线列探测器实时非均匀校正技术研究

Real-time non-uniformity correction for linear sensor

目前线列探测器仍在一些重大工程项目中广泛应用,且多采用两点定标法实现非均匀性的校正。两点定标法需要不断定标以更新校正系数。为了研究能够实现实时校正的方法,主要针对线列探测器的非均匀性特点,对目前几种简单的自适应非均匀校正算法进行分析,并给出了 VC6.0++中对 MODIS 数据的仿真结果。实验表明,恒定统计量算法中参数的选择很难确定,神经网络算法并不适于线列探测器的非均匀校正,重点分析了 SPRITE 热像仪所采用的归一法,并对其加以改进,使硬件实现更简单、快速。最后,给出了一种采用 DSP 和 CPLD 的硬件实现方案,可以完成线列探测器的实时非均匀校正。

Linear sensors are widely used in most significant applications,however,the conventional two-point correction technique is often used for non-uniform compensation,which requires two calibration points updated periodically.To find an adaptive and real-time correction algorithm,this paper analyzes several methods which are easy to implement and presents the simulation results using MODIS images with VC6.0++.Results show that the parameters for constant-statistic algorithm are hard to choose,Neural network method is not effective for linear sensors and the method used in SPRITE has a better performance.Finally,some improvements to facilitate the real-time hardware implementation are discussed and a basically structured system using CPLD and DSP is introduced.