天基红外扫描图像点目标检测算法

Algorithm of space point target detection for IR scan images

针对天基红外扫描型探测器的成像特点(时间延迟积分和半像元错列对准),提出了一种基于自适应背景预测的红外扫描图像点目标检测算法。首先,采用递归背景估计的背景预测模型,利用最速下降法求解滤波系数。其次,对背景去除后的残差图像进行自适应门限探测,并对过门限图像进行双向脉冲匹配以抑制噪声,提取目标。最后,采用蒙特卡罗方法对算法性能进行了仿真分析。实验结果表明:当信噪比大于3时,目标检测概率可达99.5%(虚警1.07×10-2)。算法实时性分析表明:算法处理能力为31.37Mb/s。

According to the imaging characters of space infrared scan sensor, such as time-delay integration and half-pixel mis-alignment, a point target detection algorithm for IR scan images was proposed based on the adaptive background prediction. Firstly, the background prediction was modeled by the recursive background estimation, and the filter coefficients were computed by the steepest descent optimization procedure. Secondly, the adaptive threshold was determined on the residual image after the background elimination. Then, the target pulse matching was performed on the thresholded image in cross- scan and in-scan direction respectively to suppress noise and extract target. Finally, the algorithm detection performance was simulated using Monte Carlo method. The experimental results show that the detection probability reaches 99.5%（probability of false alarm 1.07×10^-2） as the input SNR is no less than 3. The real-time analysis of the algorithm shows that the capability of data processing is 31.37 Mb/s.