为了实现高效自动目标检测,提出了一种可用于低信噪比条件下的空间可见光弱小目标检测算法.首先,建立空间光学图像模型,利用恒虚警率(CFAR)方法确定分割系数对单帧图像背景进行分割；然后,基于恒星结构稳定特性构建距离特征空间,并针对特征空间构造分类准则函数,使用迭代最优化分类方法提取出候选目标点；最后,依据目标运动轨迹的连续性建立空间目标轨迹关联、合并以及虚假目标轨迹删除规则,进行轨迹处理,实现空间可见光弱小目标的检测.文中还提出了单帧检测率、虚警率与序列检测率、虚警率相结合的评价方法.实验结果表明:在低信噪比条件下(SNR≤3),得到的序列检测率达到96.02％A以上,序列虚警率达到4.4％以下.该方法在低信噪比条件下显著提高了目标检测率,并有效抑制了虚警.%To realize automatic target detection, an algorithm is proposed to detect small visible optical space targets against low SNR conditions. Firstly, the single-frame image background is segmented, and the segmentation coefficient is determined by a Constant False Alarm Ratio (CFAR) method. Then, a feature space is formed based on structural stability of the star, and classification criterion function is constructed for the distance feature space. Furthermore, candidate targets are extracted by using the iterative optimization distance classification method. Finally, small visible optical space targets are detected by trajectory association based on the continuity of target motion. In addition, an e-valuation method combined with single frame detection probability, single frame false alarm probability and sequence detection probability is proposed. Experimental results indicate that the detection probability of sequence is more than 96. 02%, and the false alarm probability is less than 4. 4% whenthe SNR≤3. It concludes that the method can promote the detection probability against low SNR conditions significantly, and can remove the false alarm effectively.
展开▼
