EMPLOYING MULTI CORE ARCHITECTURE TO OPTIMIZE ON PERFORMANCE, FOR APPROACH IN METEOROLOGICAL DATA
0 likes117 views
The document discusses the design and implementation of a cloud detection application using MATLAB to optimize performance in meteorological data analysis. It highlights the importance of accurately estimating solar energy reaching the Earth's surface, as errors in cloud detection can significantly impact weather predictions and climate studies. The proposed multi-core architecture enhances computational efficiency, demonstrating a significant reduction in processing time as data size increases.
![International Journal of Advanced Research in Engineering and Technology (IJARET), ISSN 0976 –
6480(Print), ISSN 0976 – 6499(Online), Volume 6, Issue 4, April (2015), pp. 17-23 © IAEME
17
EMPLOYING MULTI CORE ARCHITECTURE TO
OPTIMIZE ON PERFORMANCE, FOR APPROACH IN
METEOROLOGICAL DATA
Swati Jain
Institute of Technology, Nirma University, Faculty Group,
Computer Engineering, Ahmedabad, India, 380025
Somil Gadhwal
Institute of Technology, Nirma University, Research Scholar,
Computer Engineering, Ahmedabad, India, 380025
ABSTRACT
Cloud detection is an important task in meteorological application. Cloud information is
especially important for now-casting purposes [1] and as an input for different satellite based
estimation of atmospheric and surface parameters [2-4]. The solar energy is the principal source of
energy in the solar system. Clouds have high reflectance and absorption property which is used to
distinguish them with land, water or sea area. There is critical demand to develop application, which
can calculate the presence of cloud by using the available satellite image processing data, so that
prediction of radiated solar energy can be optimised and energy budget can be predicted more easily.
Keywords: Absorption; Clouds; MATLAB; Remote Sensing; Reflectance.
1. INTRODUCTION
The importance of capacity and utilizing it for various purposes is realized by human being.
In order to measure the correct amount of solar energy reaching earth, the layers of cloud should be
separated while estimating the solar energy of a particular area. To estimate the amount of sunlight
reaching the earth surface can serve various purposes [5] such as
• Estimating the climate conditions.
• Predicting the weather conditions.
• Predicting the rain, type of crop to suit the above climatic condition.
• Setting up solar panels.
INTERNATIONAL JOURNAL OF ADVANCED RESEARCH IN ENGINEERING
AND TECHNOLOGY (IJARET)
ISSN 0976 - 6480 (Print)
ISSN 0976 - 6499 (Online)
Volume 6, Issue 4, April (2015), pp. 17-23
© IAEME: www.iaeme.com/ IJARET.asp
Journal Impact Factor (2015): 8.5041 (Calculated by GISI)
www.jifactor.com
IJARET
© I A E M E](https://image.slidesharecdn.com/employingmulticorearchitecturetooptimizeonperformanceforcloudmaskingapproachinmeteorologicaldata-160426104651/85/EMPLOYING-MULTI-CORE-ARCHITECTURE-TO-OPTIMIZE-ON-PERFORMANCE-FOR-APPROACH-IN-METEOROLOGICAL-DATA-1-320.jpg)
![International Journal of Advanced Research in Engineering and Technology (IJARET), ISSN 0976 –
6480(Print), ISSN 0976 – 6499(Online), Volume 6, Issue 4, April (2015), pp. 17-23 © IAEME
18
Detection of cloud is important and crucial in weather and climate studies because detection
of cloud over a region is necessary in order to support many atmospheric or weather parameters such
as aerosol optical depth, brightness temperature, fog detection etc. that provides a depth view in
climate and weather process. Any minor change in actual scenario of cloud can result in huge errors.
Researchers have tried to estimate the amount of solar energy reaching the earth surface by detecting
the amount of solar energy reflected through cloud. Using Matlab for satellite image processing, an
application is developed to efficiently and effectively calculate presence of cloud. This can be done
in several methods such as applying thresholds to greyscale images which has been used to detect
clouds since the first satellite images were produced Rossow (1989) [6]. But There are several issues
such as:-
• Threshold is never global for every satellite image; it varies according to the time of day, so
obtaining a global threshold is a huge problem.
• Also size of clouds is irregular which makes them hard to detected or point out.
Another problem are their type , clouds are of different type such as thin cloud which are at higher
height from land surface and have lower temperature , also the thick and dark cloud that are situated
at lower height that have high temperature.
Remote sensing provides an excellent way to employ meteorological studies , however the
data provided by remote sensing is huge in amount so in order to carry out any calculation, we need
an environment that gives us not only accurate results but also time taken to calculate the result
should be minimum.
So all these problems contribute to further image processing that is not only expensive in
terms of time and space but also the results obtained by them are not very satisfactory.
Increase in numbers of satellite launched and in spectrum, radiometric and temporal
resolution has results in piling up a lot of data. This is an advantage as more accurate information can
be obtained but at the cost of increasing computation requirements.
Fig. 1 Use of application in other weather prediction techniques
The paper proposes an approach by which cloud masking application can provide result in
feasible space and time complexity. Though clouds are commonly encountered phenomena, but they
are difficult to simulate because they are complex in shape and interact with light in a complex
fashion. Clouds not only limit the amount of valid land surface information in a scene, but
undetected cloudy pixels effect atmospheric correction procedures, aerosol retrievals and also
compromise the estimation of biophysical parameters [7].](https://image.slidesharecdn.com/employingmulticorearchitecturetooptimizeonperformanceforcloudmaskingapproachinmeteorologicaldata-160426104651/85/EMPLOYING-MULTI-CORE-ARCHITECTURE-TO-OPTIMIZE-ON-PERFORMANCE-FOR-APPROACH-IN-METEOROLOGICAL-DATA-2-320.jpg)
![International Journal of Advanced Research in Engineering and Technology (IJARET), ISSN 0976 –
6480(Print), ISSN 0976 – 6499(Online), Volume 6, Issue 4, April (2015), pp. 17-23 © IAEME
19
2. NEED OF CLOUD MASK
As seen in Fig 1 ,Cloud mask is an important application as any error in estimation of cloud
are amplified in further prediction ,even a minor change in prediction of temperature can cause
repeated errors in prediction of weather forecasting and climatic studies. It is considered that even
1K change in temperature can result in predicting the rainfall estimation by double of its present
condition. Since flora and fauna of India is highly dependent on rainfall, weather prediction
calculation should be carried out in a careful manner.
Weather prediction can be done using various measures, satellites are also used to capture
data in the form of multispectral images and these images are used to extract information about the
temperature of the area. Reflectivity for solar energy is more by clouds in comparison to land hence
it is possible to distinguish the cloud from land using thresholding, however it is not always possible
to obtain a global threshold. They vary according to the temperature of the surrounding area and the
also on the time i.e. day or night. If this model is used for energy budget prediction in solar energy
conservation then this error may lead to day to day energy needs. So it is important to calculate the
weather condition correctly and detecting clouds is the first step of every weather prediction
application (as shown in Fig. 1). By doing this we can separate the land surface and cloud and can
make use of reflectance values.
3. APPROACH USED
The dataset contains mask of weather condition related to cloud . In order to provide a
friendly interaction, a GUI has been created using Matlab GUI Development Environment (GUIDE).
Matlab contains worker as their computational unit. Workers are equivalent to number of cores, each
worker has their own computational space and they can communicate with each other. Collection of
worker is called pool. Matlab provides commands to open or close pool.
4. MATLAB
To Matlab provides several toolboxes such as Image processing toolbox that is used to
process images and can perform valuable computation with minimum time overhead. It also provides
Parallel computing environment.
Matlab provide Parallel processing Toolbox [8] to compute data intensive problem in
multicore processor, GPU and computer cluster. Data is parallelized using parallelized algorithms,
dedicated arrays and parallelized for loop, that gives a user freedom to efficiently distribute the data
without the burden on managing it. Parallel application can be executed in interactive or batch mode.
Matlab provides easy and efficient way to for compute image processing task involved in satellite
image processing. Several other toolboxes are GUIDE (GUI Development Environment) that allows
a user to create GUI using drag and drop facility and background function.
5. NEED OF MULTICORE ENVIRONMENT
With the continuous growth of complexity of programming, it has been difficult now to
increase the performance by scaling clock speed further. To meet this demand, modern System-On-
Chip solutions contain multiple processing cores. Reason behind using multicore programming is
• Can make use of various kind of parallelism.
• Decrease power consumption.
• Effectively hide memory latency.
• Resources are utilized more effectively](https://image.slidesharecdn.com/employingmulticorearchitecturetooptimizeonperformanceforcloudmaskingapproachinmeteorologicaldata-160426104651/85/EMPLOYING-MULTI-CORE-ARCHITECTURE-TO-OPTIMIZE-ON-PERFORMANCE-FOR-APPROACH-IN-METEOROLOGICAL-DATA-3-320.jpg)
![International Journal of Advanced Research in Engineering and Technology (IJARET), ISSN 0976 –
6480(Print), ISSN 0976 – 6499(Online), Volume 6, Issue 4, April (2015), pp. 17-23 © IAEME
20
6. CLOUD MASKING ALGORITHM
So in order to surpass all the above mention problems in cloud detection, more than one
method is used to calculate cloud mask according to their type. The algorithm is as follow [9].
7. DATASET USED
The data set used is INSAT 3D CMK data that has been provided by MOSDAC. HDF5 data
is extracted in Matlab and satellite images can be viewed by the user.
Fig. 2 Satellite Image showing cloud mask.
8. ALGORITHM
The dataset contains mask data for every half an hour interval. Fig 2 shows a pictorial view
of dataset that contains cloud mask of full disk view of earth. For every dataset belonging to
particular date, flags have been maintained for clear, cloudy, probably clear, probably cloudy. Once
the data has been loaded for binary_mask and for cmk_mask(that contains cmk mask ), a function for
cloud mask is called and it repeat the sequence for every cmk_mask data.
8.1 Data Structure used
• binary_mask - Binary mask for every states of India. That has been stored previously.
• cmk_mask - mask of cloud condition given by INSAT 3D satellite.
• Counter - Contains clear, cloudy, probably clear, probably cloudy count of a particular file.
Initially set to zero.
• day_array- contains the index of counter that is maximum for particular time of day.](https://image.slidesharecdn.com/employingmulticorearchitecturetooptimizeonperformanceforcloudmaskingapproachinmeteorologicaldata-160426104651/85/EMPLOYING-MULTI-CORE-ARCHITECTURE-TO-OPTIMIZE-ON-PERFORMANCE-FOR-APPROACH-IN-METEOROLOGICAL-DATA-4-320.jpg)
EMPLOYING MULTI CORE ARCHITECTURE TO OPTIMIZE ON PERFORMANCE, FOR APPROACH IN METEOROLOGICAL DATA
- 1. International Journal of Advanced Research in Engineering and Technology (IJARET), ISSN 0976 – 6480(Print), ISSN 0976 – 6499(Online), Volume 6, Issue 4, April (2015), pp. 17-23 © IAEME 17 EMPLOYING MULTI CORE ARCHITECTURE TO OPTIMIZE ON PERFORMANCE, FOR APPROACH IN METEOROLOGICAL DATA Swati Jain Institute of Technology, Nirma University, Faculty Group, Computer Engineering, Ahmedabad, India, 380025 Somil Gadhwal Institute of Technology, Nirma University, Research Scholar, Computer Engineering, Ahmedabad, India, 380025 ABSTRACT Cloud detection is an important task in meteorological application. Cloud information is especially important for now-casting purposes [1] and as an input for different satellite based estimation of atmospheric and surface parameters [2-4]. The solar energy is the principal source of energy in the solar system. Clouds have high reflectance and absorption property which is used to distinguish them with land, water or sea area. There is critical demand to develop application, which can calculate the presence of cloud by using the available satellite image processing data, so that prediction of radiated solar energy can be optimised and energy budget can be predicted more easily. Keywords: Absorption; Clouds; MATLAB; Remote Sensing; Reflectance. 1. INTRODUCTION The importance of capacity and utilizing it for various purposes is realized by human being. In order to measure the correct amount of solar energy reaching earth, the layers of cloud should be separated while estimating the solar energy of a particular area. To estimate the amount of sunlight reaching the earth surface can serve various purposes [5] such as • Estimating the climate conditions. • Predicting the weather conditions. • Predicting the rain, type of crop to suit the above climatic condition. • Setting up solar panels. INTERNATIONAL JOURNAL OF ADVANCED RESEARCH IN ENGINEERING AND TECHNOLOGY (IJARET) ISSN 0976 - 6480 (Print) ISSN 0976 - 6499 (Online) Volume 6, Issue 4, April (2015), pp. 17-23 © IAEME: www.iaeme.com/ IJARET.asp Journal Impact Factor (2015): 8.5041 (Calculated by GISI) www.jifactor.com IJARET © I A E M E
- 2. International Journal of Advanced Research in Engineering and Technology (IJARET), ISSN 0976 – 6480(Print), ISSN 0976 – 6499(Online), Volume 6, Issue 4, April (2015), pp. 17-23 © IAEME 18 Detection of cloud is important and crucial in weather and climate studies because detection of cloud over a region is necessary in order to support many atmospheric or weather parameters such as aerosol optical depth, brightness temperature, fog detection etc. that provides a depth view in climate and weather process. Any minor change in actual scenario of cloud can result in huge errors. Researchers have tried to estimate the amount of solar energy reaching the earth surface by detecting the amount of solar energy reflected through cloud. Using Matlab for satellite image processing, an application is developed to efficiently and effectively calculate presence of cloud. This can be done in several methods such as applying thresholds to greyscale images which has been used to detect clouds since the first satellite images were produced Rossow (1989) [6]. But There are several issues such as:- • Threshold is never global for every satellite image; it varies according to the time of day, so obtaining a global threshold is a huge problem. • Also size of clouds is irregular which makes them hard to detected or point out. Another problem are their type , clouds are of different type such as thin cloud which are at higher height from land surface and have lower temperature , also the thick and dark cloud that are situated at lower height that have high temperature. Remote sensing provides an excellent way to employ meteorological studies , however the data provided by remote sensing is huge in amount so in order to carry out any calculation, we need an environment that gives us not only accurate results but also time taken to calculate the result should be minimum. So all these problems contribute to further image processing that is not only expensive in terms of time and space but also the results obtained by them are not very satisfactory. Increase in numbers of satellite launched and in spectrum, radiometric and temporal resolution has results in piling up a lot of data. This is an advantage as more accurate information can be obtained but at the cost of increasing computation requirements. Fig. 1 Use of application in other weather prediction techniques The paper proposes an approach by which cloud masking application can provide result in feasible space and time complexity. Though clouds are commonly encountered phenomena, but they are difficult to simulate because they are complex in shape and interact with light in a complex fashion. Clouds not only limit the amount of valid land surface information in a scene, but undetected cloudy pixels effect atmospheric correction procedures, aerosol retrievals and also compromise the estimation of biophysical parameters [7].
- 3. International Journal of Advanced Research in Engineering and Technology (IJARET), ISSN 0976 – 6480(Print), ISSN 0976 – 6499(Online), Volume 6, Issue 4, April (2015), pp. 17-23 © IAEME 19 2. NEED OF CLOUD MASK As seen in Fig 1 ,Cloud mask is an important application as any error in estimation of cloud are amplified in further prediction ,even a minor change in prediction of temperature can cause repeated errors in prediction of weather forecasting and climatic studies. It is considered that even 1K change in temperature can result in predicting the rainfall estimation by double of its present condition. Since flora and fauna of India is highly dependent on rainfall, weather prediction calculation should be carried out in a careful manner. Weather prediction can be done using various measures, satellites are also used to capture data in the form of multispectral images and these images are used to extract information about the temperature of the area. Reflectivity for solar energy is more by clouds in comparison to land hence it is possible to distinguish the cloud from land using thresholding, however it is not always possible to obtain a global threshold. They vary according to the temperature of the surrounding area and the also on the time i.e. day or night. If this model is used for energy budget prediction in solar energy conservation then this error may lead to day to day energy needs. So it is important to calculate the weather condition correctly and detecting clouds is the first step of every weather prediction application (as shown in Fig. 1). By doing this we can separate the land surface and cloud and can make use of reflectance values. 3. APPROACH USED The dataset contains mask of weather condition related to cloud . In order to provide a friendly interaction, a GUI has been created using Matlab GUI Development Environment (GUIDE). Matlab contains worker as their computational unit. Workers are equivalent to number of cores, each worker has their own computational space and they can communicate with each other. Collection of worker is called pool. Matlab provides commands to open or close pool. 4. MATLAB To Matlab provides several toolboxes such as Image processing toolbox that is used to process images and can perform valuable computation with minimum time overhead. It also provides Parallel computing environment. Matlab provide Parallel processing Toolbox [8] to compute data intensive problem in multicore processor, GPU and computer cluster. Data is parallelized using parallelized algorithms, dedicated arrays and parallelized for loop, that gives a user freedom to efficiently distribute the data without the burden on managing it. Parallel application can be executed in interactive or batch mode. Matlab provides easy and efficient way to for compute image processing task involved in satellite image processing. Several other toolboxes are GUIDE (GUI Development Environment) that allows a user to create GUI using drag and drop facility and background function. 5. NEED OF MULTICORE ENVIRONMENT With the continuous growth of complexity of programming, it has been difficult now to increase the performance by scaling clock speed further. To meet this demand, modern System-On- Chip solutions contain multiple processing cores. Reason behind using multicore programming is • Can make use of various kind of parallelism. • Decrease power consumption. • Effectively hide memory latency. • Resources are utilized more effectively
- 4. International Journal of Advanced Research in Engineering and Technology (IJARET), ISSN 0976 – 6480(Print), ISSN 0976 – 6499(Online), Volume 6, Issue 4, April (2015), pp. 17-23 © IAEME 20 6. CLOUD MASKING ALGORITHM So in order to surpass all the above mention problems in cloud detection, more than one method is used to calculate cloud mask according to their type. The algorithm is as follow [9]. 7. DATASET USED The data set used is INSAT 3D CMK data that has been provided by MOSDAC. HDF5 data is extracted in Matlab and satellite images can be viewed by the user. Fig. 2 Satellite Image showing cloud mask. 8. ALGORITHM The dataset contains mask data for every half an hour interval. Fig 2 shows a pictorial view of dataset that contains cloud mask of full disk view of earth. For every dataset belonging to particular date, flags have been maintained for clear, cloudy, probably clear, probably cloudy. Once the data has been loaded for binary_mask and for cmk_mask(that contains cmk mask ), a function for cloud mask is called and it repeat the sequence for every cmk_mask data. 8.1 Data Structure used • binary_mask - Binary mask for every states of India. That has been stored previously. • cmk_mask - mask of cloud condition given by INSAT 3D satellite. • Counter - Contains clear, cloudy, probably clear, probably cloudy count of a particular file. Initially set to zero. • day_array- contains the index of counter that is maximum for particular time of day.
- 5. International Journal of Advanced Research in Engineering and Technology (IJARET), ISSN 0976 – 6480(Print), ISSN 0976 – 6499(Online), Volume 6, Issue 4, April (2015), pp. 17-23 © IAEME 21 Fig. 3 Flow of Application. 9. RESULT The output summaries the cloud condition for that particular day, weather cloudy or clear or probably cloudy. The application is executed in both serial and parallel environment and the runtime of both environment is highly comparative. It can be shown from the table that the gain over serial is less when size of data is less but it increases as the size of data increases. From the Table 1, we can conclude that the gain of parallel processing over serial in this application is nearly 83%, which can provide better time complexity and can increase the accuracy of the estimation as more and more data can be estimated in the same time. The table shows prediction of cloud at several duration of time. Fig. 4 Execution time of serial and Parallel Processing.
- 6. International Journal of Advanced Research in Engineering and Technology (IJARET), ISSN 0976 – 6480(Print), ISSN 0976 – 6499(Online), Volume 6, Issue 4, April (2015), pp. 17-23 © IAEME 22 Fig. 5 Percentage Gain of Parallel over Serial Processing. Table 1: Showing comparison of serial and parallel processing Duration (Hours) Serial (Seconds) Parallel (Seconds) Percentage Gain % 0.5 1.800 0.546 69.67 2 13.690 2.418 82.38 4 14.5230 1.7220 88.14 6 14.54700 1.6830 88.40 8 15.2390 3.2970 78.36 10 17.0000 4.3520 74.40 12 22.4000 3.5816 84.01 14 25.6370 3.6504 85.76 16 27.0974 4.0092 85.20 18 31.0286 5.6784 81.69 20 32.5262 5.5380 82.97 22 34.5232 5.8320 88.22 24 36.5480 6.3240 82.70 Average 82.45 10. CONCLUSION The application can be extended to add some more weather prediction application so that it can provide a unified platform for weather prediction application and also by using this approach, computation can be saved.
- 7. International Journal of Advanced Research in Engineering and Technology (IJARET), ISSN 0976 – 6480(Print), ISSN 0976 – 6499(Online), Volume 6, Issue 4, April (2015), pp. 17-23 © IAEME 23 REFERENCES 1. Stephens G.L., G. Vane, R.J. Boain, G.G. Mace, K. Sassen, Z. Wang, A.J. Illinworth, E.J. OConner, W.B. Rossow, S.L.Durden, S.D. Miller, R.T. Austin, A. Benedetti, C. Mitrescu. The cloud Sat Science Team 2002: The CloudSat mission and the A-Train, Bull. Amer. Meteor. Soc., 83(12), 1771-1790. 2. A Bendix J., R. Rollenbeck and W.E. Palacios, 2004. “Cloud detection in the Tropics- a suitable tool for climate-ecological studies in the high mountains of Equador, “. Intl. J. Rem. Sens., 25 (21), 4521-4540. 3. Albers S.C., J.A. McGinley, D.A.Birkenheuer and J.R. Smart 1996. The local analysis of prediction system (LAPS): Analysis of clouds, precipitation and temperature. Wea. Forcasting, 11, 273-287. 4. Sandvik A.D., 1998.” Implementation and validation of a condensation scheme in a messoscale model”. Mon. Sea. Rev., 126, 1882-1905. 5. NCAR’s Contribution To Wind And Solar Energy Prediction, https://www.rap. ucar.edu/ projects/ncars-contribution-to-wind-and-solar-energy-prediction. 6. Izak van Zyl Marais_, Johan Adam du Preez, Willem Herman Steyn.” An optimal image transform for threshold-based cloud detection using heteroscedastic discriminant analysis”, Intl. J. Rem. Sens. 7. Fernando Sedano , Pieter Kempeneers, Peter Strobl, Jan Kucera, Peter Vogt, Lucia Seebach, Jesús San-Miguel-Ayanz,” A cloud mask methodology for high resolution remote sensing data combining information from high and medium resolution optical sensors “, ELSEVIER. 8. Parallel Computing Toolbox in Matlab by Mathworks, http://www.mathworks.in/ products/parallel-computing/parallel/distarrays.html. 9. INSAT 3D Data Product Catalog by Indian Meteorological Department http://www.imd .gov.in/section/satmet/dynamic/INSAT3D_Catalog.pdf. 10. Prabhat Kumar Sinha, Vijay Kumar Yadav, Saurabha Kumar and Rajneesh Pandey, “Static Analysis of Complex Structure of Beams By Interpolation Method Approach To Matlab” International Journal of Mechanical Engineering & Technology (IJMET), Volume 5, Issue 5, 2014, pp. 28 - 44, ISSN Print: 0976 – 6340, ISSN Online: 0976 – 6359. 11. Gopichand Allaka, Prasad Raju Kalidindi, Koteswara Rao S, Manibabu Daadi, Abhay Patnala, “Design of Solid Shafts Using Matlab” International Journal of Mechanical Engineering & Technology (IJMET), Volume 3, Issue 3, 2012, pp. 645 - 653, ISSN Print: 0976 – 6340, ISSN Online: 0976 – 6359. 12. Darshana Mistry and Asim Banerjee, “Discrete Wavelet Transform Using Matlab” International journal of Computer Engineering & Technology (IJCET), Volume 4, Issue 2, 2013, pp. 252 - 259, ISSN Print: 0976 – 6367, ISSN Online: 0976 – 6375.