2016 Vol. 41, No. 4
Height gradient values change around their boundary pixels significantly since buildings in built-up areas are often higher than the surroundings. A Stereo Pair Disparity Index(SPDI) image is produced from the disparity image generated from epipolar images of a stereo pair. Built-up area results in an epipolar image are obtained by executing spatial clustering on pixels with high values in the SPDI image, and can be located in the raw image of a stereo pair. However, built-up area results that are separately detected from two images are inconsistent, and cannot be aligned because of the disparity between the homonymous points. To address this problem., the generated disparity images were aligned and overlayed to achieve reliable, consistent results. Experimental results show that built-up areas were identified from two stereo pairs accurately through our method.
Mars exploration is one of the most important major deep space exploration projects. Some problems, such as navigation accuracyandsignal real-time transmission, exist in radio navigation because of the communication latency and occlusion of celestial bodies. Autonomous optical navigation is a key technology for mars exploration and could act as an important auxiliary for radio navigation. This paper focuses on the scheme and key technologies for autonomous optical navigation in mars exploration durling the cruise and capture phase. Firstly, some concept and scheme of autonomous optical navigation have been discussed, in which the optical navigation measurement model has been introduced for defining the relationship between celestialandoptical navigation camera, and the navigation scheme for marsexploration in cruise and capture phase has been formulated too. Secondly, some key technologies focusing on navigation image processing, on-orbit calibration for optical navigation camera and optical autonomous navigation filter design have been fully presented. In further research, these key technologies could take the advantage of the theoretical basis of Deep-space Photogrammetry.
There are redundant complex issues concerning insertion and deletion processes in three-dimensional octree and R-tree index data management. Relying on GeoSOT Earth three-dimensional subdivision grids, we propose a new complex combination of the octree and R-tree indexes, the Subdivision R-tree model(Subdivision R-tree). First, GeoSOT three-dimensional subdivision octree-based grid index is used to construct a model Subdivision R-tree index structure. Subsequently, the basic design of the insertion, deletion, and query algorithm Subdivision R-tree index, is analyzed. Finally, we carry out a Subdivision R-tree indexing operation with the original data indexing performance comparison test, and discuss the threshold selection of Subdivision R-tree analysis accordingly. Test results show that the performance, especially Subdivision R-tree data update(insertionor deletion) process is better than octree. With the change of data distribution, the performance is more evident in the case that the data distribution is more concentrated, and the improvement is up to 20%.
The cartographic displacement algorithm based on the Beams model is a kind of global optimization algorithm that references the mechanics of materials. Using the model, the decay process of propagation in the displacement operation can be simulated, providing cartographically pleasing results with respect to the preservation of shape, spatial relations, and patterns of map object(s).However, the model lacks a detailed algorithm for implementation and a feasible method for setting the model's material parameters(i.e. elastic modulus, cross-sectional area, moment of intertia). Therefore, we focuses on the implementation and improvement of the algorithm. First, the basic mathematic model and solution method based on finite element method(FEM) are introduced. Second, from a point view of algorithm implementation, a detailed study of the key issues concerning the calculation and aggregation of the stiffness matrix and force vector are presented. Finally, to reduce the complexity of the parameters, we propose an adaptive parameter setting method to improve the algorithm. Supported by a constrained Delaunay triangulation(CDT), tests against a road network dataset and a building cluster dataset are carried out. The results illustrate that the improved algorithm is feasible and applicable to the displacement problems of linear object(s) and discrete polygon object clusters.
Based on a ZY-102C HR Image, three kinds of textures, the variogram texture, gray level co-occurrence matrix texture, and gradient texture, were extracted. Then, we present a new SVM classification method with multi-source data by integrating the spectral information and these three different textures. The classification result was compared with results using Maximum Likelihood Classification(MLC) and Decision Tree(DT) method. The study shows that:(1) Variogram texture and gradient texture involved in multi-source data can effectively improve image classification precision with an overall accuracy from 85.14% to 87.43% and Kappa coefficient from 0.82 to 0.85;(2) The variation function of absolute value form provides a theoretical basis for the optimal texture window analysis, and textural features based on the average step can significantly improve classification accuracy with an overall accuracy from 75.2% to 87.14% and Kappa coefficient from 0.7 to 0.87;(3) Based on multi-source data, the SVM classification method for high spatial resolution remote sensing images can effectively overcome the fragmentation problems associated with traditional image classification methods. Our results were significantly superior to MLC and DT with an overall accuracy of 89.14% and Kappa coefficient of 0.87;(4) The resource satellite data ZY-10C has a certain stability and advantages for the extraction of winter wheat.
Spatial interaction in urban agglomeration is an important driving force of urbanization in contemporary China. Most of the research based on cellular automata has focused on simulating one single city, and does not factually represent the more expansive processes of urban agglomeration. How to quantify the spatial interaction between cities and combine them in a CA model is an issue when simulating the evolutionary processes of urban agglomeration. This paper proposes an urban agglomeration simulation model based on spatial interaction(UASMBSI) that is better suited for simulating the sprawl of urban agglomeration by combining an urban flow model with a CA model. The UASMBSI is applied to simulate the urban sprawl in the Wuhan metropolitan area, with more accurate simulation results than the CA model that does not account for spatial interaction. A case study not only indicated that the spatial interaction plays an important role in simulating the sprawl of urban agglomeration, but also suggestes that the UASMBSI can represent the characteristics and rules for city development, especially the expansion of urban agglomerations.
Human action recognition plays an important role in the field such as video supervision and medical diagnosis. Current methods are based on the expansion from two-dimension artificial design features to three-dimensions, ones or extracting spatio-temporal features via trajectories. Based on deep learning methods, this paper proposes a multilayer neural network in three-dimensional space, learning rich spatio-temporal features from large amount of videos. First, we use independent subspace analysis to build a two layer stacked convolutional neural network, obtaining weights from training database. Spatio-temporal features are then quantized into visual words with K-means clustering. Non-linear support vector machine(SVM) were used to classify frequency histograms of visual words into different action groups. We apply our algorithm to Hollywood2 database, extracting spatio-temporal features from 12 human action groups. Result shows that the feature weights trained by ISA network are similar with those by Gabor filter, which have obvious selectivity of frequency and direction, robustness to phase variation, conforming to the human visual system.
A novel method based on the weighted joint covariance matrix fitting for synthetic aperture radar interferometry is presented. The method takes advantage of the coherence information of the neighboring pixel pairs and makes use of the interferometric information embedded in a joint covariance matrix, which makes it possible to estimate the interferometric phase in the presence of large coregistration errors, even up to one pixel. Benefiting from weighting on the joint covariance matrix and employing matrix fitting to estimate the interferometric phase instead of eigendecomposition, the method does not need to calculate the signal subspace dimension, thus avoiding the effect of the rank variation of the signal subspace. A fast algorithm for the interferometric phase estimation is proposed in which the closed-form solution to the interferometric phase is directly obtained, for the angle scanning search based method is time-consuming. The results of numerical simulations and real data demonstrate the validation of the proposed method.
We proposed a new image fuzzy classification method of correlation degree with a certain degree. The correlation degree basis points are obtained in each image category, and the certain degree of an image related to each category are calculated. The certain degree of an image is the weight of the corresponding image correlation degree, and the attribute of the image classification to be recognized is determined by the maximum criterion, which is the correlation degree with a weight. In this paper, to recognize the image category, the second criterion is characterized by the correlation degree with a certain degree is that the classification of the image to be recognized is determined with the principle of minimum distance between clustering centers. In several experiments with three categories of images, the experimental results which meet one of the two principles in the experiment show that the results of this method has certain advantages.
The compression of vector data is very important for reducing the space needed for data storage and improving the efficiency of data transmission and processing in WebGIS. This paper focuses on the time efficiency of vector data compression with prediction functions and proposes a vector data compression method based on sector screening that significantly reduces the quantity of candidate vertices in prediction areas to improve time efficiency. Experimental results show that the time efficiency improved by 30%-40%. Our method was compared with the conventional Douglas-Peucker method. The tests confirmed that our method can achieve a larger compression ratio when using the same compression threshold value, while obtain greater time efficiency with relatively small threshold values.
To recognize human behaviors in public areas, a new method of recognition was proposed based on deep learning. First, we pre-processed all the images in training and test samples, and utilized GMM to extract moving objects. Then, we built sample sets of various behaviors, and defined different behaviors as priori knowledge to train a deep learning network. Finally, all kinds of behaviors based on the network model of deep learning were recognized. Experimental results demonstrated our method outperforms the existed methods, and the average recognition rate is 96.82%.
Kriging interpolation is a common method applied to spatial data. Correct fitting of the variogram mode is the key to kriging interpolation. At present, experience is the guide when fitting variograms in analysis tools and development kits. An intuitional method is used to determine the parameters of a variogram and the parameters are adjusted reiteratively. These artificial operational steps obstruct the automation of kriging interpolation but also are not based on scientific theories relevant to setting parameters and option modes As a result, the predictive results are not exact. Aiming to address this problem and with the goal of automating the process, in this paper we calculate the sampling variogram and an automatic fitting algorithm for variograms is studied. Automatic kriging interpolation is realized and automatic cross validation is conducted based on the automatic fitting function in R language. The results of experiments show that a automatic method for kriging interpolation to spatial data is feasible.
The gravity field is a physical parameter reflecting the density change and dynamic characteristics of the earth under different circumstances including the solid earth tide, internal heat flow, mass exchange of solids and liquids, surface loads, and seismic tectonic movements. The time varied global gravity model has been provided by GRACE since 2002, but with the existing system error in GRACE and the need for focusing on local areas, post-processing is required when using GRACE products. In last decade, many algorithms have been shown to be effective. The ideas for these algorithms are reviewed in this article; a Gaussian filter with isotropic and non-isotropic types, the destriping filter, the empirical orthogonal functions method, wavelet analysis, and the Slepian function method. The future directions in post-processing algorithms are also discussed.
At present, Vertical components gravity is measured and solved in a local frame with the E tv s correction in airborne scalar gravimetry systems. The classic method ignores the vertical deflection effect. In this study, an improved formula is presented and the gravity measurement problem is solved in a quasi-inertial frame. The modified method improves the classical formulas for scalar systems theoretically, avoiding computing Eötvös and other approximations. The method was validated using a numerical test. The results show that the unnecessary approximations found in the traditional method will cause sub-mGal level errors. However, these errors can become almost negligible after applying a low-pass fillter with cut-off frequency at less than 1/60 s. In addition to this theoretical advantage, the modified method can acquire mGal-level horizontal gravity estimates through a precise data processing.
The vertical deformation of 22 tide stations was analyzed with a unifying datum based on GPS continuous observation data from the Chinese coastal GNSS operational observing system during the years 2009-2013. The results show that:The periodic change of the GPS height time series at tide stations has a higher correlation with the tide. The trend of vertical deformation of 22 tide stations has not only small regional consistency, but also shows a significant difference in different sea areas. By studying the vertical movement of tide stations, researchers can effectively eliminate the influence of crustal movement on regional sea level change, to obtain actual sea level change.
In highly dynamic GNSS simulation systems, since there is a contradiction between precision and calculated volume, a method of real-time high precision pseudorange calculation based on the third-order Hermite interpolation is proposed. In this method, a satellite orbit is computed by an interpolation algorithm and the sequences of the error computation is changed to optimize the pseudorange calculations. Then, by knowing the pseudoranges and derivatives for two nodes, the parameters of the pseudorange change model was computed by a third-order Hermite interpolation algorithm. The pseudorange derivative was obtained by the second order derivative Lagrange interpolation formula. The highly dynamic simulation results show that when the time interval is 1 s, the maximum pseudo-range error is 0.638 mm, and the standard deviation is 0.172 mm.
Effective spoofing detection is the premise for GNSS receiver carrying out spoofing suppression, preventing the occurrence of positioning, velocity, and timing information errors. A spoofing detection technique for GNSS receiver is proposed based on carrier phase double difference of a Spin-Antenna. After translational motion elimination and non-related differential processing, the carrier phase value of the uniform rotation of the receiver antenna can be used to realize the detection of single transmit antenna output spoofing signal by generalized likelihood ratio test. Furthermore, the impact of rotation speed, radius, and the length of data on detection performance is analyzed. The results show that increasing the length of data is the inevitable choice for improving detection performance. A non-detection zone is defined for global detection performance by comparing and limiting the decrease in the detection performance. A Monte-Carlo simulation illustrates the correctness of the method and analysis.
The most important causes of horizontal reference stations are human factors such as excessive pumping of groundwater and immoderate mineral exploitation. GPS data from Anhui Operating Reference Stations(AHCORS) between 2011-11 and 2013-9 was continuously processed the using GAMIT/GLOBK software. Then, based on these results, the 3D velocities of the reference stations were determined, and crustal horizontal movement and land vertical movement was analyzed in detail. It was found that the mean horizontal velocity of AHCORS reference stations was 34.44 mm/a in the direction of E18.01°S under ITRF2008 reference frame and 6.13 mm/a in the direction of E7.31°S with regard to the Eurasia Plate. Stations located in Huaibei Plain have moved in horizontal with 1~4 mm/a relative to No-Net-Rotation of the whole AHCORS. Tancheng-Lujiang Fault showed sinistral strike slip feature between Jiashan and Sixian, with very weak movement in the south of Jiashan. As for the vertical direction, AHCORS reference stations uplift in the south of the Huai River, but subside in the north of it. The average uplift rate of the stations located in the south of the Huai River is 2.72 mm/a, while the average subside rate of the stations located in the north of the Huai River is 2.98 mm/a.
Foundation and rail construction must meet higher accuracy for high-speed railway, in order to effectively control projection distance deformation of a GPS control network and reduce the number of projection zone for the east-west highspeed railways, this paper discusses the processing method for the oblique Mercator projection to be applied to develop uniform plane coordinate system by the basic theoriessuch as ellipsoid transformation, projection from ellipsoid to sphere and spherical cylindrical projection. Through a practical engineering example analysis, we show that the oblique Mercator projection can control the length deformation and meet the 10ppm length deformation per kilometer for control network of nonballasted-track and 25ppm for ballast track, and avoid the divisions of too many projection zones. Comparing withnormal Gaussian projection the oblique Mercator projectionhas more superiorities on the east-west high-speed railway andother route engineering control networks.
In geometrics, machine vision and other fields, the large rotation angle three-dimensional coordinate transformation often need to be applied based on the two corresponding point sets. The solvingmethods of the large rotation angle three-dimensional coordinate transformationare broadly divided into iterative methods and analytical methods. The iterative methods can obtain higher transformation accuracy. However, these methods depend on the initial value. In this contribution, quaternion is used to construct rotation matrix, a large rotation angle three-dimensional coordinate transformation formulae based on Newton iterative method is established, and a method which is widely used in guidance and control field for initial value calculation is given. Simulated and real data are used to validate the proposed algorithm, and the results are compared with those of other algorithms. The results show that the proposed algorithm makes the large rotation angle coordinate transformation model based on quaternion more robust. So the proposed algorithm is practically valuable for its stability of results, reliable accuracy and fast convergence.
GPS-derived height, due to the high efficiency and flexibility of GPS in engineering surveying, has been playing an important role under difficult environmental conditions. For applications in real projects, the geodetic height obtained by GPS should be transformed to normal height. The key issue in this transformation is to solve the height anomaly. This paper proposes a method to fit the residual height anomaly based on the EGM2008 model in combination with a separate terrain correction procedure. Datasets from an over-sea-bridge project were used to verify the performance of the proposed method. Numerical results show that, the proposed method can provide 0.16 cm accuracy for the points used for the fitting model determination and 0.96 cm accuracy for independent test points, and it can improve the precision of GPS height fitting.
According to the ASPeCt sea ice visual observation protocol, we extracted the sea ice edge from MODIS images and assessed the quality of AMSR-E SIC product using a corresponding MODIS Sea Ice Concentration(MODIS SIC). Results show that AMSR-E pixels located at the ice edges have a significantly different mean SIC value from the well-established 15% threshold. The correlation between AMSR-E and MODIS SIC is rather low with coefficients R2≤0.2 and the ASI algorithm underestimates sea ice edge concentrations in summer. Then transect analysis, applied on all the ice areas(including multi-year ice, one-year ice, new ice and open water), however, show that AMSR-E SIC and MODIS SIC have a good linear relationship with R2 of 0.82 in summer and 0.81 in winter. When AMSR-E SIC value falls between 20% and 30%, it has larger errors compared with the other range of SIC values.
Determining the exterior orientation parameters of photogrammetry image is an important task, and a lot of classic methods are in use. In the case of large attitude angle in low-altitude remote sensing image, the model of space resection algorithm based on unit quaternion has been used in many experimentations. In this paper, the stable Levenberg-Marquardt method(LM) algorithm is employed to solve the space resection problem of large attitude angle photogrammetry image. The feasibility of the proposed method is validated based on simulation data, Meanwhile, the comparing experimentations between LM algorithm andunit quaternion method show that the LM method is less sensitive to the initialization of exterior orientation parameters. The experiment also shows that the reliability of LM method is equivalent to unit quaternion method. Furthermore, if the damped factor is set suitable, LM method will be more efficient and reliable.