红外弱小目标多组独立检测方法

Multi-group independent detection for infrared small dim target

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。

A multi-group independent detection method was proposed for detecting the moving spot target in infrared sequence images. The morphological filter optimized by neural network was used to suppress background information, and single-frame target detection was realized by the adaptive threshold method. In multi-frame detection, firstly, the possible trajectory of the observation sequence was grouped and calculated, then maximum likelihood test was carried out. Since the multi-group independent detection took account of the low SNR or strong noise, the determination and elimination of trajectory were quickened in certain degree, and the performance of the multi-frame detection was increased. Here the method of multi-group independent detection was proposed, and the detection performance of grouping sequential test was analyzed. Experimental results show that the detection probability of the proposed method is higher than that of the truncated sequential test with same false alarm probability.