农业机器人田间导航参照点测距的新方法

A New Method of Distance Measurement for Agricultural Robot's Field Navigation Reference Point

传统双目视差测距法的测量值为参照点P到双目摄像机基线中心的距离,而农业机器人上的摄像头装置通常被安装在距离地面一定高度,与水平方向呈一定倾角的位置,利用传统双目视差测距法将无法得到导航参照点到机器人机身的水平距离。为此,提出了一种解决农业机器人田间导航参照点测距的新方法。通过改进的BP神经网络标定下的双目系统、改进的SIFT特征匹配,输出导航参照点在世界坐标系下的坐标X,Y),则导航参照点距机器人机身的水平距离可表达为S=(X2+Y2)(1/2)。实验表明:该方法在实际田间测试中的结果最大偏差0.479cm,最小偏差0.032cm,精度高达99%,测距过程耗时共计55ms,计算量较传统双目视差测距显著减小,具有一定的工程实用性与可行性。

The distance measured by the classical binocular parallax value is the one from the reference point P to the center of the binocular cameras baseline. But the binocular cameras on the agricultural robot are usually settled in the position that is some height from the earth and a certain angle to the horizontal. A new way to solve the measuring the distance between agricultural robot and the navigation reference point is proposed in this paper. By using the improved BP neural network and the improved SIFT to export the world coordinates( X,Y). The horizontal distance between the agricultural robot and the reference P can be expressed as S = (X2+ Y2)(1/2). Experiments show the maximum deviation is0. 479 cm,the minimum deviation is 0. 032 cm. The amount of the whole calculation is much less than the classical measuring method which is helpful to the research of agricultural robot's positioning control in the farm field.