In recent years, with the rapid development of earth observation technologies, remote sensing using the satellites has gradually entered the era of big data. Facing the current demands and characteristics of remote sensing applications, it is feasible and necessary to explore the theories and methods of high-spatial-resolution remote sensing cognition with the cooperation of visual cognition. In this context, we are inspired by Geo-informatic-Tupu and intend to study the spatial-spectral cognition of remote sensing. This paper systematically presents the theory and calculation methodology for the spatial-spectral cognition of remote sensing, and expects to standardize the processes of remote sensing information extraction, as a result to further build a sophisticated, quantitative, intelligent and integrated model for the remote sensing information interpretation. The whole methodology contains two directions' cognitive calculation, namely horizontal "bottom-up hierarchical abstraction" and longitudinal "top-down knowledge transfer". These two steps are corresponded with three principal Spatial-Spectral transformation processes, which are summarized as "extracting spatial maps based on clustering pixels' spectrum", "coordinating spatial-spectral features" and "understanding attributes through the recognition of known diagram". Our study focuses on the analysis of the involved concepts, the basic idea, the key technologies and their existing difficulties, and emphasizes on the utilization of big data and gradually the application of integrated knowledge to achieve different levels of remote sensing cognition. Through these approaches, we expect to provide a new perspective for the remote sensing interpretation with the adoption of big data resources.
In the era of big data, the rapid growth of geographic spatial temporal data has challenged the conventional application concepts, technical framework and service modes. In this paper, the concept and features of geographic spatial temporal big data is elaborated firstly. Then, the characteristics and challenges of the geographic spatial temporal big data computation are analyzed. Particularly, the theory of collaborative computing and service for the geographic spatial temporal big data is developed, which includes four levels of collaboration: data collaboration, technology collaboration, service collaboration and producing collaboration. According to the demand of the market-oriented operation and platform-based service, the technical frameworks of the geographic spatial temporal big data collaborative computing are designed. Furthermore, four common key technologies are discussed, including the remote sensing data preprocessing, the geographic spatial temporal data storage and management, the high performance computing and the visualization of geographic spatial temporal big data. Next, the remote sensing data processing system is developed, and is taken as a case to illustrate the implementation of collaborative computing and service of geographic spatial temporal big data. At last, this paper forecasts the future application mode of geographic spatial temporal big data.
Cloud and its shadow have certain degrees of impacts on the information extraction from remote sensing images. As the multi-source remote sensing data has become increasingly abundant in recent years, the cross application of the multi-source and multi-temporal remote sensing image for restoring the cloud and its shadow region, and for effectively acquiring the change information for the ground objects is an important content in studying the application of remote sensing big data. The precise detection of cloud and its shadow information is the premise and guarantee of their restoration. In general, the cloud and cloud shadow detection methods always use their spectral or spatial shape and the texture characteristics as references. However, regarding the complex and changeable spectrum and the inexpressible spatial shape characteristics, the cloud and cloud shadow information have always been difficult to be effectively detected. Based on the analysis of the spectral characteristics of thick clouds, thin clouds, snow and ice, and other feature types, a cloud and cloud shadow multi-feature collaborative detection method was proposed. (1) First of all, the cloud detection is executed. The proposed method extracts the standard thick cloud spectrum curve from the reflectance-calibrated image. Afterwards, the SAM method is used to match the cloud spectral curve from the distinguishable (red, shortwave infrared, thermal infrared) bands combination, with the absolute value of the shortwave infrared band pixel integrated to distinguish between cloud and snow, and the absolute value of the thermal infrared band pixel used to distinguish between cloud and other types of ground objects. (2) Next, the cloud shade detection is performed. Firstly, we expand the detected cloud pixel border, and produce a potential shadow mask area. Afterwards, we move the potential shadow mask along the direction of sun radiation to some distance. Thirdly, we detect the cloud shadow pixels using the brightness threshold of the near infrared band within the moved potential cloud shadow mask area. After several moves of the potential shadow mask and the implementations of cloud shadow detection based on the infrared band brightness threshold, eventually a complete cloud shadow mask is produced. The LANDSAT-8 image having the above mentioned bands is adopted in an experiment and the experimental result shows that the combination of spectral curve, "diagnosis" band and spatial relationship features can effectively detects the thin clouds, thick clouds and cloud shadows from the multispectral remote sensing image, and the overall accuracy is higher than 95%.
In order to meet the demand for quantitative information extraction from multi-sources and multi-temporal satellite data, a semi-automatically relative radiometric normalization approach was developed in this paper. The relative radiometric normalization procedure of multi-sources image is divided into two parts: the first one is sensors’ radiometric correction and the other is the normalization of radiometric discrepancy caused by external factors such as the changes of illumination. Firstly, the radiometric correction coefficients of sensors were obtained by the classes-specified regression method based on clear-sky images. Secondly, by applying the sample transferring method, the multi-sources images were semi-automatically classified, and the radiometric deviations of the corresponding sensors were adjusted. Lastly, based on the images’ classification results, the PIFs were automatically chosen by combining the NDVI difference histogram with the class restraint, and the relative radiometric normalization of images was thereby accomplished. A pair of quasi-synchronous GF1-WVF1 and Landsat 8 images, and a set of multi-sources and time-series images were adopted to demonstrate the validity of the presented approach. The results showed that, this method can effectively correct sensors’ radiometric discrepancy and achieved a higher radiometric normalization accuracy when compared with the traditional method.Meanwhile, the results from time-series dataset also demonstrated that this method can effectively reduce the fluctuation of radiation features between time-series images, and the seasonal changes of vegetation were therefore accurately presented. It can thereby offer a reference to the synergic utilization of multi-sources and time-series images.
Due to the impacts of phonological change, sensor distortion, variation of atmospheric conditions and a lot of other factors, the images acquired at different time points for the same area are affected by color differences, Color normalization tries to eliminate /reduce color differences between different images and obtain seamless mosaic result. However, the traditional band-by-band normalization methods ignore the correlation between different bands and corrected each band independently, which may lead to new color distortion. To solve this problem, this paper presents a compound cluster center based multiple linear regression color normalization method for remote sensing image. Firstly, the source image and the reference images are primarily normalized based on the mean and variation values for every band and a new feature vector is constructed. Then, the compound clusters, which are extracted by unsupervised compound classification, are used to model the variation relationship between the two images. Afterwards,the outliers in every cluster may induce suddencolor change between the images of different time, so the outliers is identified and excluded. Last, the mapping relationship between the source image and the reference image is established with respect to the centers of clusters andall bands of source image are corrected simultaneously. The proposed method has been applied to two datasets with different land cover and spatial resolution, and results show that the proposed method can obtain color consistency result. Compared with the result of traditional method, our method over performsin preserve color and overall precision.
Object-Based Image Analysis (OBIA) is becoming an important technology for the information extraction from high spatial resolution images. Multi-scale image segmentation is a key and fundamental procedure of OBIA, however, the scale selection within the multi-scale image segmentation is always difficult to achieve for the high-performance OBIA. This paper firstly generalizes the commonly used segmentation scale parameters into three aspects: the spatial parameter (the spatial distance between classes), the attribute parameter (the attribute distance or spectral difference between classes) and the merging threshold (the area or pixel number of the minimum useful object). Next, this paper proposes a spatial and spectral statistics-based scale parameter estimation method for OBIA. The main concept of this proposed method is to use the average local variogram (without considering the anisotropism of spatial distribution) or the semivariogram (considering the anisotropism of spatial distribution) to pre-estimate the optimal spatial parameter. Next, the selection of the optimal attribute parameter and the selection of the merging threshold are achieved based on the local variance histogram and the simple geometric computation, respectively. Taking the mean-shift segmentation as an example, this study uses Ikonos, Quickbird and aerial panchromatic images as the experimental data to verify the validity of the proposed scale parameter estimation method. Experiments based on the quantitative multi-scale segmentation evaluation could testify the validity of this method. This pre-estimation based scale parameter selection method is practically helpful and efficient in OBIA. The idea of this method can be further extended to be integrated into other segmentation algorithms and be adaptive to other sensor data.
The quality of image segmentation has a great impact on the results of information extraction from high spatial resolution remote sensing imagery when the object-based method is employed. During the segmentation of high spatial resolution remote sensing images, the scale parameter directly affects the construction of segmented image objects. A small scale is likely to produce broken image objects, while a large scale probably results in the mixed image objects. To solve this problem, an image segmentation framework based on a set of optimal scales is proposed in this paper. First of all, the high spatial resolution remote sensing image is processed using multi-scale segmentation methods with respect to a group of regularly distributed scales. Then the relationship between the global standard deviation of a single segmented layer and its corresponding scale is determined, from which a group of optimal scales are selected. Since the object in a layer that is segmented by a big scale parameter contains the corresponding object in a layer that is segmented by a small scale parameter, a hierarchical tree with nodes of multi-scale image objects can be created. Within this hierarchical tree, the image object of the layer that is segmented by the maximum scale is set as the root. In this manner, each image object of the layer that is segmented by the maximum scale can generate a hierarchical tree, which all together forms the image forest. Two types of features are considered when the optimal image object is selected from each hierarchical tree, which are the comprehensive evaluation index and the spectral features. The comprehensive evaluation index keeps a balance between the homogeneity and heterogeneity of the image objects. And the spectral features of the children nodes should be consistent with the parent nodes in order to dismiss the mixed image objects. Finally, the segmented result is created after the optimal image objects from all hierarchical trees are selected. In the experiment presented in this paper, the Geoeye and ZY3 images are adopted. Results show that the proposed method can effectively improve the percentage of properly segmented image objects.
Geographic Object-Based Image Analysis (GEOBIA) is widely used in high spatial resolution remote sensing image interpretation. However, the fundamental component of image segmentation severely obstructs its development, especially on the selection of segmentation parameters. To overcome these issues, we choose the widely used Fractal Network Evolution Algorithm as an example, which is provided by eCognition, and focus on looking for an approach for the scale parameter selection. Inspired by the similarity between the merging segmentation and the degrading image resolution that the unit size within both would increase, we introduce the weights of border length and the weights of object area into the metric of Local Variance proposed by Woodcock and Strahler (1987), and propose a new segmentation evaluation metric: Weighted Local Variance (WLV). Through comparing WLV with a supervised metric on a series of segmentations with limited increasing scale parameters, we found that the best segmentation result chosen by the first local maximum point of the scale-WLV curve is similar to the manual segmentation result. Then we validate WLV on two more images and expand the limited scale space to the full range, so that the segments can change from one pixel to the whole image. Results show that the segmentations chosen by WLV local maximum points could reflect the different levels in the hierarchical landscape, and the segmentation of the first levels is capable of expressing the finest homogeneous patches.
The obtaining of remotely sensed imagery may affected by design of satellite, sensors, and atmospheric conditions.Normally it is difficult to balance the temporal and spatial resolution. Which leads to the hard of information extraction from a single source remote sensing data.Obviously this may impossible to meet the application demand which ask for higher and higher resolution information on spatial and temporal. Considering the different advantages of multi-source data and the accumulation of the multi-temporal data by time,we design the classification method based on the patches for multi-source data. The patches are basic geographical units which are relatively stable on boundary and properties. With these patches, other data may reflect the spectrum performance at different time or different point of view. After calculating these features we can interpret the patches with adapt methods based on the characteristics of each land class.In the land use classification experiment of Maduo in summer of 2014, many data are collecting for cooperating classification. Long term middle resolution data cover the whole vegetation growing season and cloudless high resolution data in winter are used after solving the problem of geo-matching and multi-source compositing. Because of different advantages of these data, categories like built up, water, vegetation are interpreted separately. At last we get a high total accuracy. Not only effectively overcome the traditional perspective of insufficient data source, lack of information and complete the interpretation of the county.But also ensure the spatial and temporal resolution of land information.
Remote sensing images with high-spatial-resolution can provide a great amount of detailed information for land resource monitoring, which makes the change detection using the high-spatial-resolution remote sensing images become the focus of current remote sensing research. This article proposes a remote sensing change detection method through combining with the spatial-spectral features under the guidance of historical interpretation knowledge. Firstly, the objects of consistent spatial position are constructed by segmenting the combined image from different phases. After that, the objects’ spectral and texture features are extracted. Then, we introduce the previous land cover maps to guide the DS evidence fusion for the similarities of two different types of features. The archived class-type label in each polygon from the land cover maps is used to direct the similarity fusion with different evidence confidences. Finally, a binarization method based on the Gaussian Mixture Model is adopted to extract the change regions. Experimental results show that this method can take advantage of the historical interpretation knowledge to guide the fusion of different object-feature-similarities; moreover, to a certain extent, to effectively improve the accuracy of change detection.
Data visualization is an important service in remote sensing applications. To address the problems that it is difficult for the static pre-built map tile service to meet the requirements of professional data view, map configuration, spatial analysis and other applications, this paper presented a solution architecture for the real-time rendering and interactive visualization of remote sensing big data. Firstly, on the rendering nodes, a rendering-tile structure for image was constructed to improve the reading speed of remote sensing images. Secondly, on the visualization servers, a data-computing load balancing strategy was proposed to optimize the rendering efficiency of map tiles. Thirdly, a set of service interfaces for the interactive visualization was designed for the front ends of services. Compared with the traditional technology, this solution can not only achieve the real-time rendering and the interactive visualization of remote sensing data, but also obtain an equivalent service performance to the pre-built tile map service. Finally, based on the above solutions, an interactive visualization prototype system of remote sensing data was developed and was applied into the demonstrations of the real-time viewing of remote sensing images, the visualized computing and the visualized analysis.
Remotely sensed thematic mapping in large area is a hot and difficult topic in recent remote sensing mapping research. It is also a key factor that restricts the development of remote sensing application. The continental or global thematic mapping usually uses hundreds or thousands of scene images to cover the whole region, which causes the occurrence of overlaps in the coverage and induces time inconsistencies. How to eliminate the information redundancy between the adjacent images is becoming a crucial issue to the regional automatic thematic mapping in large area. The mapping accuracy and efficiency are the major sticking points in the large area mapping applications. In this paper, a new scheme designed for block splitting thematic mapping based on the constrained seamline networks is proposed. Firstly, the large study area is split into several small regions based on the uniform grids of Universal Transverse Mercator (UTM) Projection. Secondly, taking each uniform grid as a unit, with the lake boundary being set as the constraint, the mosaic seamlines are generated with respect to the image mosaic principles. Thus, each Landsat image is represented by a separated seamline polygon, which insures the lake boundaries would not be split. Thirdly, we combine the seamline networks of all UTM grids based on the generated seamlines between two adjacent UTM grids, and resultantly the seamline network for the whole region is built. Finally, each seamline polygon in the seamline network is taken as the valid mapping area of Landsat image, and then the lakes within each Landsat data are preserved inside the valid mapping area. The final lake mapping result is generated by combing all the lake layers. This method is tested on the Landsat 8 images for Central Asia in 2013 to generate a lake area map for the Central Asia region. 479 Landsat images are used to cover the whole study area. Except for the Aral Sea, Ala Nur, Balkash Lake and IssyKul Lake, all the other lakes lie inside the corresponding valid mapping areas. It is proved that the proposed method can effectively split the redundant area between two Landsat images; meanwhile, the lakes are not split by seamlines, so as to keep the integrity and accuracy of lakes. Compared with the lake mapping results in 2010, the number of lakes increased in 2013, while the areas of lakes decreased. The main reason of this phenomenon is that lakes in the plain deserts, such as Aral Sea, are experiencing changes of shrinking, while in the alpine regions, a lot of newly generated small lakes are emerging. The proposed method has two advantages: (1) during the image preprocessing stage, each Landsat image is analyzed to get a valid mapping area, and the lake mapping step is performed within this valid mapping area. As a result, each Landsat mapping region is restricted, which involves less computational time and editing work; (2) the valid mapping region is determined based on the lake constraint conditions; therefore, the seamline cannot cross the lake boundaries, which keeps the lake mapping results to be unique and accurate. However, this paper mainly focuses on the scheme and strategy of automatic thematic mapping, within which the detailed technologies of data flows and the seamline algorithm is still simplistic, and the accuracies and efficiencies of lake mapping is not thoroughly described. In future, its technical details and the stability of the algorithm will be improved continuously, and the continental or global scale lake mapping applications will be further studied.
The ability to acquire the remote sensing data is increasing day by day, which directly causes the remote sensing data to become diverse and massive, and the issue that the massive amount of data is being non-affordable to store has become more and more prominent. On the other hand, due to the lack of an effective and efficient method of storage management, the data that theterminal application need is difficult to found in a timely manner, therefore, is stored but useless. This paper focuses on the storage and management problems of the massive, high through put and spatially structured remote sensing data and the basic land information products. We have presented a storage and management method which uses the big data structure and can integrate both the vector and raster data. Based on the MongoDB database, the prototype system is realized and we use the data of PB rangeto test it. Eventually, we have proved that this method meets the demand for the storage and management of the remote sensing vector-raster data in the era of big data. On the basis of the study results and prototype system, the following studies need to be further explored: (1) The organization and management methods for internal data of resources, especially the objective and timely management for the vector data; (2) Real-time interactive visualization methods for different data types and storage modes of resources, achieving dynamic extraction and rendering ability based on in the heterogeneous data model; (3) To construct large data computing architecture on the heterogeneous type storage mode, and to implement multimodal computing framework to meet the needs of the remote sensing applications require.
Baiyangdian wetland is one of the very few remaining wetlands in north China, which has the functions of improving the ecological environment, maintaining the conservation of biological diversity and so on. It has the very prominent significance to researches that are focusing on the wetland changes in this area, because these researches can provide an important information support under the assistance of remote sensing to investigate the change of landscape pattern and conduct the environmental analysis. Remote sensing big data has become a trend for the development of remote sensing technology. Based on the characteristics of remote sensing big data, this paper analyzes the big data extraction technology, which is a critical part in the remote sensing applications. In addition, starting with the calculation of remote sensing information, this paper summarizes various collaborative computing problems encountered during the process of remote sensing big data calculation. Considering the application requirements of wetland water body extraction and change analysis over a long time period in Baiyangdian, this paper proposes a method based on the collaborative computing technology to extract the water body in Baiyangdian. The method firstly computes the initial NDWI threshold in its histogram and finds the possible lakes within the region, and then it computes the suitable GNDWI for every lake to implement the precise lake extraction one by one. And at last, several typical types of collaborative computing problems are analyzed in this procedure. According to the water extraction result from our analysis on a period between 1973 and 2015, we study the historical water area changes of Baiyangdian, and the results show that the water area in this region has experienced a “decrease-increase-another decrease-another increase” changing pattern.
High-resolution remote sensing image has become a major source of information for the rapid assessment of earthquake disaster, and it also brings new challenges to the study and application of seismic disaster information extraction methods that are based on the remote sensing. The existing methods of seismic disaster information extraction that are based on the remote sensing technology have some defects, such as the high dependence on the visual interpretation experience of researchers and the low accuracy of extraction results produced from the high-resolution images. This paper provides a rapid extraction method of the high-resolution remote sensing seismic disaster information with the integration of target feature library. Via building the high-resolution remote sensing disaster target feature library, this method is capable to provide services for the accumulation and application of disaster features based on the high-resolution images, thus to meet the purpose of reducing the dependence on the experience of researchers and improving the automation level and efficiency of disaster information extraction from remote sensing images. Regarding to the method framework description, this paper introduces several key technologies in the progress of earthquake disaster rapid assessment, which includes building the target feature library and conducting the method of sample matching and automatic classification. This paper takes the earthquake prone region of Ludian, Yunnan as an example. The study area is in the central area of Longtoushan town, and the earthquake disaster rapid assessment experiment is supported by the high-resolution remote sensing target feature library. Comparing our experimental results with the field survey results, it shows that the accuracy of the experimental results can meet the service requirements of the rapid assessment. It also shows that the rapid extraction of seismic disaster based on high-resolution remote sensing disaster target feature library can effectively reduce the labor workload and strongly improve the automation level of services. Generally, this method has a positive significance to the disaster emergency response.
To reduce the missing of remotely sensed data in the spatio-temporal coverage of the cloudy/rainy region and to further meet the urgent need for crop planting information at farmland parcel scale, a method of crop type identification and planting area estimation at parcel scale was developed in this paper by synergistically utilizing the multi-sources satellite imagery,with the support of remote sensing Tupu recognization theory. This method consists of three steps: firstly, based on the high resolution imagery, the objects of farmland parcel with exact boundary were extracted. Secondly, with the effective-data processing technology and the spectral indices calculation based on the multi-temporal medium resolution imagery, the fragmentary effective data was acquired and the time-series data for each object was further obtained. Finally, by constructing a multi-dimensional feature space with the help of time series analysis incorporating the crops’phenological feature, the crop types and their corresponding planting areas were mapped using the Decision Tree classifier. This method had been tested in Ningyuan county, Hunan Province, China. The results showed that, this method can precisely map the different rice types and corresponding planting areas at the farmland parcel scale. The user accuracy of the three rice types, i.e., the early double-season, single-season and late double-season rice,was 94.33%, 90.76 and 95.95%, respectively, and the overall accuracy was 92.51% with a Kappa coefficient of 0.90. The derived area accuracy of these three rice types also reached 93.37%, 91.23% and 95.42%,respectively. This experiment illustrated the effectiveness and usefulness of the proposed method and also provided a salutary lesson for the finely planting information extraction of other crops.
