最新刊期
卷 24 ， 期 9 ， 2020
-
摘要：Launched on September 15, 2019, the Interferometric Imaging Radar Altimeter (InIRA) onboard the Chinese Tiangong-2 space laboratory is the first spaceborne interferometric radar altimeter that can obtain wide-swath ocean topography measurements by adopting small incidence angles from 1° to 8° with a short baseline. InIRA achieves various technological breakthroughs, meanwhile, it also brings some challenges in data processing because no radiometer is onboard the Tiangong-2 space laboratory. Considering signal path delays is a premise for InIRA to meet its geocoding and Sea Surface Height (SSH) measurement goals, a mathematical model-based method for tropospheric path delay correction should be developed. Unlike traditional nadir-looking altimeters, which only require the propagation delay related to the velocity variation along a line path, the Tiangong-2 InIRA must consider the additional bending of radio waves for its small incidence angles. In this study, a tropospheric slant path delay correction algorithm is developed using the ray-tracing technique based on Fermat’s principle and on the numeric weather model from the European Center for Medium-range Weather Forecasts. Two calibration campaigns are conducted in March 2017 and September 2018, which recorded 12 measurement data from 9 corner reflectors. Results show that the standard deviation of the residual error range after the tropospheric slant path delay correction is approximately 6.2 cm, which indicates that a centimeter-level range accuracy is realized. Therefore, the effectiveness and reliability of the proposed algorithm in different small incidence angles are validated.关键词：Tiangong-2 Interferometric Imaging Radar Altimeter (InIRA);atmospheric path delay;slant path delay correction;Numeric Weather Model (NWM);bending of radio waves- 637
- |
- 180
- |
- 2
发布时间：2020-08-28 -
摘要：Oceanic eddy, which is an important subject in oceanic scientific research, plays an important role in the ocean thermal cycle. Because oceanic eddy is a kind of rotating three-dimensional water body, the sea surface height anomaly caused by it is very significant for eddy detection. At present, the satellite altimeter is the main sensor to detect the height anomaly of eddy. However, the satellite altimeter can only obtain the one-dimensional sea surface height information of eddy along the satellite track, and the resolution of the data fusion product is also rough. It is difficult for an altimeter to map mesoscale or sub-mesoscale (15—300 km) oceanic processes, since a 200—300 km gap usually exists between two successive tracks. Therefore, an altimeter is normally used to study large-scale (>300 km) oceanic processes, the sub-mesoscale or small-scale eddies cannot be detected using conventional radar altimeters. Interferometric Synthetic Aperture Radar (InSAR) not only has very high spatial resolution, but also can measure height based on interferometric phase, so it has great potential in eddy height anomaly detection. However, the height measurement accuracy of the existing InSAR system is generally in the meter level, which cannot meet the centimeter precision needed for eddy detection. The imaging altimeter based on SAR interferometry is expected to be the next generation of satellite altimeter. The Interferometric Imaging Radar Altimeter (InIRA) on Tiangong-2 is the first spaceborne Ku-band interferometric Synthetic Aperture Radar (SAR) that is specially designed for ocean surface topography altimetry. InIRA provides a large number of images for the observation and investigation of sub-mesoscale or small-scale eddies. This altimeter not only acquires the SAR image of oceanic eddies, but also obtains the interferometric phase of the complex images of two antennas and identifies the ocean surface height anomaly of eddies, thereby providing a new possibility for eddy detection. In order to prove the ability of InIRA in ocean eddy detection, this paper carries out the research of oceanic eddy detection based on InIRA interferometric data. An eddy detection method based on InIRA complex image is proposed by calculating the relative height anomaly of the ocean surface. The interferometric data of InIRA are processed to detect eddies after a series of procedures, including image coregistration, flat-earth phase removal, system parameter calibration, and phase noise suppression. Results show that the interferometric phase change corresponds to the ocean surface height anomaly induced by the eddy, and the relative height anomaly of the eddy is 23 cm. The moderate resolution imaging spectroradiometer Sea Surface Temperature (SST) and chlorophyll-a (CHL) data are used to verify the eddy identified from the InIRA images. The core of the eddy and the spiral arms emanated from the core are about 2.5°—3.0° cooler than the surrounding water. The CHL concentration in the core of the eddy and the spiral arms is about 1.0—1.5 mg·m-3 higher than the adjacent water. And the center of the eddy with lowest SST corresponds to the highest CHL concentration. The findings indicate that the proposed method can realize the detection of oceanic eddy and preliminarily prove the ability of Tiangong-2 InIRA in detecting eddy, which reflects the great potential and application value of Tiangong-2 InIRA data in the investigation of sub-mesoscale oceanic dynamic environment.关键词：oceanic eddies;Tiangong-2;near-nadir cross-track interferometric SAR;eddy detection- 636
- |
- 182
- |
- 1
发布时间：2020-08-28 -
摘要：Remote sensing image scene classification is an important means of remote sensing image interpretation, which has important application value in land and resources investigation, ecological environment monitoring, disaster assessment, target interpretation and so on. In recent years, deep learning has become a research hotspot in the field of remote sensing scene classification, and data set is the basis for its development. Most of the existing remote sensing scene classification datasets are true color images with single scale and less spectral information. And other hyperspectral data sets have relatively small data coverage. The data of Tiangong-1 Hyperspectral Imager has the characteristics of high spatial resolution, high hyperspectral resolution and wide coverage. It can be used for comprehensive feature extraction and analysis of spectral spatial information, which can provide more abundant data sources for remote sensing image classification application research, and make up for the deficiency of spectral information and limited application of common remote sensing scene classification data sets. In this study, based on the high-quality data acquired by Tiangong-1 Hyperspectral Imager, Tiangong-1 Hyperspectral Remote Sensing Scene Classification data set (TG1HRSSC) is produced through radiation correction, geometric correction, spatial clipping, band screening, and data quality analysis and control. The dataset, which contains the 204 hyperspectral multiresolution image data of nine typical scenes (e.g., city, farmland, forest, pond culture, desert, lake, river and airport), is published and shared in the Space Application Data Promoting Service Platform for China Manned Space Engineering (http://www.msadc.cn [2019-09-10]). The dataset includes one band of 5 m resolution full spectrum, 54 bands of 10 m resolution visible and near-infrared spectrum, and 52 bands of 20 m resolution short-wave infrared spectrum. In addition, this paper describes and analyzes the data set from four aspects: scene distribution, time distribution, spectral distribution and scale distribution. In order to test the application effect of data classification, three classical convolution neural networks (VGG-VD-16, AlexNet and GoogleLeNet) are selected to train the data sets by transfer learning. The overall classification accuracy is 91.52 ± 0.60, 90.47 ± 0.23 and 89.12 ± 0.34, respectively. Results show that the scene classification of the dataset is effective. In following research, the network model can be designed to make full use of the multi-spectral characteristics of the data to achieve more accurate scene classification, and to improve the generalization ability of existing models by using the characteristics of multi-scale data. The data set (TG1HRSSC) has the advantages of hyperspectral, high spatial resolution and multi-scale. The abundant spectral information and fine spatial information provide data support for the research of target recognition of fine ground objects, remote sensing scene classification, remote sensing semantic understanding and other applications, which has unique value and application prospects.关键词：Tiangong-1;Hyperspectral Imager;scene classification;data set;deep learning- 1390
- |
- 879
- |
- 9
发布时间：2020-08-28 -
摘要：Port cities play a vital role in the implementation of the Belt and Road initiative. Remote sensing data can provide effective information support for the dynamic monitoring of port cities. In this paper, 58 groups of high-quality multispectral image data obtained by the wide-band imaging spectrometer of Tiangong-2 at different times are selected to construct the thematic dataset of 25 port cities along the Belt and Road. This dataset can provide important support for the remote sensing monitoring in resources and environment of the port cities and its surrounding area along the Belt and Road.Firstly, The raw data of wide-band imaging spectrometer undergo through a series of processing steps, such as format analyzing, field of view combination, homogeneity correction, image framing, absolute radiometric correction, geometric correction and format packaging, to generate standard data products and its auxiliary files. Then, image quality evaluation indexes, including signal to noise ratio, information entropy, clarity, contrast, radiation uniformity, etc. are adopted to assess data quality and to select high quality data products. Calibration and data quality control methods are used to improve the accuracy of data products. At last, several typical convolution neural network algorithms are used to verify the application potential of the data set in scene classification and recognition.350 typical port and non-port image samples are extracted from the data set in this article and be marked manually. After data enhancement, 250 samples of them are used to train several typical convolutional neural networks, they are AlexNet, VGG16, ResNet18 and MobileNet-v2, and the remaining samples are used for verification. The average recognition accuracy of the four algorithms is 91%, in which Resnet18 and MobileNet-v2 networks have the highest accuracy rate (93%), Resnet18 network has more advantages than other networks in precision rate, VGG16 and MobileNet-v2 network have the highest recall rate. Therefore, the dataset is suitable for the common convolutional neural networks and has good application effect in the scene classification and recognition of port city.The multispectral remote sensing image data set constructed in this paper comes from the wide-band imaging spectrometer of Tiangong-2. It has the characteristics of wide spectral range, high spatial resolution and diverse observation time, and is a beneficial extension for the existing data set of the port city along the Belt and Road. After strictly data processing and data quality control, the image data has standard format and reliable quality. Some of them has been applied in many fields, such as ocean, land, ecology and environment. In addition, it has a good application effect in the scene classification and recognition of port city. This data set is a valuable data resource for the remote sensing applications of the port city and its surrounding areas along the Belt and Road.关键词：Tiangong-2;Wide-band Imaging Spectrometer;the Belt and Road;port city;remote sensing image data set- 668
- |
- 157
- |
- 0
发布时间：2020-08-28
-
摘要：With the rapid development of remote sensing satellite imaging technology, remote sensing satellite video provides a new way to acquire moving vehicle target information, and it has become a new data source of vehicle information for intelligent transportation systems. However, in satellite video images, the vehicle is only a few to a dozen pixels and has less contrast with the background. Obtaining the vehicle’s local detail features is difficult. Many problems will arise if the traditional vehicle detection method in monitoring videos is directly applied on satellite videos. Thus, a method that can efficiently exploit and utilize the latest satellite video datasets is urgently needed.On the basis of an analysis of the difference between moving target detection of remote sensing satellite video and traditional monitoring video, a method of moving vehicle detection for remote sensing satellite video automatically constrained by the region of interest was proposed. First, part of the video data is predetected by using the interframe difference method. Then, all the detection results are superimposed together. Morphological processing was perform to obtain the Region Of Interest (ROI) of moving vehicles. Second, moving vehicles were detected based on the improved Gaussian background difference method under the constraint of ROI.Skybox-1 satellite video data were used to qualitatively and quantitatively analyze the accuracy and efficiency of moving vehicle detection. Most of the vehicles were successfully detected and marked out, thereby indicating that the method can be used to detect vehicles in satellite video data. The detection accuracy of our method is more than 93% in all cases, thus indicating that our method has an extremely low false alarm rate. The detection rate is between 70% and 80%, which indicates that the method can accurately detect most of the vehicles in the satellite video data. In addition, the quality of the test is stable at more than 0.84. We can conclude that the method can ensure a high detection accuracy and an optimal detection rate; therefore, the quality of the method is excellent. In this paper, we take the automatic extraction of the moving area as a pretreatment step, which means, after users wait for a few seconds, the program will detect vehicles in the satellite video data set at a near-real-time rate. The method can efficiently exploit and utilize the latest satellite video datasets.The experimental results showed that the proposed method can effectively reduce the number of pseudo-moving targets caused by dynamic background changes, with a high detection rate, high detection quality, very low false alarm rate, and high operating efficiency. Therefore, the detection of moving vehicle targets in a satellite video can be realized effectively.关键词：remote sensing satellite video;Skybox-1;moving vehicles;constraint by region of interest;frame difference;background difference- 745
- |
- 201
- |
- 2
发布时间：2020-08-28 -
摘要：As an important element of farmland ecosystems, Crop Residues Cover (CRC) provides a barrier against water erosion and improves soil structure and organic matter content. Timely and accurate estimation of CRC at regional scale is essential for understanding the ecosystem condition and interactions with the surrounding environment. Satellite remote sensing is an effective method of regional CRC estimation. Tillage indices based on multi-spectral satellite imagery data are commonly used in CRC estimation. However, this method is ineffective in high coverage areas due to “saturation”. Previous studies have shown that a combination of image spectral and textural information can solve saturation problems to a certain extent. Sentinel-2 is a new satellite mission that can provide observations at multi-spectral bands with spatial resolutions of 10, 20, and 60 m. Sentinel-2 can provide more information about texture compared with the commonly used multi-spectral satellite Landsat-8 Operational Land Imager. Therefore, exploring the potential of combining spectral and textural information from Sentinel-2 data is an important task in CRC estimation.The objectives of this study are to (1) analyze correlation between field measured CRC and satellite-derived variables such as Sentinel-2 band reflectance, tillage indices, and gray-level co-occurrence matrix statistics in different windows, and (2) determine the optimal CRC estimation method from optimal subset regression with various combinations of tillage indices and image textural features.The results showed that the Normalized Difference Tillage Index (NDTI), B12_CO (contrast of band12, B12 in window 5×5), and B12_DI (dissimilarity of B12 in window 5×5) were significantly correlated with the measured CRC with correlation coefficient R values of 0.765, –0.641, –0.553. The estimation model based on NDTI outperformed the models based on other single variables. The model that combined the spectral and textural information in an optimal window (R=0.869， RMSE=11.0%) provided a more precise result than that based solely on spectral information (R=0.775 and RMSE=14.5%). The results demonstrated that a combination of spectral and textural information can improve the accuracy of CRC estimation.关键词：Sentinel-2;crop residue coverage;crop residue indices;gray-level co-occurrence matrix;texture window;optimal subset regression method;Landsat OLI- 807
- |
- 183
- |
- 8
发布时间：2020-08-28 -
摘要：Many characteristics of remote sensing images, such as large scale, complex illumination and occlusion, dense, multiple scales, various posture targets, and the lack of a large number of labeled images for training depth networks, pose great challenges to the integrity and accuracy of remote sensing image segmentation. In deep convolutional networks for segmentation, resolution is significantly reduced by multiple pooling, thereby reducing the prediction accuracy of pixel class.On the basis of the deep convolutional coding-decoding network, an end-to-end remote sensing image segmentation model with full residual connection and multiscale feature fusion is proposed in this paper. First, the features in the encoder are merged into the corresponding layers of the decoder, and the residual unit is added to the corresponding convolution layer. The full residual connection constructed by the operation enables the model as a whole to effectively enhance feature fusion and be easier to train. Second, the feature pyramid module, which aggregates multiscale context information, is used on the high-level feature map of the fifth stage of the encoder before feature fusion, thus enabling the model to effectively deal with multiscale changes of the target and improve the segmentation performance.Experiments were conducted on the ISPRS Vaihingen and Road Detection datasets. The proposed model was evaluated from the two aspects of average IOU and average F1-score. A comparison between the current advanced models and the results in the literature shows that the proposed model is better than the comparison models. The average IOU on the two datasets is 85% and 84%, and the average F1 value is 92% and 93%, respectively.An end-to-end remote sensing image segmentation model with full residual connection and multiscale feature fusion is proposed in this paper. The proposed model achieves better results than the current advanced image semantic segmentation model on the two datasets. The segmentation targets are more complete, continuous, and have fewer misclassifications and leakages. The proposed model also achieves better results than the comparative model in road segmentation of remote sensing images from different sources, thereby further verifying the robustness of the model.关键词：remote sensing image segmentation;deep convolutional neural network- 775
- |
- 210
- |
- 5
发布时间：2020-08-28
Technology and Methodology
-
摘要：Building extraction plays a significant role in land use analysis like urban planning. Classical methods based on hand-crafted features fail to derive prominent building extraction results due to the limited representation capacity of the hand-crafted features. In this paper, we achieve building extraction in pixel level based on a deep Convolutional Neural Network (CNN) with an encoder-decoder structure. In contrast to the hand-crafted features that require professional knowledge and have a poor representation capacity, convolutional neural networks are equipped with a high representation capacity and able to learn highly abstract and distinguishing features from data. The encoder is used to derived a space compressed representation of the input raw image. This compressed representation is also called a feature of the input image and it is assumed to be abstract and distinguishing. The decoder uses the feature as input and recover the space resolution to the size of the input image. Thereby, the encoder-decoder network achieves pixel-wise building extraction in an end-to-end way from the raw image to the building extraction result.Applying the encoder-decoder network to building extraction will cause a Marginal Phenomenon (MP). Specifically, the prediction accuracy near the edges of a patch is usually lower than that near the central area. Marginal phenomenon will lead to the reduction of building extraction accuracy. To alleviate this effect, we propose the usage of the Field of View Enhancement (FoVE) method. The FoVE method includes two parts: enlarging the patch size and cropping patches with overlaps when making predictions. Therefore, the FoVE method contains two hyper-parameters, which are patch size and overlapping size. Extensive experiments on two building extraction datasets are conducted to analyze the impact of the two hyper-parameters through the Precision-Recall Curves (PRC) and some interesting conclusions are derived from the the analysis: (1) Enlarging the input patch size when making prediction can effectively improve the building extraction performance while the improvement saturates as the overlapping size increases; (2) Cropping patches with an overlap when making prediction can improve the building extraction performance while the improvement saturates as the input patch size increases; (3) The FoVE can effectively improve building extraction accuracy but this improvement from the FoVE has a limit; (4) The convolutional neural network for building extraction plays the key role and further attentions should be focused on the network design.In addition to the numerical analysis of the FoVE experimental results, we attempt to explain why FoVE works and why it has a limit. We blame them on the Field of View (FoV) and that is reason why the method is call FoVE. FoV plays an important role in building extraction and a larger FoV is beneficial to building extraction. Firstly, the marginal phenomenon is caused by the lack of context information of the marginal pixels. FoVE improves the overall accuracy through abandoning the unreliable predictions of the marginal pixels. Secondly, enlarging input patches can enlarge the FoV of each pixel and thus improves the accuracy. Thirdly, the the improvement from FoVE has a limit because that when the field of view is large enough, the improvement derived from more contextual information can be ignore.关键词：remote sensing;building extraction;convolutional neural network;deep learning;aerial imagery- 828
- |
- 359
- |
- 15
发布时间：2020-08-28 -
摘要：Target vibration generates special phase modulation of Synthetic Aperture Radar (SAR) echo signals called micro-Doppler effect, which can provide favorable information for micro-motion target detection and recognition. Mining such favorable information is of great significance to improve the performance of SAR systems. At present, research on SAR vibration target detection is still insufficient. The existing detection algorithms have many problems, such as large computational complexity, weak anti-clutter and anti-noise ability, and inability to adapt to multiple vibration targets.To solve SAR real-time detection of multiple vibration targets under a strong clutter and noise background, this paper proposes a novel vibration target detection algorithm. The detection algorithm uses the Displaced Phase Center Antenna (DPCA) cancellation technique to suppress ground clutter and accumulates DPCA signal along the azimuth direction to improve the anti-noise ability. The frequency spectrum of the SAR azimuth echo of a vibrating target has a high similarity with pulse sequence. Therefore, the pulse repetition interval transform (PRI transform) method of detecting repeated pulse sequences is chosen in this paper to realize the detection of vibration targets. The detection algorithm is a two-step process. The first step is to find the range cell positions of the vibration targets (called the aim range cells). This step mainly uses DPCA technology to eliminate clutter and accumulates the DPCA signal along the azimuth direction. Then, the vibration target range cells are determined by setting an appropriate threshold. The second step is to determine the number of vibration targets in the aim range cells obtained in the first step and estimate their vibration frequencies. This step mainly performs pseudo-pulse processing on the azimuth spectrum of the target range cell and then performs PRI transformation on the pseudo-pulse train to detect the vibration targets and estimate the vibration frequencies. The algorithm can detect the vibration targets under strong clutter and noise conditions, and it has high frequency estimation precision and a small amount of calculation. Even when multiple vibration targets are present in a single range cell, the algorithm is still applicable.Simulation results prove the correctness and high efficiency of the proposed algorithm. Results show that under the condition of weak Signal-to-Noise Ratio (SNR), the first step of the detection can determine the vibration target range cells well, and when the SNR>-40 dB, the detection probability of the aim range cell is higher than 95%. In the second step, all the vibration targets in each aim range cell are detected successfully, and the vibration frequency of each vibration target is estimated accurately. Compared with the GLRT algorithm and the Hough transform algorithm, the proposed algorithm has the advantages of small computational complexity and high detection speed. The total computation time of two-step detection does not exceed 0.6 s, which shows that the proposed algorithm is suitable for real-time detection. The quasidata detection results prove that detecting vibration targets in the actual scene is feasible.The algorithm outperforms previous ones in that it involves a fairly small amount of computation and exhibits better anti-clutter and anti-noise performance. Hence, this algorithm has high practical value for remote sensing vibration targets.关键词：remote sensing;Synthetic Aperture Radar (SAR);vibration target;SFM signal;DPCA;PRI transformation;CFAR- 638
- |
- 250
- |
- 2
发布时间：2020-08-28
Doctor's Voice
- Postal code：100190
- Tel：010-58887052 Email：nrsb@aircas.ac.cn
- Technical support is provided by Beijing Founder electronics co., LTD 京ICP备20021838号-9
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
0