years is mainly divided into three roughly directions: based on area grayscale, based on image feature with based on explaining similar registration Algorithm.Registration Algorithm based on area grayscale is simple to operate, but to very responsive with reference to grey scale change between image and training sample.Carry out on the expert system that similar registration Algorithm needs to be based upon picture automatic interpretation based on explaining, do not make a breakthrough progress so far.The registration Algorithm that feature based is relevant extracts image feature, compressed image information in feature space, then mates feature, thus the image reducing grey scale change rings, and enhances robustness, is therefore more applicable to the needs of automatic batch essence registration.
But existing main stream approach is nearly all carry out calculating and processing for single width image or single wave band, but common remotely-sensed data is all multi-wavelength data mostly, its all band is not just fully used like this, cause the serious waste of image information, cause the result precision of smart registration not high.
Summary of the invention
The object of the invention is to overcome the deficiencies in the prior art, a kind of automatic precision method for registering based on multiband remote sensing image is provided, makes full use of each band class information of multiband remote sensing image, realize the full-automatic smart registration of image.
For achieving the above object, the automatic precision method for registering of multiband remote sensing image of the present invention comprises the following steps:
S1: for the multiband remote sensing image needing to carry out registration collected, according to pre-set level choose wherein several wave bands as training sample;
S2: respectively feature point extraction is carried out for corresponding baseline sample in every width training sample and reference images;
S3: respectively the unique point of every width training sample is mated with the unique point of corresponding baseline sample, obtain M
iindividual matching double points, i represents the sequence number of training sample, i=1,2 ..., K, K represent the quantity of training sample;
S4: obtain matching double points to step S3 and carry out Outlier match detection, concrete grammar is:
S4.1: the matching result that each training sample is obtained with corresponding baseline sample, using the unique point coordinate of training sample in each matching double points as input, using the unique point coordinate of baseline sample as output, training obtains coordinate transformation model corresponding to this width training sample;
S4.2: for every width training sample, adopts corresponding coordinate transformation model to carry out coordinate transform to the unique point of training sample in matching double points, obtains the coordinate transforming of this unique point;
S4.3: for every width training sample, calculates the error R between the coordinate of each training sample unique point in matching double points and coordinate transforming
k, the span of k is k=1,2 ..., M
i, then calculate total root-mean-square error RMSE of training sample unique point in all matching double points; If the error R of a kth unique point
k>=d × RMSE, d be default be more than or equal to 1 constant, then the matching double points of correspondence is deleted, otherwise retains;
S5: to the matching double points of training samples all after Outlier match check processing, the occurrence number that statistics identical match point is right, matching double points occurrence number being greater than predetermined threshold value, as common control pair, obtains common matching double points set;
S6: coordinate exchange model is generated to set according to common control, coordinate transform is carried out and error of calculation R to all training sample unique points in other matching double points beyond set to common control, sort from small to large according to error R, according to preset ratio screening several matching double points above;
S7: the common control obtained according to step S5 screens to set and step S6 the matching double points obtained and generates final registration model, treats registering images carry out registration according to this registration model.
The automatic precision method for registering of multiband remote sensing image of the present invention, wave band is screened as training sample from multiband remote sensing image, then feature point extraction and Feature Points Matching are carried out to baseline sample corresponding in every width training sample and reference images, respectively Outlier match detection is carried out to the matching double points of every width training sample, by Exceptional point to deletion, the occurrence number that identical match point is right is added up from remaining matching double points, occurrence number is selected to be greater than the matching double points of predetermined threshold value as common control pair, obtain common control to set, according to common control, coordinate transformation model is generated to set again, coordinate transform is carried out to training sample unique points all in other matching double points, the matching double points less according to preset ratio screening error, finally according to common control, final registration model is generated to the matching double points with screening, treat registering images and carry out registration.Present invention utilizes each band class information of multiband remote sensing image, the registration model precision of foundation is higher, and the automatic precision that can realize remote sensing image corrects, and effectively eliminates the error between image, ensures Geometrical consistency.
Accompanying drawing explanation
Fig. 1 is the process flow diagram of the automatic precision method for registering that the present invention is based on multiband remote sensing image;
Fig. 2 is Moravec operator computing method schematic diagram.
Embodiment
Below in conjunction with accompanying drawing, the specific embodiment of the present invention is described, so that those skilled in the art understands the present invention better.Requiring particular attention is that, in the following description, when perhaps the detailed description of known function and design can desalinate main contents of the present invention, these are described in and will be left in the basket here.
Embodiment
Fig. 1 is the process flow diagram of the automatic precision method for registering that the present invention is based on multiband remote sensing image.As shown in Figure 1, the automatic precision method for registering that the present invention is based on multiband remote sensing image comprises the following steps:
S101: multiband remote sensing image is screened:
For the multiband remote sensing image collected, according to pre-set level choose wherein several wave bands as training sample.Adopt information entropy as the index of basis for selecting herein, concrete grammar is: the information entropy calculating each wave band image, the amount of image information comprised in this wave band image of the larger expression of information entropy is abundanter, therefore information entropy is greater than the wave band image of predetermined threshold value as training sample.The computing formula of information entropy H is:
Wherein, p
nrepresent that gray-scale value is the pixel proportion in the picture of n.
Except the wave band image filtered out, for the situation that some fixing type of ground objects of survey region is outstanding, the Indices distribution plan that can also add default index as one of training sample, for feature point detection provides more abundant image information.
S102: feature point extraction:
The every width training sample respectively step S101 obtained and carry out feature point extraction as baseline sample corresponding in the reference images of registration reference.
Because remote sensing image has possessed certain precision after system compensation or rough registration, consider system effectiveness, used Moravec (Mo Laweike) operator to carry out feature point extraction in the present embodiment, simple and quick, it is convenient to realize.Fig. 2 is Moravec operator computing method schematic diagram.As shown in Figure 2, centered by pixel (x, y), get the window of a w × w, this pixel and surrounding four direction in calculation window (be respectively level, vertical, 45 °, 135 °) the autocorrelation function of pixel gray-scale value, computing formula is:
Wherein w (x, y) represents square two-value window, if pixel is in this window, its value is 1, otherwise is 0; (u, v) represents four direction.Choose the interest value of the minimum value in four direction auto-correlation function value as this pixel.Then set threshold value, traversal image, is greater than the pixel alternatively point of this threshold value by interest value.The window of a fixed size is finally set, allow this window travel through imaged image, after each moving window using candidate point maximum for interest value in window as unique point.
S103: Feature Points Matching:
Respectively the unique point of every width training sample is mated with the unique point of corresponding baseline sample, thus obtain M
iindividual matching double points.Each matching double points comprises the unique point of a training sample and the unique point of baseline sample.
Normalized-cross-correlation function method is adopted to carry out the coupling of unique point in this example, its main thought is the normalized-cross-correlation function of the match window respectively in Calculation Basis sample centered by unique point in template window and training sample, when feature point pair matching, its normalized-cross-correlation function reaches maximum and is greater than threshold value.The related coefficient of normalized-cross-correlation function method after past average and normalized becomes insensitive to picture contrast, light and shade, thus makes that normalized-cross-correlation function is more reliable, adaptability is stronger.
When carrying out Feature point correspondence coupling, because the size of remote sensing image data in practical application is often comparatively large, in view picture image, global search can be consuming time very long.In the prior art, because the remote sensing image data after system compensation comprises geo-localisation information, therefore first can position each unique point in baseline sample and obtain unique point coordinate, in training sample centered by this unique point coordinate, in the window of pre-sizing, search for the matching characteristic point of this unique point.Adopt and greatly can increase matching efficiency in this way, save time.
S104: Outlier match check processing:
In the Feature Points Matching of step S103, what unavoidably there will be registration coordinate and its actual coordinate generation substantial deviation mismatches situation, therefore needs to carry out Outlier match detection to the matching result of every width training sample and baseline sample.Based on RANSAC theory of algorithm, the present invention utilizes total root-mean-square error to weigh matching effect, calculate total root-mean-square error by actual coordinate and the coordinate after coordinate transform, after rejecting the obvious irrational individuality of error, correct matching result can be obtained.Outlier match detects and comprises the following steps:
S4.1: obtain coordinate transformation model:
For the matching result of every width training sample with corresponding baseline sample, using training sample unique point coordinate in each matching double points as input, using baseline sample unique point coordinate as output, training obtains coordinate transformation model corresponding to this width training sample.
Adopt quadratic polynomial coordinate transformation model in the present embodiment, its expression formula is:
Wherein, n represents the quantity of matching double points, n=M in this step
i, the coordinate of (t, s) representation feature point in training sample, the coordinate of (ω, λ) representation feature point in baseline sample, polynomial coefficient vector A
t={ a
11, a
12..., a
21, a
22and A
s={ b
11, b
12..., b
21, b
22least square method can be utilized to solve:
Wherein, T={t
1, t
2..., t
n, S={s
1, s
2..., s
n,
S4.2: coordinate transform is carried out to the unique point in training sample:
For every width training sample, coordinate transformation model is adopted to carry out coordinate transform, M in note training sample to coupling centering training sample unique point
ithe coordinate of individual unique point is (t
k, s
k), the span of k is k=1,2 ..., M
i, calculate at the projection coordinate (t of each unique point in benchmark image according to the registration model of step S104
k', s
k').
S4.3: calculate the error of each unique point and total root-mean-square error:
The error R of unique point k
kcomputing formula be:
Calculate total root-mean-square error RMSE of all unique points of coupling centering training sample, computing formula is:
S4.4: detect Outlier match and process:
If the error R of a kth unique point
k>=d × RMSE, d be default be more than or equal to 1 constant, then the matching double points of correspondence is deleted, otherwise retains, thus by Exceptional point to detecting and removing.
S105: screening common control pair:
The matching double points obtained between different-waveband is concentrated, and there is the situation that some matching double points repeatedly occurs.Because these points all show the characteristic of unique point in different situations, so these common controls are in the middle of image, relative to other matching double points, there is better representativeness, also comprise higher reliability.Based on these common controls pair, then filter out the higher control point set of precision.
The concrete grammar obtaining common control right is: in the matching result to images to be matched all after Outlier match check processing, the occurrence number that statistics identical match point is right, matching double points occurrence number being greater than predetermined threshold value, as common control pair, obtains common control to set.
S106: screening matching double points:
According to the common control that step S105 obtains, coordinate transformation model is generated to set, then coordinate transform is carried out and error of calculation R sorts from small to large according to error R to the training sample unique point in other match points beyond set to common control, according to preset ratio screening several matching double points above.
S107: generate final registration model:
Generating final registration model according to the common control that step S105 obtains to screening with step S106 the matching double points obtained, treating registering images according to this registration model and carrying out registration.
In order to verify beneficial effect of the present invention, test with the Landsat TM5 data instance of some.The root-mean-square error RMSE that each time data processing obtains is 4.8232m, i.e. 0.16 pixel, and much smaller than 1 pixel, visible the present invention is relative to general method for registering, and precision has had and significantly improves.Calculate average out to 6.05s consuming time, each time, time dimension used was held in 5 ~ 7 seconds, consuming time basicly stable, and the impact by many factors differences such as the quality of image own, gray scales is very little, is very suitable for the batch processing of remotely-sensed data.
Although be described the illustrative embodiment of the present invention above; so that those skilled in the art understand the present invention; but should be clear; the invention is not restricted to the scope of embodiment; to those skilled in the art; as long as various change to limit and in the spirit and scope of the present invention determined, these changes are apparent, and all innovation and creation utilizing the present invention to conceive are all at the row of protection in appended claim.
Claims (6)
1. an automatic precision method for registering for multiband remote sensing image, is characterized in that, comprise the following steps:
S1: for the multiband remote sensing image collected, according to pre-set level choose wherein several wave bands as training sample;
S2: respectively feature point extraction is carried out for corresponding baseline sample in every width training sample and reference images;
S3: respectively the unique point of every width training sample is mated with the unique point of corresponding baseline sample, obtain M
iindividual matching double points, i represents the sequence number of training sample, i=1,2 ..., K;
S4: obtain matching double points to step S3 and carry out Outlier match detection, concrete grammar is:
S4.1: the matching result that each training sample is obtained with corresponding baseline sample, using the unique point coordinate of training sample in each matching double points as input, using the unique point coordinate of baseline sample as output, training obtains coordinate transformation model corresponding to this width training sample;
S4.2: for every width training sample, adopts corresponding coordinate transformation model to carry out coordinate transform to the unique point of training sample in matching double points, obtains the coordinate transforming of this unique point;
S4.3: for every width training sample, calculates the error R between the coordinate of each training sample unique point in matching double points and projection coordinate
k, the span of k is k=1,2 ..., M
i, then calculate total root-mean-square error RMSE of training sample unique point in all matching double points; If the error R of a kth unique point
k>=d × RMSE, d be default be more than or equal to 1 constant, then the matching double points of correspondence is deleted, otherwise retains;
S5: to the matching double points of training samples all after Outlier match check processing, the occurrence number that statistics identical match point is right, matching double points occurrence number being greater than predetermined threshold value, as common control pair, obtains common matching double points set;
S6: coordinate exchange model is generated to set according to common control, coordinate transform is carried out and error of calculation R to all training sample unique points in other match points beyond set to common control, sort from small to large according to error R, according to preset ratio screening several matching double points above;
S7: the common control obtained according to step S5 screens to set and step S6 the matching double points obtained and generates final registration model, treats registering images carry out registration according to this registration model.
2. automatic precision method for registering according to claim 1, is characterized in that, in described step S1, the method choosing training sample is: the information entropy calculating each wave band image, information entropy is greater than the wave band image of predetermined threshold value as image subject to registration.
3. automatic precision method for registering according to claim 1, is characterized in that, in described step S1, affects sample and also comprises Indices distribution plan.
4. automatic precision method for registering according to claim 1, is characterized in that, in described step S2, feature point extraction adopts Moravec operator to extract unique point.
5. automatic precision method for registering according to claim 1, is characterized in that, in described step S3, Feature Points Matching adopts normalized-cross-correlation function to mate.
6. automatic precision method for registering according to claim 1, is characterized in that, described coordinate transformation model adopts quadratic polynomial model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510188808.1A CN104809724A (en) | 2015-04-21 | 2015-04-21 | Automatic precise registration method for multiband remote sensing images |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510188808.1A CN104809724A (en) | 2015-04-21 | 2015-04-21 | Automatic precise registration method for multiband remote sensing images |
Publications (1)
Publication Number | Publication Date |
---|---|
CN104809724A true CN104809724A (en) | 2015-07-29 |
Family
ID=53694529
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510188808.1A Pending CN104809724A (en) | 2015-04-21 | 2015-04-21 | Automatic precise registration method for multiband remote sensing images |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104809724A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105741295A (en) * | 2016-02-01 | 2016-07-06 | 福建师范大学 | High-resolution remote sensing image registration method based on local invariant feature point |
CN105844587A (en) * | 2016-03-17 | 2016-08-10 | 河南理工大学 | Low-altitude unmanned aerial vehicle-borne hyperspectral remote-sensing-image automatic splicing method |
CN108428220A (en) * | 2018-03-05 | 2018-08-21 | 武汉大学 | Satellite sequence remote sensing image sea island reef region automatic geometric correction method |
CN109272041A (en) * | 2018-09-21 | 2019-01-25 | 联想(北京)有限公司 | The choosing method and device of characteristic point |
CN110889432A (en) * | 2019-10-29 | 2020-03-17 | 北京迈格威科技有限公司 | Feature point matching method and device, computer equipment and storage medium |
CN111536991A (en) * | 2020-04-26 | 2020-08-14 | 哈尔滨工业大学 | Remote sensing satellite multi-mode imaging load simulation device |
CN111581407A (en) * | 2020-04-20 | 2020-08-25 | 国家卫星气象中心(国家空间天气监测预警中心) | Method, device and medium for constructing global geographic positioning reference image database |
CN112053350A (en) * | 2020-09-04 | 2020-12-08 | 国家卫星气象中心(国家空间天气监测预警中心) | Splicing and cutting processing method and device for remote sensing data of region of interest and computer equipment |
CN112598717A (en) * | 2020-12-14 | 2021-04-02 | 珠海欧比特宇航科技股份有限公司 | Full-spectrum registration method and medium for hyperspectral satellite images |
CN115035340A (en) * | 2022-06-13 | 2022-09-09 | 电子科技大学 | Remote sensing image classification result verification method |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6515939B1 (en) * | 1985-09-04 | 2003-02-04 | The United States Of America As Represented By The Secretary Of The Navy | Pulse sampled optical fiber hydrophone array (U) |
CN102436652A (en) * | 2011-08-31 | 2012-05-02 | 航天恒星科技有限公司 | Automatic registering method of multisource remote sensing images |
CN103020945A (en) * | 2011-09-21 | 2013-04-03 | 中国科学院电子学研究所 | Remote sensing image registration method of multi-source sensor |
CN104182949A (en) * | 2014-08-18 | 2014-12-03 | 武汉大学 | Image inking and fusing method and system based on histogram feature point registration |
-
2015
- 2015-04-21 CN CN201510188808.1A patent/CN104809724A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6515939B1 (en) * | 1985-09-04 | 2003-02-04 | The United States Of America As Represented By The Secretary Of The Navy | Pulse sampled optical fiber hydrophone array (U) |
CN102436652A (en) * | 2011-08-31 | 2012-05-02 | 航天恒星科技有限公司 | Automatic registering method of multisource remote sensing images |
CN103020945A (en) * | 2011-09-21 | 2013-04-03 | 中国科学院电子学研究所 | Remote sensing image registration method of multi-source sensor |
CN104182949A (en) * | 2014-08-18 | 2014-12-03 | 武汉大学 | Image inking and fusing method and system based on histogram feature point registration |
Non-Patent Citations (1)
Title |
---|
丘文涛: "多源图像配准技术的研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105741295A (en) * | 2016-02-01 | 2016-07-06 | 福建师范大学 | High-resolution remote sensing image registration method based on local invariant feature point |
CN105844587A (en) * | 2016-03-17 | 2016-08-10 | 河南理工大学 | Low-altitude unmanned aerial vehicle-borne hyperspectral remote-sensing-image automatic splicing method |
CN105844587B (en) * | 2016-03-17 | 2019-03-19 | 河南理工大学 | A kind of low latitude UAV system target in hyperspectral remotely sensed image method for automatically split-jointing |
CN108428220A (en) * | 2018-03-05 | 2018-08-21 | 武汉大学 | Satellite sequence remote sensing image sea island reef region automatic geometric correction method |
CN109272041A (en) * | 2018-09-21 | 2019-01-25 | 联想(北京)有限公司 | The choosing method and device of characteristic point |
CN110889432B (en) * | 2019-10-29 | 2022-07-29 | 北京迈格威科技有限公司 | Feature point matching method and device, computer equipment and storage medium |
CN110889432A (en) * | 2019-10-29 | 2020-03-17 | 北京迈格威科技有限公司 | Feature point matching method and device, computer equipment and storage medium |
CN111581407B (en) * | 2020-04-20 | 2023-09-12 | 国家卫星气象中心(国家空间天气监测预警中心) | Method, device and medium for constructing global geographic positioning reference image database |
CN111581407A (en) * | 2020-04-20 | 2020-08-25 | 国家卫星气象中心(国家空间天气监测预警中心) | Method, device and medium for constructing global geographic positioning reference image database |
CN111536991A (en) * | 2020-04-26 | 2020-08-14 | 哈尔滨工业大学 | Remote sensing satellite multi-mode imaging load simulation device |
CN112053350A (en) * | 2020-09-04 | 2020-12-08 | 国家卫星气象中心(国家空间天气监测预警中心) | Splicing and cutting processing method and device for remote sensing data of region of interest and computer equipment |
CN112053350B (en) * | 2020-09-04 | 2024-03-26 | 国家卫星气象中心(国家空间天气监测预警中心) | Method and device for splicing and cutting remote sensing data of region of interest and computer equipment |
CN112598717A (en) * | 2020-12-14 | 2021-04-02 | 珠海欧比特宇航科技股份有限公司 | Full-spectrum registration method and medium for hyperspectral satellite images |
CN112598717B (en) * | 2020-12-14 | 2024-05-17 | 珠海欧比特卫星大数据有限公司 | Full spectrum registration method and medium for hyperspectral satellite images |
CN115035340A (en) * | 2022-06-13 | 2022-09-09 | 电子科技大学 | Remote sensing image classification result verification method |
CN115035340B (en) * | 2022-06-13 | 2024-05-14 | 电子科技大学 | Remote sensing image classification result verification method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104809724A (en) | Automatic precise registration method for multiband remote sensing images | |
CN110705457B (en) | Remote sensing image building change detection method | |
CN110136170B (en) | Remote sensing image building change detection method based on convolutional neural network | |
US11150231B2 (en) | Method for measuring a mudflat elevation by remotely sensed water content | |
CN110929607B (en) | Remote sensing identification method and system for urban building construction progress | |
CN109493320B (en) | Remote sensing image road extraction method and system based on deep learning, storage medium and electronic equipment | |
CN103279759B (en) | A kind of vehicle front trafficability analytical procedure based on convolutional neural networks | |
Patil et al. | Image classification tool for land use/land cover analysis: A comparative study of maximum likelihood and minimum distance method | |
Li | Remotely sensed images and GIS data fusion for automatic change detection | |
Skidmore | Accuracy assessment of spatial information | |
Sisodia et al. | A comparative analysis of remote sensing image classification techniques | |
CN103235810B (en) | Remote sensing image reference mark data intelligence search method | |
CN107239795A (en) | SAR image change detecting system and method based on sparse self-encoding encoder and convolutional neural networks | |
CN111814597A (en) | Urban function partitioning method coupling multi-label classification network and YOLO | |
CN113936214A (en) | Karst wetland vegetation community classification method based on fusion of aerospace remote sensing images | |
CN105354841A (en) | Fast matching method and system for remote sensing images | |
Zhang et al. | Reconstruction of GF-1 soil moisture observation based on satellite and in situ sensor collaboration under full cloud contamination | |
CN107967454B (en) | Double-path convolution neural network remote sensing classification method considering spatial neighborhood relationship | |
CN103839075A (en) | SAR image classification method based on united sparse representation | |
CN101561882B (en) | Sub-pixel spatial mapping method based on spatial correlation | |
Putri | Analysis of land cover classification results using ann, svm, and rf methods with r programming language (case research: Surabaya, Indonesia) | |
CN104157009A (en) | Multi-source remote sensing image quality quantification comparison and selection method | |
Yang et al. | Urban roads network detection from high resolution remote sensing | |
CN104809336A (en) | Method for sampling region factor by considering spatial correlation | |
Talebi et al. | Automated classification of urban areas for storm water management using aerial photography and LiDAR |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
EXSB | Decision made by sipo to initiate substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20150729 |