论文部分内容阅读
针对红外序列图像中运动弱小点目标的检测问题,提出了一种多组独立检测方法。首先,利用神经网络优化的形态学算子进行背景抑制,基于自适应门限实现单帧目标检测。在多帧检测中,先将可能的航迹观测序列进行分组累加,然后进行似然比检验。由于多组独立检测考虑了信噪比过低或者强噪声干扰的影响,一定程度上加速了航迹的确认和删除,提高了多帧检测的性能。基于此提出了多组独立检测方法,并对算法性能进行了详细分析。实测数据结果证明:在相同虚警概率情况下,多组独立检测法的检测性能优于截断序贯处理算法的检测性能。
Aiming at the detection of moving weak dot targets in infrared sequence images, a multi-group independent detection method is proposed. Firstly, the background is suppressed by the morphological operator optimized by neural network, and the single-frame target detection is achieved based on the adaptive threshold. In multi-frame detection, the possible track observation sequences are grouped together first, and then the likelihood ratio test is performed. Since multiple sets of independent tests consider the impact of low signal-to-noise ratio or strong noise interference, the confirmation and deletion of tracks are accelerated to a certain extent, and the performance of multi-frame detection is improved. Based on this, several sets of independent detection methods are proposed, and the performance of the algorithm is analyzed in detail. The experimental results show that the detection performance of multiple sets of independent detection methods is superior to the detection performance of truncated sequential processing algorithms under the same false alarm probability.