红外图像处理算法硬件实现验证系统

A Verification System for the Hardware Implementation of IR Image Processing Algorithm

由红外探测器采集得到的原始红外图像需要经过各种硬件算法处理才能使用,而这些硬件算法受限于FPGA和DSP的系统资源上限往往需要进行必要的优化和精简,导致处理效果相较于上位机端同功能的软件算法大打折扣。提出了一种软硬件算法对比平台,通过全双工千兆网络联系上位机和硬件系统,定性对比移植后的硬件算法和上位机软件算法之间的差异,以此作为改善硬件移植后的算法的依据,并对硬件移植后算法的性能进行评价。主要做法为:首先通过千兆网络将探测器输出的原始红外视频源传输到上位机预存,进而上位机将预存的原始红外视频发给硬件处理板进行硬件算法处理并回接处理结果,最后上位机将预存的原始红外视频通过同功能软件算法处理后与硬件处理后的红外视频逐帧进行差值处理,定性得出硬件算法优劣。

Original infrared images obtained by the infrared detector needs to be processed by a variety of hardware algorithms. However, these hardware algorithms are limited by the system resources in the FPGA and DSP and often need necessary optimization and streamlining, which leads to great processing effect reduction compared to the software algorithm with same function in host computer side. This paper presents a hardware and software algorithm comparison platform, which connect the host computer and hardware systems through a full-duplex Gigabit network and make a qualitative comparison between the transplanted hardware algorithm and the PC software algorithm. Based on the comparison we can get the basis of the improvement for the transplanted hardware algorithm and evaluate the performance of it. Firstly, the original IR images obtained by the IR detector is transferred to the host computer through the Gigabit network, and then the host computer will transmit the stored original IR video to the hardware processing board to be processed by hardware algorithm and receive back the processed results. Finally, host computer will process the original IR image using software algorithm with same function and compare it with the hardware algorithm results by means of frame-by-frame difference processing, and qualitatively determine the pros and cons of the hardware algorithm.