结合点云和影像的标靶自动识别与中心提取

Automatic Target Recognition and Center Positioning Combining Point Clouds and Optical Images

针对多站地面激光点云拼接中标靶自动识别困难的问题,提出了结合光学影像和点云深度信息的标靶自动识别方法,以及模板自匹配的标靶中心提取方法,实现了标靶的自动识别和中心提取。经室内、室外两种环境真实数据的验证,该方法能自动识别布置在室内、室外场景中的标靶;对自动识别的标靶进行中心提取,精度均在2 mm之内,与人工提取结果一致,证明了方法的有效性和稳定性。该方法为基于标靶的点云自动拼接提供了良好的技术支撑。

In this paper, a method is presented to achieve automatic target recognition for terrestrial laser scanner point clouds registration. In the approach, targets are recognized by combining optical images and range information in point clouds, and target centers are acquired by using self-matching method on target intensity images. The whole process is accomplished automatically. Experiments using indoor and outdoor real datasets show that all targets are recognized without manual participation in indoor and outdoor scenes, and the accuracy of target centers is within 2 mm, which is consistent with manual results＇. The effectiveness and robustness of the proposed method is verified by the experiments. The method provides technical support for automatic point clouds registration.