基于四旋翼飞行器的光流法动态目标检测

Dynamic Target Detection Based on Optical Flow Method of Four-rotor Aircraft

针对四旋翼飞行器避障算法中对于运动的目标检测完整性不够,准确性及鲁棒性不高的问题,提出了一种基于金字塔Lucas Kanade光流的单目视觉避障算法,提高了光流检测到的运动目标的完整性和鲁棒性。通过引入图像动态模型建立光流约束条件,克服了因为亮度改变引起的约束方程不成立问题。实验结果表明算法在飞行器低速飞行且遮挡和亮度变化条件下的跟踪效果良好,并且显著减少了匹配搜索的时间,节省了飞行器的飞行时间。

Against the imperfect and instability for testing moving target in four-rotor aircraft obstacle avoiding algorithm, one kind of obstacle avoiding algorithm was put forward based on pyramid Lucas Kanade optical flow which had improved the optical flow obstacle avoiding detecting moving object into integrity and accura- cy. Through referencing image non-stationary model and constructing optical flow constraint, it had overcome the problem that changing brightness led to invalid construction. The experimental results showed that our algorithm worked quite effectively when four-rotor aircraft flying with low speed flight and changing brightness. It also could reduce the matching and searching time significantly, saved the aircraft flight time, and reduced the energy consumption