An approach for urban traffic volume compression based on similarity attributes.

With the development and integration of ITS-related control and management systems, the ITS data management and applications are becoming the critical requirements recently. Due to the limitation of the traditional database systems, the more efficient approach used for the ITS data compression is needed to form the fundamental of the ITS data storage, restoring and real-time management. A data compression model is presented in the paper based on the similarity attributes existing in the urban traffic volumes. The similarity attributes are extracted explicitly using the set of three-layer feed-forward neural networks while the activities of hidden layer units of the network are stored instead of the original data. Based on the discussed compression model, the volumes can be restored with the high fidelity and the compression ratio up to 15:1. Finally the implementation of data compression and restoring is undertaken to demonstrate the full application.