| With the development of web technologies, some Content-based Image Retrieval system, namely, "finding images through image", are introduced on the Internet. In the field of stone industry, Stone Image Retrieval system is an application based on Internet environment, which can recognize unknown image through the image characteristics.The current research on Content-based Image Retrieval focuses on how to make the comparison more accurate, which is concern about the accuracy. However, when a great deal of access concurrently occur, Stone Image Retrieval system would be inefficient because of the large volume of calculation and the inefficient access to the database.This thesis focuses on the speed of "finding images through image". Firstly, the existing system performance is analyzed, secondly, constraints of the performance are found out, and finally, a corresponding optimizing solution is given out. The solution optimizes the system from the following aspects such as Memory database, multi-threaded technology and high-performance clusters. In the thesis, the principle and application of Memory database are studied, the Oracle TimesTen is attempted to visit in the .Net environment by ODBC interface, then, high-performance clusters are created in Stone Image Retrieval system, by using loading balance algorithms, as well as .Net Remoting, moreover, multi-threaded technology is added it into the system. At last, a series of tests has been done to illustrate that how the techniques upgrade the performance.Experiments show that: through the introduction of the memory database, it becomes faster by visiting the binary data type of image characteristics, the average retrieval time reduce an order of magnitude; reasonable designed multi-threaded programs makes CPU resources more efficiently used by image retrieval calculate; the cluster can enhance the stone image retrieval system of concurrent performance, then, shorten the concurrent request for the user's average waiting time. |
PDF Full Text Request