CN103530381B - A kind of parallel optimization method towards remote sensing image neighborhood processing - Google Patents
A kind of parallel optimization method towards remote sensing image neighborhood processing Download PDFInfo
- Publication number
- CN103530381B CN103530381B CN201310487801.0A CN201310487801A CN103530381B CN 103530381 B CN103530381 B CN 103530381B CN 201310487801 A CN201310487801 A CN 201310487801A CN 103530381 B CN103530381 B CN 103530381B
- Authority
- CN
- China
- Prior art keywords
- remote sensing
- sensing image
- image block
- neighborhood
- long
- 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.)
- Expired - Fee Related
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/51—Indexing; Data structures therefor; Storage structures
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to a kind of parallel optimization method towards remote sensing image neighborhood processing.The inventive method utilizes octree structure to index for the remote sensing image block data foundation of neighborhood each other in parallel cluster environment;Set up the relationship map between local remote sensing image block coordinate system and long-range neighborhood remote sensing image block coordinate system;Design the buffering of long-range adjacent region data access request, the access request of discrete pixels data is aggregated into continuous print large data access request;After teledata returns, again process the local pixel in buffer queue, utilize described mapping, the pixel value of the long-range pixel that quick obtaining is asked, and then optimize the Parallel Implementation of remote sensing image neighborhood processing algorithm.Present invention, avoiding the explicit message transmission between remote sensing image partition neighborhood, reduce the complexity of neighborhood type Remote Sensing Data Processing Parallel algorithm, improve the network transmission efficiency of data, provide a kind of effective realization means for the remote sensing image parallel processing under cluster environment.
Description
Technical field
The present invention relates to the processing method of a kind of remote sensing image, be specifically related to a kind of parallel towards remote sensing image neighborhood processing
Optimization method.
Background technology
Parallel computation is the conventional means improving remote sensing image data treatment effeciency, and its basic thought is " dividing and rule ",
I.e. utilize the remote sensing image processing method feature to pixel operation similitude, big remote sensing image is carried out piecemeal, each
After image block has processed, then it is spliced to form final result, and then reaches what image overall treatment efficiency improved
Purpose.But, along with people's constantly extending and the most deep remote sensing application, remote sensing image processing method also becomes more
Come the most complicated.According to current pixel point in remote sensing image processing procedure, the dependence of other pixel and position relatively are closed
System, can be divided into a process, neighborhood processing and Global treatment, the wherein meter of neighborhood processing method by remote sensing image processing method
Calculate input data and include current pixel and all pixels of a certain appointment neighborhood centered by current pixel.Due to neighborhood
Processing method needs the participation of other pixels on image, this be processing method parallelization realize bring great difficulty,
On the one hand interdepending between pixel will cause the frequent communication between parallel image block, affect process performance;The opposing party
The realization of face parallel method needs programming personnel explicitly to define message transmission, and programming is complicated and does not possess flexibility.
For the remote sensing image parallel processing of neighborhood operation, current implementation is around reducing leading between parallel node more
Believing that cost proposes, mainly having two kinds: one is to use the two-dimensional block decomposition strategy with overlay region at data catabolic phase,
This method causes the redundant storage of data, and for different types of neighborhood operation, needs to reformulate different
Overlay strategy, it is achieved mechanism underaction;Two is to utilize statistical model or experience to be simulated adjacent region data, but this
Method causes loss of significance to varying degrees, have impact on the accuracy of final result.
Summary of the invention
The technical problem to be solved is to provide a kind of adjacent towards remote sensing image for above-mentioned the deficiencies in the prior art
The parallel optimization method that territory processes, the method can be avoided explicit message transmission between remote sensing image block, reduce data and process multiple
Miscellaneous degree, it is possible to aggregate remote neighborhood territory pixel access request, improves network transmission efficiency.
The present invention solves the technical scheme that above-mentioned technical problem used: a kind of parallel towards remote sensing image neighborhood processing
Optimization method, it is characterised in that: comprise the steps:
Step one: initialize, sets buffer queue critical value and remote sensing image block processes end condition;
Step 2: obtain the blocking information of remote sensing image in parallel cluster environment, by method of partition to remote sensing image block
Set up partition neighborhood index;
Step 3: with the upper left angle point of each remote sensing image block as the origin of coordinates, with through origin of coordinates level to the right
Direction is X-axis positive axis, with through origin of coordinates vertically downward direction for Y-axis positive axis, set up local remote sensing respectively
Image block and the plane right-angle coordinate of long-range neighborhood remote sensing image block, set up local remote sensing image block coordinate simultaneously
Relationship map between system and remotely neighborhood remote sensing image block coordinate system;
Step 4: carry out neighborhood operation according to the local pixel in secondary ordered pair remote sensing image block, when pending this locality picture
When element needs to access long-range neighborhood territory pixel, the neighborhood operation of this this locality pixel wouldn't be processed, simultaneously by this this locality pixel
Coordinate value in local remote sensing image block coordinate system and all long-range neighborhood territory pixel that associates with this this locality pixel
Coordinate value in ground remote sensing image block coordinate system preserves to buffer queue as a structure, then proceedes to process remaining
Local pixel;
Step 5: when the structure quantity in buffer queue reaches buffer queue critical value or the place to remote sensing image block
Reason reaches end condition, the coordinate in local remote sensing image block coordinate system of the long-range neighborhood territory pixel in polymerization buffer queue
Value, calculates poly-according to the relationship map of local remote sensing image block coordinate system with long-range neighborhood remote sensing image block coordinate system
Buffer queue medium-long range neighborhood territory pixel coordinate value in long-range neighborhood remote sensing image block coordinate system after conjunction, finds out simultaneously and needs
Long-range neighborhood remote sensing image block to be accessed, then utilizes the partition neighborhood set up in described step 2 index to parse remote
The store path of journey neighborhood territory pixel, sets up the long-range number between local remote sensing image block and long-range neighborhood remote sensing image block
According to access request, other processing equipments these remote data access request being distributed in cluster, wait access request knot
Fruit returns, and access request results is by long-range neighborhood territory pixel coordinate in long-range neighborhood remote sensing image block coordinate system herein
The data pair that the pixel value of value and this long-range neighborhood territory pixel is formed;
Step 6: after the access request results in described step 5 returns, utilize local remote sensing image block coordinate system with
Remotely the relationship map between neighborhood remote sensing image block coordinate system sets up long-range neighborhood territory pixel at local remote sensing image block
The relationship map between coordinate value and the pixel value of this long-range neighborhood territory pixel in coordinate system;
Step 7: the long-range neighborhood territory pixel drawn according to step 6 coordinate value in local remote sensing image block coordinate system with
Relationship map between the pixel value of this long-range neighborhood territory pixel is simultaneously according to the principle of buffer queue first in first out, the most right
Step 4 is deposited and carries out neighborhood operation into the local pixel in described buffer queue, when the institute preserved in described buffer queue
The neighborhood operation having local pixel empties described buffer queue after being disposed;
Step 8: return step 4, until the neighborhood operation of all local pixel in this this locality remote sensing image block processes
Complete.
Preferably, the method for partition in described step 2 is octree structure method, i.e. to each remote sensing image block, and profit
With four edges circle of remote sensing image block, its neighborhood is divided into upper and lower, left and right, upper left, lower-left, upper right, bottom right eight
Individual region, utilizing octree structure is that the adjacent remote sensing image block being distributed in described eight regions sets up index, described rope
Draw and include the area identification of remote sensing image block, store path and block size.
Preferably, in described step 3, local remote sensing image block coordinate system and long-range neighborhood remote sensing image block coordinate system
Between the step of relationship map be: for the remote sensing image block Block that size is M row * N row, with its phase
Adjacent remote sensing image block BlockrStore path and size be respectively PathrAnd MrRow * NrRow, r ∈ U, D, L,
R, LU, LD, RU, RD}, wherein U is under upper, D is, L is left, and R be the right side, and LU is upper left, and LD is
Lower-left, RU is upper right, and RD is bottom right, according to geometrical relationship, then has:
P in formula (1) (x, y) represent abscissa under local remote sensing image block coordinate system XOY be x, ordinate be the picture of y
Vegetarian refreshments, as p (x, y) ∈ Blockr, this pixel coordinate (x in long-range neighborhood remote sensing image block coordinate systemr, yr) by
Formula (2) is given:
Preferably, in described step 5: the long-range neighborhood territory pixel in polymerization buffer queue is at local remote sensing image block coordinate
The processing mode of the coordinate value in system is: reject the long-range neighborhood territory pixel repeated in buffer queue at local remote sensing image
Coordinate value in piecemeal coordinate system, and according to remote sensing image block index and same long-range neighborhood remote sensing image will be under the jurisdiction of divide
The long-range neighborhood territory pixel of block carries out classification set.
In described step 6, after the access request results in step 5 returns, set up Key-Value structural array, its
In Key be long-range neighborhood territory pixel in local remote sensing image block coordinate system coordinate value, Value is long-range neighborhood picture
The pixel value of element.
Compared with prior art, it is an advantage of the current invention that:
1, avoid the explicit message transmission between remote sensing image block, reduce neighborhood type Remote Sensing Data Processing method parallel
The complexity realized.The present invention defines the index structure towards neighborhood for remote sensing image block, and by setting up this
Relationship map between ground remote sensing image block coordinate system and remotely neighborhood remote sensing image block coordinate system and setting up according to it
Long-range neighborhood territory pixel coordinate value in local remote sensing image block coordinate system and this long-range neighborhood territory pixel pixel value it
Between relationship map so that the local remote sensing image block of neighborhood and long-range neighborhood remote sensing image block are by unified each other
Coordinate computation, carries out fast transparent access, in the address procedures of data and transmitting procedure to long-range neighborhood remote sensing image pixel
Need not explicitly transmit message, be greatly reduced the difficulty of neighborhood type remote sensing image data processing method Parallel Implementation, carry
High efficiency.
2, achieve long-range neighborhood image block pixel and access the polymerization of information, improve the efficiency of network transmission.This
The buffer queue that long-range neighborhood territory pixel accesses is established in bright, and according to remote sensing image block index and the offset address of pixel
Long-range neighborhood territory pixel in buffer queue is carried out merger, multiple long-range neighborhood territory pixel access request small-sized, that be interrupted are gathered
Minority is large-scale in synthesis, the request of continuous print image data, it is to avoid frequently set up time loss needed for network path connects with
And the repetitive requests of long-range neighborhood territory pixel data, improve the efficiency of network transmission.
Accompanying drawing explanation
Fig. 1 is the flow chart of parallel optimization method in the embodiment of the present invention towards remote sensing image neighborhood processing.
Fig. 2 is the neighborhood zoning plan that the embodiment of the present invention illustrates remote sensing image block in example.
Fig. 3 is that the embodiment of the present invention illustrates remote sensing image block Octree neighborhood index structure figure in example.
Fig. 4 is the relationship map signal that the embodiment of the present invention illustrates local coordinate and long-range neighborhood coordinate system in example
Figure.
Detailed description of the invention
Below in conjunction with accompanying drawing embodiment, the present invention is described in further detail.
The parallel optimization method towards remote sensing image neighborhood processing that the present embodiment provides, it comprises the steps, sees figure
Shown in 1:
Step one: initialize, sets buffer queue critical value and remote sensing image block processes end condition;
Step 2: obtain the blocking information of remote sensing image in parallel cluster environment, by method of partition to remote sensing image block
Set up partition neighborhood index;
In this step, described method of partition is octree structure method, i.e. to each remote sensing image block, utilizes remote sensing shadow
As its neighborhood is divided into upper and lower, left and right, upper left, lower-left, upper right, region, eight, bottom right by four edges circle of piecemeal,
Utilizing octree structure is that the adjacent remote sensing image block being distributed in described eight regions sets up index, and described index includes distant
The sense area identification of image block, store path and block size;
Step 3: with the upper left angle point of each remote sensing image block as the origin of coordinates, with through origin of coordinates level to the right
Direction is X-axis positive axis, with through origin of coordinates vertically downward direction for Y-axis positive axis, set up local remote sensing respectively
Image block and the plane right-angle coordinate of long-range neighborhood remote sensing image block, set up local remote sensing image block coordinate simultaneously
Relationship map between system and remotely neighborhood remote sensing image block coordinate system;
In this step, set up the pass between local remote sensing image block coordinate system and long-range neighborhood remote sensing image block coordinate system
The step that system maps is: for the remote sensing image block Block that size is M row * N row, the remote sensing being adjacent
Image block BlockrStore path and size be respectively PathrAnd MrRow * NrRow, r ∈ U, D, L, R, LU,
LD, RU, RD}, wherein U is under upper, D is, L is left, and R be the right side, and LU is upper left, and LD is lower-left, RU
For upper right, RD is bottom right, according to geometrical relationship, then has:
P in formula (1) (x, y) represent abscissa under local remote sensing image block coordinate system XOY be x, ordinate be the picture of y
Vegetarian refreshments, as p (x, y) ∈ Blockr, this pixel coordinate (x in long-range neighborhood remote sensing image block coordinate systemr,yr) by public affairs
Formula (2) is given:
Step 4: carry out neighborhood operation according to the local pixel in secondary ordered pair remote sensing image block, when pending this locality picture
When element needs to access long-range neighborhood territory pixel, the neighborhood operation of this this locality pixel wouldn't be processed, simultaneously by this this locality pixel
Coordinate value in local remote sensing image block coordinate system and all long-range neighborhood territory pixel that associates with this this locality pixel
Coordinate value in ground remote sensing image block coordinate system preserves to buffer queue as a structure, then proceedes to process remaining
Local pixel;
Step 5: when the structure quantity in buffer queue reaches buffer queue critical value or the place to remote sensing image block
Reason reaches end condition, the coordinate in local remote sensing image block coordinate system of the long-range neighborhood territory pixel in polymerization buffer queue
Value, calculates poly-according to the relationship map of local remote sensing image block coordinate system with long-range neighborhood remote sensing image block coordinate system
Buffer queue medium-long range neighborhood territory pixel coordinate value in long-range neighborhood remote sensing image block coordinate system after conjunction, finds out simultaneously and needs
Long-range neighborhood remote sensing image block to be accessed, then utilizes the partition neighborhood set up in described step 2 index to parse remote
The store path of journey neighborhood territory pixel, sets up the long-range number between local remote sensing image block and long-range neighborhood remote sensing image block
According to access request, other processing equipments these remote data access request being distributed in cluster, wait access request knot
Fruit returns, and access request results is by long-range neighborhood territory pixel coordinate in long-range neighborhood remote sensing image block coordinate system herein
The data pair that the pixel value of value and this long-range neighborhood territory pixel is formed;
In this step, the coordinate value in local remote sensing image block coordinate system of the long-range neighborhood territory pixel in polymerization buffer queue
Processing mode be: reject in buffer queue the long-range neighborhood territory pixel repeated in local remote sensing image block coordinate system
Coordinate value, and according to remote sensing image block index and the long-range neighborhood of same long-range neighborhood remote sensing image block will be under the jurisdiction of
Pixel carries out classification set;
Step 6: after the access request results in described step 5 returns, utilize local remote sensing image block coordinate system with
Remotely the relationship map between neighborhood remote sensing image block coordinate system sets up long-range neighborhood territory pixel at local remote sensing image block
The relationship map between coordinate value and the pixel value of this long-range neighborhood territory pixel in coordinate system;Set up Key-Value structure number
Group, Key therein is long-range neighborhood territory pixel coordinate value in local remote sensing image block coordinate system, and Value is long-range
The pixel value of neighborhood territory pixel;
Step 7: the long-range neighborhood territory pixel drawn according to step 6 coordinate value in local remote sensing image block coordinate system with
Relationship map between the pixel value of this long-range neighborhood territory pixel is simultaneously according to the principle of buffer queue first in first out, the most right
Step 4 is deposited and carries out neighborhood operation into the local pixel in described buffer queue, when the institute preserved in described buffer queue
The neighborhood operation having local pixel empties described buffer queue after being disposed;
Step 8: return step 4, until the neighborhood operation of all local pixel in this this locality remote sensing image block processes
Complete.
Below using wide 32768 pixels, the Tiff form that high by 32768, pixel depth is as 8bit gray scale remote sensing image as
Example, the parallel optimization method towards remote sensing image neighborhood processing providing above-described embodiment remarks additionally, neighborhood
Computing is carried out by approach of mean filter, and wherein window masterplate is 3*3, the size system of each image block in this explanation example
One is 1024*1024, is analyzed using the 34th remote sensing image block Block34 as local remote sensing image block,
Towards the parallel optimization method of remote sensing image neighborhood processing in the present embodiment, comprise the steps:
Step one: initialize, sets in buffer queue the critical value of structure quantity as W=100, at remote sensing image block
The end condition of reason is all local remote sensing image block pixels of traversal the 34th remote sensing image block Block34;
Step 2: the host node from parallel cluster obtains the blocking information of remote sensing image, including the store path of each piecemeal
And block size, by method of partition, remote sensing image block is set up partition neighborhood and index, as in figure 2 it is shown, with the 30th
Four edges circle of four remote sensing image block Block34 according to orientation by this locality remote sensing image block Block34Neighborhood divide
Composition is upper and lower, left and right, upper left, lower-left, upper right, region, eight, bottom right, these eight region correspondences is compiled respectively
Code is U, D, L, R, LU, LD, RU, RD;Width, height and remote sensing image according to original remote sensing image divide
The dimension information of block, the partition neighborhood that can calculate the 34th remote sensing image block Block34 is respectively the first remote sensing shadow
As piecemeal Block1, the second remote sensing image block Block2, the 3rd remote sensing image block Block3, the 33rd remote sensing shadow
As piecemeal Block33, the 35th remote sensing image block Block35, the 65th remote sensing image block Block65, the 6th
16 remote sensing image block Block66With the 67th remote sensing image block Block67, as it is shown on figure 3, utilize Octree
Structure is that the adjacent remote sensing image block being distributed in described eight regions sets up index, and described index includes remote sensing image block
Area identification, store path and block size;
Step 3: with the upper left angle point of the 34th remote sensing image block Block34 for origin of coordinates O, with through O water
Flat direction to the right is X positive axis, and vertically downward direction is Y-axis positive axis, sets up local remote sensing image block plane
Rectangular coordinate system XOY, simultaneously with the upper left angle point of remaining remote sensing image block for origin of coordinates Or, wherein r ∈ 1,2,
3,33,35,65,66,67}, with through Or level direction to the right for X positive axis, vertically downward direction is
Y-axis positive axis, sets up long-range neighborhood remote sensing image block rectangular coordinate system, sets up local remote sensing image block coordinate simultaneously
Relationship map between system and remotely neighborhood remote sensing image block coordinate system, according to geometrical relationship:
Coordinate p (x, y) the ∈ Block of a pixel in as local remote sensing image block coordinate systemr, r ∈ U, D,
When L, R, LU, LD, RU, RD}, this pixel coordinate (x in long-range neighborhood remote sensing image block coordinate systemr,yr)
Be given by formula (2):
Owing to the size of each image block is unified for 1024*1024, herein NrAnd MrValue be 1024, p (x, y)
As the coordinate of a pixel in local remote sensing image block coordinate system with in long-range neighborhood remote sensing image block coordinate system
The coordinate relation of this pixel is p (x, y)=pr(1024+x, 1024+y), r ∈ 1,2,3,33,35,65,66,67},
(x, y) at its long-range neighborhood remote sensing image block for pixel coordinate p under i.e. local remote sensing image block coordinate system XOY
BlockrCoordinate system XrOrYrUnder corresponding points coordinate be pr(1024+x, 1024+y), as it is shown on figure 3, with the 30th
Point (-1 ,-1) under the XOY coordinate system that four remote sensing image block Block34 upper left angle points are set up is dividing with the first remote sensing image
Block Block1The X that upper left angle point is set up1O1Y1Coordinate under coordinate system is (1023,1023), shown in Figure 4;
Step 4: carry out neighborhood operation according to the local pixel in secondary ordered pair remote sensing image block, when pending this locality picture
When element needs to access long-range neighborhood territory pixel, the neighborhood operation of this this locality pixel wouldn't be processed, simultaneously by this this locality pixel
Coordinate value in local remote sensing image block coordinate system and all long-range neighborhood territory pixel that associates with this this locality pixel
Coordinate value in ground remote sensing image block coordinate system preserves to buffer queue as a structure, such as definition buffer queue
PixelRequestQueue is List<RemotePixelElement>, if currently pending local remote sensing image pixel
Coordinate is p (0,0), needs to access 5 long-range neighborhood remote sensing image pixel coordinates to the neighborhood processing of p (0,0), is respectively
P (-1 ,-1), p (-1,0), p (-1,1), p (0 ,-1) and p (1 ,-1), create RemotePixelElement structure,
RemotePixelElement structure comprises local remote sensing image pixel coordinate information and sits local remote sensing image pixel
Mark information carries out processing required long-range neighborhood remote sensing image pixel coordinate information, is buffered in by RemotePixelElement
In buffer queue PixelRequestQueue, then proceed to process remaining local pixel;
Step 5: when in buffer queue PixelRequestQueue, the structure quantity of storage reaches buffer queue critical value
W=100 or the process to remote sensing image block have traveled through all local remote sensing of the 34th remote sensing image block Block34
During image block pixel, the element in PixelRequestQueue carrying out classified finishing, that is polymerized in buffer queue is remote
Journey neighborhood territory pixel coordinate value in local remote sensing image block coordinate system, according to local remote sensing image block coordinate system with remote
The relationship map of journey neighborhood remote sensing image block coordinate system calculate polymerization after buffer queue medium-long range neighborhood territory pixel remotely
Coordinate value in neighborhood remote sensing image block coordinate system, finds out the long-range neighborhood remote sensing image block needing to access simultaneously, picks
Except the long-range neighborhood territory pixel repeated in buffer queue coordinate value in local remote sensing image block coordinate system, according to distant
The long-range neighborhood territory pixel being under the jurisdiction of same long-range neighborhood remote sensing image block is carried out classification set by sense image block index, so
The rear store path utilizing the partition neighborhood set up in described step 2 index to parse long-range neighborhood territory pixel, sets up local distant
Remote data access request between sense image block and remotely neighborhood remote sensing image block, please by these remote data access
Seeking other processing equipments being distributed in cluster, wait that access request results returns, access request results is the most adjacent herein
Between territory pixel coordinate value and the pixel value of this long-range neighborhood territory pixel in long-range neighborhood remote sensing image block coordinate system
Data pair;
Step 6: after the access request results in described step 5 returns, utilize local remote sensing image block coordinate system with
Remotely the relationship map between neighborhood remote sensing image block coordinate system sets up long-range neighborhood territory pixel at local remote sensing image block
The relationship map between coordinate value and the pixel value of this long-range neighborhood territory pixel in coordinate system, i.e. sets up Key-Value structure
Array RemotePixelMappingList, for each element in array, Key is that long-range neighborhood territory pixel is distant in this locality
Coordinate value in sense image block coordinate system, Value is the pixel value of long-range neighborhood territory pixel;
Step 7: the long-range neighborhood territory pixel drawn according to step 6 coordinate value in local remote sensing image block coordinate system with
Relationship map between the pixel value of this long-range neighborhood territory pixel is simultaneously according to the principle of buffer queue first in first out, the most right
Step 4 is deposited and carries out neighborhood operation, computing into the local pixel in described buffer queue PixelRequestQueue
The RemotePixelMappingList that the long-range neighborhood remote sensing image pixel of Cheng Suoxu all can create in step 8 obtains,
Empty after the neighborhood operation of all local pixel preserved in described buffer queue PixelRequestQueue is disposed
Element in described buffer queue PixelRequestQueue;
Step 8: return step 4, until the neighborhood operation of all local pixel in this this locality remote sensing image block processes
Complete.
Claims (5)
1. the parallel optimization method towards remote sensing image neighborhood processing, it is characterised in that: comprise the steps:
Step one: initialize, sets buffer queue critical value and remote sensing image block processes end condition;
Step 2: obtain the blocking information of remote sensing image in parallel cluster environment, by method of partition to remote sensing image block
Set up partition neighborhood index;
Step 3: with the upper left angle point of each remote sensing image block as the origin of coordinates, with through origin of coordinates level to the right
Direction is X-axis positive axis, with through origin of coordinates vertically downward direction for Y-axis positive axis, set up local remote sensing respectively
Image block and the plane right-angle coordinate of long-range neighborhood remote sensing image block, set up local remote sensing image block coordinate simultaneously
Relationship map between system and remotely neighborhood remote sensing image block coordinate system;
Step 4: carry out neighborhood operation according to the local pixel in secondary ordered pair remote sensing image block, when pending this locality picture
When element needs to access long-range neighborhood territory pixel, the neighborhood operation of this this locality pixel wouldn't be processed, simultaneously by this this locality pixel
Coordinate value in local remote sensing image block coordinate system and all long-range neighborhood territory pixel that associates with this this locality pixel
Coordinate value in ground remote sensing image block coordinate system preserves to buffer queue as a structure, then proceedes to process remaining
Local pixel;
Step 5: when the structure quantity in buffer queue reaches buffer queue critical value or the place to remote sensing image block
Reason reaches end condition, the coordinate in local remote sensing image block coordinate system of the long-range neighborhood territory pixel in polymerization buffer queue
Value, calculates poly-according to the relationship map of local remote sensing image block coordinate system with long-range neighborhood remote sensing image block coordinate system
Buffer queue medium-long range neighborhood territory pixel coordinate value in long-range neighborhood remote sensing image block coordinate system after conjunction, finds out simultaneously and needs
Long-range neighborhood remote sensing image block to be accessed, then utilizes the partition neighborhood set up in described step 2 index to parse remote
The store path of journey neighborhood territory pixel, sets up the long-range number between local remote sensing image block and long-range neighborhood remote sensing image block
According to access request, other processing equipments these remote data access request being distributed in cluster, wait access request knot
Fruit returns, and access request results is by long-range neighborhood territory pixel coordinate in long-range neighborhood remote sensing image block coordinate system herein
The data pair that the pixel value of value and this long-range neighborhood territory pixel is formed;
Step 6: after the access request results in described step 5 returns, utilize local remote sensing image block coordinate system with
Remotely the relationship map between neighborhood remote sensing image block coordinate system sets up long-range neighborhood territory pixel at local remote sensing image block
The relationship map between coordinate value and the pixel value of this long-range neighborhood territory pixel in coordinate system;
Step 7: the long-range neighborhood territory pixel drawn according to step 6 coordinate value in local remote sensing image block coordinate system with
Relationship map between the pixel value of this long-range neighborhood territory pixel is simultaneously according to the principle of buffer queue first in first out, the most right
Step 4 is deposited and carries out neighborhood operation into the local pixel in described buffer queue, when the institute preserved in described buffer queue
The neighborhood operation having local pixel empties described buffer queue after being disposed;
Step 8: return step 4, until the neighborhood operation of all local pixel in this this locality remote sensing image block processes
Complete.
Parallel optimization method towards remote sensing image neighborhood processing the most according to claim 1, it is characterised in that:
Method of partition in described step 2 is octree structure method, i.e. to each remote sensing image block, utilizes remote sensing image to divide
Its neighborhood is divided into upper and lower, left and right, upper left, lower-left, upper right, region, eight, bottom right by four edges circle of block, utilizes
Octree structure is that the adjacent remote sensing image block being distributed in described eight regions sets up index, and described index includes remote sensing shadow
As the area identification of piecemeal, store path and block size.
The most according to claim 1 towards the parallel optimization method of remote sensing image neighborhood processing, it is characterised in that: institute
State in step 3, the relationship map between local remote sensing image block coordinate system and remotely neighborhood remote sensing image block coordinate system
Step be: for size be M row * N row remote sensing image block Block, the remote sensing image being adjacent divides
Block BlockrStore path and size be respectively PathrAnd MrRow * NrRow, r ∈ U, D, L, R, LU, LD,
RU, RD}, wherein U is under upper, D is, L is left, and R be right, and LU is upper left, and LD is lower-left, and RU is the right side
On, RD is bottom right, according to geometrical relationship, then has:
P in formula (1) (x, y) represent abscissa under local remote sensing image block coordinate system XOY be x, ordinate be the picture of y
Vegetarian refreshments, as p (x, y) ∈ Blockr, this pixel coordinate (x in long-range neighborhood remote sensing image block coordinate systemr,yr) by public affairs
Formula (2) is given:
The most according to claim 1 towards the parallel optimization method of remote sensing image neighborhood processing, it is characterised in that: institute
State in step 5: the coordinate value in local remote sensing image block coordinate system of the long-range neighborhood territory pixel in polymerization buffer queue
Processing mode is: reject the long-range neighborhood territory pixel repeated in buffer queue in local remote sensing image block coordinate system
Coordinate value, and index according to remote sensing image block and the long-range neighborhood picture of same long-range neighborhood remote sensing image block will be under the jurisdiction of
Element carries out classification set.
The most according to claim 1 towards the parallel optimization method of remote sensing image neighborhood processing, it is characterised in that: institute
State in step 6, after the access request results in step 5 returns, set up Key-Value structural array, Key therein
For long-range neighborhood territory pixel coordinate value in local remote sensing image block coordinate system, Value is the pixel of long-range neighborhood territory pixel
Value.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310487801.0A CN103530381B (en) | 2013-10-17 | 2013-10-17 | A kind of parallel optimization method towards remote sensing image neighborhood processing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310487801.0A CN103530381B (en) | 2013-10-17 | 2013-10-17 | A kind of parallel optimization method towards remote sensing image neighborhood processing |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103530381A CN103530381A (en) | 2014-01-22 |
CN103530381B true CN103530381B (en) | 2016-08-17 |
Family
ID=49932390
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310487801.0A Expired - Fee Related CN103530381B (en) | 2013-10-17 | 2013-10-17 | A kind of parallel optimization method towards remote sensing image neighborhood processing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103530381B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109299060A (en) * | 2018-11-22 | 2019-02-01 | 华东计算技术研究所(中国电子科技集团公司第三十二研究所) | Distributed file method and system based on image rectangular blocks |
CN111654646B (en) * | 2020-06-09 | 2023-06-13 | 苏州精濑光电有限公司 | Image synthesis method, device, system and storage medium |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101159061A (en) * | 2007-11-16 | 2008-04-09 | 清华大学 | Great neighborhood image parallel processing method |
CN101350098A (en) * | 2008-08-19 | 2009-01-21 | 黄建军 | Method, system and equipment for processing image |
CN101799762A (en) * | 2010-04-07 | 2010-08-11 | 中国科学院对地观测与数字地球科学中心 | Quick parallelization programming template method for remote sensing image processing algorithm |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7212938B2 (en) * | 2002-09-17 | 2007-05-01 | M7 Visual Intelligence, Lp | Method of using a self-locking travel pattern to achieve calibration of remote sensors using conventionally collected data |
-
2013
- 2013-10-17 CN CN201310487801.0A patent/CN103530381B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101159061A (en) * | 2007-11-16 | 2008-04-09 | 清华大学 | Great neighborhood image parallel processing method |
CN101350098A (en) * | 2008-08-19 | 2009-01-21 | 黄建军 | Method, system and equipment for processing image |
CN101799762A (en) * | 2010-04-07 | 2010-08-11 | 中国科学院对地观测与数字地球科学中心 | Quick parallelization programming template method for remote sensing image processing algorithm |
Non-Patent Citations (2)
Title |
---|
一种适用于机群系统的细粒度遥感图像镶嵌并行算法;安兴华等;《清华大学学报》;20021231;第42卷(第10期);第1389-1392页 * |
基于邻域信息的遥感图像模糊聚类及并行算法设计;龚雪晶等;《计算机应用》;20071231;第27卷(第10期);第2512-2514页 * |
Also Published As
Publication number | Publication date |
---|---|
CN103530381A (en) | 2014-01-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106709067B (en) | Multisource heterogeneous space data circulation method based on Oracle database | |
CN109100730B (en) | Multi-vehicle cooperative rapid map building method | |
WO2019170164A1 (en) | Depth camera-based three-dimensional reconstruction method and apparatus, device, and storage medium | |
CN103077549B (en) | A kind of real-time large-scale terrain the Visual Implementation method based on kd tree | |
CN108763287A (en) | On a large scale can traffic areas driving map construction method and its unmanned application process | |
CN107766471A (en) | The organization and management method and device of a kind of multi-source data | |
CN103426165A (en) | Precise registration method of ground laser-point clouds and unmanned aerial vehicle image reconstruction point clouds | |
CN104657968B (en) | Automatic vehicle-mounted three-dimensional laser point cloud facade classification and outline extraction method | |
CN105469445B (en) | A kind of step-length changeably drawing generating method | |
CN110637461A (en) | Densified optical flow processing in computer vision systems | |
CN112305559A (en) | Power transmission line distance measuring method, device and system based on ground fixed-point laser radar scanning and electronic equipment | |
CN104504742A (en) | Three-dimensional military mark plotting method based on electronic sand table system | |
CN111914615A (en) | Fire-fighting area passability analysis system based on stereoscopic vision | |
CN114299464A (en) | Lane positioning method, device and equipment | |
US20230274466A1 (en) | Point cloud polar coordinate coding method and device | |
WO2022237821A1 (en) | Method and device for generating traffic sign line map, and storage medium | |
CN112270259A (en) | SAR image ship target rapid detection method based on lightweight convolutional neural network | |
CN110381442A (en) | A kind of group robot Target Searching Method based on implicit information interactive mode | |
CN103530381B (en) | A kind of parallel optimization method towards remote sensing image neighborhood processing | |
CN110651475B (en) | Hierarchical data organization for compact optical streaming | |
CN101937577A (en) | Method capable of generating shadow with boundary pixel oversampling effect | |
CN107784001A (en) | Parallel spatial querying method based on CUDA | |
CN103809937B (en) | A kind of intervisibility method for parallel processing based on GPU | |
CN116721229B (en) | Method, device, equipment and storage medium for generating road isolation belt in map | |
Cai et al. | Occupancy grid mapping without ray-casting for high-resolution LiDAR sensors |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20160817 Termination date: 20161017 |
|
CF01 | Termination of patent right due to non-payment of annual fee |