On 29 May 2009 at 14.00 clock, the MSE project "Object-based classification for the` cultural-historical survey 'of a landscape "is presented. Those interested are cordially invited to this event at the University Northwestern Switzerland in Muttenz.
Summary
An object-based classification consists of a segmentation and a subsequent classification of the segments. The option of calculating new features at the segment level, using this method also objects in high resolution Aerial photo data is classified. The resulting, classified objects can be compared in the analysis result in a change in another. If there is already classified objects, or from other sources derived data in a GIS are available, they can be used as the segmentation of a new image and be examined for a change up. Also allows the object-based classification approach to model change classes over several data sets (multi-date classification).
parts of these methods and processes were implemented in this work with the software ERDAS IMAGINE Objective and evaluated. We investigated the extraction of trails, roads and main roads in a test area in Andermatt Swiss Image-orthophotos. In the extraction of the trails and tracks the images were segmented into lines. A subsequent filtering based on a pixel-based classification and the use of other operators to filter, already allowed a satisfactory result without that the line segments were classified as such. In narrow paths and line segments, there are few characteristics that are suitable to carry out a successful classification. In the extraction of main roads, on land segments, some useful features are calculated, which were better classify ermöglichten.Zur change the analysis results a number of classifications visually compared. Despite, some many misclassifications, the results can be useful for finding changes. In addition, the results of a data set with a pixel-based classification of another data set were compared in order to extract segments that indicate a change.
Summary
An object-based classification consists of a segmentation and a subsequent classification of the segments. The option of calculating new features at the segment level, using this method also objects in high resolution Aerial photo data is classified. The resulting, classified objects can be compared in the analysis result in a change in another. If there is already classified objects, or from other sources derived data in a GIS are available, they can be used as the segmentation of a new image and be examined for a change up. Also allows the object-based classification approach to model change classes over several data sets (multi-date classification).
parts of these methods and processes were implemented in this work with the software ERDAS IMAGINE Objective and evaluated. We investigated the extraction of trails, roads and main roads in a test area in Andermatt Swiss Image-orthophotos. In the extraction of the trails and tracks the images were segmented into lines. A subsequent filtering based on a pixel-based classification and the use of other operators to filter, already allowed a satisfactory result without that the line segments were classified as such. In narrow paths and line segments, there are few characteristics that are suitable to carry out a successful classification. In the extraction of main roads, on land segments, some useful features are calculated, which were better classify ermöglichten.Zur change the analysis results a number of classifications visually compared. Despite, some many misclassifications, the results can be useful for finding changes. In addition, the results of a data set with a pixel-based classification of another data set were compared in order to extract segments that indicate a change.
cutting paths 2003 and 2004
0 comments:
Post a Comment