A new tool for the remote sensing of groundwater tables: satellite images of pastoral wells

There have been many approaches to the extraction of roads. Even though the complete automatic interpretation of aerial or satellite images is still remote, it is possible to obtain sound results from some images under some conditions. In this work we will show the importance of texture and second order statistics in the recognition of roads from satellite and aerial images. Since this type of images is in general registered, the images can be combine with other information from a GIS. In this work vector layers for roads networks are used in combination with raster aerial or satellite images. Several results with high-resolution satellite and aerial images are presented. Shadows and other obstacles caused some mistakes and they present a problem that remains to be tackled. Despite all this, the importance of texture for the extraction of roads is proven. Future work toward a complete automation introducing new information layers from a GIS is also discussed.

A semi-automatic road extraction method from high-resolution (1-m) satellite images is presented. As IKONOS, a high-resolution (1-m) satellite has been launched and several companies have plans to launch high-resolution satellites, extraction of man-made objects from high-resolution satellite images has been main concern of many scientists. The method consists of three phases; 1) NUBS (Non Uniform B-Spline) curve is formed by given seed points. 2) A road candidate area is made by straightening image along the NUBS curve. 3) Finally, road is extracted by a tracking algorithm which uses adaptive least squares correlation match method and linearity. Due to straightening image, the tracking algorithm extracts roads accurately even though there are road gaps, and the size of a matrix for least squares correlation match can be reduced. We test our method on high-resolution (1-m) satellite (IKONOS) image. The test result reveals our method is robust and can be one of the feasible solutions of mapping from high-resolution (1-m) satellite images.

Satellite images can be analyzed and used for a variety of purposes. In the future, satellite image analysis will become more important since the number of satellites launches, and the amount of satellite data increase every year. Under these circumstances, there are some problems to be solved. One is the existence of low-resolution satellite images. To analyze the lower resolution of satellite images there are some technical issues such as reduction of noise, misclassification of object recognition. Therefore, high-resolution images are necessary. However, high-resolution satellite images are expensive, and its images may not be available in the past satellite images. Super-resolution which is introduced in image processing is a method to solve these problems. Convolutional neural network (CNN)-based methods are effective, and there is a need for models that can perform super-resolution with higher accuracy. In this paper, we propose a method for super-resolving satellite images, based on the improved the RCAN (residual channel attention network) model with SRM (style-based recalibration module). The proposed method improves the PSNR (peak signal to noise ratio) by 0.0181 dB compared to the conventional RCAN model.

This dissertation investigates robust signal processing and machine learning techniques, with the objective of improving the robustness of two applications against various threats, namely Global Navigation Satellite System (GNSS) based positioning and satellite imaging. GNSS technology is widely used in different fields, such as autonomous navigation, asset tracking, or smartphone positioning, while the satellite imaging plays a central role in monitoring, detecting and estimating the intensity of key natural phenomena, such as flooding prediction and earthquake detection. Considering the use of both GNSS positioning and satellite imaging in critical and safety-of-life applications, it is necessary to protect those two technologies from either intentional or unintentional threats. In the real world, the common threats to GNSS technology include multipath propagation and intentional/unintentional interferences. This thesis investigates methods to mitigate the influence of such sources of error, with the final objective of designing secure and resilient GNSS receivers for robust GNSS positioning. Generally speaking, there are two kinds of positioning technology in GNSS receiver: two-step (2SP) and Direct Position Estimation (DPE) methods. The 2SP method is the conventional positioning method, where 3 blocks are typically required: acquisition block, tracking block and navigation solution block. To improve robustness against common threats mentioned above, this thesis built robust signal processing methods in each block: 1) the use of robust statistics to design enhanced correlation schemes that are resilient to jamming interferences in acquisition and tracking blocks, named as Robust Interference Mitigation (RIM); 2) the use of Deep Learning to learn signal correlation in tracking block in multipath-rich environments, where propagation channel models are too complex to be of practical use; and 3) the design of a variational-based Robust Kalman Filter that can mitigate outliers and provide position, velocity, and time (PVT) estimates in challenging environments. The DPE method is a new positioning technique, aiming to acquire PVT solution directly instead of using Doppler-shift and time-delay of satellite signal as intermediate variables. However, given that the input to DPE is still the received satellite signal, the influence of the three kinds of aforementioned threats is still relevant. This thesis therefore investigates the use of robust statistics to mitigate jamming signals in DPE receivers, similarly as done for the 2SP method earlier. To evaluate the loss of efficiency of using RIM in DPE schemes, the Cramér-Rao Bound (CRB) is considered, showing a bounded loss. Regarding the application to satellite imaging technology, this thesis investigates two problems in multispectral and hyperspectral imaging, respectively. In terms of multispectral imaging, the low frequency of high-resolution spatial image occurrence is a common drawback that many researchers aimed at solving. This thesis proposes a Bayesian filtering method to fuse multi-temporal and multi-spectral images (from high-resolution/low-availability and low-resolution/high-availability images) for high-resolution spatial estimates to extend the availability of high-resolution imaging. The novelty of this method is to consider historical high spatial resolution images and, to address the computational cost increase due to high-dimensional state-space model resulting from multi-spectral images. To cope with the high-dimensionality, the state space is partitioned into several low dimensional subspaces, with one filter or smoother dealing with the low-dimensional subspaces and interacting among them. The parallel processing is also introduced to speed up the whole processing under the assumption that all pixels in the coarse spatial resolution images are independent. As a result of this work, the fusing method improved robustness against low temporal frequency of high-resolution spatial images, while maintaining an affordable computational complexity. As a use case, this thesis considered Land Remote-Sensing Satellite (Landsat) images as high-resolution spatial images and Moderate Resolution Imaging Spectroradiometer (MODIS) images as coarse-resolution images, with the objective of water and flood monitoring. In terms of hyperspectral imaging, this thesis focuses on improving hyperspectral unmixing method. In hyperspectral unmixing, the most common model due to its tractability is the linear mixing model, however, it might fail in modeling the nonlinear interactions between different materials. This drawback is addressed in this thesis by considering an Auto-encoder (AEC) based model. This thesis proposed an AEC based method, imposing a particular structure to both the encoder and the decoder networks in order to account for the physical properties of the problem, that is the fact that the encoder should invert the mixing process. Overall, this thesis leverages and develops new tools in robust statistics and Bayesian filtering to effectively and efficiently deal with model mismatches and, ultimately, provide improved data fusion methodologies that are applicable to a variety of problems, such as GNSS or remote sensing.--Author's abstract

Detecting buildings from very high resolution aerial and satellite images is very important for mapping, urban planning, and land use analysis. Although it is possible to manually locate buildings from very high resolution images; this operation may not be robust and fast. Therefore, automated systems to detect buildings from very high resolution aerial and satellite images are needed. Unfortunately, the solution is not straightforward due to the diverse characteristics and uncontrolled imaging conditions of the scenes. To overcome these difficulties, herein we propose a novel solution to detect buildings from very high resolution grayscale aerial and panchromatic Ikonos satellite images using structural features and probability theory. For this purpose, we extract structural features from the given test image using a steerable filter set. Extracted structural features indicate geometrical properties of objects in the image. Using them, we estimate the probability density function (pdf) which indicates locations of buildings to be detected. Our extensive tests on a large and diverse data set indicate the robustness and practical usefulness of our method.

Abstract. High resolution satellite images are widely used to produce and update a digital map since they became widely available. It is well known that the accuracy of digital map produced from satellite images is decided largely by the accuracy of geometric modelling. However digital maps are made by a series of photogrammetric workflow. Therefore the accuracy of digital maps are also affected by the quality of satellite images, such as image interpretability. For satellite images, parameters such as Modulation Transfer Function(MTF), Signal to Noise Ratio(SNR) and Ground Sampling Distance(GSD) are used to present images quality. Our previous research stressed that such quality parameters may not represent the quality of image products such as digital maps and that parameters for image interpretability such as Ground Resolved Distance(GRD) and National Imagery Interpretability Rating Scale(NIIRS) need to be considered. In this study, we analyzed the effects of the image quality on accuracy of digital maps produced by satellite images. QuickBird, IKONOS and KOMPSAT-2 imagery were used to analyze as they have similar GSDs. We measured various image quality parameters mentioned above from these images. Then we produced digital maps from the images using a digital photogrammetric workstation. We analyzed the accuracy of the digital maps in terms of their location accuracy and their level of details. Then we compared the correlation between various image quality parameters and the accuracy of digital maps. The results of this study showed that GRD and NIIRS were more critical for map production then GSD, MTF or SNR.

Semi-automatic mapping of linear-trending bedforms using ‘Self-Organizing Maps’ algorithm

With the growing number of high-resolution satellite images, the traditional image retrieval method has become a bottleneck in the massive application of high-resolution satellite images because of the low degree of automation. However, there are few studies on the automation of satellite image retrieval. This paper presents an automatic high-resolution satellite image accurate retrieval method based on effective coverage (EC) information, which is used to replace the artificial screening stage in traditional satellite image retrieval tasks. In this method, first, we use a convolutional neural network to extract the EC of each satellite image; then, we use an effective coverage grid set (ECGS) to represent the ECs of all satellite images in the library; finally, the satellite image accurate retrieval algorithm is proposed to complete the process of screening images. The performance evaluation of the method is implemented in three regions: Wuhan, Yanling, and Tangjiashan Lake. The large number of experiments shows that our proposed method can automatically retrieve high-resolution satellite images and significantly improve efficiency.

The detection of the road is one of an area of satellite image classification. The satellite image classification plays a vital role in various area of monitoring different resources available on the earth surface. Here, the high-resolution satellite data from Google earth is acquired from a different region of Mumbai, Maharashtra, India region for detection of road. This research paper used two different algorithms i.e. radial basis function neural network and Naive Bayes classifiers for the detection of reading features from the high-resolution satellite image. Both algorithms are implemented using the Matlab simulation toolbox. Radial Basis Function and Naive Bayes is a supervised classification technique applied on High-Resolution Satellite Image. Extraction of Road from the satellite image is a very difficult task because in the rural areas there are many unstructured roads which may consist of mud and concrete. After applying the algorithms on the image high-resolution satellite, the accuracy of classifiers is calculated using confusion matrix and Kappa coefficient. The accuracy of Naive Bayes found to be 91% with Kappa Value 0.698 and the accuracy of radial basis function found to be 99% with a Kappa value of 0.9831. The accuracy calculation using confusion matrix and Kappa value shows that the radial basis function neural network classifier is better than Naive Bayes classifiers for the detection of the road using high-resolution satellite image.

This study focused on characterising aquifer systems based on water-level changes observed systematically at 159 paired groundwater monitoring wells throughout Korea. Using spectral analysis, principal component analysis (PCA), and cross-correlation analysis with linear regression, aquifer conditions were identified from the comparison of water-level changes in shallow alluvial and deep bedrock monitoring wells. The spectral analysis could identify the aquifer conditions (i.e., unconfined, semi-confined and confined) of 58.5% of bedrock wells and 42.8% of alluvial wells: 93 and 68 wells out of 159 wells, respectively. Even among the bedrock wells, 50 wells (53.7%) exhibited characteristics of the unconfined condition, implying significant vulnerability of the aquifer to contaminants from the land surface and shallow depths. It appears to be better approach for deep bedrock aquifers than shallow alluvial aquifers. However, significant portions of the water-level changes remained unclear for categorising aquifer conditions due to disturbances in data continuity. For different aquifer conditions, PCA could show typical pattern and factor scores of principal components. Principal component 1 due to wet-and-dry seasonal changes and water-level response time was dominant covering about 55% of total variances of each aquifer conditions, implying the usefulness of supplementary method of aquifer characterisation. Cross-correlation and time-lag analysis in the water-level responses to precipitations clearly show how the water levels in shallow and deep wells correspond in time scale. No significant differences in time-lags was found between shallow and deep wells. However, clear time-lags were found to be increasing from unconfined to confined conditions: from 1.47 to 2.75 days and from 1.78 to 2.75 days for both shallow alluvial and deep bedrock wells, respectively. In combination of various statistical methods, three types of water-level fluctuation patterns were identified from the water-level pairs: Type I of identical aquifer systems (77.8%), Type II of the different aquifer systems with different recharge flow paths (9.5%), and Type III of unmatched aquifer system pairs and correlations (12.7%). Type I and II could be used as verification of aquifer condition in the paired monitoring system. However, Type III shows the complexity of water-level fluctuation in different aquifer conditions. This study showed that confined or not-confined conditions are not directly related to the depth of wells in the aquifer. Therefore, the utilisation of groundwater as a water-supply source should be carefully designed, tested for its hydrogeologic conditions, and managed to ensure sustainable quantity and quality.

The application of remote sensing monitoring techniques plays a crucial role in evaluating and governing the vast amount of ecological construction projects in China. However, extracting information of ecological engineering target through high-resolution satellite image is arduous due to the unique topography and complicated spatial pattern on the Loess Plateau of China. As a result, enhancing classification accuracy is a huge challenge to high-resolution image processing techniques. Image processing techniques have a definitive effect on image properties and the selection of different parameters may change the final classification accuracy during post-classification processing. The common method of eliminating noise and smoothing image is majority filtering. However, the filter function may modify the original classified image and the final accuracy. The aim of this study is to develop an efficient and accurate post-processing technique for acquiring information of soil and water conservation engineering, on the Loess Plateau of China, using SPOT image with 2.5 m resolution. We argue that it is vital to optimize satellite image filtering parameters for special areas and purposes, which focus on monitoring ecological construction projects. We want to know how image filtering influences final classified results and which filtering kernel is optimum. The study design used a series of window sizes to filter the original classified image, and then assess the accuracy of each output map and image quality. We measured the relationship between filtering window size and classification accuracy, and optimized the post-processing techniques of SPOT5 satellite images. We conclude that (1) smoothing with the majority filter is sensitive to the information accuracy of soil and water conservation engineering, and (2) for SPOT5 2.5 m image, the 5×5 pixel majority filter is most suitable kernel for extracting information of ecological construction sites in the Loess Plateau of China.

Choung, Y.-J. and Jung, D. 2021. Comparison of machine and deep learning methods for mapping sea farms using high-resolution satellite image. In: Lee, J.L.; Suh, K.-S.; Lee, B.; Shin, S., and Lee, J. (eds.), Crisis and Integrated Management for Coastal and Marine Safety. Journal of Coastal Research, Special Issue No. 114, pp. 420–423. Coconut Creek (Florida), ISSN 0749–0208. Previous research had shown that the supervised machine learning approach performed better than unsupervised machine learning for mapping sea farms using a high-resolution satellite image. The present work compares a support vector machine (SVM), which represents the supervised machine learning approach, and a deep neural network (DNN), which represents the deep learning approach, for mapping sea farms using KOMPSAT-3 satellite images acquired in the South Sea of South Korea. First, coastal maps were generated from the image source given by SVM and DNN. Next, the above-water and underwater farms were detected separately from both the maps based on the minimum and maximum thresholds. Finally, the detection accuracy of both the above-water and underwater farms from both coastal maps was assessed. Statistical results showed that deep learning (DNN) provided better performance than machine learning (SVM) for detecting above-water farms from the given high-resolution satellite image, while both DNN and SVM yielded the same performance for underwater farms. However, a few errors occurred in the detection because of the limitations of the pixel-based classification approaches. In future research, the deep learning algorithm combined with object-based classification, such as the convolutional neural network, can be used to detect sea farms from the given high-resolution image more accurately.

Automated 3D reconstruction based on satellite images has become a research hotspot at the interdisciplinary of photogrammetry and computer vision. The 3D results based on satellite images will play a key role in the understanding of global 3D information, monitoring of national geographic and urban construction, with the inherent advantage of satellite images in global coverage. Researchers have devoted substantial effort to develop state-of-the-art 3D reconstruction methods for two-view satellite images and multi-view satellite images. However, it is still a challenging task to obtain complete and accurate 3D results with satellite images due to the difference in shooting angles between satellite images, exposure differences and building occlusions in urban scenes. In this paper, we execute theoretical analyses and experimental evaluations about the popular 3D reconstruction methods towards satellite images following the order of two views to multiple views: (1) The advanced dense matching methods aimed at satellite images are reviewed theoretically and evaluated experimentally. (2) The state-of-the-art 3D reconstruction based on two-view satellite images are analysed in detail and experimentally evaluated with two-view WorldView-3 satellite images. (3) The popular fusion methods of multi-view DSM are analysed theoretically and assessed on multi-view WorldView-3 satellite images. This review will be helpful for researchers dedicated to enhancing the accuracy and completeness of the results of 3D reconstruction from urban satellite images.

A fully automatic and high-precision cartographic mapping of terrain features such as forests, rivers or roads from multispectral satellite or aerial images is a challenging problem in remote sensing, largely due to the fact that it requires an adequate representation of irregular and discontinuous objects. Being able to provide sub-meter resolution multispectral images, high-resolution satellites such as QuickBird or Ikonos broaden the application possibilities of satellite imagery and offer the possibility of making them sensors of choice for a variety of environmental applications. Their high spatial and radiometric resolution facilitates visual interpretation. Temporal resolution of image databases can be largely increased, due to the known revisit time and pointing capabilities of the satellite platform, which facilitates large-scale change-detection and monitoring of selected areas in order to keep natural environment databases up-to-date. Certain analyses involving spectral change detection and dynamically obtained maps can be performed more easily and in a more automated fashion. Finally, high-resolution satellite images offer the potential for being ever more competitive in terms of price with the aerial images. This chapter presents methods based on Self-Organizing Maps (Kohonen et al., 1996) developed in efforts to fully automate the generation of hydrographic maps from remotely sensed imagery. The complexities of generating cartographic representations of hydrological objects, such as rivers and lakes, from satellite and aerial images consists generally of two categories of tasks: the first involves the extraction process of the required linear or a real feature while the second involves generation of a suitable representation in a form appropriate for cartographic map presentation. The presented approach applies the technology of Self-Organizing Maps (SOM) at both stages of the hydrological mapping process, i.e., the detection of water bodies from multispectral images and the subsequent tracing of hydrological systems or networks. The first task can be approached by applying a classification technique or through a scene analysis method. A number of different methods have been reported in the literature. Conventional image processing techniques typically apply edge detection algorithms (Ma & Manjunath, 2000) in efforts to define water regions. A similar problem of road detection from satellite images was discussed in (Auclair Fortier et al., 2001). A rule-based approach to segmentation of satellite images was presented in (Ton et al., 1991). Selection of

Seasonal landslide mapping and estimation of landslide mobilization rates using aerial and satellite images