Catálogo de publicaciones - libros

We present a global approach that enables the production of 3D soccer sequences from non-calibrated video cameras. Our system can produce a 3D animated model of the scene from a single non-calibrated moving camera (a TV sequence for example). The results presented here are very encouraging even with a single camera approach and will probably improve with the future introduction of multiple images that will help resolving occlusion issues and integrating into a single model information coming from various locations on the field. The key point of our approach is that it doesn’t need any camera calibration and it still works when the camera parameters vary along the process. Details on the registration and tracking processes are given as well as the description of the “Virtual Reality” system used for displaying the resulting animated model.

The ortho-rectification of satellite images is normally a time-consuming and expensive task. It can strongly benefit from automatic or semi-automatic procedures in order to avoid field work for ground control survey. This article presents an automated method to automatically georeference satellite images acquired by the ASTER sensor. The method is based on automatic matching of images and Digital Elevation Models (DEM). First a DEM is extracted from the two stereo image bands, which then is matched to the SRTM (Shuttle Radar Topography Mission) DEM in order to correct its position in a map reference system. This allows for the simultaneous correction of image geo-location, which then is followed by the ortho-rectification. The method was applied to an ASTER image from North Portugal and assessed using topographic map data. It was possible to geo-reference images in hilly terrain with positional accuracy better than one pixel.

An efficient sorting mail system is mainly based on an accurate optical recognition of the envelopes addresses. However, the location of the address block (ABL) should be done before the OCR recognition process. The location step is very crucial as it has a great impact on the global performance of the system. Actually, a good location step leads to a better recognition rate. The limit of current methods depends on modular linear architectures used for ABL. Their performances depend on each independent module performance. We are presenting in this paper a new approach for ABL based on the hierarchical graph coloring and on the pyramidal organization of data that present the advantage to guarantee a good coherence between different modules and that reduces both the computation time and the rejection rate. The proposed method gives very satisfying rate of 98% of good location on a set of 750 envelope images.

This paper presents a new method for detecting poultry skin tumors in hyperspectral reflectance images. We employ the principal component analysis (PCA), discrete wavelet transform (DWT), and kernel discriminant analysis (KDA) to extract the independent feature sets in hyperspectral reflectance image data. These features are individually classified by a linear classifier and their classification results are combined using product rule. The final classification result based on the proposed method shows the better performance in detecting tumors compared with previous works.

This paper presents real-time fabric defect detection based in intelligent techniques. Neural networks (NN), fuzzy modeling (FM) based on product-space fuzzy clustering and adaptive network based fuzzy inference system (ANFIS) were used to obtain a clearly classification for defect detection. Their implementation requires thresholding its output, and based in previous studies a confusion matrix based optimization is used to obtain the threshold. Experimental results for real fabric defect detection were obtained from the experimental apparatus presented in the paper, that showed the usefulness of the three intelligent techniques, although the NN has a faster performance. Online implementation of the algorithms showed they can be easily implemented with commonly available resources and may be adapted to industrial applications without great effort.