A Real Time Video Processing Based Surveillance System for Early Fire and Flood Detection

The intelligent surveillance system based on video content analysis has been the core trend in the security service industry. Due to the high complexity in techniques such as real time processing and image contents analysis/understanding, a well developed product is not available until now in this field. However, except the existing functions such as remote controlling, danger warning by artificial monitoring, video recording in a fix schedule or activated by scene action, etc., an intelligent surveillance system should include more advanced functions. This paper presents a simple and effective method for automatic disaster detection in the early stage by real time video analysis. Via analyzing the specific feature vectors including spatial-temporal spectra variation, color/grayscale histogram concentration, etc., a fully automatic process is developed to substantially improve the performance, especially in early detection thus to reduce the loss caused by natural disaster, of the existing security surveillance systems. The experimental results show that the proposed system not only has good performance but also alleviate the system cost than the previous systems thus open the future trend in the related security industry products.