This paper presents two algorithms needed to perform a dense 3D-reconstruction from video streams recorded with uncalibrated cameras. Our algorithm for camera self-calibration makes extensive use of the constant focal...
详细信息
ISBN:
(纸本)9789898111210
This paper presents two algorithms needed to perform a dense 3D-reconstruction from video streams recorded with uncalibrated cameras. Our algorithm for camera self-calibration makes extensive use of the constant focal length. Furthermore, a fast dense reconstruction can be performed by fusion of tessellations obtained from different sub-sequences (LIFT). Moreover, we will present our system for performing the reconstruction in a projective coordinate system. Since critical motions are common in the majority of practical situations, care has been taken to recognize and deal with them.
Observations and decisions in computer vision are inherently uncertain. The rigorous treatment of uncertainty has therefore received a lot of attention, since it not only improves the results compared to ad hoc method...
详细信息
The performance of classifiers is commonly evaluated by classification rate and false alarm rate (FAR). Many applications like traffic monitoring, surveillance and other security relevant tasks suffer from the problem...
详细信息
ISBN:
(纸本)9783540715900
The performance of classifiers is commonly evaluated by classification rate and false alarm rate (FAR). Many applications like traffic monitoring, surveillance and other security relevant tasks suffer from the problem balancing the performance criteria in an appropriate way. In this contribution, we propose a kernel classification scheme with high performance in discriminating classes and rejecting clutter objects. Especially, it determines a class membership assessment. The classification scheme consists of two kernel classification stages and a maximum decision module as combiner. For tests, we use targets taken from the MSTAR synthetic aperture radar (SAR) dataset and clutter objects extracted from SAR scenes by a screening process. The dependency on parameter variations is shown and receiver operator characteristic (ROC) curves are given. The results confirm the high classification performance at low FARs. The integration into an operational demonstration system is in progress.
Model-based object recognition in range imagery typically involves matching the image data to the expected model data for each feasible model and pose hypothesis. Since the matching procedure is computationally expens...
详细信息
ISBN:
(纸本)0819457760
Model-based object recognition in range imagery typically involves matching the image data to the expected model data for each feasible model and pose hypothesis. Since the matching procedure is computationally expensive, the key to efficient object recognition is the reduction of the set of feasible hypotheses. This is particularly important for military vehicles, which may consist of several large moving parts such as the hull, turret, and gun of a tank, and hence require an eight or higher dimensional pose space to be searched. The presented paper outlines techniques for reducing the set of feasible hypotheses based on an estimation of target dimensions and orientation. Furthermore, the presence of a turret and a main gun and their orientations are determined. The vehicle parts dimensions as well as their error estimates restrict the number of model hypotheses whereas the position and orientation estimates and their error bounds reduce the number of pose hypotheses needing to be verified. The techniques are applied to several hundred laser radar images of eight different military vehicles with various part classifications and orientations. On-target resolution in azimuth, elevation and range is about 30 cm. The range images contain up to 20% dropouts due to atmospheric absorption. Additionally some target retro-reflectors, produce outliers due to signal crosstalk. The presented algorithms are extremely robust with respect to these and other error sources. The hypothesis space for hull orientation is reduced to about 5 degrees as is the error for turret rotation and gun elevation, provided the main gun is visible.
Segmentation is a fundamental task in patternrecognition and basis for high level applications like scene reconstruction, change detection, or object classification. The performance of these tasks suffers from a dist...
详细信息
ISBN:
(纸本)9783540715900
Segmentation is a fundamental task in patternrecognition and basis for high level applications like scene reconstruction, change detection, or object classification. The performance of these tasks suffers from a distorted segmentation. In this contribution an adaptive diffusion-based segmentation method is proposed suppressing perturbations in the segmentation with focusing on small regions with high contrast to their surrounding. The algorithm determines in each step the diffusion tensor. It is re-weighted with respect to an assessment stage. A comparative study uses high-resolution remote sensing data.
Change detection plays an important role in different military areas as strategic reconnaissance, verification of armament and disarmament control and damage assessment. It is the process of identifying differences in...
详细信息
ISBN:
(纸本)0819457949
Change detection plays an important role in different military areas as strategic reconnaissance, verification of armament and disarmament control and damage assessment. It is the process of identifying differences in the state of an object or phenomenon by observing it at different times. The availability of spaceborne reconnaissance systems with high spatial resolution, multi spectral capabilities, and short revisit times offer new perspectives for change detection. Before performing any kind of change detection it is necessary to separate changes of interest from changes caused by differences in data acquisition parameters. In these cases it is necessary to perform a pre-processing to correct the data or to normalize it. Image registration and, corresponding to this task, the ortho-rectification of the image data is a further prerequisite for change detection. If feasible, a 1-to-1 geometric correspondence should be aspired for. Change detection on an iconic level with a succeeding interpretation of the changes by the observer is often proposed;nevertheless an automatic knowledge-based analysis delivering the interpretation of the changes on a semantic level should be the aim of the future. We present first results of change detection on a structural level concerning urban areas. After preprocessing, the images are segmented in areas of interest and structural analysis is applied to these regions to extract descriptions of urban infrastructure like buildings, roads and tanks of refineries. These descriptions are matched to detect changes and similarities.
This paper is dedicated to modifying some steps needed to generate a dense 3D-reconstruction from a video stream. Since critical motions are common in the majority of practical situations, care was taken in our work t...
详细信息
Regarding computer vision as optimal decision making under uncertainty, a new optimization paradigm is introduced, namely, maximizing the product of the likelihood function and the posterior distribution on scene hypo...
详细信息
One of the main challenges in video-based multi-target tracking is the consistent maintenance of object identities over time. We present a novel approach to that challenge that integrates tracking and detection in a s...
详细信息
Ground target classification in high-resolution SAR data has become increasingly important over the years. Kernel machines like the Support Vector Machine (SVM) and the Relevance Vector Machine (RVM) afford a great ch...
详细信息
ISBN:
(纸本)3540408614
Ground target classification in high-resolution SAR data has become increasingly important over the years. Kernel machines like the Support Vector Machine (SVM) and the Relevance Vector Machine (RVM) afford a great chance to solve this problem. But it is not possible to customize these kernel machines. Therefore the main objective of this work has been the development of a mechanism that controls the classification quality versus the computational effort. The investigations have been carried out with usage of the MSTAR public target dataset. The result of this work is an extended RVM, the RVMG. A single parameter is controlling the robustness of the system. The spectrum varies from a machine 15 times faster and of 10% lower quality than the SVM, goes to a 5 times faster and equal quality machine, and ends with a machine a little bit faster than the SVM and of better quality than the Lagrangian Support Vector Machine (***).
暂无评论