论文部分内容阅读
近岸红外目标检测由于背景复杂和实时性的要求,是当前研究的一个难点。基于上下文的视觉注意机制是解决这一难题的有效途径。首先根据近岸红外目标存在于海面这一先验知识,利用灰度、纹理特征及位置关系,分割出海面;其次利用基于相位谱的视觉注意机制定位显著区域;最后综合海面和显著区域位置,便可检测出近岸红外目标。实验证明,这种算法高效、准确,能够满足目标检测阶段的实时性要求。
Near-shore infrared target detection due to the background of complex and real-time requirements, is a difficult study. Context-based visual attention mechanism is an effective way to solve this problem. Firstly, based on the a priori knowledge of near-shore infrared target existing in the sea surface, the sea surface is segmented by using the gray-level, texture features and the positional relationship. Secondly, the visual attention mechanism based on phase spectrum is used to locate the salient regions. Finally, Can detect near-shore infrared target. Experiments show that this algorithm is efficient and accurate and can meet the real-time requirements of the target detection phase.