红外桥梁目标识别在多DSP系统上的并行实现

The Parallel Implementation for Bridge Recognition in Infrared Images Based on Multi-DSPs System

针对多总线多DSP实时图像识别系统 ,研究了桥梁目标识别的并行算法 .算法采用流水线技术 ,在其设计过程中重点考虑将算法划分成独立的模块化的子任务 ,并将任务均衡地分配给各处理器 .为进一步提高算法速度及流水线效率 ,对算法进行了优化 .实验结果表明本并行算法具有成本低。

A parallel algorithm for bridge recognition operated in a real time image processing system is developed based on multiple bus and multi DSPs. The pipelining technology is used in the algorithm. It is intensively considered that the algorithm is partitioned into independent and modular sub tasks, and these sub tasks are equally assigned to every DSP. The algorithm is optimized for speeding up and improving efficiency. The experiment results show that the cost of the parallel!algorithm is the best optimum with high efficiency