设计了一种基于FPGA+DSP+ASIC结构的红外图像小目标实时检测系统。该系统将红外图像的非均匀性校正、多级滤波等计算强度高、具有数据流结构特点的成熟子算法用硬件方式实现;将需要较大存储空间、抽象层次高的算法任务分配给DSP完成,充分发挥各类型器件的优势。测试表明,算法和硬件平台做到了很好的吻合,系统在处理速度、可重构性等方面达到良好的平衡和统一。
This paper presented a design of hardware platform based on FPGA + DSP + ASIC for small target real-time detection in infrared image. In the system,realized the algorithms such as UNC and MLF which had data-flow structure by FGPA or ASIC. On the other hand, DSPs realized other algorithms which needed lots of memory space and whose abstracting level was high. In this way, could make use of the devices' advantage and avoid the disadvantage. The test shows the hardware platform and the algorithm match very well. The system reaches a good balance among the processing speed and stability and reconfiguration.