顾及畸变的地质图像三角网分区地理配准及精度评价

Geographic alignment and accuracy evaluation of triangular network partitions of scanned geologic images considering aberrations

中文关键词: 地质图校正  地理配准  图像分割  精度评价  Delaunay三角网中图分类号:P283.2文献标志码:A文章编号:1674-7801(2025)02-0362-09

英文关键词: geological map correction  geographic alignment  image segmentation  precision evaluation  Delau-nay triangular mesh.

中文摘要:

      针对大范围扫描地质图地理配准中存在的地图畸变复杂、转换参数描述与控制点选取困难的问题，提出了一种边缘检测和分区配准相结合的扫描地质图配准算法。该算法使用 Canny检测和 Bessel函数进行图廓降噪，再由最小二乘法拟合图廓并校正畸变，最后根据图像变形情况使用 Delaunay三角网分割图像，逐一解算各区域配准参数，拼接重建精确的地理配准图像。结果表明:(1)图像畸变校正后，40个图廓点横向、纵向平均偏差分别降低 52.90%、77.81%；(2)相比传统的相似多项式、仿射变换、三阶多项式、样条函数配准，本文方法点要素配准精度分别提高了 22.64%、22.07%、14.26%、1.76%，线要素配准精度提高了 49.73%、51.25%、37.01%、26.54%，表明本文方法配准精度更高，实现了顾及畸变的高保真精确地理配准。

英文摘要:

      Aiming at the problems of complex map distortion, difficult description of transformation parametersand selection of control points in geographic alignment of large-scale scanned geological maps, an alignment algo-rithm for scanned geological maps combining edge detection and zonal alignment is proposed. The algorithm usesCanny detection and Bessel function for contour noise reduction; then the least squares method is used to fit the con-tour and correct the distortion; finally, according to the deformation of the image, the image is segmented by Delau-nay triangular mesh, and the alignment parameters of each region are solved one by one to reconstruct the accurate geo-aligned image by splicing. The results show that: (1) After the image distortion correction, the average horizon-tal and vertical deviation of 40 contour points are reduced by 52.90% and 77.81%, respectively; (2) Compared withthe traditional similar polynomials, affine transforms, third-order polynomials, and spline functions, the accuracy ofthe point elements of this paper's method improves by 22.64%, 22.07%, 14.26%, and 1.76%, and the accuracy ofline elements improves by 49.73%, 49.73%, and 1.76%, respectively. The accuracy is improved by 49.73%,51.25%, 37.01%, and 26.54%, which indicates that the method in this paper has higher alignment accuracy andachieves high-fidelity accurate geographic alignment considering aberrations.