CN110083446A - A kind of GPU parallel with remote sensing image real-time processing method and system under zero I/O mode - Google Patents
A kind of GPU parallel with remote sensing image real-time processing method and system under zero I/O mode Download PDFInfo
- Publication number
- CN110083446A CN110083446A CN201910327088.0A CN201910327088A CN110083446A CN 110083446 A CN110083446 A CN 110083446A CN 201910327088 A CN201910327088 A CN 201910327088A CN 110083446 A CN110083446 A CN 110083446A
- Authority
- CN
- China
- Prior art keywords
- real
- image
- gpu
- processing
- parallel
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/48—Program initiating; Program switching, e.g. by interrupt
- G06F9/4806—Task transfer initiation or dispatching
- G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/20—Processor architectures; Processor configuration, e.g. pipelining
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/50—Indexing scheme relating to G06F9/50
- G06F2209/5017—Task decomposition
Abstract
The invention discloses a kind of GPU parallel with remote sensing image real-time processing method and system under zero I/O mode, comprising the following steps: to input data carry out Real-time modeling set；Data after Real-time modeling set are subjected to image processing parallel and under zero I/O mode using GPU；Real-time modeling set and image processing are run using real-time processing frame.The invention has the advantages that: overturn conventional process mode, complete image processing result of the currently processed node independent of previous step processing node in process chain, realize real-time treatment effect, above procedure does not need intermediate achievement and generates and produce directly across grade, to improve the production efficiency of conventional orthography production, production cost is saved, and the mode greatly improves the production capacity of the DOM of software using simply.
Description
Technical field
The present invention relates to remote sensing image high-performance treatments fields, it particularly relates to which a kind of GPU is parallel and under zero I/O mode
Remote sensing image real-time processing method and system.
Background technique
With the continuous development of remote sensing technology, remote sensing image data amount is growing day by day.In remote sensing image treatment process, production
Process is complicated, and step is various, and a large amount of input-output operation generates a large amount of intermediate data, is unfavorable for data management and storage benefit
With the promotion of rate, and in the whole process, requires manual intervention and control, to remote sensing image production technology, hardware storage device
There are pressure and test with the everyways such as producers；Traditional remote sensing image production consumed resource is big, and calculating cycle is long, and after
Continuous treatment process depends on the achievement of previous step output to carry out the tune ginseng and operation processing of current business as input, causes entirely
Production process takes a long time, and the degree of automation is low, and the output of a large amount of intermediate achievements occupies the problems such as storage hard disk space is huge,
As shown in figure 5, satellite image, which is inlayed, there are following problems in data handling procedure: 1, intermediate data bring a large amount of IO expense with
Memory space occupies；2, not good preview mechanism bring is handled repeatedly；3, complicated treatment process is unable to satisfy emergency field
Scape.
At present to improve image processing efficiency, graphics processor (GPU) is commonly used to accelerate computing capability, in remote sensing image
In GPU parallelization processing technique and implementation method, by GPU carry out non-graphic general-purpose computations need to by DirectX or
OpenGL similar figure API and GPU is linked up, and this programming mode is excessively complicated, and can not effectively play the calculating of GPU
Performance；The analysis of GPU parallel computing is with application, and 2006, the publication of CUDA transported the GPU of NVIDA at GPU high-performance
The representative of calculation, so, CUDA can only support the GPU of NVIDIA；Remote sensing image algorithm design studies based on OpenCL show with
OpenCL is the advantage that development platform can play GPU calculated performance.
For the problems in the relevant technologies, currently no effective solution has been proposed.
Summary of the invention
For above-mentioned technical problem in the related technology, the present invention propose a kind of GPU parallel with remote sensing shadow under zero I/O mode
As real-time processing method and system, it can accelerate and change original two aspect of production method in GPU, improve image processing efficiency,
And occupancy memory capacity few as far as possible, to reach solution problem above.
To realize the above-mentioned technical purpose, the technical scheme of the present invention is realized as follows:
A kind of GPU parallel with remote sensing image real-time processing method under zero I/O mode, comprising the following steps:
Real-time modeling set is carried out to input data；
Data after Real-time modeling set are subjected to image processing parallel and under zero I/O mode using GPU；
Real-time modeling set and image processing are run using real-time processing frame.
Further, the real-time processing frame specifically includes:
Parallel trigger is submitted into multitask；
Using CPU parallel processing, input image is read in into memory；
Individual task is divided into subtask using CPU, CPU carries out data dispatch and task distributes；
The stream compression of memory to video memory is carried out using CPU, GPU receives multiple instructions parallelization triggering, and reads in simultaneously more
The subtask image blocks of task；
Pixel-level parallelization operation is carried out using GPU, calculates the image blocks of subtask；
The stream compression of memory to hard disk is carried out using CPU.
Further, the Real-time modeling set specifically includes:
Achievement image capturing range is resampled to the image of low resolution；
Image calculation processing is carried out using image processor GPU for low resolution parameter；
It will priori data needed for treated real-time processing links of image statistics；
It is stored for data after statistics.
Further, zero I/O mode includes:
Using GPU to each processing node in block data successively calculation processing chain, its achievement is stored in video memory；
After the processing nodes all in process chain calculate, the achievement final in video memory is circulated to interior
It deposits.
Another aspect of the present invention, provide a kind of GPU parallel with remote sensing image real time processing system under zero I/O mode, packet
It includes:
Module is run, for running Real-time modeling set and image processing using real-time processing frame；
Modeling module, for carrying out Real-time modeling set to input data；
Image processing module, for carrying out data after Real-time modeling set at image parallel and under zero I/O mode using GPU
Reason.
Further, the operation module includes:
Trigger module, for parallel trigger to be submitted in multitask；
Input image is read in memory for utilizing CPU parallel processing by parallel processing module；
Division module, for individual task to be divided subtask using CPU, CPU carries out data dispatch and task distributes；
First stream compression, for carrying out the stream compression of memory to video memory using CPU, GPU receives multiple instructions parallelization
Triggering, and the subtask image blocks of multitask are read in simultaneously；
First computing module calculates the image blocks of subtask for carrying out Pixel-level parallelization operation using GPU；
Second stream compression, for carrying out the stream compression of memory to hard disk using CPU.
Further, in the step S2 Real-time modeling set specifically includes the following steps:
Resampling module, for achievement image capturing range to be resampled to the image of low resolution；
Third computing module, for carrying out image calculation processing using image processor GPU for low resolution parameter；
Statistical module, for will priori data needed for treated real-time processing links of image statistics；
Memory module, for being stored for data after statistics.
Further, zero I/O mode includes:
4th computing module, for using GPU to each processing node in block data successively calculation processing chain, by its at
Fruit is stored in video memory；
Circulate module, is used for after the processing nodes all in process chain calculate, will be final described in video memory
Achievement circulates to memory.
Beneficial effects of the present invention: so that remote sensing image processing can reach live effect, and do not need intermediate achievement and
It is produced directly across grade, to improve the production efficiency of conventional orthography production, saves production cost, and the mode uses letter
It is single, greatly improve the production capacity of the DOM of software.
Detailed description of the invention
It in order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, below will be to institute in embodiment
Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the invention
Example, for those of ordinary skill in the art, without creative efforts, can also obtain according to these attached drawings
Obtain other attached drawings.
Fig. 1 be the GPU described according to embodiments of the present invention parallel with remote sensing image real-time processing method under zero I/O mode
Flow chart；
Fig. 2 be the GPU described according to embodiments of the present invention parallel with remote sensing image real time processing system under zero I/O mode
Schematic diagram；
Fig. 3 is the process chain flow chart described according to embodiments of the present invention；
Fig. 4 is that the GPU described according to embodiments of the present invention is parallel and zero I/O mode schematic diagram；
Fig. 5 is to inlay flow chart of data processing figure according to satellite image.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, those of ordinary skill in the art's every other embodiment obtained belong to what the present invention protected
Range.
Shown in as shown in Figure 1, Figure 3 and Figure 4, according to embodiments of the present invention a kind of GPU parallel with remote sensing under zero I/O mode
Image real-time processing method, comprising the following steps:
Real-time modeling set is carried out to input data；
Wherein, input remote sensing image is handled according to involved in process chain node, counts subsequent processing by Real-time modeling set
Required prior information, and process chain includes: just to penetrate correction, fusion, demotion, wave band calculating, even color, the processing node such as inlaying；Packet
It includes quick obtaining prior information (low resolution), such as: geographical coordinate, grayscale information and the prior information standard for inlaying the acquisitions such as line
Really, the responsible data that can be used as subsequent operation.
Data after Real-time modeling set are subjected to image processing parallel and under zero I/O mode using GPU；
Wherein, GPU uses GPU to carry out hardware-accelerated parallel, plays the graphics capability of GPU；Including being carried out by GPU
Single task Pixel-level parallel processing；Zero I/O mode is that the stream compression between all processing nodes is all completed in memory, no centre
Achievement generates, for one of the feature quickly handled, wherein quickly processing is according to Real-time modeling set information, by input data by place
Chain is managed, under conditions of not exporting intermediate achievement, processing to output result image；The quick processing further relates to application task scheduling
The feature that system is handled, and task scheduling system is the nucleus number and image block number according to CPU, it is automatic to calculate parallel
Number of tasks starts task by parallel task number；The data preparation for handling the stream compression between node, handling node, gives at CPU
Reason；Calculating in processing node is routed automatically to idle GPU equipment, by GPU according to current GPU occupation rate and video memory occupation rate
Responsible completion parallel.
Image processing is carried out parallel and under zero I/O mode using GPU to specifically include: quickly being produced true resolution ratio and is just being penetrated
Image, from input to output, a step is at figure；
It judges whether to inlay operation；
When carrying out damascene process, image capturing range after computation modeling carries out piecemeal for achievement image capturing range；According to edge
Embedding network data, calculates the corresponding weight of image capturing range and raw video range after piecemeal, and by resolution ratio needed for actual production with
And priori data obtained by Modeling Calculation, the processing such as correction, fusion is just being penetrated under the conditions of GPU hardware, it will be considered that after weight
Achievement is write out；
When without damascene process, piecemeal is carried out for achievement image capturing range；The resolution ratio according to needed for actual production with
And priori data obtained by Modeling Calculation, raw video is handled by block；Achievement image is write out by block.
Specifically, piecemeal is specially image block, whole scape remote sensing image to be processed is inputted；Logic partitioning image is defeated
Out；Specific step is as follows: according to achievement geographic range, image wave band number and the locating depth calculated in Real-time modeling set, in conjunction with current meter
The internal storage state of calculation machine, the starting point of calculating logic piecemeal size and each piecemeal；
Block-by-block is quickly handled
Logic partitioning image is inputted, by task scheduling system, by the output of Real-time modeling set and image block, by processing
Chain is quickly handled；With the output of piecemeal processing result；
Specific step is as follows:
Step 1. is according to logic partitioning image and the current utilization of GPU, automatic to calculate parallel task number, i.e., primary place
The batch number of reason, a task handle a logic partitioning；
Each task in Batch is assigned on idle GPU by step 2.CPU；
Node required input data currently processed in task and model information are transferred to video memory by memory by step 3.；
Step 4.GPU realizes parallel processing in such a way that each verification answers a pixel to calculate, and input piecemeal image is pressed
Modeling information processing is completed, and piecemeal result is stored to video memory；
Step 5. continues to execute next processing node, using the achievement of step 4 as input, executes step step 3-step
4, until link processings all in process chain are completed；
After the completion of step 6. will be handled, the result video stream being stored in video memory is gone to memory by CPU；
Step 7. writes out the corresponding piecemeal image of current Batch task；
Step 8. continues to execute step 1-step 7, until all tasks are completed.
Real-time modeling set and image processing are run using real-time processing frame.
Wherein, processing frame includes that Real-time modeling set, task schedule and CPU cooperate with processing with GPU in real time；
CPU is suitble to do logical operation and the circulation of data operates, and each GPU has thousands of a processing cores, is suitble to
Do the parallel processing of Pixel-level.Therefore, data are given to CPU from file to memory, from the interior circulation scheduler task for being stored to video memory
It does, and really handles work and give GPU and do, for the waste for reducing computing resource, it is necessary to according to the computing capability of the two
Reasonable cooperative mode is determined with execution feature, and logical operation is put into CPU, and numerical value calculating is put into GPU, reduces various interactions
Expense, and then improve the execution efficiency of program.
Data are made into Real-time modeling set, information required for each processing links is obtained from model, rather than from
It is obtained in the achievement of previous step, then must export entity result without previous step, it is hard by memory-also just by the circulation of data
Disk has been optimized to memory-video memory.
In one particular embodiment of the present invention, the real-time processing frame specifically includes:
Parallel trigger, A, B, C are submitted into multitask；
Using CPU parallel processing, input image is read in into memory；
Individual task is divided into the subtask of certain amount and granularity using CPU, A1, A2 ... An, B1, B2,
...Bn,C1,C2,...Cn；, CPU carries out data dispatch and task distributes；
The stream compression of memory to video memory is carried out using CPU, GPU receives multiple instructions parallelization triggering, and reads in simultaneously more
The subtask image blocks of task, A1, B1, C1 correspond to image blocks；
Pixel-level parallelization operation is carried out using GPU, realizes that the image blocks of subtask calculate；
The stream compression of memory to hard disk is realized using CPU.
In one particular embodiment of the present invention, the Real-time modeling set specifically includes:
Achievement image capturing range is resampled to the image of low resolution by setting parameter；
Carry out image under the conditions of GPU hardware for low resolution parameter just penetrates the calculation processings such as correction, fusion；
Histogram will need to be counted when such as doing the processing of even color by priori data needed for treated real-time processing links of image statistics
The information such as figure, mean value, variance；It does when inlaying, need to calculate image inlays the information such as line；
Information after counting, is stored in specified path, does priori data preparation for subsequent real-time processing.
Specifically, whole scape image Real-time modeling set；By the remote sensing image input to be processed of whole scape, by each processing node prior information
Output；Specific step is as follows:
Step 1. is according to whole scape image geography information and user setting silhouette target coordinate system, calculating achievement image geography model
It encloses；
Step 2. will input whole scape image and be resampled to low resolution (8 times)；
Step 3. carries out image processing using the achievement of step 2 as input data, by calculate node current in process chain；
Step 4. is on the basis of step 3 achievement, prior information needed for counting subsequent processing；
Model file is written in prior information and process chain calculate node parameter etc. by step 5.；
Step 6. continues to execute next processing node, using the result of step 5 as input, executes step 3-step again
5, until calculate nodes all in process chain execute completion, after the completion of execution, comprising respectively calculating section in process chain in model file
The modeling information of point.
In one particular embodiment of the present invention, zero I/O mode includes:
Using GPU to each processing node in block data successively calculation processing chain, its achievement is stored in video memory；
After the processing nodes all in process chain calculate, the achievement final in video memory is circulated to interior
It deposits.
Specifically, achievement is retained in video memory, is continued to execute after GPU completes currently processed node calculating to block data
Next processing node calculates；
Until calculate nodes all in process chain are finished；
End result in video memory is circulated to memory.
As shown in Figure 2, Figure 3 and Figure 4, another aspect of the present invention, provide a kind of GPU parallel with remote sensing shadow under zero I/O mode
As real time processing system, comprising:
Module is run, for running Real-time modeling set and image processing using real-time processing frame；
Modeling module, for carrying out Real-time modeling set to input data；
Image processing module, for carrying out data after Real-time modeling set at image parallel and under zero I/O mode using GPU
Reason.
The image processing module specifically includes:
Judgment module inlays operation for judging whether to；
When carrying out damascene process, image capturing range after computation modeling carries out piecemeal for achievement image capturing range；According to edge
Embedding network data, calculates the corresponding weight of image capturing range and raw video range after piecemeal, and by resolution ratio needed for actual production with
And priori data obtained by Modeling Calculation, it is handled using image processor GPU, it will be considered that the achievement after weight is write out；
When without damascene process, piecemeal is carried out for achievement image capturing range；The resolution ratio according to needed for actual production with
And priori data obtained by Modeling Calculation, raw video is handled by block；Achievement image is write out by block.
Image processing module includes all calculation processing nodes in process chain, such as: just penetrating correction, fusion, demotion, wave band
Calculating, even color such as inlay at the processing node.
In one particular embodiment of the present invention, the operation module includes:
Trigger module, for parallel trigger to be submitted in multitask；
Input image is read in memory for utilizing CPU parallel processing by parallel processing module；
Division module, for individual task to be divided subtask using CPU, CPU carries out data dispatch and task distributes；
First stream compression, for carrying out the stream compression of memory to video memory using CPU, GPU receives multiple instructions parallelization
Triggering, and the subtask image blocks of multitask are read in simultaneously；
First computing module calculates the image blocks of subtask for carrying out Pixel-level parallelization operation using GPU；
Second stream compression, for carrying out the stream compression of memory to hard disk using CPU.
Further, in the step S2 Real-time modeling set specifically includes the following steps:
Resampling module, for achievement image capturing range to be resampled to the image of low resolution；
Third computing module, for carrying out image calculation processing using image processor GPU for low resolution parameter；
Statistical module, for will priori data needed for treated real-time processing links of image statistics；
Memory module, for being stored for data after statistics.
In one particular embodiment of the present invention, zero I/O mode includes:
4th computing module, for using GPU to each processing node in block data successively calculation processing chain, by its at
Fruit is stored in video memory；
Circulate module, is used for after the processing nodes all in process chain calculate, will be final described in video memory
Achievement circulates to memory.
Specifically, achievement is retained in video memory, is continued to execute after GPU completes currently processed node calculating to block data
Next processing node calculates；
Until calculate nodes all in process chain are finished；
End result in video memory is circulated to memory.
In order to facilitate understanding above-mentioned technical proposal of the invention, below by way of in specifically used mode to of the invention above-mentioned
Technical solution is described in detail.
When specifically used, GPU according to the present invention parallel with remote sensing image real-time processing method under zero I/O mode,
As shown in figure 4, by taking 1 number of high score as an example, it is right by image processing efficiency of the GPU parallel and under " zero IO " mode
Than conventional manufacturing procedures image processing efficiency, this technology is verified to the promotion effect of conventional DOM production efficiency.
Input data:
Input data is the raw video after three correction of sky, including panchromatic and multispectral data, and input data is 16 locating depths
Single band (panchromatic) and 16 locating depth, 4 wave band (multispectral), 229 scapes pair, input data amount are about 234G.
Outputting result requirement:
Outputting result is the even color of 3 wave band, 8 locating depth and inlays rear achievement by standard map sheet, and achievement resolution ratio is 2 meters, standard drawing
Width is 100,000 map sheets.
Software parameter:
It is 4 that parallel task number, which is arranged, in software；Achievement format is tfw format.
Standalone hardware environment:
Computer model: Dell PRECISION Tower 7810；
Processor:Xeon(R)CPU E5-2630v3@2.40GHz 2.40GHz；
Memory: 64.00G；System type: 32 thread of win7 (64x) 2 processor；
Video card: NVIDIA Tesla K40c；Storage: solid state hard disk.
The quick treatment effeciency of table 1 statistics
| Processing step | Used time | Data volume |
| Real-time modeling set | 2:24:00 | 10G |
| Statistical information | 0:20:00 | 5G |
| Quickly processing | 7:04:00 | 306G |
| It is total | 9:28:00 | 321G |
2 conventional treatment Efficiency Statistics of table
| Processing step | Total used time | Data volume |
| Just penetrating correction | 7:08:00 | 201G |
| Fusion | 24:20:00 | 600G |
| Wave band calculates | 8:23:00 | 458G |
| Even color+demotion | 40:00:00 | 250G |
| Inlay line computation | 28:00:00 | 50G |
| It inlays | 8:00:00 | 306G |
| It is total | 115:23:00 | 1865G |
Comparison discovery, orthography production of the GPU parallel and under " zero IO " mode, production efficiency, which has, significantly to be mentioned
It rises, the occupancy of memory space is also greatly lowered, greatly improves production efficiency wherein, different software treatment effeciency is not
Together.
In conclusion by means of above-mentioned technical proposal of the invention, so that remote sensing image processing can reach live effect,
And do not need intermediate achievement and produced directly across grade, to improve the production efficiency of conventional orthography production, save production
Cost, and the mode greatly improves the production capacity of the DOM of software using simply.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Within mind and principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (8)
1. a kind of GPU parallel with remote sensing image real-time processing method under zero I/O mode, which comprises the following steps:
Real-time modeling set is carried out to input data；
Data after Real-time modeling set are subjected to image processing parallel and under zero I/O mode using GPU；
Real-time modeling set and image processing are run using real-time processing frame.
2. GPU according to claim 1 parallel with remote sensing image real-time processing method under zero I/O mode, which is characterized in that
The real-time processing frame specifically includes:
Parallel trigger is submitted into multitask；
Using CPU parallel processing, input image is read in into memory；
Individual task is divided into subtask using CPU, CPU carries out data dispatch and task distributes；
The stream compression of memory to video memory is carried out using CPU, GPU receives multiple instructions parallelization triggering, and reads in multitask simultaneously
Subtask image blocks；
Pixel-level parallelization operation is carried out using GPU, calculates the image blocks of subtask；
The stream compression of memory to hard disk is carried out using CPU.
3. GPU according to claim 1 parallel with remote sensing image real-time processing method under zero I/O mode, which is characterized in that
The Real-time modeling set specifically includes:
Achievement image capturing range is resampled to the image of low resolution；
Image calculation processing is carried out using image processor GPU for low resolution parameter；
It will priori data needed for treated real-time processing links of image statistics；
It is stored for data after statistics.
4. GPU according to claim 1-3 is parallel with remote sensing image real-time processing method under zero I/O mode, special
Sign is that zero I/O mode includes:
Using GPU to each processing node in block data successively calculation processing chain, its achievement is stored in video memory；
After the processing nodes all in process chain calculate, the achievement final in video memory is circulated to memory.
5. a kind of GPU parallel with remote sensing image real time processing system under zero I/O mode characterized by comprising
Modeling module, for carrying out Real-time modeling set to input data；
Image processing module, for data after Real-time modeling set to be carried out image processing parallel and under zero I/O mode using GPU；
Module is run, for running Real-time modeling set and image processing using real-time processing frame.
6. GPU according to claim 5 parallel with remote sensing image real time processing system under zero I/O mode, which is characterized in that
The operation module includes:
Trigger module, for parallel trigger to be submitted in multitask；
Input image is read in memory for utilizing CPU parallel processing by parallel processing module；
Division module, for individual task to be divided subtask using CPU, CPU carries out data dispatch and task distributes；
First stream compression, for carrying out the stream compression of memory to video memory using CPU, GPU receives multiple instructions parallelization triggering,
And the subtask image blocks of multitask are read in simultaneously；
First computing module calculates the image blocks of subtask for carrying out Pixel-level parallelization operation using GPU；
Second stream compression, for carrying out the stream compression of memory to hard disk using CPU.
7. GPU according to claim 5 parallel with remote sensing image real time processing system under zero I/O mode, which is characterized in that
Real-time modeling set in the step S2 specifically includes the following steps:
Resampling module, for achievement image capturing range to be resampled to the image of low resolution；
Third computing module, for carrying out image calculation processing using image processor GPU for low resolution parameter；
Statistical module, for will priori data needed for treated real-time processing links of image statistics；
Memory module, for being stored for data after statistics.
8. special according to the described in any item GPU of claim 5-7 parallel with remote sensing image real time processing system under zero I/O mode
Sign is that zero I/O mode includes:
4th computing module, for, to each processing node in block data successively calculation processing chain, its achievement being deposited using GPU
It is stored in video memory；
Circulate module, is used for after the processing nodes all in process chain calculate, by the achievement final in video memory
It circulates to memory.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910327088.0A CN110083446A (en) | 2019-04-23 | 2019-04-23 | A kind of GPU parallel with remote sensing image real-time processing method and system under zero I/O mode |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910327088.0A CN110083446A (en) | 2019-04-23 | 2019-04-23 | A kind of GPU parallel with remote sensing image real-time processing method and system under zero I/O mode |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110083446A true CN110083446A (en) | 2019-08-02 |
Family
ID=67416107
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910327088.0A Pending CN110083446A (en) | 2019-04-23 | 2019-04-23 | A kind of GPU parallel with remote sensing image real-time processing method and system under zero I/O mode |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110083446A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111080520A (en) * | 2019-12-10 | 2020-04-28 | 西安中科星图空间数据技术有限公司 | Satellite image fast embedding method and device based on main frame search |
| CN113012083A (en) * | 2021-03-03 | 2021-06-22 | 上海景运信息技术有限公司 | Data processing method for image zero IO in remote sensing ortho image processing process |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080186313A1 (en) * | 2007-02-06 | 2008-08-07 | Samsung Electronics Co., Ltd. | Method and apparatus for high-speed image processing based on graphic processing unit |
| CN102609971A (en) * | 2012-01-11 | 2012-07-25 | 南京大学 | Quick rendering system using embedded GPU (Graphics Processing Unit) for realizing 3D-GIS (Three Dimensional-Geographic Information System) |
| CN103281376A (en) * | 2013-05-31 | 2013-09-04 | 武汉大学 | Method for automatic caching construction of massive timing sequence remote-sensing images in cloud environment |
| CN103617626A (en) * | 2013-12-16 | 2014-03-05 | 武汉狮图空间信息技术有限公司 | Central processing unit (CPU) and ground power unit (GPU)-based remote-sensing image multi-scale heterogeneous parallel segmentation method |
| US20150206309A1 (en) * | 2014-01-21 | 2015-07-23 | University Of Rochester | System and method for real-time image registration |
| CN105683914A (en) * | 2013-11-22 | 2016-06-15 | 英特尔公司 | Method and apparatus to improve performance of chained tasks on a graphics processing unit |
| CN109300079A (en) * | 2017-07-25 | 2019-02-01 | 北京吉威时代软件股份有限公司 | The real-time light and color homogenization preview technology of remote sensing image based on GPU |
-
2019
- 2019-04-23 CN CN201910327088.0A patent/CN110083446A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080186313A1 (en) * | 2007-02-06 | 2008-08-07 | Samsung Electronics Co., Ltd. | Method and apparatus for high-speed image processing based on graphic processing unit |
| CN102609971A (en) * | 2012-01-11 | 2012-07-25 | 南京大学 | Quick rendering system using embedded GPU (Graphics Processing Unit) for realizing 3D-GIS (Three Dimensional-Geographic Information System) |
| CN103281376A (en) * | 2013-05-31 | 2013-09-04 | 武汉大学 | Method for automatic caching construction of massive timing sequence remote-sensing images in cloud environment |
| CN105683914A (en) * | 2013-11-22 | 2016-06-15 | 英特尔公司 | Method and apparatus to improve performance of chained tasks on a graphics processing unit |
| CN103617626A (en) * | 2013-12-16 | 2014-03-05 | 武汉狮图空间信息技术有限公司 | Central processing unit (CPU) and ground power unit (GPU)-based remote-sensing image multi-scale heterogeneous parallel segmentation method |
| US20150206309A1 (en) * | 2014-01-21 | 2015-07-23 | University Of Rochester | System and method for real-time image registration |
| CN109300079A (en) * | 2017-07-25 | 2019-02-01 | 北京吉威时代软件股份有限公司 | The real-time light and color homogenization preview technology of remote sensing image based on GPU |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111080520A (en) * | 2019-12-10 | 2020-04-28 | 西安中科星图空间数据技术有限公司 | Satellite image fast embedding method and device based on main frame search |
| CN113012083A (en) * | 2021-03-03 | 2021-06-22 | 上海景运信息技术有限公司 | Data processing method for image zero IO in remote sensing ortho image processing process |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104050632B (en) | Method and system for the processing of multisample pixel data | |
| CN106462393B (en) | System and method for unified Application Programming Interface and model | |
| US9355492B2 (en) | System, method, and computer program product for utilizing a wavefront path tracer | |
| CN113178014B (en) | Scene model rendering method and device, electronic equipment and storage medium | |
| US20130243329A1 (en) | Parallel object detection method for heterogeneous multithreaded microarchitectures | |
| CN103761215B (en) | Matrix transpose optimization method based on graphic process unit | |
| CN103019671B (en) | The framework programmed method of the generic programming of data-oriented intensity remote sensing image processing | |
| CN110134516A (en) | Finance data processing method, device, equipment and computer readable storage medium | |
| CN107341761A (en) | A kind of calculating of deep neural network performs method and system | |
| US9305324B2 (en) | System, method, and computer program product for tiled deferred shading | |
| CN103871019B (en) | Method and apparatus for processing path image to promote rasterisation | |
| WO2023087893A1 (en) | Object processing method and apparatus, computer device, storage medium and program product | |
| CN106202145A (en) | A kind of preprocessing of remote sensing images system of Data-intensive computing | |
| CN110083446A (en) | A kind of GPU parallel with remote sensing image real-time processing method and system under zero I/O mode | |
| CN102880509A (en) | Compute unified device architecture (CUDA) based grid digital elevation model (DEM) neighborhood analysis system and method | |
| CN110135569A (en) | Heterogeneous platform neuron positioning three-level flow parallel method, system and medium | |
| CN115509764B (en) | Real-time rendering multi-GPU parallel scheduling method and device and memory | |
| CN105468439A (en) | Adaptive parallel algorithm for traversing neighbors in fixed radius under CPU-GPU (Central Processing Unit-Graphic Processing Unit) heterogeneous framework | |
| Emeliyanov et al. | GPU-based tracking algorithms for the ATLAS high-level trigger | |
| Zhang | Speeding up large-scale geospatial polygon rasterization on GPGPUs | |
| CN109493410A (en) | A kind of real-time rendering method of gigabit grade pixel image | |
| CN105701760A (en) | Histogram real-time generation method of geographic raster data optional polygon area | |
| CN112948123B (en) | Spark-based grid hydrological model distributed computing method | |
| CN109410136A (en) | Even color method and processing unit based on most short transmission path | |
| DE102021114013A1 (en) | TECHNIQUES FOR EFFICIENT SCANNING OF AN IMAGE |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190802 |
|
| RJ01 | Rejection of invention patent application after publication |