Co-registration of aerial imagery and Light Detection and Ranging (LiDAR) data is quite challenging because the different imaging mechanisms produce significant geometric and radiometric distortions between the two multimodal data sources. To address this problem, we propose a robust and effective coarse-to-fine registration method that is conducted in two stages utilizing spatial constraints and Gabor structural features. In the first stage, the LiDAR point cloud data is transformed into an intensity map that is used as the reference image. Then, coarse registration is completed by designing a partition-based Features from Accelerated Segment Test (FAST) operator to extract the uniformly distributed interest points in the aerial images and thereafter performing a local geometric correction based on the collinearity equations using the exterior orientation parameters (EoPs). The coarse registration aims to provide a reliable spatial geometry relationship for the subsequent fine registration and is designed to eliminate rotation and scale changes, as well as making only a few translation differences exist between the images. In the second stage, a novel feature descriptor called multi-Scale and multi-Directional Features of odd Gabor (SDFG) is first built to capture the multi-scale and multi-directional structural properties of the images. Then, the three-dimensional (3D) phase correlation (PC) of the SDFG descriptor is established to detect the control points (CPs) between the aerial and LiDAR intensity image in the frequency domain, where the image matching is accelerated by the 3D Fast Fourier Transform (FFT) technique. Finally, the obtained CPs not only are employed to refine the EoPs, but also are used to achieve the fine registration of the aerial images and LiDAR data. We conduct experiments to verify the robustness of the proposed registration method using three sets of aerial images and LiDAR data with different scene coverage. Experimental results show that the proposed method is robust to geometric distortions and radiometric changes. Moreover, it achieves the registration accuracy of less than 2 pixels for all cases, which outperforms the current four state-of-the-art methods, demonstrating its superior registration performance.

航空影像和光探测与测距 (LiDAR) 数据的联合配准非常具有挑战性，因为不同的成像机制会在两个多模态数据源之间产生显着的几何和辐射失真。为了解决这个问题，我们提出了一种鲁棒有效的粗到细配准方法，该方法利用空间约束和 Gabor 结构特征分两个阶段进行。在第一阶段，将 LiDAR 点云数据转换为强度图，用作参考图像。然后，粗配准是通过设计基于分区的特征从加速分段测试 (FAST) 算子中提取均匀分布的航拍图像中的兴趣点，然后使用外部方向参数 (EoP) 基于共线性方程执行局部几何校正来完成的. 粗配准旨在为后续的精细配准提供可靠的空间几何关系，旨在消除旋转和尺度变化，以及使图像之间仅存在少量平移差异。在第二阶段，首先构建一个称为奇伽柏多尺度和多方向特征（SDFG）的新特征描述符来捕获图像的多尺度和多方向结构特性。然后，建立 SDFG 描述符的三维 (3D) 相位相关 (PC) 以检测频域中航空和 LiDAR 强度图像之间的控制点 (CP)，其中图像匹配通过 3D 快速傅立叶变换加速(FFT) 技术。最后，获得的 CP 不仅用于细化 EoP，还用于实现航拍图像和 LiDAR 数据的精细配准。我们使用三组航拍图像和具有不同场景覆盖范围的 LiDAR 数据进行实验，以验证所提出的配准方法的鲁棒性。实验结果表明，所提出的方法对几何失真和辐射变化具有鲁棒性。此外，它在所有情况下都实现了小于 2 个像素的配准精度，优于当前的四种最先进的方法，