The Research On Multi-Resolution Data Storage And Region Query Processing Algorithm In Sensor Network

These years, wireless sensor network is focused by many researchers due to its large value and deep influence to the people.With the large energy cost ratio in the total node energy cost, or the special purpose of many queries .etc, sometimes users would like to do queries with different resolution. So, it is necessary to establish multi-resolution data storage architecture.In the past multi-resolution data storage methods, most ones are based on the compression technology by transformation idea, they get important features of original vector, for example discrete cosine transform, wavelet transform. However, it makes that the storage cost in high-level node is very large.Aiming at the multi-resolution storage problem, firstly we propose a multi-resolution data storage model MDS: sensor data are ranked and approximated in low-level sensor nodes using spatial correlations inside the region unit; then, results are dealt with approximation and clustering technologies in high-level nodes using spatial correlations between region units. These operations produce approximation results with resolution from low to high. Simulation experiment shows that the storage pressure in high-level nodes is alleviated effectively.When answering aggregate query, by using previous data storage mechanism with transformation based compression methods, nodes in the query region need to send all stored characteristic message about the query region to the sink node which will do query processing in centralized way. It makes the communication cost reaches to the communication cost on snapshot query with the same region.Based on MDS, we propose a new region query processing method which optimizes the query process and gets lower communication cost by simulation.Finally, aiming at the historical data storage problem, we define the memory released weight factor, and propose a weight-based historical data storage method, it ensures that different historical timestamp could gets data compression operation with different levels. Simulation experiment shows that it can effectively alleviate the conflict between the reduction of data storage and the reduction of query accuracy about the historical data on the whole.