空间邻近目标的存在对红外传感器的信号处理提出了超分辨的新要求。通过对红外焦平面的目标成像进行建模,推导了基于最小二乘准则的超分辨目标函数,针对传统最优化方法,对起始估计位置要求高、对高维目标函数计算复杂的缺点,引入粒子群优化算法,优化超分辨目标函数,联合估计出目标在红外焦平面的投影位置和辐射强度,实现对空间邻近目标的红外超分辨。结果表明,在模型最小二乘准则下,基于粒子群优化的超分辨算法性能优于传统的最陡下降法,具备更强的超分辨能力。

The closely spaced objects (CSOs) create blur imprints on infrared focal plane,which make it necessary for information processing to have a super-resolution. Objects′ imaging on infrared focal plane is modeled,and a super-sesolution objective function based on least square criterion is presented. Traditional optimization methods have to be carefully initialized otherwise they will get poor estimation performance and suffer from large computation load. So a particle swarm optimization (PSO) algorithm is introduced to optimize the super-resolution objective function,and jointly estimate the projection position and radiant intensity of targets on the focal plane,then realize the super-resolution of CSOs. Simulation results show that the least square-based PSO algorithm gains superior performances than that of the traditional steepest descent method and possesses the better capability of super-resolution.