CN102799898A - Efficient target identification processing method of high-resolution remote sensing image in context - Google Patents
Efficient target identification processing method of high-resolution remote sensing image in context Download PDFInfo
- Publication number
- CN102799898A CN102799898A CN2012102260343A CN201210226034A CN102799898A CN 102799898 A CN102799898 A CN 102799898A CN 2012102260343 A CN2012102260343 A CN 2012102260343A CN 201210226034 A CN201210226034 A CN 201210226034A CN 102799898 A CN102799898 A CN 102799898A
- Authority
- CN
- China
- Prior art keywords
- image data
- destination object
- target
- border
- image
- 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
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses an efficient target identification processing method of a high-resolution remote sensing image in a context. The efficient target identification processing method comprises the special steps as follows: 1. graying original image data, and setting an empirical value of a target object, so as to obtain a gray threshold value and a minimum bounding rectangle of the target object; 2. carrying out encoding on the minimum bounding rectangle by using a decimal Morton code, so as to establish a linear quad-tree; 3. extracting and generating a border polygon of the target object; 4. cutting an original image target data block by using the border polygon of the target object; 5. dynamically establishing a dynamic virtual clustering machine, and distributing the image target data block to a corresponding calculation node and calculating according to demands; and 6. collecting results of the calculation nodes in each level by a cluster main control calculation node, and returning to be at a corresponding position of the original image data according to an address code in the step 2 according to the step in order. With the adoption of the efficient processing method provided by the invention, the image data calculation is automatically blocked and distributed; the man-machine interaction is reduced; and meanwhile, the image treatment efficiency is further improved by using a thought of local calculation.
Description
Technical field
The invention belongs to remote sensing image applied technical field category, relate generally to high-resolution remote sensing image Target Recognition high-efficient treatment method under a kind of overall background.
Background technology
The continuous development of remote sensing technology be accompanied by remote sensing image in the space, the raising of space-time, resolution, cause the data volume of each scape image to heighten, the also corresponding increase of calculated load.According to the regularity and the correlativity characteristics of image data storage, its algorithm also has the characteristics of consistance, neighborhood property, row succession, thereby has created good condition for the image parallel computation.In addition, the development of Along with computer technology, storage, the applied research about remote sensing image data both at home and abroad turned to distributed type assemblies calculating, high performance parallel process etc. techniques.This provides technical foundation for further studying efficient processing of big data remote sensing image.
Current; The research of efficiently processing aspect of relevant remote sensing image; A large amount of work has been done by lot of domestic and foreign experts and scholars, research institution and commercial company; The tissue that its research mainly concentrates on image data and the improvement and the lifting of concrete Processing Algorithm aspect, the major technique that the remote sensing image high-efficient treatment method adopts at present have with parallel data and are treated to the High-Performance Computing Cluster treatment technology on basis and with large-scale distributed grid computing technology or the cloud computing technology that is treated to the basis.The thought of wherein studying in cluster system data parallel research and adopting Local treatment is less to the data distributing method of the division of image data and cluster computing node; Though existing method can promote the performance of image processing to a certain extent; If but machine imposes reasonable mechanism in due course, its space that aspect efficient processing performance, will have a certain upgrade.The inventive method is based on the few relatively characteristics of remote sensing image destination number under the overall background, proposes a kind ofly towards the cluster high-efficient treatment method of remote sensing image based on data parallel, can significantly reduce the computing time with the target extraneous areas, improves Target Recognition efficient.
Summary of the invention
The objective of the invention is deficiency to prior art; The existing cluster computing technique data parallel thought of utilization; Adopt a kind of active data to divide and the computing node organizational form, proposed high-resolution remote sensing image Target Recognition high-efficient treatment method under a kind of overall background.
The technical solution adopted for the present invention to solve the technical problems is following:
Step (1). the gray processing of raw video data is handled, the empirical value of target setting object, the gray threshold and the minimum outsourcing rectangle of acquisition destination object;
The raw video data are carried out gray processing through the weighted mean value method handle, the image analytic degree through image data calculates the big or small of destination object with empirical value and draws the gray threshold of destination object then; And will handle the grey scale pixel value of back image data and the gray threshold of destination object and compare, finally obtain the minimum outsourcing rectangle (MBR) of all destination objects.
Concrete comparison procedure is following:
Thought quadtree structure the processed image data is divided into blocks, the image data is OK
, column
number of pixels contained respectively
and
,
If
and
is odd number; When row when then dividing, row;
,
; If cut-off rule is through destination object when dividing; Promptly divide a grey scale pixel value of being expert at or being listed as in the gray threshold scope of destination object; Then the sub-block at this destination object place is no longer divided, and this sub-block is called the minimum outsourcing rectangle corresponding with destination object; Otherwise continue to divide.
Described destination object is often referred to the user's interest ground object target, like the included steamer in sea target recognition field, naval vessels, maritime patrol aircraft.
Step (2). adopt decimal system Morton sign indicating number to the minimum outsourcing rectangle structure linear quadtree of encoding;
With the address code of decimal system Morton sign indicating number (abbreviation
sign indicating number) as linear quadtree; Carry out after block divides at the image data of step (1), the method through from bottom to top makes up linear quadtree to minimum outsourcing rectangle;
Step (3). polygonal extraction in destination object border and generation;
All minimum outsourcing rectangles that step (1) is marked off; Press the from left to right scanning one by one of ranks order; If scanning draws in this minimum outsourcing rectangle a plurality of destination objects are arranged; Then the pixel to a plurality of destination objects that can merge in this minimum outsourcing rectangle merges the border polygon that then produces corresponding destination object that can not merge; In the polygonal extraction in destination object border and generating, must be through the vector quantization process of a raster data, this vector quantization process comprises following basic step:
1) polygonal Boundary Extraction: adopt high-pass filtering with the grating image binaryzation or with particular value sign frontier point;
2) follow the trail of the boundary line: to each border segmental arc by a node to its another adjacent node searching; Common frontier point to each known approach axis; Remove self approach axis external demand along the next adjacent frontier points of other seven direction searchs, up to being linked to be the border segmental arc.
3) topological relation generates: for the border segmental arc of vector representation, judge each polygonal spatial relationship on itself and the former figure, form complete topological structure then through the mathematics geometric relationship, and the getting in touch of foundation and attribute data.
4) remove unnecessary point and the slick and sly curve that belongs to Polygonal Boundary: because node searching is that individual element carries out, must remove the unnecessary point that causes thus, to reduce data redundancy through matched curve; Adopt interpolation algorithm that the curve that belongs to Polygonal Boundary is carried out smooth treatment simultaneously.
Benefit algorithm commonly used includes linear iteration technique, segmentation cubic polynomial interpolation method, positive axis of a parabola average weighted method, inclined shaft para-curve average weighted method, splines method of interpolation.
Step (4). use the border polygon cutting raw video target data block of destination object;
The minimum outsourcing rectangle of all destination objects that employing step (1) obtains cuts the raw video data as the vector polygon of corresponding destination object, and the target block quantity of statistics raw video data cutting generation; Because the border polygon can influence the merging of image data at cutting process,, be convenient to inlaying of result after the target processing with complete therefore to polygonal border extension 2~3 pixels of object boundary.
Step (5). adopt load balancing thought and quad-tree structure thought to organize dynamic construction dynamic virtual cluster machine, the silhouette target data block that step (4) is obtained is distributed to corresponding computing node and enhancing, fusion, classification and the cloud of image data is monitored calculate on request;
The Enhancement Method of described image data comprises: the contrast based on the spatial transform method strengthens, histogram strengthens, based on convolution enhancing, the sharpening enhancing of airspace filter method; And based on high pass, the LPF image Enhancement Method of frequency domain method.
The fusion of image data comprises: IHS converter technique, main composition converter technique, high-pass filtering fusion method, Brovery conversion fusion method.
The classification of image data comprises: supervised classification, unsupervised classification.
The monitoring of the cloud of image data comprises: based on the method for spectrum, based on the method for texture; Wherein spectrographic technique has: ISCCP method, APLOOL method, CO slice method, CLAVR method etc.; Texture method has co-occurrence matrix method, the laws texture energy method based on statistical model, and based on the method for structure with based on the method for field model.
Step (6). cluster master control computing node is collected the result of each level computing node, and returns the relevant position of raw video data according to the order of sequence according to the address code of step (2).
The beneficial effect that the present invention compared with prior art has is:
(1) realizes that image data calculates automatic piecemeal and distribution processor, reduce man-machine interaction.Traditional calculates parallel processing based on cluster; Comprise tasks in parallel and data parallel; Wherein the parallel of data do not have the accurate data partition strategy usually, and the data distribution processing also relatively arbitrarily, and the time of data parallel processing simultaneously also depends primarily on the maximum duration of data block calculation cost.Given this, the present invention has mainly used computer cluster efficient parallel processing power and the load balance process mechanism based on data parallel.Concrete grammar shows that mainly image data is divided and organizing of cluster computing node all adopted quad-tree structure thought; The thought of using Local treatment has simultaneously reduced the calculating in the uncorrelated zone that has nothing to do with target in the image processing process, has greatly improved the efficient of target processing.The present invention under overall background in the image data target processing application prospect very big; Especially the application aspect fields such as civilian traffic, military affairs such as marine Ship Target identification is more extensive; Can be applied to identification of general high resolution image datum target and classification processing through expansion, have great using value such as emergency response such as disaster and fast processing.
(2) to the few relatively characteristics of institute's recognition objective quantity in the Target Recognition process under the overall background, the thought that adopts localization to calculate is proposed.Its most outstanding characteristics are only to calculate the minimum outsourcing rectangle MBR relevant with Target Recognition through using the position relation of Morton sign indicating number record object, have shortened this method greatly in the image data processing time.Size through the estimation destination object is provided with the gray threshold of piecemeal automatically, for the user definition threshold parameter is offered reference, thereby provide outsourcing rectangle more accurately with satisfy realistic objective the zone that will calculate.
Description of drawings
Fig. 1 is a process flow diagram of the present invention;
Fig. 2 is four fork number partitioned organization and Morton coding rule figure of the present invention;
Fig. 3 is that data of the present invention are divided and Morton coding synoptic diagram;
Fig. 4 is cluster computing node of the present invention and image data piece distribution synoptic diagram.
Embodiment
Below in conjunction with accompanying drawing and embodiment the present invention is described further.
As shown in Figure 1, high-resolution remote sensing image Target Recognition high-efficient treatment method under the overall background, concrete performing step is following:
Step (1). the gray processing of raw video data is handled, the empirical value of target setting object, the gray threshold and the minimum outsourcing rectangle of acquisition destination object;
The raw video data are carried out gray processing through the weighted mean value method handle, the image analytic degree through image data calculates the big or small of destination object with empirical value and draws the gray threshold of destination object then; And will handle the grey scale pixel value of back image data and the gray threshold of destination object and compare, finally obtain the minimum outsourcing rectangle (MBR) of all destination objects.
Concrete comparison procedure is as shown in Figure 2:
Thought quadtree structure the processed image data is divided into blocks, the image data row
, column
number of pixels contained respectively
and
,
If
and
is odd number; When row when then dividing, row;
,
; If cut-off rule is through destination object when dividing; Promptly divide a grey scale pixel value of being expert at or being listed as in the gray threshold scope of destination object; Then the sub-block at this destination object place is no longer divided, and this sub-block is called the minimum outsourcing rectangle corresponding with destination object; Otherwise continue to divide.
Described destination object is often referred to the user's interest ground object target, like the included steamer in sea target recognition field, naval vessels, maritime patrol aircraft.
Step (2). adopt decimal system Morton sign indicating number to the minimum outsourcing rectangle structure linear quadtree of encoding;
With the address code of decimal system Morton sign indicating number (abbreviation
sign indicating number) as linear quadtree; Carry out after block divides at the image data of step (1), the method through from bottom to top makes up linear quadtree to minimum outsourcing rectangle;
Step (3). polygonal extraction in destination object border and generation;
All minimum outsourcing rectangles that step (1) is marked off; Press the from left to right scanning one by one of ranks order; If scanning draws in this minimum outsourcing rectangle a plurality of destination objects are arranged; Then the pixel to a plurality of destination objects that can merge in this minimum outsourcing rectangle merges the border polygon that then produces corresponding destination object that can not merge; In the polygonal extraction in destination object border and generating, must be through the vector quantization process of a raster data, this vector quantization process comprises following basic step:
1) polygonal Boundary Extraction: adopt high-pass filtering with the grating image binaryzation or with particular value sign frontier point;
2) follow the trail of the boundary line: to each border segmental arc by a node to its another adjacent node searching; Common frontier point to each known approach axis; Remove self approach axis external demand along the next adjacent frontier points of other seven direction searchs, up to being linked to be the border segmental arc.
3) topological relation generates: for the border segmental arc of vector representation, judge each polygonal spatial relationship on itself and the former figure, form complete topological structure then through the mathematics geometric relationship, and the getting in touch of foundation and attribute data.
4) remove unnecessary point and the slick and sly curve that belongs to Polygonal Boundary: because node searching is that individual element carries out, must remove the unnecessary point that causes thus, to reduce data redundancy through matched curve; Adopt interpolation algorithm that the curve that belongs to Polygonal Boundary is carried out smooth treatment simultaneously.
Benefit algorithm commonly used includes linear iteration technique, segmentation cubic polynomial interpolation method, positive axis of a parabola average weighted method, inclined shaft para-curve average weighted method, splines method of interpolation.
Step (4). use the border polygon cutting raw video target data block of destination object;
The minimum outsourcing rectangle of all destination objects that employing step (1) obtains cuts the raw video data as the vector polygon of corresponding destination object, and the target block quantity of statistics raw video data cutting generation; Because the border polygon can influence the merging of image data at cutting process,, be convenient to inlaying of result after the target processing with complete therefore to polygonal border extension 2~3 pixels of object boundary.
Step (5). adopt load balancing thought and quad-tree structure thought to organize dynamic construction dynamic virtual cluster machine, the silhouette target data block that step (4) is obtained is distributed to corresponding computing node and enhancing, fusion, classification and the cloud of image data is monitored calculate on request;
The Enhancement Method of described image data comprises: the contrast based on the spatial transform method strengthens, histogram strengthens, based on convolution enhancing, the sharpening enhancing of airspace filter method; And based on high pass, the LPF image Enhancement Method of frequency domain method.
The fusion of image data comprises: IHS converter technique, main composition converter technique, high-pass filtering fusion method, Brovery conversion fusion method.
The classification of image data comprises: supervised classification, unsupervised classification.
The monitoring of the cloud of image data comprises: based on the method for spectrum, based on the method for texture; Wherein spectrographic technique has: ISCCP method, APLOOL method, CO slice method, CLAVR method etc.; Texture method has co-occurrence matrix method, the laws texture energy method based on statistical model, and based on the method for structure with based on the method for field model.
Step (6). cluster master control computing node is collected the result of each level computing node, and returns the relevant position of raw video data according to the order of sequence according to the address code of step (2).
Embodiment
The first step: for the gray processing of image data is handled; The gray threshold that rule of thumb is worth the target setting object carries out block with rule to the length and width of image data by quad-tree structure thought and divides; Through repeatedly circulation and stratification process, obtain the minimum outsourcing rectangle of destination object at last.Results is as shown in Figure 2; The image size is
; Through once dividing, then each subblock sizes is
; Divide subblock sizes again for
, the rest may be inferred.
Second step: adopt decimal system Morton sign indicating number that target block is encoded and make up quaternary tree; At first also encode by the line scanning of advancing according to the natural number order to first step results; After accompanying drawing 2 is divided, can get the Morton coding shown in the accompanying drawing 3 left figure, form the quaternary tree of a reversal of the natural order of things again by bottom-up structure shown in the left side, see accompanying drawing 3 right figure; Wherein the black leaf node is represented the destination object region, will be distributed to the image data zone of each computing node below being.
The 3rd step: the minimum outsourcing rectangle that is directed against the destination object of the first step; Employing from left to right scans by the ranks order; The picture dot of the grey scale pixel value through merging given threshold range can further obtain destination object border polygons all in this MBR; In sub-block 1, the process through overscanning merges can obtain 3 destination object border polygons simultaneously.
The 4th step: the destination object border polygon that utilized for the 3rd step obtained cuts the raw video data, can obtain the consistent image data piece of based target quantity.Shown in accompanying drawing 2, to given target sizes threshold range, adopt its MBR to cut to qualified target, obtain the black leaf node constructed like the right figure of Fig. 3, Morton encodes and is respectively (1,3,8,9,10,12,21) totally 7 destination objects.Wherein Morton be encoded to 1 through repeatedly the circulation, can cut and produce 3 destination objects.
The 5th step: adopt load balancing thought and organize the cluster computing node to make up the Virtual Cluster machine according to quad-tree structure thought, the image data piece that the 3rd step was obtained is distributed to corresponding computing node and calculates on request.Earlier destination number and each target information amount are added up; Organize the cluster computing node to form quad-tree structure then; Specifically organize shown in accompanying drawing 4; Method for organizing is:
,
is level (layer) number of quaternary tree Virtual Organization structure.Based on the thinking of load balancing, destination number Num computing node has following distribution situation:
(3) the rest may be inferred;, branch on the basis of
grade data, remaining data mean allocation is arrived current
level computing node before distributing.
The 6th step: cluster master control computing node is collected the result of each level computing node, and returns the relevant position of raw video data according to the order of sequence according to the address code of step (2); Destination object result's collection is collected and merging in time by bottom-up step by step, through the too much operation of wheel, arrives the master control computing node at last.Shown in accompanying drawing 4, seven calculating destination objects are organized into a quasi-equilibrium quad-tree structure, because last computing node only distributes a computational data, therefore can not distribute computing node to return back to the upper level computing node is two-level node.
Claims (1)
1. high-resolution remote sensing image Target Recognition high-efficient treatment method under the overall background is characterized in that comprising the steps:
Step (1). the gray processing of raw video data is handled, the empirical value of target setting object, the gray threshold and the minimum outsourcing rectangle of acquisition destination object;
The raw video data are carried out gray processing through the weighted mean value method handle, the image analytic degree through image data calculates the big or small of destination object with empirical value and draws the gray threshold of destination object then; And will handle the grey scale pixel value of back image data and the gray threshold of destination object and compare, finally obtain the minimum outsourcing rectangle of all destination objects;
Concrete comparison procedure is following:
Thought quadtree structure the processed image data is divided into blocks, the image data is OK
, column
number of pixels contained respectively
and
,
If
or
is odd number; When row when then dividing, row;
,
; If cut-off rule is through destination object when dividing; Promptly divide a grey scale pixel value of being expert at or being listed as in the gray threshold scope of destination object; Then the sub-block at this destination object place is no longer divided, and this sub-block is called the minimum outsourcing rectangle corresponding with destination object; Otherwise continue to divide;
Step (2). adopt decimal system Morton sign indicating number to the minimum outsourcing rectangle structure linear quadtree of encoding;
With the address code of decimal system Morton sign indicating number as linear quadtree, to carry out after block divides at the image data of step (1), the method through from bottom to top makes up linear quadtree to minimum outsourcing rectangle;
Step (3). polygonal extraction in destination object border and generation;
All minimum outsourcing rectangles that step (1) is marked off; Press the from left to right scanning one by one of ranks order; If scanning draws in this minimum outsourcing rectangle a plurality of destination objects are arranged; Then the pixel to a plurality of destination objects that can merge in this minimum outsourcing rectangle merges the border polygon that then produces corresponding destination object that can not merge; In the polygonal extraction in destination object border and generating, must be through the vector quantization process of a raster data, this vector quantization process comprises following basic step:
1) polygonal Boundary Extraction: adopt high-pass filtering with the grating image binaryzation or with particular value sign frontier point;
2) follow the trail of the boundary line: to each border segmental arc by a node to its another adjacent node searching; Common frontier point to each known approach axis; Remove self approach axis external demand along the next adjacent frontier points of other seven direction searchs, up to being linked to be the border segmental arc;
3) topological relation generates: for the border segmental arc of vector representation, judge each polygonal spatial relationship on itself and the former figure, form complete topological structure then through the mathematics geometric relationship, and the getting in touch of foundation and attribute data;
4) remove unnecessary point and the slick and sly curve that belongs to Polygonal Boundary: because node searching is that individual element carries out, must remove the unnecessary point that causes thus, to reduce data redundancy through matched curve; Adopt interpolation algorithm that the curve that belongs to Polygonal Boundary is carried out smooth treatment simultaneously;
Interpolation algorithm commonly used includes linear iteration technique, segmentation cubic polynomial interpolation method, positive axis of a parabola average weighted method, inclined shaft para-curve average weighted method, splines method of interpolation;
Step (4). use the border polygon cutting raw video target data block of destination object;
The minimum outsourcing rectangle of all destination objects that employing step (1) obtains cuts the raw video data as the vector polygon of corresponding destination object, and the target block quantity of statistics raw video data cutting generation; Simultaneously to polygonal border extension 2~3 pixels of object boundary, be convenient to inlaying of result after the target processing with complete;
Step (5). adopt load balancing thought and quad-tree structure thought to organize dynamic construction dynamic virtual cluster machine, the silhouette target data block that step (4) is obtained is distributed to corresponding computing node and enhancing, fusion, classification and the cloud of image data is monitored calculate on request;
The Enhancement Method of described image data comprises: the contrast based on the spatial transform method strengthens, histogram strengthens, based on convolution enhancing, the sharpening enhancing of airspace filter method; And based on high pass, the LPF image Enhancement Method of frequency domain method;
The fusion of image data comprises: IHS converter technique, main composition converter technique, high-pass filtering fusion method, Brovery conversion fusion method;
The classification of image data comprises: supervised classification, unsupervised classification;
The monitoring of the cloud of image data comprises: based on the method for spectrum, based on the method for texture; Wherein spectrographic technique has: ISCCP method, APLOOL method, CO slice method, CLAVR method etc.; Texture method has co-occurrence matrix method, the laws texture energy method based on statistical model, and based on the method for structure with based on the method for field model;
Step (6). cluster master control computing node is collected the result of each level computing node, and returns the relevant position of raw video data according to the order of sequence according to the address code of step (2).
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2012102260343A CN102799898A (en) | 2012-06-28 | 2012-06-28 | Efficient target identification processing method of high-resolution remote sensing image in context |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2012102260343A CN102799898A (en) | 2012-06-28 | 2012-06-28 | Efficient target identification processing method of high-resolution remote sensing image in context |
Publications (1)
Publication Number | Publication Date |
---|---|
CN102799898A true CN102799898A (en) | 2012-11-28 |
Family
ID=47198997
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2012102260343A Pending CN102799898A (en) | 2012-06-28 | 2012-06-28 | Efficient target identification processing method of high-resolution remote sensing image in context |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102799898A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105787911A (en) * | 2016-03-21 | 2016-07-20 | 中国林业科学研究院资源信息研究所 | Image erosion and expansion processing method based on topology fractal algorithm |
CN106156138A (en) * | 2015-04-10 | 2016-11-23 | 高德软件有限公司 | The foundation of the spatial index of point of interest, Perimeter method and apparatus |
CN106796615A (en) * | 2014-09-15 | 2017-05-31 | 欧特克公司 | Use the parallel processing of Self-absorption Correction Factor |
CN106845452A (en) * | 2017-02-23 | 2017-06-13 | 中国公路工程咨询集团有限公司 | High score Remote Sensing Data Processing method based on the Big Dipper |
CN107533630A (en) * | 2015-01-20 | 2018-01-02 | 索菲斯研究股份有限公司 | For the real time machine vision of remote sense and wagon control and put cloud analysis |
CN107545557A (en) * | 2016-06-23 | 2018-01-05 | 爱威科技股份有限公司 | Egg detecting method and device in excrement image |
CN107545572A (en) * | 2016-06-23 | 2018-01-05 | 爱威科技股份有限公司 | The treating method and apparatus of target in image |
CN109102480A (en) * | 2018-07-06 | 2018-12-28 | 航天星图科技(北京)有限公司 | A kind of Gram-Schmdit fusion method suitable for distributed structure/architecture |
CN114529489A (en) * | 2022-03-01 | 2022-05-24 | 中国科学院深圳先进技术研究院 | Multi-source remote sensing image fusion method, device, equipment and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1584932A (en) * | 2004-06-10 | 2005-02-23 | 西安煤航遥感信息有限责任公司 | Optimizing method for image transfigure border side tracking |
CN101706950A (en) * | 2009-07-13 | 2010-05-12 | 中国科学院遥感应用研究所 | High-performance implementation method for multi-scale segmentation of remote sensing images |
CN102005034A (en) * | 2010-12-01 | 2011-04-06 | 南京大学 | Remote sensing image segmentation method based on region clustering |
CN102013015A (en) * | 2010-12-02 | 2011-04-13 | 南京大学 | Object-oriented remote sensing image coastline extraction method |
-
2012
- 2012-06-28 CN CN2012102260343A patent/CN102799898A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1584932A (en) * | 2004-06-10 | 2005-02-23 | 西安煤航遥感信息有限责任公司 | Optimizing method for image transfigure border side tracking |
CN101706950A (en) * | 2009-07-13 | 2010-05-12 | 中国科学院遥感应用研究所 | High-performance implementation method for multi-scale segmentation of remote sensing images |
CN102005034A (en) * | 2010-12-01 | 2011-04-06 | 南京大学 | Remote sensing image segmentation method based on region clustering |
CN102013015A (en) * | 2010-12-02 | 2011-04-13 | 南京大学 | Object-oriented remote sensing image coastline extraction method |
Non-Patent Citations (1)
Title |
---|
曾志等: "一种基于分块的遥感影像并行处理机制", 《浙江大学学报(理学版)》 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106796615A (en) * | 2014-09-15 | 2017-05-31 | 欧特克公司 | Use the parallel processing of Self-absorption Correction Factor |
CN107533630A (en) * | 2015-01-20 | 2018-01-02 | 索菲斯研究股份有限公司 | For the real time machine vision of remote sense and wagon control and put cloud analysis |
CN106156138A (en) * | 2015-04-10 | 2016-11-23 | 高德软件有限公司 | The foundation of the spatial index of point of interest, Perimeter method and apparatus |
CN106156138B (en) * | 2015-04-10 | 2019-08-09 | 高德软件有限公司 | The foundation of the spatial index of point of interest, Perimeter method and apparatus |
CN105787911A (en) * | 2016-03-21 | 2016-07-20 | 中国林业科学研究院资源信息研究所 | Image erosion and expansion processing method based on topology fractal algorithm |
CN105787911B (en) * | 2016-03-21 | 2019-02-12 | 中国林业科学研究院资源信息研究所 | A kind of Image erosion and expanding processing based on topological fractal algorithm |
CN107545572A (en) * | 2016-06-23 | 2018-01-05 | 爱威科技股份有限公司 | The treating method and apparatus of target in image |
CN107545557A (en) * | 2016-06-23 | 2018-01-05 | 爱威科技股份有限公司 | Egg detecting method and device in excrement image |
CN106845452A (en) * | 2017-02-23 | 2017-06-13 | 中国公路工程咨询集团有限公司 | High score Remote Sensing Data Processing method based on the Big Dipper |
CN106845452B (en) * | 2017-02-23 | 2020-07-14 | 中国公路工程咨询集团有限公司 | Beidou-based high-resolution remote sensing data processing method |
CN109102480A (en) * | 2018-07-06 | 2018-12-28 | 航天星图科技(北京)有限公司 | A kind of Gram-Schmdit fusion method suitable for distributed structure/architecture |
CN109102480B (en) * | 2018-07-06 | 2022-02-22 | 中科星图股份有限公司 | Gram-Schmidt fusion method suitable for distributed architecture |
CN114529489A (en) * | 2022-03-01 | 2022-05-24 | 中国科学院深圳先进技术研究院 | Multi-source remote sensing image fusion method, device, equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102799898A (en) | Efficient target identification processing method of high-resolution remote sensing image in context | |
CN102708369B (en) | Sea ice parameter extraction method on basis of satellite image | |
CN101751449B (en) | Spatial overlap analysis method and system used in geographic information system | |
Chen et al. | Multiscale grid method for detection and reconstruction of building roofs from airborne LiDAR data | |
CN111243091B (en) | Massive DEM pyramid slice parallel construction method based on distributed system | |
Wang et al. | 3D-CenterNet: 3D object detection network for point clouds with center estimation priority | |
CN103413133A (en) | Automatically-extracting power line method in random laser point cloud data | |
CN104361110A (en) | Mass electricity consumption data analysis system as well as real-time calculation method and data mining method | |
CN102625126A (en) | Prediction-based three-dimensional mesh coding method | |
WO2021036550A1 (en) | Visual task-based method and system for evaluating compression quality of point cloud data | |
CN110633262B (en) | Spark-based map intersection area calculation method and system | |
CN101853485A (en) | Non-uniform point cloud simplification processing method based on neighbor communication cluster type | |
CN112785710A (en) | Rapid unitization method, system, memory and equipment for OSGB three-dimensional model building | |
CN103871088A (en) | Method and system for partitioning compression of spatial statistical data based on sparse characteristic of image | |
CN110120083B (en) | Water quality forecast distribution map drawing method and device and electronic equipment | |
Yang et al. | An efficient rendering method for large vector data on large terrain models | |
CN107679127A (en) | Point cloud information parallel extraction method and its system based on geographical position | |
CN109241369B (en) | Rainfall isopleth construction method based on grid stretching method | |
CN111009034A (en) | Three-dimensional model monomer method, system, storage medium and equipment | |
CN104166715A (en) | Vxworks platform electronic chart engine | |
Yu et al. | Saliency computation and simplification of point cloud data | |
CN116011061A (en) | Three-dimensional reconstruction model monomer segmentation method, system and terminal for multi-target building | |
CN103458032B (en) | The method and system of a kind of spatial data accessing rule dynamic statistics and Information Compression | |
CN112905729B (en) | Thermodynamic diagram generation method and device for track data, electronic equipment and storage medium | |
CN113850254A (en) | Building vector outline simplifying method, model and model establishing method based on deep learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C12 | Rejection of a patent application after its publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20121128 |