Search

The Self-Organizing Map (SOM) is a powerful tool for exploratory data analysis which has been employed in a wide range of data mining applications. We present a novel approach to reveal the inherent hierarchical structure of data using multiple SOMs together with heuristics which optimize the stability. In particular, we address shortcomings of the Growing Hierarchical Self-Organizing Map (GHSOM) regarding the decision which areas in the hierarchical structure need to be represented by a finer granularity and which areas do not. We introduce the Tension and Mapping Ratio extension to exploit specific characteristics of the SOM based on the topology preservation. As a main result, in contrast to the GHSOM, the inherent hierarchical structure of the data is revealed without requiring the user to define a threshold parameter which controls the map sizes of the individual SOMs. We evaluate our approach using data from real-world data mining projects in the music domain.

In this paper we propose an algorithm for counting moving objects in outdoor environments. Similar to other approaches the proposed algorithm uses a traditional surveillance system approach: background subtraction followed by noise removing, tracking and object labeling. The novelty of the proposed algorithm is that each processing step uses a stereo camera. In fact, the algorithms use both rectified and depth images obtained with a stereo camera. Moving objects are extracted from the scene using a novel background subtraction approach. In outdoor environment the images can be degraded by noise. We considered two types of noise, namely periodic movements due to the wind and the cast shadow. In this paper novel algorithms for detecting these types of noise are proposed. Using them, the noisy pixels are removed. The resulting segmented pixels are grouped together by clustering connected regions and then tracked during their movements. The obtained blobs, which correspond to the moving objects, are thus obtained with high accuracy. The final part of the algorithm is blob classification, which identifies the moving objects. The knowledge of the moving objects can be used to build more complex counting applications than that based on just counting blobs. Two simple applications based on the proposed algorithm are worked out and discussed in this paper, namely one that counts the people moving on the left and on the right of the video scene and one that counts the cars moving in the same way.