Program Optimization Of Iris Recognition Algorithm Based On DM642

Along with the progress of technology, the people pay more attention to the biometric recognition technology. A personal identification system based on iris patterns is regarded as one of the most promising biometric recognition technology because of its excelsior accuracy and advantage.The developments of iris collection device and embedded iris recognition system have the same important effect as the iris recognition algorithm. The realization and optimization of iris recognition algorithm based on DM642 hardware platform are discussed and has very high practical value and realistic significance.In this paper we firstly introduce resumptively the basic principle and the process of iris recognition including iris image collection, the estimation of iris image quality, pre-processing of iris image and feature extraction of iris and pattern match. Meanwhile, the system composing of iris recognition system based on PC and embedded iris recognition system are introduced according to different platform. The DM642 embedded platform is introduced in detail.Secondly, the transplant of iris recognition algorithm based on DM642 platform is introduced. And performance difference of X86 and the DM642 hardware platform and the DM642 platform application procedure software construction are also introduced. Then the entire transplant process and the key point of iris recognition algorithm are discussed.All kinds of code optimization measures and the memory structure optimization based on DM642 and their applications on iris recognition algorithm optimization are introduced in detail. Some high-level optimized methods such as linear assembly and software pipelining and skills of application are researched. Because DM642 is a fixed-point DSP chip, the optimization of floating-point operation based on fixed-point chip will be the emphases of the paper. Two kinds of optimization methods of fixed-point simulation and the block floating point came true concretely and the fixed-point simulations turn to the optimal method of floating-point operation of iris recognition algorithm as its well result.Finally, some performances of optimized iris recognition algorithm are tested and some problems of optimization are summarized. The results of optimization are analyzed by target of time, veracity of location and rate of pattern match. The well and true result of optimization of iris recognition algorithm is proved.