基于FPGA+DSP架构的红外图像线性目标跟踪

IR Linear Object Tracking Based on FPGA+DSP System

为满足电力巡线中对输电线实时跟踪的需要,利用FPGA+DSP的成熟架构,提出了一种新的红外线性目标跟踪方法,根据输电线的特征,将传统的二维跟踪转换为一维搜索跟踪。首先利用FPGA的并行计算能力对图像进行预处理滤除部分噪声的影响,然后利用DSP对图像作Sobel卷积求得图像的梯度并归一化和阈值化,最后依据红外线性目标的方向特征对特定方向的梯度求重心得到跟踪坐标。工程应用表明,该方法能够实时稳定地跟踪红外线性目标并具有较强的抗干扰能力。

In order to meet the needs of real-time tracking of power lines in power line inspection,the infrared image（IR）post-processing technique and linear object tracking are researched based on FPGA＋DSP system in this paper.According to the characteristics of the power line,the traditional 2Dtracking is converted to one dimensional search and tracking.Firstly,a Gaussian smoothing algorithm is deployed to preprocess the IR images by FPGA＇s parallel processing characteristic,and then it utilizes the extremely strong digital computing capability of DSP to obtain the normalized gradient of the image by Sobel convolution.Finally,object＇s coordinate position is obtained by calculating gradient barycenter with directional feature of linear object.The practical application shows that it can track the linear target stably with strong anti-jamming capability.