The sub-project "Spatial integration" is realised by the three project partners Jena-Optronik (DJO), Jena, Germany, the Zentrum für Bild- und Signalverarbeitung e.V. (ZBS), Ilmenau, Germany and Desotron, Soemmerda, Germany. Jena-Optronik is responsible for the classification and integration of all software components in the scope of the sub-project "Spatial Integration", The ZBS is working on the software for the sub-project "Structure Integration" and Desotron will develop a software package in order to accomplish the goals of the sub-project "Geometric Integration".

 

Goals of the sub-project "ENVILAND - Spatial Integration

Specific Tasks - Spatial Integration

The sub-project "ENVILAND - Spatial Integration" is being accomplished by JenaOptronik. Two additional software packages will be integrated in the form of two further sub-projects. The sub-project "Structure Integration" is based on the cooperation with the Zentrum für Bild- und Signalverarbeitung e.V., Ilmenau, Germany. The second software package will be provided in the scope of the sub-project "Geometric Integration" by Desotron, Sömmerda, Germany.

Main goal of the sub-project "Spatial Integration" is the analysis of different methods of land use mapping resulting in a highly automated land use classification process. Experiments are being conducted for this automated process with the goal of simulating land use mapping and change detection without any interference by the user. Data input are image data (SAR, optical data) and the formal description of the classification scheme. The results of all ENVILAND sub-projects will be brought together to enable the experiments that need to be carried out for this simulation. These experiments should finally lead to an improvement of the classification accuracy and to a statistical and geometric stability of the segmentation results that allows reliable change detection. The experiments should provide the necessary knowledge for an evaluation of the feasibility of an automated classification with the required properties. A second goal is the identification of further needs in research and development for the creation of program systems the enable an easy way of continuous mapping of the surface. Correspondingly, quantitative aspects like a high number of land cover categories are of less priority. The long-term goal for this procedure development could be the further development of the results from ENVILAND to modules in a ground segment that operationally and continuously delivers classification results as some kind of "standard product".

 

Specific Tasks - Structure Integration

The registration of image data from different sources with different resolution that were being taken at different times and recording conditions can be done by modeling the recording conditions or from the image data themselves. Methods that look at pass points and homologous points are being used successfully in combination with the respective models of the recording technique. If automated procedures are used, disadvantages such as ambiguity, aspect dependence, and other recording conditions can occur from the definition of the homologous point's properties at an iconic level and with a very restricted environment. Better and more stable and reliable results can be expected from the partly integral properties of structure information (e.g. from regions and region borders) because of their noise immunity and shape specifics. Outlines, which are harder to classify such as covered outlines and highlight outlines will not occur in the expected scenes or will only be of little importance. A second important issue of the structure integration is the segmentation of regions for supporting the classification involving the temporal information and scale properties. The integration of structure information requires the realization of the following sub-goals:

• Iconic image data improvement in consideration of the specific properties of the data source (noise) and the further processing goals
  
• Automatic and stable extraction of primary structure information
  
• Linking of elementary structure data by using parameters that are being taken from the goals of image data interpretation and further boundary conditions of the image generation
  
• Formulation of algorithms that retrieve homogeneous structures despite of different sensor types and resolutions
  
• Formulation of algorithms that creates segments of an image which are being deduced from structures
  
• Formulation of position-, scale-, and rotation-invariant approximation approaches, derivation of quality standards
  
• Homogenisation and regulation of structure data, e.g. by using morphological operators for shape describing list data
  
• Formulation of concepts for the efficient description of structure components and their properties
  
• Detection and identification of structure properties
  
• Formulation of methods for the use of the identified structure data for geometric corrections and classification tasks
  

 

Specific Tasks - Geometric Integration

The exact geometric overlay of the single data files is of utmost importance for the analysis of time series (e.g. change detection). For that reason the sub-goal geometric integration is trying to get consolidated findings about following issues:

• complete modelling of the image geometrics of SAR sensors, optical matrix sensors, optical push-broom sensors, and optical whisk-broom sensors
• determination of the parameter of a sensor model for the respective sensor types from ground control data
  
• statistical adjustment of the model parameters and the pass point coordinates
  
• model determination including a statistical adjustment in the case of several overlapping data sets of different types (SAR, matrix, push-broom, whisk-broom) where the pass points are available at a random combination of image and ground points
  
• use of calculated models for the exact overlay of overlapping data sets in image or ground coordinates
  
• three-dimensional point measurements for the determination of coordinates of homologous points from an image series at a random combination of sensors types including the statistical adjustment for over-determined cases
  
• morphing of the overlapping images from a series in a ground- or image coordinate system of choice with the conservation of the conformity of the homologous pass points or pass structures