CN104036286A - Multispectral remote-sensing image classification processing method based on Hadoop - Google Patents
Multispectral remote-sensing image classification processing method based on Hadoop Download PDFInfo
- Publication number
- CN104036286A CN104036286A CN201410201066.7A CN201410201066A CN104036286A CN 104036286 A CN104036286 A CN 104036286A CN 201410201066 A CN201410201066 A CN 201410201066A CN 104036286 A CN104036286 A CN 104036286A
- Authority
- CN
- China
- Prior art keywords
- classification
- task
- jobs node
- node
- jobs
- 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.)
- Granted
Links
Abstract
The invention relates to the cross field of remote-sensing classification and a computer distributed computation technology and particularly relates to a multispectral remote-sensing image classification processing method based on Hadoop. Different points lie in that the method includes the following steps: step a): firstly carrying out construction of a classification model on a single task sub-tracker; step b): then using a heartbeat message communication mechanism of a distributed framework to send the classification model to different task trackers and the task trackers starting concurrently carrying out an image block classification task after receiving a classification model file; step c) at last reducing classification results of different sub-blocks into one complete classification result map. The multispectral remote-sensing image classification processing method based on the Hadoop improves significantly the classification speed during classification of mass remote-sensing image data.
Description
Technical field
The present invention relates to the crossing domain of Classification in Remote Sensing Image and computer distribution type computing technique, relate in particular to a kind of multi-spectral Images Classification disposal route based on Hadoop.
Background technology
Along with the develop rapidly of remote sensing technology, remote sensing image is widely used in the aspects such as agricultural, forestry, geology, mineral products, the hydrology, ocean, environmental monitoring, and classification of remote-sensing images is as the element task flow process in remote sensing field, also progressive constantly.In recent years, remote sensing image data amount is explosive increase, forms GB level, TB level, the development trend of PB level.To classification of remote-sensing images, work has brought huge pressure for this, how fast and accurately Remote sensing image classification to be become to remote sensing image application and is badly in need of the problem solving.
In a large amount of practical applications, classification work is general or adopt the artificial semi-supervised sorting technique of extracting the supervised classification method of classification model or automatically choosing sample, after all needing to create disaggregated model according to priori, then image classification is processed.For example, based on SVM(support vector machine, Support Vector Machine) forest classification method is without the artificial classification samples that gathers, and realized automatically the sample process of classification of machine.Traditional has utilized digital elevation model (Digital Elevation Model) based on SVM forest classification method, 250 meters of resolution images of Modis sensor production, to the TM carrying on Landsat satellite series, 30 meters of resolution images that ETM+ sensor produces carry out subsidiary classification, SVM forest classification method selects accurate classification samples by the complex calculation of several data, accurate classification samples enters svm classifier device successively, svm classifier device is trained, until svm classifier device reaches enough niceties of grading, finally use the svm classifier device of training gained to image classification.Machine is automatically sampled and is trained the characteristic of classification to reduce the time that sample is chosen, and has saved manpower and materials.
But the structure of traditional svm classifier method, based on stand-alone environment, is limited to the computing power of CPU and the size of internal memory, can only serial processing remote sensing image file, still need to consume long time.And in processing procedure easily because mechanical disorder causes program crashing, even basic remote sensing image data also can be impaired, the classification effectiveness of stand-alone environment has been difficult to the demand of satisfying magnanimity classification of remote-sensing images Processing tasks.
Summary of the invention
In order to overcome above-mentioned weak point, the present invention proposes a kind of multi-spectral Images Classification disposal route based on Hadoop, and the method has significantly improved the classification speed when magnanimity remote sensing image data is classified.
The present invention solves the problems of the technologies described above adopted technical scheme: a kind of multi-spectral Images Classification disposal route based on Hadoop, and its difference is, its method comprises the following steps:
Step a), first in single work child node, carry out the structure of disaggregated model;
Step b), then utilize the heartbeat message communication mechanism of Distributed Architecture that disaggregated model is sent to each jobs node, jobs node starts the concurrent image blocks classification task of carrying out after receiving disaggregated model file;
Step c), by each piecemeal classification results reduction, be finally a complete classification results figure.
The invention has the beneficial effects as follows: the inventive method is to improve on the basis of traditional Remote Image Classification, utilize distributed treatment and storage capacity, realize classification of remote-sensing images task one-to-many, the tupe of multi-to-multi, the method that the present invention proposes is guaranteeing under the prerequisite of nicety of grading, not only greatly improve the processing speed of classification of remote-sensing images, more can overcome the issuable classification disruption of mechanical disorder in classic method, the most supervised classifications in the compatible remote sensing of the present invention field or semi-supervised sorting technique, with the form of a kind of " total-minute-total ", realized the high efficiency of Remote Image Classification.
The present invention utilizes the distributed concurrent processing storage capacity of the data of PC cluster, improve the speed of magnanimity classification of remote-sensing images, the classification of tradition stand-alone environment is processed easily because the reasons such as mechanical disorder may cause tasks interrupt, even part jobs node fault paralysis in PC cluster in the present invention, task management center still can be given classification task efficiently in order other jobs nodes and be processed.Consider that classification of remote-sensing images requirements of process spends a small amount of time to global image collecting sample train classification models, the present invention is arranged in disaggregated model training process on single jobs node, utilize heartbeat mechanism to set up the information interaction between jobs node and task management center, JobClient, again disaggregated model is distributed to other jobs nodes, realize many jobs nodes parallel sorting, guaranteeing that under the prerequisite of nicety of grading, classification speed is obviously accelerated.
Accompanying drawing explanation
Fig. 1 is the multi-spectral Images Classification disposal route schematic diagram based on Hadoop;
Fig. 2 is single node svm classifier model extraction schematic flow sheet;
Fig. 3 is distributed image classification treatment scheme schematic diagram;
Fig. 4 is the multi-spectral Images Classification disposal route data flow diagram based on Hadoop;
Fig. 5 is that the multi-spectral Images Classification based on Hadoop is processed PC aggregated structure figure.
Embodiment
In order to realize above technical scheme, the present invention need to solve following particular problem: how in Distributed Architecture, to train unified disaggregated model, how jobs node transmits unified disaggregated model, the task division problem of remote sensing image, training classifier and use sorter to how concurrent the carrying out in computer cluster of this dual serial process of image classification, how transmission tasks information etc. in framework.In the inventive method, work child node is also jobs node, and the jobs node of carrying out reduction task is called as reduction task node.
Referring to Fig. 1-Fig. 5, embodiment of the present invention method is example by the forest classification method of take based on SVM, sets forth and how to realize distributed image classification.Generally speaking, the multi-spectral Images Classification disposal route of the embodiment of the present invention based on Hadoop comprises the following steps:
Step a), first in single work child node, carry out the structure of disaggregated model;
Step b), then utilize the heartbeat message communication mechanism of Distributed Architecture that disaggregated model is sent to each jobs node, jobs node starts the concurrent image blocks classification task of carrying out after receiving disaggregated model file;
Step c), by each piecemeal classification results reduction, be finally a complete classification results figure.
preferably,described step a) comprises following concrete steps:
Step a1), import image to be sorted and classification related data into any one jobs node (TaskTracker), jobs node often will send heartbeat message to task management center (JobTracker) at regular intervals, to information such as whether duty, Data Update information, the jobs node of task management center feedback jobs node are normal;
Step a2), task management center continue to receive the heartbeat message that jobs node transmits, and judges whether the data in jobs node have renewal; If certain jobs node data has renewal, a disaggregated model extraction procedure bag will be copied to this jobs node in task management center, and indicate this jobs node to extract disaggregated model to new image data; Meanwhile to engineering center (JobClient), initiate a message, inform that engineering center (JobClient) divides new data;
Step a3), jobs node receives disaggregated model extraction procedure bag, start the local decompress(ion) of a Java Virtual Machine and new image carried out to disaggregated model and extract; Engineering center (JobClient) receives jobs node Data Update message, and new image is carried out to logic partitioning, and logic partitioning file is stored;
Step a4), jobs node completes disaggregated model and extracts, and by heartbeat message, informs task management center.
preferably,described step b) comprises following concrete steps:
B1), task management center receives the heartbeat message that jobs node sends, and then judges whether engineering center (JobClient) has completed the division work of this image; When disaggregated model extraction work and image division work have all completed, task management center starts, according to logical partitioning, new image is divided into a plurality of classification block tasks (Task), and be responsible for each subtask of scheduling and move on jobs node, dispatching distribution task is carried out by sending heartbeat rreturn value to jobs node in task management center, and heartbeat rreturn value has comprised tasks carrying instruction, task Jar file, disaggregated model file and data to be sorted position information;
B2), jobs node receives the heartbeat rreturn value of returning at task management center, judges whether oneself is assigned with task, if the task of being assigned to, the task Jar file of just heartbeat rreturn value being passed back localization; Simultaneously, according to the Data Position information to be sorted in heartbeat rreturn value, this jobs node captures all references data voluntarily it is directly copied to local working folder from the data of other jobs node storages, then creates a new Java Virtual Machine and moves each image blocks classification task.
preferably,described step c) comprises following concrete steps:
Step c1), task management center receives the jobs node complete message of classifying, at certain idle jobs node, create reduction (Reduce) task, and inform all jobs node positional informations that completed classification task of reduction task node;
Step c2), reduction task node received the jobs node positional information of classification task, by this information, from completing the jobs node of classification task, transfer classification results, and be complete grouped data by the classification results reduction of this image;
Step c3), reduction task node completes reduction task, storage grouped data, and storage information is submitted to task management center by heartbeat communication.
Concrete, Fig. 1 is the multispectral image classification processing method schematic diagram based on Hadoop, the task that image to be sorted is set up disaggregated model completes on single jobs node, generate single width image global classification model, the task of image being classified according to disaggregated model realizes in Distributed Calculation cluster, image is divided into the image blocks that logic size is 64M, and allocation schedule, to each jobs node parallel sorting, can effectively reduce classification required time.
Concrete, Fig. 2 is the multi-spectral Images Classification disposal route PC aggregated structure figure based on Hadoop, whole framework is comprised of data switching center, task management center (JobTracker), engineering center (JobClient) and N jobs node.
Data switching center connects task management center, engineering center and All Jobs node, in framework, wantonly two nodes (task management center and engineering center also belong to architectural node) can be interconnected by data switching center, and data switching center is comprised of the network switch of some.
Task management center (JobTracker) is responsible for each subtask of scheduling and is operated on jobs node, and real-time monitoring task node working condition, if find that there is the subtask of failure, just rerun it, generally task management central part is deployed on independent machine.
Engineering center (JobClient) carries out logical partitioning to image, acquiescence is divided into image the image blocks of some 64M, each task Dou Hui engineering center is packaged into Java file by reference program and configuration parameter and stores, and task management center is submitted to in path, engineering center is generally arranged in user side.
Jobs node (TaskTracker) is the fundamental unit of computing and storage in framework, remotely-sensed data can be entered by any one jobs node (Ye Keyou engineering center enters), jobs node is responsible for the training of whole scape remotely-sensed data disaggregated model, and image blocks classification task.In framework, jobs node can have N (N >=2), and for mass remote sensing data classification, jobs node quantity and classification speed are proportional.
Between task management center and jobs node, all pass through " heartbeat mechanism " and intercom mutually between engineering center and task management center, heartbeat mechanism is a simple circulation, regularly sends heartbeat to take over party.Send heartbeat message, tell heartbeat take over party, whether jobs node survives, and serves as message channel between the two simultaneously.Heartbeat take over party receives up-to-date heartbeat message, by heartbeat rreturn value, makes corresponding action command.
Concrete, the forest classification method that Fig. 3 be take based on SVM is example, has illustrated that after remotely-sensed data input jobs node, jobs node extracts the flow process of image classification model.Remote sensing image can be entered by any jobs node, and jobs node is submitted the message of Data Update to task management center by heartbeat.After task management center receives heartbeat message, record jobs node Data Update information, by heartbeat rreturn value, communicate by letter with jobs node, indication jobs node carries out disaggregated model extraction to new data.
Jobs node receives the indication of extracting disaggregated model, file system by the jar file of localized operation from shared-file system copy job node place, jobs node unzips to local working directory jar file, starts a Java Virtual Machine moving model extraction procedure.Jobs node completes after disaggregated model extracts, by heartbeat, model memory address is submitted to task management center.
Concrete, Fig. 4 has illustrated to realize the method for remote sensing image efficient classification in Distributed Architecture.After task management center receives the heartbeat message that jobs node sends, record the information of jobs node Data Update, except sending heartbeat rreturn value to jobs node, Hai Huixiang engineering center sends heartbeat message, and heartbeat message has recorded the remotely-sensed data of new typing in the storage information of each jobs node; Engineering center receives the heartbeat that send at task management center, obtains remotely-sensed data memory address to be sorted, size of data K, and data are given tacit consent to division, is divided into Z image blocks, and except last block size is S, all image blocks sizes are 64M.Relation between image blocks size and complete image size of data meets following formula: K=64 * Z+S.After engineering center completes division, by the storage of image division information, and inform that by heartbeat rreturn value task management center has completed the division work of image to be sorted.After task management center confirms that image has been divided, whether the disaggregated model of inquiring about this image extracts complete.If image is divided and disaggregated model has one not complete, heartbeat will be continued to monitor in task management center; If image is divided and corresponding disaggregated model all completes, task management center starts the classification task of this image.
Utilize JobTracker(task management center in Hadoop) job scheduler, transfer to job scheduler to dispatch classification task, and it carried out to initialization.Create a manipulating object that represents classification task, its encapsulation and recorded information, for tracking task conditions and process.Job scheduler obtains image division information, then for each image blocks creates a subtask (Map), for view picture image creates a reduction task (Reduce).
Jobs node regularly sends heartbeat to task management center, a task can be distributed for normal idle jobs node in task management center, by heartbeat rreturn value and jobs node, communicates.
After jobs node is assigned with task, the jar file of localized operation is copied in the file system at this jobs node place.Meanwhile, the needed all files of sort program (comprising this image classification model) is copied to local disk.Afterwards, the content in jar file is unziped to local working directory, create a new Java Virtual Machine and move each classification task.
After jobs node completes classification subtask, by classification results storage, by heartbeat, submit sorting result information to, task management center gives record.
The reduction task of this image, in minute distribution sort subtask, also can be distributed to a jobs node in task management center.Reduction task node obtains sorting result information corresponding to task management center image by heartbeat mechanism, and the jobs node of carrying out classification subtask according to the sorting result information of real-time update from other those pulls classification results data.
Reduction task node often pulls a classification results data, just classification results is stitched together, until that view picture classification results figure splices is complete, exports complete classification image, is stored in local directory, and completes message to task management center submission task.
Fig. 5 has illustrated variation and the processing of remotely-sensed data stream in multi-spectral Images Classification disposal route based on Hadoop, and remotely-sensed data is entered by any jobs node, and jobs node goes out corresponding disaggregated model to Extraction of Image.Engineering center divides image, obtains image division information.The corresponding disaggregated model of PC cluster utilization, parallel computation, by each image blocks classification, last reduction is complete classification results.
Example of the present invention is realized under PC cluster environment, the experiment proved that, when processing magnanimity multi-spectral Images Classification, the present invention is guaranteeing under the prerequisite suitable with the precision of traditional classification method acquired results, improved work efficiency, improve traditional classification method automaticity, can be used in mass remote sensing data Fast Classification, as Landsat, Modis, Alos, Cerbus, resource series satellite etc.
Compared to traditional Remote Image Classification, the present invention substitutes unit processing environment with PC cluster processing environment, even if breaking down, the Partial Jobs node in PC cluster quits work, task also can be assigned on other normal jobs nodes and process, and has reduced the issuable risk of mechanical disorder.This brand-new mass remote sensing data tupe is along with the increase of process image data amount, with the advantage of traditional remote sensing image disposal route in efficiency can be further obviously, and there is suitable dirigibility, can embed processing procedure and the method for all " classifying after first model " form, as image supervised classification, based on neural network classification method etc.
Remote Image Classification of the present invention can be selected sorting technique based on support vector machine, the sorting technique based on decision tree, the sorting technique based on artificial neural network etc., these select sample monitoring sorting technique comparatively speaking automatically, can realize the feature of computer intelligence sampling train classification models, can save a large amount of manpower and materials that artificial sampling process consumes, more be applicable to mass remote sensing data classification, in conjunction with the present invention, can have given play to larger advantage.
Claims (4)
1. the multi-spectral Images Classification disposal route based on Hadoop, is characterized in that, its method comprises the following steps:
Step a), first in single work child node, carry out the structure of disaggregated model;
Step b), then utilize the heartbeat message communication mechanism of Distributed Architecture that disaggregated model is sent to each jobs node, jobs node starts the concurrent image blocks classification task of carrying out after receiving disaggregated model file;
Step c), by each piecemeal classification results reduction, be finally a complete classification results figure.
2. the multi-spectral Images Classification disposal route based on Hadoop as claimed in claim 1, is characterized in that, described step a) comprises following concrete steps:
Step a1), import image to be sorted and classification related data into any one jobs node (TaskTracker), jobs node often will send heartbeat message to task management center (JobTracker) at regular intervals, to information such as whether duty, Data Update information, the jobs node of task management center feedback jobs node are normal;
Step a2), task management center continue to receive the heartbeat message that jobs node transmits, and judges whether the data in jobs node have renewal; If certain jobs node data has renewal, a disaggregated model extraction procedure bag will be copied to this jobs node in task management center, and indicate this jobs node to extract disaggregated model to new image data; Meanwhile to engineering center (JobClient), initiate a message, inform that engineering center (JobClient) divides new data;
Step a3), jobs node receives disaggregated model extraction procedure bag, start the local decompress(ion) of a Java Virtual Machine and new image carried out to disaggregated model and extract; Engineering center (JobClient) receives jobs node Data Update message, and new image is carried out to logic partitioning, and logic partitioning file is stored;
Step a4), jobs node completes disaggregated model and extracts, and by heartbeat message, informs task management center.
3. the multi-spectral Images Classification disposal route based on Hadoop as claimed in claim 1, is characterized in that, described step b) comprises following concrete steps:
B1), task management center receives the heartbeat message that jobs node sends, and then judges whether engineering center (JobClient) has completed the division work of this image; When disaggregated model extraction work and image division work have all completed, task management center starts, according to logical partitioning, new image is divided into a plurality of classification block tasks (Task), and be responsible for each subtask of scheduling and move on jobs node, dispatching distribution task is carried out by sending heartbeat rreturn value to jobs node in task management center, and heartbeat rreturn value has comprised tasks carrying instruction, task Jar file, disaggregated model file and data to be sorted position information;
B2), jobs node receives the heartbeat rreturn value of returning at task management center, judges whether oneself is assigned with task, if the task of being assigned to, the task Jar file of just heartbeat rreturn value being passed back localization; Simultaneously, according to the Data Position information to be sorted in heartbeat rreturn value, this jobs node captures all references data voluntarily it is directly copied to local working folder from the data of other jobs node storages, then creates a new Java Virtual Machine and moves each image blocks classification task.
4. the multi-spectral Images Classification disposal route based on Hadoop as claimed in claim 1, is characterized in that, described step c) comprises following concrete steps:
Step c1), task management center receives the jobs node complete message of classifying, at certain idle jobs node, create reduction (Reduce) task, and inform all jobs node positional informations that completed classification task of reduction task node;
Step c2), reduction task node received the jobs node positional information of classification task, by this information, from completing the jobs node of classification task, transfer classification results, and be complete grouped data by the classification results reduction of this image;
Step c3), reduction task node completes reduction task, storage grouped data, and storage information is submitted to task management center by heartbeat communication.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410201066.7A CN104036286B (en) | 2014-05-14 | 2014-05-14 | A kind of multi-spectral Images Classification processing method based on Hadoop |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410201066.7A CN104036286B (en) | 2014-05-14 | 2014-05-14 | A kind of multi-spectral Images Classification processing method based on Hadoop |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104036286A true CN104036286A (en) | 2014-09-10 |
| CN104036286B CN104036286B (en) | 2017-06-30 |
Family
ID=51467051
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410201066.7A Active CN104036286B (en) | 2014-05-14 | 2014-05-14 | A kind of multi-spectral Images Classification processing method based on Hadoop |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104036286B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105096334A (en) * | 2015-09-16 | 2015-11-25 | 中国科学院地理科学与资源研究所 | Mine area monitoring method and system |
| CN105491108A (en) * | 2015-11-19 | 2016-04-13 | 浪潮集团有限公司 | System and method for processing remote sensing images |
| CN106991656A (en) * | 2017-03-17 | 2017-07-28 | 杭州电子科技大学 | A kind of distributed geometric correction system and method for mass remote sensing image |
| CN107358260A (en) * | 2017-07-13 | 2017-11-17 | 西安电子科技大学 | A kind of Classification of Multispectral Images method based on surface wave CNN |
| CN111092943A (en) * | 2019-12-13 | 2020-05-01 | 中国科学院深圳先进技术研究院 | Multi-cluster remote sensing method and system of tree structure and electronic equipment |
| CN112070062A (en) * | 2020-09-23 | 2020-12-11 | 南京工业职业技术大学 | Hadoop-based crop waterlogging image classification detection and implementation method |
| US11455233B2 (en) * | 2017-01-24 | 2022-09-27 | Transform Sr Brands Llc | Performance utilities for mobile applications |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102790706A (en) * | 2012-07-27 | 2012-11-21 | 福建富士通信息软件有限公司 | Safety analyzing method and device of mass events |
-
2014
- 2014-05-14 CN CN201410201066.7A patent/CN104036286B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102790706A (en) * | 2012-07-27 | 2012-11-21 | 福建富士通信息软件有限公司 | Safety analyzing method and device of mass events |
Non-Patent Citations (2)
| Title |
|---|
| YUANQING LIN 等: "《Large-scale image classification : Fast feature extraction and SVM training》", 《COMPUTER VISION AND PATTERN RECOGNITION（CVPR），2011 IEEE CONFERENCE ON》 * |
| 李波: "《基于Hadoop的海量图像数据管理》", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105096334A (en) * | 2015-09-16 | 2015-11-25 | 中国科学院地理科学与资源研究所 | Mine area monitoring method and system |
| CN105096334B (en) * | 2015-09-16 | 2018-02-09 | 中国科学院地理科学与资源研究所 | A kind of mining area monitoring method and system |
| CN105491108A (en) * | 2015-11-19 | 2016-04-13 | 浪潮集团有限公司 | System and method for processing remote sensing images |
| US11455233B2 (en) * | 2017-01-24 | 2022-09-27 | Transform Sr Brands Llc | Performance utilities for mobile applications |
| CN106991656A (en) * | 2017-03-17 | 2017-07-28 | 杭州电子科技大学 | A kind of distributed geometric correction system and method for mass remote sensing image |
| CN106991656B (en) * | 2017-03-17 | 2019-09-06 | 杭州电子科技大学 | A kind of mass remote sensing image distribution geometric correction system and method |
| CN107358260A (en) * | 2017-07-13 | 2017-11-17 | 西安电子科技大学 | A kind of Classification of Multispectral Images method based on surface wave CNN |
| CN107358260B (en) * | 2017-07-13 | 2020-09-29 | 西安电子科技大学 | Multispectral image classification method based on surface wave CNN |
| CN111092943A (en) * | 2019-12-13 | 2020-05-01 | 中国科学院深圳先进技术研究院 | Multi-cluster remote sensing method and system of tree structure and electronic equipment |
| CN111092943B (en) * | 2019-12-13 | 2022-09-20 | 中国科学院深圳先进技术研究院 | Multi-cluster remote sensing method and system of tree structure and electronic equipment |
| CN112070062A (en) * | 2020-09-23 | 2020-12-11 | 南京工业职业技术大学 | Hadoop-based crop waterlogging image classification detection and implementation method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104036286B (en) | 2017-06-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104036286A (en) | Multispectral remote-sensing image classification processing method based on Hadoop | |
| CN107577805B (en) | Business service system for log big data analysis | |
| CN101957863B (en) | Data parallel processing method, device and system | |
| CN109948428A (en) | The GPU cluster deep learning edge calculations system of facing sensing information processing | |
| Yu et al. | RTIC-C: A big data system for massive traffic information mining | |
| CN109842563A (en) | Content delivery network flow dispatching method, device and computer readable storage medium | |
| CN107241305A (en) | A kind of network protocol analysis system and its analysis method based on polycaryon processor | |
| CN110308984A (en) | It is a kind of for handle geographically distributed data across cluster computing system | |
| CN103116525A (en) | Map reduce computing method under internet environment | |
| CN104519112A (en) | Intelligent selecting framework for staged cloud manufacturing services | |
| Gong et al. | The design and benchmarking of a cloud-based platform for processing and visualization of traffic data | |
| CN108090186A (en) | A kind of electric power data De-weight method on big data platform | |
| Shtern et al. | Towards a multi-cluster analytical engine for transportation data | |
| CN103207804A (en) | MapReduce load simulation method based on cluster job logging | |
| KR101643886B1 (en) | System and method for analyzing large-scale high resolution satellite image | |
| CN103530297A (en) | Method and device capable of automatically carrying out website analysis | |
| Shinkuma et al. | System design for predictive road-traffic information delivery using edge-cloud computing | |
| Gugnani et al. | Rule based classification on a multi node scalable hadoop cluster | |
| Park et al. | Design of containerized marine knowledge system based on IoT-Cloud and LoRaWAN | |
| DE102020119518A1 (en) | METHOD AND DEVICE FOR MULTIPLE ASYNCHRONOUS CONSUMERS | |
| Tummala et al. | Data management challenges in coastal applications | |
| CN104270458A (en) | Method and device for achieving data distribution management of cloud storage system | |
| CN104182522A (en) | Secondary indexing method and device on basis of circulation bitmap model | |
| Hilgendorf | Efficient industrial big data pipeline for lossless transfer of vehicular data | |
| CN104506663B (en) | A kind of intelligent cloud computing operation management system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |