Presentation is loading. Please wait.

Presentation is loading. Please wait.

11/26/081 AUTOMATIC SOLAR ACTIVITY DETECTION BASED ON IMAGES FROM HSOS NAOC, HSOS YANG Xiao, LIN GangHua

Similar presentations


Presentation on theme: "11/26/081 AUTOMATIC SOLAR ACTIVITY DETECTION BASED ON IMAGES FROM HSOS NAOC, HSOS YANG Xiao, LIN GangHua"— Presentation transcript:

1 11/26/081 AUTOMATIC SOLAR ACTIVITY DETECTION BASED ON IMAGES FROM HSOS NAOC, HSOS YANG Xiao, LIN GangHua

2 11/26/082 OUTLINE Purpose for Automatic Detection -- Space Weather Forecast Purpose for Automatic Detection -- Space Weather Forecast Automatic Solar Activities Detection Based on Images from HSOS Automatic Solar Flare Detection Automatic Solar Flare Detection Conclusions Conclusions

3 11/26/083 OUTLINE Purpose for Automatic Detection -- Space Weather Forecast Purpose for Automatic Detection -- Space Weather Forecast Automatic Solar Activities Detection Based on Images from HSOS Automatic Solar Flare Detection Automatic Solar Flare Detection Conclusions Conclusions

4 11/26/084 Purpose for Automatic Detection -- Space Weather Forecast (1) ‏ The term “space weather” refers to adverse conditions on the Sun that may affect space-borne or ground-based technological systems and can endanger human health or life. The importance of space weather is increasing day after day because of the way solar activities affect life on Earth and it will continue to increase as we rely more and more on different communication and power system. The term “space weather” refers to adverse conditions on the Sun that may affect space-borne or ground-based technological systems and can endanger human health or life. The importance of space weather is increasing day after day because of the way solar activities affect life on Earth and it will continue to increase as we rely more and more on different communication and power system. Ground based systems: Induced electric fields and currents can disrupt the normal operation of high voltage power transmission grids, pipelines, telecommunications cables, metallic oil and gas pipelines and railway signaling. Ground based systems: Induced electric fields and currents can disrupt the normal operation of high voltage power transmission grids, pipelines, telecommunications cables, metallic oil and gas pipelines and railway signaling. Communications systems: Wireless communications systems suffer from interruption of service like frequency jamming and dropped communications due to radio bursts caused by solar microwave emissions. Communications systems: Wireless communications systems suffer from interruption of service like frequency jamming and dropped communications due to radio bursts caused by solar microwave emissions. Space based systems: Adverse space weather conditions can cause anomalies and system failures and increased drag on the movement of satellites and spacecraft leading to slow-downs, changes in orbits and shorter life-times of missions. Space based systems: Adverse space weather conditions can cause anomalies and system failures and increased drag on the movement of satellites and spacecraft leading to slow-downs, changes in orbits and shorter life-times of missions.

5 11/26/085 Purpose for Automatic Detection -- Space Weather Forecast (2) ‏ Necessity of applying automatic detection There are an increasing number of space missions and ground based observatories providing continuous observation of the Sun at many different wavelengths. We are becoming “data rich” but without automated data analysis and knowledge extraction techniques, we continue to be “knowledge poor”. There are an increasing number of space missions and ground based observatories providing continuous observation of the Sun at many different wavelengths. We are becoming “data rich” but without automated data analysis and knowledge extraction techniques, we continue to be “knowledge poor”. A long standing problem in solar physics is establishing a correlation between the occurrence of solar activity (e.g., solar flares and coronal mass ejection(CMEs)) and solar features (sunspots, active regions and filaments) observed in various wavelengths. A long standing problem in solar physics is establishing a correlation between the occurrence of solar activity (e.g., solar flares and coronal mass ejection(CMEs)) and solar features (sunspots, active regions and filaments) observed in various wavelengths. An efficient prediction system requires the successful integration of solar physics, machine learning and maybe solar imaging. An efficient prediction system requires the successful integration of solar physics, machine learning and maybe solar imaging. There is no machine learning algorithm that is known to provide the “best” learning performance especially in the solar domain. In most cases, empirical studies must be carried out to compare the performances of these algorithms before the final decision on which learning algorithm to use can be made. There is no machine learning algorithm that is known to provide the “best” learning performance especially in the solar domain. In most cases, empirical studies must be carried out to compare the performances of these algorithms before the final decision on which learning algorithm to use can be made.

6 11/26/086 OUTLINE Purpose for Automatic Detection -- Space Weather Forecast Purpose for Automatic Detection -- Space Weather Forecast Automatic Solar Activities Detection Based on Images from HSOS Automatic Solar Flare Detection Automatic Solar Flare Detection Conclusions Conclusions

7 11/26/087 Solar Multi-channel Telescope

8 11/26/088 Full solar disk vector magnetograph at Huairou

9 11/26/089 Automatic Solar Activities Detection Based on Images from HSOS Real-time detecion of solar flares in Hα full-disk images Real-time detecion of solar flares in Hα full-disk images Automatic detection, classification and tracking of filaments in Hα full-disk images Automatic detection, classification and tracking of filaments in Hα full-disk images Automatic detection of sunspots using magnetic full-disk images Automatic detection of sunspots using magnetic full-disk images

10 11/26/0810 OUTLINE Purpose for Automatic Detection -- Space Weather Forecast Purpose for Automatic Detection -- Space Weather Forecast Automatic Solar Activities Detection Based on Images from HSOS Automatic Solar Flare Detection Automatic Solar Flare Detection Conclusions Conclusions

11 11/26/0811 AUTOMATIC SOLAR FLARE DETECTION A solar flare is an intense, abrupt release which occurs in areas on the Sun where the magnetic field is changing due to flux emergence or sunspot motion. A solar flare is an intense, abrupt release which occurs in areas on the Sun where the magnetic field is changing due to flux emergence or sunspot motion. Methods for automatic flare detection: a combination of region-based and edge- based segmentation methods, neural network technique, RBF, SVM, etc. Methods for automatic flare detection: a combination of region-based and edge- based segmentation methods, neural network technique, RBF, SVM, etc. Feature analysis and preprocessing. Feature analysis and preprocessing.

12 11/26/0812 NEURAL NETWORKS A method for the automatic detection of solar flares from Hα images using the multi-layer perceptron(MLP) with back-propagation training rule. A method for the automatic detection of solar flares from Hα images using the multi-layer perceptron(MLP) with back-propagation training rule.

13 11/26/0813 RADIAL BASIS FUNCTION(RBF) (1)‏

14 11/26/0814 RADIAL BASIS FUNCTION(RBF) (2)‏ After computing the optimal weights, the RBF network can be used as a classifier to segment the test data into the corresponding classes, with - 1 indicating a non-flare state and 1 indicating a flare state. After computing the optimal weights, the RBF network can be used as a classifier to segment the test data into the corresponding classes, with - 1 indicating a non-flare state and 1 indicating a flare state.

15 11/26/0815 Support Vector Machine(SVM)‏

16 11/26/0816 Feature Analysis and Preprocessing QU et al. (2003) used nine features for solar flare detection. QU et al. (2003) used nine features for solar flare detection. Feature 1: mean brightness of the frame. Feature 1: mean brightness of the frame. Feature 2: standard deviation of brightness. Feature 2: standard deviation of brightness. Feature 3: variation of mean brightness between consecutive images. Feature 3: variation of mean brightness between consecutive images. Feature 4: absolute brightness of a key pixel. Feature 4: absolute brightness of a key pixel. Feature 5: radial positon of the key pixel. Feature 5: radial positon of the key pixel. Feature 6: contrast between the key pixel and the minimum value of ite neighbors in a 7 by 7 window. Feature 6: contrast between the key pixel and the minimum value of ite neighbors in a 7 by 7 window. Feature 7: mean brightness of a 50 by 50 window, whe the key pixel is on the center. Feature 7: mean brightness of a 50 by 50 window, whe the key pixel is on the center. Feature 8: standard deviation of the pixels in the aforementioned 50 by 50 window. Feature 8: standard deviation of the pixels in the aforementioned 50 by 50 window. Feature 9: difference of the mean brightness of the 50 by 50 window between the current and the previous images. Feature 9: difference of the mean brightness of the 50 by 50 window between the current and the previous images.

17 11/26/0817 Three steps in the experiments of solar flare detection: (a) Preprocessing to obtain the nine features of solar flares. (a) Preprocessing to obtain the nine features of solar flares. (b) MLP, RBF and SVM training and testing program used for solar flare detection (b) MLP, RBF and SVM training and testing program used for solar flare detection (c) Region growing and edge detection methods for obtaining the flare properties. (c) Region growing and edge detection methods for obtaining the flare properties.

18 11/26/0818 CLASSIFICATION PERFORMANCE Through experiments, SVM is found to be the best for the solar-flare detection because it offers the best classification result and the training and testing speed are relatively fast. The second choise is RBF. MLP is not a well-controlled learning machine. Through experiments, SVM is found to be the best for the solar-flare detection because it offers the best classification result and the training and testing speed are relatively fast. The second choise is RBF. MLP is not a well-controlled learning machine.

19 11/26/0819 SOLAR FLARE DETECTION Automatic procedure to detect and characterize flares.

20 11/26/0820 OUTLINE Purpose for Automatic Detection -- Space Weather Forecast Purpose for Automatic Detection -- Space Weather Forecast Automatic Solar Activities Detection Based on Images from HSOS Automatic Solar Flare Detection Automatic Solar Flare Detection Conclusions Conclusions

21 11/26/0821 Conclutsions Despite the recent advances in solar imaging, machine learning and data mining have not been widely applied to solar data. Despite the recent advances in solar imaging, machine learning and data mining have not been widely applied to solar data. It is necessary to select well-performanced and appropriate algorithms to the study in the solar domain. It is necessary to select well-performanced and appropriate algorithms to the study in the solar domain.

22 11/26/0822 Thank you!


Download ppt "11/26/081 AUTOMATIC SOLAR ACTIVITY DETECTION BASED ON IMAGES FROM HSOS NAOC, HSOS YANG Xiao, LIN GangHua"

Similar presentations


Ads by Google