Peifeng Su
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Peifeng Su has a total of 19 co-authors. Here are the top co-authors by number of shared publications:
Petri Pellikka
Department of Geosciences and Geography, University of Helsinki, Helsinki, Finland
4 shared publications- Jorma Joutsensaari
Department of Technical Physics, University of Eastern Finland, Kuopio, Finland
2 shared publications - Lubna Dada
PSI Center for Energy and Environmental Sciences, Paul Scherrer Institute, Villigen, Switzerland
2 shared publications - Markku Kulmala
Aerosol and Haze Laboratory, Beijing Advanced Innovation Center for Soft Matter Science and Engineering, Beijing University of Chemical Technology, Beijing, China
2 shared publications - Martha Arbayani Zaidan
University of Helsinki
2 shared publications Sasu Tarkoma
Department of Computer Science, Faculty of Science, University of Helsinki, 00014 Helsinki, Finland
2 shared publications
Latest Publications
- Open Access
- 13 March 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 16 (19391404) , 3353-3363
Image-to-Image Training for Spatially Seamless Air Temperature Estimation With Satellite Images and Station Data
Air temperature at approximately 2 m above the ground ( $T_{a}$ ) is one of the most important environmental and biophysical parameters to study various earth surface processes. $T_{a}$ measured from meteorological stations is inadequate to study its spatio-temporal patterns since the stations are unevenly and sparsely distributed. Satellite-derived land surface temperature (LST) provides global coverage, and is generally utilized to estimate $T_{a}$ due to the close relationship between LST and $T_{a}$ . However, LST products are sensitive to cloud contamination, resulting in missing values in LST and leading to the estimated $T_{a}$ being spatially incomplete. To solve the missing data problem, we propose a deep learning method to estimate spatially seamless $T_{a}$ from LST that contains missing values. Experimental results on 5-year data of mainland China illustrate that the image-to-image training strategy alleviates the missing data problem and fills the gaps in LST implicitly. Plus, the strong linear relationships between observed daily mean $T_{a}$ ( $T_{mean}$ ), daily minimum $T_{a}$ ( $T_{min}$ ), and daily maximum $T_{a}$ ( $T_{max}$ ) make the estimation of $T_{mean}$ , $T_{min}$ , and $T_{max}$ simultaneously possible. For mainland China, the proposed method achieves results with $R^{2}$ of 0.962, 0.953, 0.944, mean absolute error (MAE) of 1.793, 2.143, and 2.125 $^\circ$ C, and root mean square error (RMSE) of 2.376, 2.808, and 2.823 $^\circ$ C for $T_{mean}$ , $T_{min}$ , and $T_{max}$ , respectively. Our study provides a new paradigm for estimating spatially seamless ground-level parameters from satellite products. Code and more results are available at https://github.com/cvvsu/LSTa
- 13 March 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 16 (19391404) , 3353-3363
Open AccessImage-to-Image Training for Spatially Seamless Air Temperature Estimation With Satellite Images and Station Data
Air temperature at approximately 2 m above the ground ( $T_{a}$ ) is one of the most important environmental and biophysical parameters to study various earth surface processes. $T_{a}$ measured from meteorological stations is inadequate to study its spatio-temporal patterns since the stations are unevenly and sparsely distributed. Satellite-derived land surface temperature (LST) provides global coverage, and is generally utilized to estimate $T_{a}$ due to the close relationship between LST and $T_{a}$ . However, LST products are sensitive to cloud contamination, resulting in missing values in LST and leading to the estimated $T_{a}$ being spatially incomplete. To solve the missing data problem, we propose a deep learning method to estimate spatially seamless $T_{a}$ from LST that contains missing values. Experimental results on 5-year data of mainland China illustrate that the image-to-image training strategy alleviates the missing data problem and fills the gaps in LST implicitly. Plus, the strong linear relationships between observed daily mean $T_{a}$ ( $T_{mean}$ ), daily minimum $T_{a}$ ( $T_{min}$ ), and daily maximum $T_{a}$ ( $T_{max}$ ) make the estimation of $T_{mean}$ , $T_{min}$ , and $T_{max}$ simultaneously possible. For mainland China, the proposed method achieves results with $R^{2}$ of 0.962, 0.953, 0.944, mean absolute error (MAE) of 1.793, 2.143, and 2.125 $^\circ$ C, and root mean square error (RMSE) of 2.376, 2.808, and 2.823 $^\circ$ C for $T_{mean}$ , $T_{min}$ , and $T_{max}$ , respectively. Our study provides a new paradigm for estimating spatially seamless ground-level parameters from satellite products. Code and more results are available at https://github.com/cvvsu/LSTa
- 4 February 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Geoscience and Remote Sensing Letters
- Vol. 19 (1545598X) , 1-5
Retrieval of Multiple Atmospheric Environmental Parameters From Images With Deep Learning
Retrieving atmospheric environmental parameters such as atmospheric horizontal visibility and mass concentration of aerosol particles with a diameter of 2.5 or 10 ${\mu }\text{m}$ or less (PM 2.5 , PM 10 , respectively) from digital images provides new tools for horizontal environmental monitoring. In this study, we propose a new end-to-end convolutional neural network (CNN) for the retrieval of multiple atmospheric environmental parameters (RMEPs) from images. In contrast to other retrieval models, RMEP can retrieve a suite of atmospheric environmental parameters including atmospheric horizontal visibility, relative humidity (RH), ambient temperature, PM 2.5 , and PM 10 simultaneously from a single image. Experimental results demonstrate that: 1) it is possible to simultaneously retrieve multiple atmospheric environmental parameters; 2) spatial and spectral resolutions of images are not the key factors for the retrieval on the horizontal scale; and 3) RMEP achieves the best overall retrieval performance compared with several classic CNNs such as AlexNet, ResNet-50, and DenseNet-121, and the results are based on experiments on images extracted from webcams located in different continents (test $R^{2}$ values are 0.63, 0.72, and 0.82 for atmospheric horizontal visibility, RH, and ambient temperature, respectively). Experimental results show the potential of utilizing webcams to help monitor the environment. Code and more results are available at https://github.com/cvvsu/RMEP .
- 4 February 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Geoscience and Remote Sensing Letters
- Vol. 19 (1545598X) , 1-5
Retrieval of Multiple Atmospheric Environmental Parameters From Images With Deep Learning
Retrieving atmospheric environmental parameters such as atmospheric horizontal visibility and mass concentration of aerosol particles with a diameter of 2.5 or 10 ${\mu }\text{m}$ or less (PM 2.5 , PM 10 , respectively) from digital images provides new tools for horizontal environmental monitoring. In this study, we propose a new end-to-end convolutional neural network (CNN) for the retrieval of multiple atmospheric environmental parameters (RMEPs) from images. In contrast to other retrieval models, RMEP can retrieve a suite of atmospheric environmental parameters including atmospheric horizontal visibility, relative humidity (RH), ambient temperature, PM 2.5 , and PM 10 simultaneously from a single image. Experimental results demonstrate that: 1) it is possible to simultaneously retrieve multiple atmospheric environmental parameters; 2) spatial and spectral resolutions of images are not the key factors for the retrieval on the horizontal scale; and 3) RMEP achieves the best overall retrieval performance compared with several classic CNNs such as AlexNet, ResNet-50, and DenseNet-121, and the results are based on experiments on images extracted from webcams located in different continents (test $R^{2}$ values are 0.63, 0.72, and 0.82 for atmospheric horizontal visibility, RH, and ambient temperature, respectively). Experimental results show the potential of utilizing webcams to help monitor the environment. Code and more results are available at https://github.com/cvvsu/RMEP .
- Open Access
- 25 January 2022
- Published by Copernicus GmbH in Atmospheric Chemistry and Physics
- Vol. 22 (2) , 1293-1309
New particle formation event detection with Mask R-CNN
Atmospheric new particle formation (NPF) is an important source of climate-relevant aerosol particles which has been observed at many locations globally. To study this phenomenon, the first step is to identify whether an NPF event occurs or not on a given day. In practice, NPF event identification is performed visually by classifying the NPF event or non-event days from the particle number size distribution surface plots. Unfortunately, this day-by-day visual classification is time-consuming and labor-intensive, and the identification process renders subjective results. To detect NPF events automatically, we regard the visual signature (banana shape) which has been observed all over the world in NPF surface plots as a special kind of object, and a deep learning model called Mask R-CNN is applied to localize the spatial layouts of NPF events in their surface plots. Utilizing only 358 human-annotated masks on data from the Station for Measuring Ecosystem–Atmosphere Relations (SMEAR) II station (Hyytiälä, Finland), the Mask R-CNN model was successfully generalized for three SMEAR stations in Finland and the San Pietro Capofiume (SPC) station in Italy. In addition to the detection of NPF events (especially the strongest events), the presented method can determine the growth rates, start times, and end times for NPF events automatically. The automatically determined growth rates agree with the manually determined growth rates. The statistical results validate the potential of applying the proposed method to different sites, which will improve the automatic level for NPF event detection and analysis. Furthermore, the proposed automatic NPF event analysis method can minimize subjectivity compared with human-made analysis, especially when long-term data series are analyzed and statistical comparisons between different sites are needed for event characteristics such as the start and end times, thereby saving time and effort for scientists studying NPF events.
- 25 January 2022
- Published by Copernicus GmbH in Atmospheric Chemistry and Physics
- Vol. 22 (2) , 1293-1309
Open AccessNew particle formation event detection with Mask R-CNN
Atmospheric new particle formation (NPF) is an important source of climate-relevant aerosol particles which has been observed at many locations globally. To study this phenomenon, the first step is to identify whether an NPF event occurs or not on a given day. In practice, NPF event identification is performed visually by classifying the NPF event or non-event days from the particle number size distribution surface plots. Unfortunately, this day-by-day visual classification is time-consuming and labor-intensive, and the identification process renders subjective results. To detect NPF events automatically, we regard the visual signature (banana shape) which has been observed all over the world in NPF surface plots as a special kind of object, and a deep learning model called Mask R-CNN is applied to localize the spatial layouts of NPF events in their surface plots. Utilizing only 358 human-annotated masks on data from the Station for Measuring Ecosystem–Atmosphere Relations (SMEAR) II station (Hyytiälä, Finland), the Mask R-CNN model was successfully generalized for three SMEAR stations in Finland and the San Pietro Capofiume (SPC) station in Italy. In addition to the detection of NPF events (especially the strongest events), the presented method can determine the growth rates, start times, and end times for NPF events automatically. The automatically determined growth rates agree with the manually determined growth rates. The statistical results validate the potential of applying the proposed method to different sites, which will improve the automatic level for NPF event detection and analysis. Furthermore, the proposed automatic NPF event analysis method can minimize subjectivity compared with human-made analysis, especially when long-term data series are analyzed and statistical comparisons between different sites are needed for event characteristics such as the start and end times, thereby saving time and effort for scientists studying NPF events.
- Open Access
- 13 September 2021
- Published by Copernicus GmbH in EGUsphere
- Vol. 2021, 1-21
New Particle Formation Events Detection with Deep Learning
Atmospheric new particle formation (NPF) is an important source of climate-relevant aerosol particles which has been observed at many locations globally. To study this phenomenon, the first step is to identify whether an NPF event occurs or not on a given day. In practice, NPF event identification is performed visually by classifying the NPF event or non-event days from the particle number size distribution surface plots. Unfortunately, this day-by-day visual classification is time-consuming, labor-intensive, and the identification process renders subjective results. To detect NPF events automatically, we regard the visual signature (banana shape) which has been observed all over the world in NPF surface plots as a special kind of object, and a deep learning model called Mask R-CNN is applied to localize the spatial layouts of NPF events in their surface plots. Utilizing only 358 human-annotated masks on data from the Station for Measuring Ecosystem and Atmospheric Relations (SMEAR) II station (Hyytiälä, Finland), the Mask R-CNN model was successfully generalized for three SMEAR stations in Finland and the San Pietro Capofiume (SPC) station in Italy. In addition to the detection of NPF events (especially the strongest events), the presented method can determine the growth rates, start times, and end times for NPF events automatically. The automatically determined growth rates agree with the growth rates determined by the maximum concentration and mode fitting methods. The statistical results valid the potential of applying the proposed method on different sites, which will improve the automatic level for NPF events detection and analysis. Furthermore, the proposed automatic NPF event analysis method provides more consistent results compared with human-made analysis, especially when long-term data series are analyzed and statistically comparisons between different sites are needed for event characteristics such as the start and end times, thereby saving time and effort of scientists studying NPF events.
- 13 September 2021
- Published by Copernicus GmbH in EGUsphere
- Vol. 2021, 1-21
Open AccessNew Particle Formation Events Detection with Deep Learning
Atmospheric new particle formation (NPF) is an important source of climate-relevant aerosol particles which has been observed at many locations globally. To study this phenomenon, the first step is to identify whether an NPF event occurs or not on a given day. In practice, NPF event identification is performed visually by classifying the NPF event or non-event days from the particle number size distribution surface plots. Unfortunately, this day-by-day visual classification is time-consuming, labor-intensive, and the identification process renders subjective results. To detect NPF events automatically, we regard the visual signature (banana shape) which has been observed all over the world in NPF surface plots as a special kind of object, and a deep learning model called Mask R-CNN is applied to localize the spatial layouts of NPF events in their surface plots. Utilizing only 358 human-annotated masks on data from the Station for Measuring Ecosystem and Atmospheric Relations (SMEAR) II station (Hyytiälä, Finland), the Mask R-CNN model was successfully generalized for three SMEAR stations in Finland and the San Pietro Capofiume (SPC) station in Italy. In addition to the detection of NPF events (especially the strongest events), the presented method can determine the growth rates, start times, and end times for NPF events automatically. The automatically determined growth rates agree with the growth rates determined by the maximum concentration and mode fitting methods. The statistical results valid the potential of applying the proposed method on different sites, which will improve the automatic level for NPF events detection and analysis. Furthermore, the proposed automatic NPF event analysis method provides more consistent results compared with human-made analysis, especially when long-term data series are analyzed and statistically comparisons between different sites are needed for event characteristics such as the start and end times, thereby saving time and effort of scientists studying NPF events.
- Open Access
- 12 October 2020
- Published by MDPI AG in Remote Sensing
- Vol. 12 (20) , 3319
Band Ranking via Extended Coefficient of Variation for Hyperspectral Band Selection
Hundreds of narrow bands over a continuous spectral range make hyperspectral imagery rich in information about objects, while at the same time causing the neighboring bands to be highly correlated. Band selection is a technique that provides clear physical-meaning results for hyperspectral dimensional reduction, alleviating the difficulty for transferring and processing hyperspectral images caused by a property of hyperspectral images: large data volumes. In this study, a simple and efficient band ranking via extended coefficient of variation (BRECV) is proposed for unsupervised hyperspectral band selection. The naive idea of the BRECV algorithm is to select bands with relatively smaller means and lager standard deviations compared to their adjacent bands. To make this simple idea into an algorithm, and inspired by coefficient of variation (CV), we constructed an extended CV matrix for every three adjacent bands to study the changes of means and standard deviations, and accordingly propose a criterion to allocate values to each band for ranking. A derived unsupervised band selection based on the same idea while using entropy is also presented. Though the underlying idea is quite simple, and both cluster and optimization methods are not used, the BRECV method acquires qualitatively the same level of classification accuracy, compared with some state-of-the-art band selection methods
- 12 October 2020
- Published by MDPI AG in Remote Sensing
- Vol. 12 (20) , 3319
Open AccessBand Ranking via Extended Coefficient of Variation for Hyperspectral Band Selection
Hundreds of narrow bands over a continuous spectral range make hyperspectral imagery rich in information about objects, while at the same time causing the neighboring bands to be highly correlated. Band selection is a technique that provides clear physical-meaning results for hyperspectral dimensional reduction, alleviating the difficulty for transferring and processing hyperspectral images caused by a property of hyperspectral images: large data volumes. In this study, a simple and efficient band ranking via extended coefficient of variation (BRECV) is proposed for unsupervised hyperspectral band selection. The naive idea of the BRECV algorithm is to select bands with relatively smaller means and lager standard deviations compared to their adjacent bands. To make this simple idea into an algorithm, and inspired by coefficient of variation (CV), we constructed an extended CV matrix for every three adjacent bands to study the changes of means and standard deviations, and accordingly propose a criterion to allocate values to each band for ranking. A derived unsupervised band selection based on the same idea while using entropy is also presented. Though the underlying idea is quite simple, and both cluster and optimization methods are not used, the BRECV method acquires qualitatively the same level of classification accuracy, compared with some state-of-the-art band selection methods