A. Marazzi

A flexible environment for earthquake rapid damage detection and assessment

This paper outlines a system architecture able to reduce the effects of a devastating seismic event by providing a rapid and reliable damage detection and estimation of the extent and location of the suffered area. This result has been accomplished by the integration of data access and standardization techniques, image processing tools, GIS technology, analytical modeling and communication tools. A two-phase operating model has been conceived. In the pre-event era, images and data about building and infrastructures are collected and analyzed exploiting GIS capabilities. Immediately after the occurrence of the earthquake, the system must be ready to receive near-real time satellite imagery of the affected area to be compared with the pre-event imagery data set. A correspondence between the integrated databases, within the GIS environment, and the real-time imagery is also established. The post-event imagery is then compared with the pre-event one, by means of different digital processing techniques to exploit the different resolution of satellite images.

Automatic selection of the number of clusters in multidimensional data problems

When processing multidimensional remote sensing data, one of the main problem is the choice for the appropriate number of clusters; despite of the great number of good algorithms for clustering, each of them works properly only when the appropriate number of clusters is selected. As adaptive versions of the K-means, competitive learning (CL) algorithms also have a similar crucial problem; various efforts to improve the performance of CL were made with the introduction of frequency sensitive competitive learning (FSCL) and rival penalised competitive learning (RPCL). We present an improvement of the RPCL algorithm well adapted to work with every kind of real clustering data problems. The basic idea of this new algorithm is to introduce a competition also between the weights. The algorithm was tested on multiband images with different weights initial position, giving similar results.

A mixed fractal/wavelet based approach for characterization of textured remote sensing images

One of the problems encountered in the field of remote sensing image characterization, is the choice for the right features. The addition of textures as a discriminating parameter is a good help in the step of segmentation of different zones. The authors present an approach that is a mix between a wavelet multiscale analysis and a fractal characterization, in order to exploit both the main characteristic of the two approaches and to override the limitations of the two techniques. The chain was applied to different textured images showing an improvement respect to other methods based on wavelet transform and fractal approach alone.

Satellite data analysis for earthquake damage assessment

This paper moves from the results of RADATT (rapid damage assessment telematic tool), a project funded by the European Commission -- DG XIII. The final goal of the developed system architecture was to sensibly reduce the effects of a devastating seismic event by providing the responsible agencies a rapid and reliable damage detection and estimation of the extent and location of the suffered area. This result has been accomplished by the integration, within a single user interface environment, of data access and standardization techniques, image processing tools, GIS technology, analytical modeling and communication tools. A two-phase operating model has been conceived. In the pre-event era, images and data about building and infrastructures are collected and analyzed exploiting GIS capabilities. Immediately after the occurrence of the earthquake, the system must be ready to receive near- real time satellite imagery of the affected area to be compared with the pre-event imagery data set. A correspondence between the integrated databases, within the GIS environment, and the real-time imagery is also established. The post-event imagery is then compared with the pre-event one, by means of different digital processing techniques to exploit the different resolution of satellite images. In this paper the capability of this quick change detection analysis, given the availability of the pre-event information in the GIS environment, is discussed.

Rain pattern detection by means of packet wavelets

The recent advances in microwave telecommunications and the need for more precise weather forecasting systems are two of the many fields where it is extremely important to be able to analyze quickly, precisely and, possibly, in a fully or partially automatic way, the data obtained by systems like meteorological radars. This work presents a wavelet packet based algorithm, combined with a C-means classifier, for rain patterns detection and tracking from this data. The use of this kind of classification chain is motivated by the high efficiency and low computational load of the wavelet transform algorithm and by the observation that a large class of natural textures can be modeled as quasi-periodic signal, whose dominant frequencies are located in the middle frequency channels, easily provided by this transform. The chain was applied to a radar data sequence of a rain event of Northern Italy. The storm dynamics were studied at different meso-scales.

An integrated approach for high-compression of videoconference sequences

We propose a system for very low bit rate video transmission, leading to good quality images with 200:1 compression ratios. The algorithm first locates the head of the speaker using active snakes. A new form of internal energy is defined allowing a robust and fast head tracking. The most important face characteristics are extracted using a new algorithm. Fractional pixel block matching is used for motion estimation and compensation. The compensated difference image is decomposed in different psicovisual importance areas that are independently processed. The areas are segmented into regions the borders of which are coded using a differential chain code.

Object tracking in complex scenes by means of the correspondence based method

The problem of tracking targets in complex scenes can be solved with correspondence based methods by finding out an appropriate set of features for the identification of targets in motion, establishing a correspondence between the representation of the same objects at subsequent times by feature matching, tracking the target motion behaviour (history) in time. The authors propose a new solution to this problem for a nonmodel based system of tracking. The system presented is able to localize and to track a generic target in motion in a real scene without any information about it.

Tracking the evolution of rain patterns by means of modal matching

The usefulness of the modal matching approach for shape analysis to meteorological data interpretation is shown. A tracking chain for the detection and the analysis of rain patterns or the rainfall area of a given rain event is developed and tested on actual meteorological radar sequences. A morphing procedure to guess the evolution of the structures of interest between two radar frames is also introduced.

Intelligent image interpretation for high-compression high-quality sequence coding

Video transmission at very low bit rate has got growing attention in recent years. In this paper we propose an approach for the compression of 144 x 176 pixels Q-CIF video conference sequences. The compression ratio well exceeds 200 : 1 (thus leading to bit-rates under 10 kbit/s for 10 frames/s) with very good psycovisual quality of the reconstructed images. The algorithm integrates and improves different feature extraction and image coding techniques. At first the speakers head is detected by means of active snakes. A new form of internal energy is defined that allows a very robust and fast head tracking. After head detection, internal facial features (i.e., eyes, nose, and mouth) are located by means of a new algorithm. The image is decomposed into different parts with different psicovisual relevance. This information is used to guide in an intelligent way the subsequent processing of the motion compensated difference image. The important areas are coded more accurately while the less relevant areas are coded in a coarser way. This approach grants very high compression while the image quality remains high. Subpixel block matching is used to obtain a motion compensated difference image. This image is segmented into homogeneous regions that are then coded by means of a technique based on differential chain code.

G.5 – Tracking by Cooccurrence Matrix

Publisher Summary This chapter proposes a new method for solving the target tracking problem by exploiting the so called co-occurrence matrix. The co-occurrence matrix explicitly defines how the actual targets and their predictions match and give important hints about solving the problem of occlusion and splitting among objects. In this method, the first feature of a moving target is its shape, as obtained by observing the frames and extracting the differences between the image with the moving objects and the same image without the objects in motion. A reference image, built by means of the Discrete Gray Level Follower algorithm is needed to obtain a binary image with black regions on a white background. The binary image gives information on the regions of motion. The algorithm proposed is based on the assumption that all the motion phenomena have a temporal correlation. Given the co-occurrence matrix and the position of the centroid and the area of each blob, the algorithm is able to classify tile blobs of the new image in four classes. The algorithm is composed by three different levels of analysis. After the co-occurrence analysis, next step is to analyze the correspondences found and give a label to each blob of the image. Then the trajectory of each target can be tracked during its motion along the frame sequence. The system presented is able to localize and track a generic target in motion in a real scene.