基于可重构计算技术的图像分类系统研究

Reconfigurable computing technology based research on the image classification system

考虑到图像分类方法是计算密集型算法,设计使用可重构方法对图像分类系统加速.再由纹理参数本身的特性,利用基于差异的动态重构方法以有效利用芯片.实验表明,所设计的图像分类系统与通用处理器相比,显著减少了分类时间;同时在不增加运行时间的前提下,与传统实现方法相比,能更有效地利用芯片面积.

In view of the high computation intensity computing technology was adopted to accelerate of image classification algorithms, a reconfigurable calculation. Based on texture＇s characteristics, a difference-based reconfigurable method was introduced to reduce the utilization ratio of the chip area. Results show that the system has better speed performance when compared with a PC and uses FPGA effectively without extra time when compared with traditional realization methods.