CN108090865B - On-orbit real-time streaming processing method and system for optical satellite remote sensing images - Google Patents
On-orbit real-time streaming processing method and system for optical satellite remote sensing images Download PDFInfo
- Publication number
- CN108090865B CN108090865B CN201711354960.8A CN201711354960A CN108090865B CN 108090865 B CN108090865 B CN 108090865B CN 201711354960 A CN201711354960 A CN 201711354960A CN 108090865 B CN108090865 B CN 108090865B
- Authority
- CN
- China
- Prior art keywords
- processing
- data
- time
- module
- real
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- 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
- 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
- G06F9/5044—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 considering hardware capabilities
-
- 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
- G06F9/505—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 considering the load
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
- Image Input (AREA)
Abstract
本发明提供一种光学卫星遥感影像在轨实时流式处理方法及系统，包括构建单处理模组的处理流水线，分别测定数据流入、数据处理、数据流出的速度，记录相应耗时；估计要达到实时处理性能所需要的处理模组数N，制定并配置数据动态分包策略，根据不同处理算法需要实现配置，包括初步划分为连续的标准景，根据需要支持标准景内部进一步划分为连续的数据段，根据需要支持数据段内更进一步地划分为并列的格网、超像素或像素；根据硬件能力确定负载均衡策略，部署定制化流式处理架构以实现光学卫星遥感影像在轨实时处理。该方法可以充分利用在轨处理模组的处理能力与处理算法本身的特性，最优化在轨数据处理流程，提高遥感信息获取效率和实时性。
The invention provides a method and system for in-orbit real-time streaming processing of optical satellite remote sensing images, including constructing a processing pipeline of a single processing module, respectively measuring the speed of data inflow, data processing and data outflow, and recording the corresponding time-consuming; The number of processing modules N required for real-time processing performance, formulate and configure data dynamic subcontracting strategies, and implement configurations according to different processing algorithms, including initial division into continuous standard scenes, and support for further division of standard scenes into continuous data as needed The data segment can be further divided into parallel grids, superpixels or pixels according to the needs; the load balancing strategy is determined according to the hardware capabilities, and the customized streaming processing architecture is deployed to realize the real-time processing of optical satellite remote sensing images in orbit. The method can make full use of the processing capability of the on-orbit processing module and the characteristics of the processing algorithm itself, optimize the on-orbit data processing flow, and improve the efficiency and real-time performance of remote sensing information acquisition.
Description
Technical Field
The invention belongs to the field of high-performance processing of space remote sensing images, and relates to an on-orbit real-time processing method and system of multi-processing module optical satellite remote sensing images based on intelligent data subpackaging and load balancing.
Background
The traditional optical remote sensing satellite information acquisition link comprises four main links of satellite acquisition data-satellite data downloading-ground centralized processing-information extraction, and with the rapid development of a high-resolution earth observation satellite, explosively increased remote sensing data brings huge pressure to the data downloading link and a ground processing system. In general, the time delay from satellite data generation to ground effective information acquisition is tens of minutes, and the information acquisition timeliness is low. In order to fully utilize the limited transmission bandwidth and the satellite transit time window and shorten the information acquisition delay, a secondary information acquisition link of satellite acquisition and processing data and effective information downloading and distribution needs to be established, namely, massive original data needs to be processed in an on-orbit real-time manner so as to directly generate and download effective information. Because the on-orbit processing platform has volume and power consumption limitations and limited performance, on one hand, a plurality of groups of embedded processing platforms with low power consumption and high performance need to be adopted to cooperatively process data to realize on-orbit real-time processing; on the other hand, it is necessary to fully parallelize the computation and fully exploit the performance of the computation storage means.
Disclosure of Invention
The invention aims to solve the problem of providing a streaming processing technical scheme suitable for in-orbit real-time processing of optical satellite remote sensing images aiming at a plurality of groups of low-power-consumption embedded in-orbit processing modules.
The technical scheme of the invention provides an on-orbit real-time streaming processing method for optical satellite remote sensing images, which comprises the following steps:
wherein, TIN，TProcAnd TOUTRespectively measuring the time consumed by data inflow, data processing and data outflow in the step 1;
step 3, formulating and configuring a data dynamic subpackage strategy according to the result of the step 2, realizing configuration according to different processing algorithm requirements, and including preliminarily dividing into continuous standard scenes, supporting the interior of the standard scenes to be further divided into continuous data segments according to requirements, and further dividing into parallel grids, super pixels or pixels in the data segments according to requirements;
step 4, determining a load balancing strategy according to the hardware capacity;
step 5, deploying a customized streaming processing framework to realize on-orbit real-time processing of the optical satellite remote sensing image, wherein the on-orbit real-time processing comprises deploying N processing modules based on the processing module number N obtained in the step 2; and (4) performing real-time subpackaging and distribution on continuously inflowing input according to the data subpackaging strategy formulated in the step (3) and the load balancing strategy determined in the step (4) so as to realize parallel processing among a plurality of processing modules and realize on-orbit real-time processing of the optical satellite remote sensing images.
Moreover, the configuration in step 3 is implemented according to different processing algorithm requirements, including but not limited to the strategies of dividing according to a standard scene, dividing according to a grid, dividing according to a super-pixel, and dividing according to a pixel.
And, the load balancing policy in step 4 includes, but is not limited to, a policy of sequentially and evenly distributing data packets and a policy of weighting and distributing data packets according to different processing capabilities of modules according to different hardware capabilities and processing algorithm requirements.
And the strategy of weighted distribution according to the processing capacities of different modules is to set weights according to the processing capacities of the processing modules, and the modules with strong processing capacities and large weights are distributed with more to-be-processed data packets.
Moreover, the processing capability weights of the single processing module1 to the single processing module N are respectively W1，W2，…，WNThen, there are:
generating C data packets in unit time, the number of data packets processed by each processing module is Wi·。
The invention also provides an on-orbit real-time streaming processing system for the optical satellite remote sensing image, which comprises the following modules:
the first module is used for constructing a processing assembly line of the single processing module, respectively measuring the speed of data inflow, data processing and data outflow and recording corresponding consumed time;
a second module for estimating a number of processing modules N required to achieve real-time processing performance as follows,
wherein, TIN，TProcAnd TOUTRespectively measuring the time consumed by data inflow, data processing and data outflow in the step 1;
the third module is used for formulating and configuring a data dynamic subpackage strategy according to the result obtained by the second module, realizing configuration according to different processing algorithm requirements, and comprises the steps of preliminarily dividing the data into continuous standard scenes, further dividing the interior of the standard scenes into continuous data sections according to requirements, and further dividing the interior of the data sections into parallel grids, super pixels or pixels according to requirements;
a fourth module for determining a load balancing policy according to the hardware capability;
the fifth module is used for deploying a customized streaming processing architecture to realize on-orbit real-time processing of the optical satellite remote sensing image, and comprises a processing module number N obtained based on the second module and N processing modules; and performing real-time subpackaging and distribution on continuously inflowing input according to a data subpackaging strategy formulated by the third module and a load balancing strategy determined by the fourth module so as to realize parallel processing among a plurality of processing modules and realize on-orbit real-time processing of the optical satellite remote sensing images.
Moreover, the third module implements configurations according to different processing algorithm requirements, including but not limited to strategies of dividing according to a standard scene, dividing according to a grid, dividing according to super-pixels, and dividing according to pixels.
And the load balancing strategy in the fourth module includes, but is not limited to, strategies of sequentially and evenly distributing data packets and weighting and distributing according to different module processing capacities according to different hardware capacities and processing algorithm requirements.
And the strategy of weighted distribution according to the processing capacities of different modules is to set weights according to the processing capacities of the processing modules, and the modules with strong processing capacities and large weights are distributed with more to-be-processed data packets.
Moreover, the processing capability weights of the single processing module1 to the single processing module N are respectively W1，W2，…，WNThen, there are:
generating C data packets per unit timeThen the number of data packets processed by each processing module is Wi·。
The streaming processing framework suitable for the on-orbit processing of the optical satellite remote sensing image can fully utilize the processing capacity of the on-orbit processing module and the characteristics of the processing algorithm, optimize the on-orbit data processing flow, realize the on-orbit real-time processing of the optical satellite remote sensing image of 'data inflow-data processing-data outflow', and improve the acquisition efficiency and the real-time performance of remote sensing information. The processing algorithm applicable to the invention comprises and is not limited to optical satellite sensor correction, target detection, cloud detection and the like, and has important market value.
Drawings
FIG. 1 is a schematic flow chart of an embodiment of the present invention;
FIG. 2 is a schematic diagram of a data processing flow according to an embodiment of the present invention;
fig. 3 is a schematic diagram of data dynamic packetization according to an embodiment of the present invention.
Detailed Description
The technical scheme of the invention is explained in detail in the following by combining the drawings and the embodiment.
The invention provides an in-orbit real-time streaming processing architecture of an optical satellite remote sensing image, which comprises the steps of determining the time consumed by data inflow, data processing and data outflow in a single processing module aiming at the characteristics of an in-orbit processing algorithm and the characteristics of processing hardware of a specific optical satellite remote sensing image; further calculating the number of processing modules required for achieving the real-time processing performance; and combining the characteristics of an on-orbit processing algorithm to formulate and configure a data dynamic subpackage strategy so as to generate a customized streaming architecture combining software and hardware; and deploying a customized streaming framework to realize on-orbit real-time processing of the optical satellite remote sensing image.
As shown in fig. 1, an in-orbit real-time streaming processing method for optical satellite remote sensing images provided in an embodiment includes the following steps:
Here, the processing module refers to a basic data processing unit including functional modules for processing, storing, communicating, and the like, and includes but is not limited to a processing board and an SoC system using an FPGA and a DSP as core devices. Based on the processing module, the incoming data can be processed through programming, and the processing result is transmitted out. In the processing process, for a specific data block, three main steps of data inflow, data processing and data outflow must be executed in a strict sequence; for a plurality of data blocks, the three steps can be processed by constructing a pipeline stack, that is, for data A, B, C which flows in successively, data B can be processed while data C flows in, and the processing result of data a is output. Due to the inconsistency of the time consumption of the three steps of data inflow, data processing and data outflow, the execution efficiency of the whole process depends on the step (usually data processing) which takes the longest time.
The invention realizes on-track parallel real-time processing by reasonably sub-packaging the input data and uniformly distributing the input data to the multi-processing modules. Aiming at specific processing algorithms such as optical satellite sensor correction, target detection, cloud detection and the like, the ratio of the processing rate of the target algorithm to the data input and output rate needs to be measured first, and then the minimum number of processing modules needed to be used is obtained.
Combining the hardware characteristics of the processing module, the target algorithm needs to be efficiently mapped to the processing module, and the average values of the consumed time of data inflow, data processing and data outflow are obtained by executing a plurality of continuous tests on the single-processing module pipeline and are respectively marked as TIN，TProcAnd TOUT。
And 2, calculating the number of the processing modules required for achieving the real-time processing performance, namely the total number of the single processing modules. The ratio of the data processing speed to the data inflow/outflow speed (the slowest of the two) can be calculated and rounded up. The invention further proposes a preferred scheme, and the number of the processing modules is N, and the value is obtained by the following formula:
wherein T isIN，TProcAnd TOUTAre respectively asAnd executing a plurality of groups of tests on the single processing module to obtain the average value of the consumed time of data inflow, data processing and data outflow.
The step is independent mathematical calculation based on the result of the step 1, and is independent of processing environments such as specific software and hardware.
And 3, formulating and configuring a data dynamic subpackage strategy according to the result of the step 2, wherein the strategy comprises the strategies of dividing according to a standard scene, dividing according to a grid, dividing according to a super pixel, dividing according to a pixel and the like.
In the field of optical satellite remote sensing image data processing, a square or nearly square image product is output according to the imaging width of a satellite, and the image product is called as a standard scene. As shown in fig. 2, the left side is the image data to be processed, which are sequentially transmitted (flowed) from top to bottom, and are initially divided into continuous standard scenes; inside the standard scene, the standard scene can be further divided into continuous data segments (segment 0, segment 1, segment 2, segment 3, etc.); further, the data segments may be divided into parallel grids, superpixels, or pixels. The right side is a parallel multi-processing module, and the processing module is internally composed of a control core (such as an embedded CPU, a DSP and the like) suitable for processing scalar operations such as control, branching, circulation, iteration and the like and a parallel core (such as an embedded GPU, an FPGA and the like) suitable for processing single instruction multiple data vector operations. In fig. 2, the left divided data segment 0, 1, 2, 3 is distributed to modules 0, 1, 2, 3(Module0, Module1, Module2, Module3) for processing, and the data segment is further divided into grids and processed with the control core and the parallel core.
The method adopts different sub-packaging strategies for different types of optical remote sensing image processing algorithms, sub-packages the inflow data according to the whole scene granularity for the algorithm needing the whole scene processing, such as the correction of an optical satellite sensor, and distributes the data to different processing modules one by one for processing; for algorithms such as target identification and cloud detection, a standard scene needs to be further divided into a plurality of approximately equal data segments, the data segments are sequentially distributed to different data processing modules section by section for processing, the data segments are further divided into grids, superpixels or pixels in the processing modules, and a control core and a parallel core are called to perform collaborative parallel computing according to the main operation type of a specific processing algorithm (different operation cores are suitable for processing different types of operations).
Based on the processing module number N required for realizing real-time calculation obtained in the step 2, the specific data subpackaging strategy is formulated by combining the characteristics of different types of processing algorithms. The sub-packaging strategy is closely related to the characteristics of the processing algorithm, and is not related to the configuration and the computing capacity of a single processing module. For example, for an algorithm needing to consider the relation between data segments, the data packet is distributed to N processing modules for processing by taking the whole scene as a unit; for the algorithm capable of processing independently among the data segments, the data packets are distributed to N processing modules by taking the data segments as units, and the data segments are further divided into grids which are distributed to different cores in a single module for processing.
And 4, determining a load balancing strategy according to the hardware capacity.
The load balancing strategy comprises the strategies of sequentially and evenly distributing data packets and weighting and distributing according to different module processing capacities according to different hardware capacities and processing algorithm requirements without limitation.
The embodiment sets the weight according to the processing capacity of the processing module, and the module group with strong processing capacity and large weight is allocated with more data packets to be processed, so that each processing module is fully utilized, the overall execution efficiency is improved, and node overload or node starvation is avoided. The processing capability weights of the single processing module1 to the single processing module N are respectively W1，W2，…，WNThen, there are:
generating C data packets in unit time, the number of data packets processed by each processing module is Wi·。
And 5, deploying a customized streaming processing framework to realize on-orbit real-time processing of the optical satellite remote sensing image. As shown in fig. 3, based on the number N of processing modules obtained in step 2, N processing modules (1-N) are deployed; the data distribution module performs real-time sub-packaging and distribution on continuously flowing input according to the data sub-packaging strategy formulated in the step 3 and the load balancing strategy determined in the step 4 so as to realize parallel processing among a plurality of processing modules; inside the processing module, scalar operation and vector operation are cooperatively processed by the control core and the parallel core according to the characteristics of the processing algorithm, so that the on-orbit real-time processing of the optical satellite remote sensing image is realized.
Taking a time line for realizing real-time cloud detection processing production of the optical satellite remote sensing image by applying the method, from the original point, the camera spends 0.403 seconds to generate data of 'data 0 segment 0', and the data segment is distributed to the module0 for processing and consumes 1.439 seconds; while the camera continues to spend 0.403 seconds generating "data 0 segment 1" data, which is assigned to module1 for processing, taking 1.501 seconds; and so on. Therefore, through a reasonable data sub-packaging and balancing strategy, the data can be immediately processed after being generated, and the whole system does not have the phenomena of waiting and data accumulation. And 1.417 seconds after the data generation is finished, the data processing is finished, and the real-time cloud detection processing of the optical satellite remote sensing image is realized.
In specific implementation, the method provided by the invention can realize automatic operation flow based on software technology, and can also realize a corresponding system in a modularized mode.
The embodiment of the invention also provides an on-orbit real-time streaming processing system for the optical satellite remote sensing image, which comprises the following modules:
the first module is used for constructing a processing assembly line of the single processing module, respectively measuring the speed of data inflow, data processing and data outflow and recording corresponding consumed time;
a second module for estimating a number of processing modules N required to achieve real-time processing performance as follows,
wherein, TIN，TProcAnd TOUTRespectively measuring the time consumed by data inflow, data processing and data outflow in the step 1;
the third module is used for formulating and configuring a data dynamic subpackage strategy according to the result obtained by the second module, realizing configuration according to different processing algorithm requirements, and comprises the steps of preliminarily dividing the data into continuous standard scenes, further dividing the interior of the standard scenes into continuous data sections according to requirements, and further dividing the interior of the data sections into parallel grids, super pixels or pixels according to requirements;
a fourth module for determining a load balancing policy according to the hardware capability;
the fifth module is used for deploying a customized streaming processing architecture to realize on-orbit real-time processing of the optical satellite remote sensing image, and comprises a processing module number N obtained based on the second module and N processing modules; and performing real-time subpackaging and distribution on continuously inflowing input according to a data subpackaging strategy formulated by the third module and a load balancing strategy determined by the fourth module so as to realize parallel processing among a plurality of processing modules and realize on-orbit real-time processing of the optical satellite remote sensing images.
The specific implementation of each module can refer to the corresponding step, and the detailed description of the invention is omitted.
The specific embodiments described herein are merely illustrative of the spirit of the invention. Various modifications or additions may be made to the described embodiments or alternatives may be employed by those skilled in the art without departing from the spirit or ambit of the invention as defined in the appended claims.
Claims (8)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711354960.8A CN108090865B (en) | 2017-12-15 | 2017-12-15 | On-orbit real-time streaming processing method and system for optical satellite remote sensing images |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711354960.8A CN108090865B (en) | 2017-12-15 | 2017-12-15 | On-orbit real-time streaming processing method and system for optical satellite remote sensing images |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108090865A CN108090865A (en) | 2018-05-29 |
| CN108090865B true CN108090865B (en) | 2021-06-15 |
Family
ID=62176588
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711354960.8A Active CN108090865B (en) | 2017-12-15 | 2017-12-15 | On-orbit real-time streaming processing method and system for optical satellite remote sensing images |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108090865B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112165572A (en) * | 2020-09-10 | 2021-01-01 | Oppo（重庆）智能科技有限公司 | Image processing method, device, terminal and storage medium |
| CN113283412B (en) * | 2021-07-26 | 2021-09-21 | 广州乐盈信息科技股份有限公司 | On-orbit real-time streaming processing method for optical satellite remote sensing image |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006115283A (en) * | 2004-10-15 | 2006-04-27 | Mitsubishi Electric Corp | Data compressor and data expanding device |
| JP2007293558A (en) * | 2006-04-25 | 2007-11-08 | Hitachi Ltd | Target recognition program and target recognition apparatus |
| CN102521687A (en) * | 2011-12-01 | 2012-06-27 | 中国资源卫星应用中心 | Miniaturized universal platform for preprocessing remote-sensing satellite data |
| CN103020620A (en) * | 2012-12-14 | 2013-04-03 | 武汉大学 | Remote-sensing image ortho-rectification method based on central processing unit (CPU) and graphics processing unit (GPU) coprocessing |
| CN103796018A (en) * | 2014-01-14 | 2014-05-14 | 齐齐哈尔大学 | Remote-sensing image real-time compression and progressive transmission system |
| CN103809969A (en) * | 2014-01-15 | 2014-05-21 | 中国公路工程咨询集团有限公司 | Remote-sensing image data parallel resampling method based on pre-fragmentation in cloud environment |
| CN106909644A (en) * | 2017-02-21 | 2017-06-30 | 中国科学院电子学研究所 | A kind of multistage tissue and indexing means towards mass remote sensing image |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2013020143A1 (en) * | 2011-08-04 | 2013-02-07 | University Of Southern California | Image-based crack quantification |
| CN105430378A (en) * | 2015-11-26 | 2016-03-23 | 航天恒星科技有限公司 | An image quality evaluation system and method |
| CN106846292B (en) * | 2016-12-14 | 2019-08-27 | 中国资源卫星应用中心 | A Method for On-orbit Detection of Invalid Pixels |
| CN107391258B (en) * | 2017-07-20 | 2020-09-22 | 武汉大学 | Software and hardware integrated portable remote sensing image real-time processing system |
-
2017
- 2017-12-15 CN CN201711354960.8A patent/CN108090865B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006115283A (en) * | 2004-10-15 | 2006-04-27 | Mitsubishi Electric Corp | Data compressor and data expanding device |
| JP2007293558A (en) * | 2006-04-25 | 2007-11-08 | Hitachi Ltd | Target recognition program and target recognition apparatus |
| CN102521687A (en) * | 2011-12-01 | 2012-06-27 | 中国资源卫星应用中心 | Miniaturized universal platform for preprocessing remote-sensing satellite data |
| CN103020620A (en) * | 2012-12-14 | 2013-04-03 | 武汉大学 | Remote-sensing image ortho-rectification method based on central processing unit (CPU) and graphics processing unit (GPU) coprocessing |
| CN103796018A (en) * | 2014-01-14 | 2014-05-14 | 齐齐哈尔大学 | Remote-sensing image real-time compression and progressive transmission system |
| CN103809969A (en) * | 2014-01-15 | 2014-05-21 | 中国公路工程咨询集团有限公司 | Remote-sensing image data parallel resampling method based on pre-fragmentation in cloud environment |
| CN106909644A (en) * | 2017-02-21 | 2017-06-30 | 中国科学院电子学研究所 | A kind of multistage tissue and indexing means towards mass remote sensing image |
Non-Patent Citations (2)
| Title |
|---|
| ZY-3卫星轨道拟合与预报精度分析;向夏芸 等;《测绘通报》;20151231(第1期);第8-14页 * |
| 大型遥感影像数据库的空间无缝数据组织;王密 等;《武汉大学学报信息科学版》;20011031;第26卷(第5期);第419-424页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108090865A (en) | 2018-05-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102492477B1 (en) | Matrix multiplier | |
| CN107301456B (en) | Implementation method of multi-core acceleration of deep neural network based on vector processor | |
| US9099866B2 (en) | Apparatus, methods and systems for parallel power flow calculation and power system simulation | |
| CN108875958A (en) | Use the primary tensor processor of outer product unit | |
| KR20180052063A (en) | Convolution neural network system and operation method thereof | |
| KR102610842B1 (en) | Processing element and operating method thereof in neural network | |
| CN111931917A (en) | Forward computing implementation method and device, storage medium and electronic device | |
| Parthasarathy et al. | Defer: Distributed edge inference for deep neural networks | |
| KR20220015813A (en) | Method and apparatus for performing deep learning operations. | |
| Schmidt et al. | A generic VHDL template for 2D stencil code applications on FPGAs | |
| CN113010845B (en) | Computing device, method and related product for performing matrix multiplication | |
| CN112199636A (en) | Fast convolution method and device suitable for microprocessor | |
| CN113486298A (en) | Model compression method and matrix multiplication module based on Transformer neural network | |
| CN113592066B (en) | Hardware acceleration method, device, equipment and storage medium | |
| CN108090865B (en) | On-orbit real-time streaming processing method and system for optical satellite remote sensing images | |
| CN111861793A (en) | Distribution and utilization electric service distribution method and device based on cloud edge cooperative computing architecture | |
| CN108537334A (en) | A kind of acceleration array design methodology for CNN convolutional layer operations | |
| CN116962176A (en) | Data processing method, device and system of distributed cluster and storage medium | |
| CN108304926A (en) | A kind of pond computing device and method suitable for neural network | |
| CN111831354A (en) | Data precision configuration method, device, chip array, equipment and medium | |
| TW202316325A (en) | Parallel depth-wise processing architectures for neural networks | |
| CN108875957B (en) | Primary tensor processor and the system for using primary tensor processor | |
| CN109190755A (en) | Matrix conversion device and method towards neural network | |
| CN113052299A (en) | Neural network memory computing device based on lower communication bound and acceleration method | |
| US11682099B2 (en) | Hardware accelerator for integral image computation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20240108 Address after: No. 120, Haixiang Middle Road, Fengcheng Street, Haiyang City, Yantai City, Shandong Province, 265100 Patentee after: Land sea space (Yantai) Information Technology Co.,Ltd. Address before: 430072 Hubei Province, Wuhan city Wuchang District of Wuhan University Luojiashan Patentee before: WUHAN University |