基于前景—背景可区分性评价因子的运动目标多源协同检测

在多源信息融合中,对不同源信息及处理结果的可信度度量是影响多源信息融合精确性的关键.针对可见光和热红外动目标检测问题,提出了基于F-BDEF的运动目标多源协同检测算法.F-BDEF即前景-背景可区分性评价因子,是一种无基准的运动分割质量评价因子,用来评价不同信息源(可见光/热红外)运动检测结果的好坏.实验表明:与现有融合检测算法比较,该算法具有较高的检测精度,能较好得解决光照突变、阴影、鬼影、低对比度夜晚场景等问题.

Multi-modal cooperative moving objects detection based on F-BDEF

How to evaluate the reliabilities of different image sensors and their processing results is an important problem in the field of multi-modal fusion. In this paper, we focus on multi-modal fusion moving objects detection, in which visible light and infrared image sensors are adopted. An evaluation factor named F-BDEF (Foreground-background Distinguishability Evaluation Factor) is proposed to evaluate the reliabilities of the detection results of two sensors. Then a multi-modal fusion moving objects detection based on F-BDEF is proposed, in which F-BDEF is used to distinguish between false positive alarm and false negative alarm, and to choose the more accurate detection region from visible light result and infrared result. The experiments on several datasets showed that the proposed detection method received more accurate results and could overcome many disturbances, such as sudden change of illumination, shadow, ghost, low-contrast night scene.