Research And Implementation Of Remote Sensing Image’s Closed Sequential Pattern Mining Algorithm

Data mining in remote sensing image has been a research area of broad application perspective. With the data sets of remote sensing image database, mining of the remote sensing image becomes the major approach for spatial data mining. As a result of the rapid change of image acquisition and storage technology, we could conveniently get access to a large amount of remote sensing image data. Being a new and very potential research area, data mining in image is a technology for mining hidden knowledge or patterns from image data and relations between/inside images. By synthesizing the GIS data from satellite, a kind of the approaches in this area uses Apriori algorithm and sequence pruning algorithm to find the association rule between different properties and different objects. This leads us to the integration of sequence mining algorithm into the remote sensing image mining.As the core algorithm for data mining in remote sensing image, the performance of the sequential pattern mining algorithm has always been the bottleneck for the whole approach. To overcome the bad performance of Apriori and prefixSpan algorithm on large data sets, this paper proposed a modified closed sequential pattern mining approach for remote sensing image.This paper uses a closed sequential mining algorithm based on BIDE and integrates it in the remote sensing image data mining. This algorithm can do the closed sequential check directly without maintaining the candidates for the closed sequence and reduce the search space quickly. We tested every module in this algorithm on remote sensing data sets and proved the algorithm of high efficiency.BIDE algorithm on closed sequence checking and search space reduction process requires a lot of character matching and support computing operations for larger scale data sets. Both operations consume a lot of time overhead. In order to reduce both operating time overhead, the paper proposes a position-based closed sequential mining algorithms. The algorithm records the information of every event’s position,makes use of the information to get frequent 1-sequential,and checks them.We reduce scaning shadow database and save time. For data sets of different sizes, this paper compared the position-based closed sequential pattern mining algorithm to the BIDE algorithm in experiments. The experimental results show that the former time performance has been improved remarkably.