| Multi-core parallel computing technology is the current hot researching field of computer science. In the near future, along with sustained growth of the number of core in a chip, multi-core computation will become a widely popular computing model, and significantly improve the calculation ability of computers, with great development potential and broad application of space. Presently, the CPU basic frequency's promotion receives the serious restriction as a result of the technique of production and the radiation question, the CPU performance improvement has mainly turned toward the multi-core system development. However, if wants to obtain the high efficiency which truly the multi-core processor brings, software's development must follow hardware's step, current multi-core processor software overall lag in hardware. The traditional single thread serial computation software can only cause the multi-core processor idle, when the algorithm design and the software development can use the multi-core processor's characteristic fully, its superiority can manifest truly.This article launches the research regarding the multi-core parallel computing in the image analysis domain's application, the discussion algorithm design and parallel class realizes. The topic research's content includes:(1)In the foundation of analyzing parallel computation and multi-core architecture, analyze the main factors, which influence parallel on multi-core, and performance bottleneck, probe into the latest programming technology, which are suitable for multi-core;(2)In the foundation of introducing edge detection, image segmentation and image matching, analyze the parallelism of the image processing fundamental arithmetic, and according to this kind of parallel construction support image analysis function's parallel computing class, and tests its performance;(3) Analyze the code of function class, design image analysis parallel computing class based-on multi-core architecture and test the performance, according to the code in the framework of multi-core performance, make out corresponding adjustment. Optimize the class code from source code, code parallel and the correctness check;(4)After optimization, make out the extended analysis according to Amdahl law and Gustafson law, obtain the external performance evaluation. Presently, the image monitoring technology is used popularly in military, environment, geology, meteorology, public security, transportation and industry applications, image gathering tends to the high accuracy the great width and the high frequency day by day, simultaneously the supervisory system also gradually develops by the artificial interpret to the intelligent processing direction. So the image analysis high performance computation urgently demands. This article research adapts to the current computer software and hardware technology trend of development, is helpful in enhancing the image analysis real-time and the precision. The topic characteristic and innovation mainly have two points: First, the image analysis function class by the traditional serial arithmetic transformation is on the IA high performance multi-core parallel computing; Second, uses the brand-new technology to carry on the parallel code the performance analysis, realizes calculates software's algorithm structure to optimization, the complier optimization and the source code level optimizes, enhances the extension which its operation efficiency and the adaptation to hardware develop.On Intel construction multi-core CPU is the brand-new technology which in the recent two years just appeared on the market. The research adapted to this system's application software technology is just started, the correlation data and the application achievement might use for reference were few.The research based on multi-core structure is quite scarce currently. This thesis carries on a preliminary exploration to image analysis parallel computation class. Many topics need to be expanded for further research. |
PDF Full Text Request