水下三维声纳目标在线运动监测与识别

随着声学探测在海洋资源开发中的广泛应用,水声成像技术已成为水下目标监测的重要手段,文章提出了一种基于三维声纳技术的在线运动目标识别方法。通过对三维声学图像进行网格搜索和三角面片连接,进行单帧三维声学图像的多层实时重建,实现单帧图像内目标的重建、聚类与标示。结合GPS定位仪和姿态仪信息,修正位移和姿态变化引起的运动误差,利用反向投影和最近点搜索方法查找相邻图像帧之间两两匹配的控制点对,进行相邻图像帧的快速配准。根据配准矩阵将相邻图像帧的的各个目标转换到同一全局坐标系中,提取有效的声学目标特征变化相对值,并评估特征权重,实现相邻图像帧之间运动目标的在线检测与识别。通过室内水池和湖试实验,结果表明该方法能有效地实现三维声学图像在线运动目标实时识别。

On-line motion recognition for underwater 3-D sonar object

Aiming at the real-time detection requirement of underwater moving object, an on-line motion recognition method for underwater 3-D sonar object is proposed. Multi-layer reconstruction and object clus-tering of single sonar frame are fulfilled by grid searching and triangle connection. Motion error is correct-ed by GPS and POSE instrument due to the change of displacement and pose. 3-D sonar imaging registra-tion is achieved by reverse projection and the closest point searching. According to the registration matrix, all sonar frames are translated into a global coordinate system, and sonar object features are extracted. Ef-fective moving object detection is achieved by estimating feature weight. Experimental results in the indoor pool and lake show that the system can efficiently achieve 3-D sonar data on-line moving object recogni-tion.