摘要
对于具有旋转和缩放的图像,为了解决景象匹配技术在硬件平台上的实时实现问题,分别从软件和硬件两个方面对系统进行设计:软件方面,改进传统的傅里叶梅林变换(FMT),并减少细节信息对计算结果的影响;硬件方面,采用DSP TMS320C6416为核心处理器,利用其片内存储空间大,运算速度快的特点,并利用数字信号处理支持库中提供的一系列高效手工优化函数,如滤波器和傅里叶变换,快速完成景象匹配所涉及的复杂算法,使得算法能够实时实现。实验结果表明:基于定点DSPTMS320C6416的图像处理系统可以实时完成旋转和缩放下的景象匹配技术,其处理时间小于40 ms,旋转测量精度满足要求。
In order to solve the problem of rotated and scaled scene matching in the hardware platform of real-time implementation, software and hardware design were presented for the real-time scene matching system. The traditional Fourier-Mellin transform （FMT） had been improved to suit the hardware platform, and reduce the influence of details information on the calculation results. On the other hand, using DSP TMS320C6416 processor as the core, which possessed the property of high processing speed and a large on-chip memory, it also provided a set of high efficient manual optimization functions commonly used in signal processing applications, such as filtering and FFF, it made the involved complex algorithms realize on the hardware real-time. Experimental results show that DSP TMS320C6416 based real-time scene matching processing can be realized, the processing time is less than 40 ms and the rotation measurement accuracy meets the requirement.
出处
《红外与激光工程》
EI
CSCD
北大核心
2011年第8期1576-1581,共6页
Infrared and Laser Engineering
基金
国家自然科学基金青年基金(60902067)
