The surface matching is useful to solve the problem of comparing two surfaces obtained from two data sets of the same terrain. The data sets of points could be get from different sources such as Aerial and Satellite imagery, SAR, IFSAR, UDAR etc. The solution of surface matching problem is based on the some characterized contours existed on the surfaces as line features, which have to be expressed by parametric representation in frequency domain. Line features could be determined in the 2-D image or object space. For finding corresponding pair of point belonging to corresponding interest lines on the both surfaces, the image and object line features-based matching method has been used. Surface matching accuracy is estimated basing on the coordinates differences between the original and transformed surfa

One of the basic processes in photogrammetry consists of identification and measurements of conjugate (homologous) points located within image overlapping. In analytical photogrammetry this process is solved manually by an observer. In digital photogrammetry this process is solved automatically by software and it is called image matching. This process has considerable importance for automation of orientation or aerial triangulation of photographs. The accuracy of image matching process influences the accuracy of determination of image orientation elements and computed point coordinates. This article presents the author's idea concerning matching of digital images with regard to correlation between neighbouring pixels. First, the problem of correlation between point co-ordinates will be examined in analytical photogrammetry, what will simplify considerations related to digital photo gram metry.

One of the major tasks in digital photogrammetry is image matching technique for finding corresponding points in a stereopair. Area-based matching has been acknowledged as being more precise than feature-based matching. Least squares multi-point matching (LSMM) is one of the Global Image Matching (GIM) which was developed from the Least-squares Single point matching (LSSM) so called area-based matching. LSMM method has been more reliable than LSSM one because the relationship between the different neighbouring points is considered in simultaneous computation. LSMM is just for the simultaneous determination of the horizontal parallaxes at the node points of the regular rectangular nets for the purpose of the establishing the DEM. This paper undertakes a trial of improving the accuracy of LSMM by consideration of the correlated pixels and terrain height differences.

Statistical moments have been used in different applications as in shape analysis of object, pattern recognition, edge detection texture analysis etc. The idea is to use the moments as features of high level for surface matching. The essential goal of surface matching is to determine transformation parameters between two surfaces generated in TIN or DEM without identical points. Statistical moments are considered as features that are applied to solve that goal, One of the main problems with using statistical moments for surface matching and for other applications is a very expensive computation time. To overcome this difficulty many algorithms have already been proposed. New approach of efficient computation of inertial moments for surface matching is proposed in the paper. The approach is based on Green's theorem that allows for transforming double integral into a line integral. In the consequence computation time of inertial moments of a single TIN-model (triangle) is reduced by a factor 4 as compared with time consumed by the use of direct method of double integral. The direct computation using line integral, that does not involve any approximation, ensures preservation of the accuracy of computed moments.

Today, the new era with Very High Resolution Satellite (VHRS) imageries as IKON OS, QuickBird, EROS, Orb View etc., provides orthophoto in large scale of 1 :5 OOO, to update existing maps, to compile general-purpose or thematic maps. Orthophotomap in the scale of I :5 OOO with Ground Sampling Distance of 0.5 m is one of three important sources for establishing GIS together with a Digital Elevation Model of ±LO m accuracy in height and a topographic map in the scale of 1: IO OOO. Therefore, the accuracy of products of VHRS imageries affects reliability of GIS. Nevertheless, the accuracy of products of processing VHRS imageries is at first dependent on chosen geometrical sensor models. The understanding of geometrical sensor models of VHRS imageries is very important for improving processing of VHRS imageries. The polynomial models are to provide a simple, generic set of equations to represent the indirect relationship between the ground and its image. The polynomial models or replacement sensor models must not only model the ground-to-image relationship accurately. Physical (or parametrical) model describes dir~ctly strict geometrical relations between the terrain and its image, using satellite's orbital parameters and basing on the co-linearity condition. In such model, the above-mentioned multi-source distorting factors are taken into consideration. In this paper a review of practical accuracy of geometrical models of VHRS imageries taken from different research institutions in the world in last years has been presented.

The purpose of surface matching is to determine transformation parameters without known corresponding points for two data sets of spatial point coordinates obtained with use of different sensors. Instead of different features such as points of interest, lines, surface patches in the TIN (Triangle Irregular Network) or DEM model are used. The paper presents an approach of using inertial moments of TIN models generated from two data sets of same terrain for surface matching. The inertial moments could easily be calculated for each triangle in the TIN using formulae given. Three moment invariants/,./,,,,"/""'' that are used as the features of high level for surface matching are defined.

The paper presents the line moments of edge contour detected in an image as the high level features which are useful for surface matching. It has been proved that line moments do not depend on scale and rotation in transformation and they are sensitive to small changes of line erroneously extracted. Therefore, line moments are the useful tools in the process of feature-based matching, which can be used for merging (comparing) two surfaces derived with different sensors for the same terrain scene. In order to receive a line in an image, the edge pixels of terrain contour have to be detected and then linked into a line. The paper also focuses on the problem of using wavelet transform for automatic detection of edge pixels. The suggestion of 3-D line moments for surface matching has been presented in the section 5.

Surface matching is a fundamental task that must be solved whenever we want to merge data sets of same physical surface obtained with different sources. In general, the points in two sets are in different reference systems, not identical with different accuracy, distribution and density. To perform surface matching the different target functions are proposed. Proposed target function in section 1.4 is based on the condition of equality of triangle areas (TIN). This target function can be used for surface patches of pattern set S 1 generated also in squares (DEM). Conception of target function provided in section 1.5 is based on the combination of two conditions. The first of them is a condition of fitting two normal vectors of surface patches generated by square model (DEM) in set S 1 (as a pattern) and of triangle created with three points in set S2 (as a candidate). The second is one of four target functions upper represented.

Przy porównaniu dwóch powierzchni terenu, które są reprezentowane dwoma zbiorami danych punktów otrzymanych z różnych sensorów tego samego obszaru wykorzystano sieć neuronową Hopfiel'da HNN (Hopfield' s Neural Network) dla uzyskania rozwiązania optymalnego, tzn. kiedy funkcja energii tej sieci zdąża do minimum. Funkcję tę wyznaczono z sumy warunków zapewniających jednoznaczność cech należących do odpowiednich elementów powierzchni oraz z funkcji celu. Stabilny stan sieci HNN jest określony wówczas, gdy funkcja energii dąży do minimum. W tak.im przypadku zadanie dopasowania dwóch powierzchni jest najlepsze.

Automatic generation of DEM based on image matching technique is an important part in digital photograrnmetry. DEM has been found widely in practical applications as in the fields of surveying, civil engineering, road design, agriculture, military, etc. For large DEM project, accuracy, efficiency, and economy are the three main factors to be taken into account. Accuracy is the first important factor to be considered. The effect of modeling methods on DEM accuracy have been tested. The characteristics of terrain surface are considered as a difficult and traditional topic in generating of DEM. Terrain surface may be open or covered with different objects. For generation of accurate DEM, objects on terrain surface must be detected and filtered. There are many methods for detecting and filtering objects which have been found on the terrain surface. Some of them are basing on mutual location of measured points along definite profiles (geometric analysis) or on images converted. This article presents formulas for automatically detecting and filtering covered objects on the base of the geometric analysis of measured points mutually located along determined profiles in X and Y-direction.

The edge extraction seem to be an important step in matching of aerial images for 3D city and also in analysis and interpretation of satellite imagery for middle scale mapping production. Since twenty years the problem of edge and line extraction is still actual. The degree of full automation for edge and line extraction is in the way to advance. In practice there are several groups of methods using for extraction of edges and lines. The methods based on the differential geometry are often useful to solve this tasks. This paper presents a new algorithm approach to find the coefficients of function fitting to an edge in 2D image that is written by Fouriers expansion. Basing on the gradients of edge pixel that is Gausians line-spread function, the fitting function to edge should be constituted. For improving reliability of linking edge pixels into line it is proposed to apply three thresholds in which one is based on the finding correlation coefficients of grey levels between the neighbouring pixels in the image window.

Since 2000 when first imageries of Space Imaging of one metre resolution satellite products appeared on the World market, many institutions started using them for cartographic production such as orthophotomaps on a large scale. A choice of the mathematic sensor models of imageries for their orthorectification in producing orthophotomaps is one of the main investigation directions. In order to restitute the functional relation between imageries and their ground space, the use of sensor models is required. They can be grouped into two classes, the generalized sensor models (geometric or replacement sensor models) and physical or parametric models. The paper presents a brief overview of the geometric models such as RPC (Rational Polynomial Coefficients). Their properties, and in particular their advantages and disadvantages are discussed. Also the parametric models, developed by the authors are presented in this paper. They are based on time-dependent collinearity equation of the mathematic relation between ground space and its imageries through parameters describing the sensor position in satellite orbit and position of the orbit in the geocentric system.

The coastal zone and estuaries of Quang Ninh and Hai Phong have great potential not only for economic development but also for protection and conservation of biodiversity and ecosystem. Nowadays, due to industrial, agricultural and anthropogenic activities signs of water pollution in the region have been found. The level of surface water pollution can be determined by traditional methods through observatory stations. However, a traditional approach to determine water contamination is discontinuous, and thereby makes pollution assessment of the entire estuary very difficult. Nowadays, remote sensing technology has been developed and widely applied in many fields, for instance, in monitoring water environments. Remote sensing data combined with information from in-situ observations allow for extraction of polluted components in water and accurate measurements of pollution level in the large regions ensuring objectivity. According to results obtained from Spot-5 imagery of Quang Ninh and Hai Phong, the extracted pollution components, like BOD, COD and TSS can be determined with the root mean square error, the absolute mean error and the absolute mean percentage error (%): ±4.37 (mg/l) 3.86 (mg/l), 27%; ±55.32 (mg/l), 48.30 (mg/l), 14%; and ±32.90 (mg/l), 23.38 (mg/l), 28%; respectively. Obtained outcomes guarantee objectivity in assessing water contaminant levels in the investigated regions and show the advantages of remote sensing applications in Resource and Environmental Monitoring in relation to Water – Air – Land.