无人机遥感图像几何畸变校正全过程控制方法研究

Research on the Whole Process Control Method for Geometric Distortion Correction of Unmanned Aerial Vehicle Remote Sensing Images

摘要:

几何畸变是一种常见的像素点突变行为，在无人机遥感图像中，像素点几何畸变行为的表现能力越强，遥感主机对无人机图像的控制能力越弱。为准确校正像素点几何畸变行为，提升遥感主机对无人机图像的控制能力，针对无人机遥感图像几何畸变校正全过程控制方法展开研究。提取无人机遥感图像的几何描述符，并定义畸变像素点的尺度空间，完成对畸变像素点的配准处理。针对几何畸变像素点实施重采样，通过计算畸变校正参数的方式，求解校正处理函数，实现对无人机遥感图像几何畸变的校正。利用畸变校正交点，确定图像几何畸变区域的覆盖范围，再联合控制对象标定条件，完成对无人机遥感图像几何畸变校正的全过程控制。实验结果表明，所提控制方法的应用，可以避免几何畸变节点处的像素值出现明显变化的情况，畸变前后的像素差不超过5.0pt，能够实现对无人机遥感图像中像素点几何畸变行为的精准校正，符合实际应用需求。

Abstract:

Geometric distortion is a common mutation behavior of pixels, and the more powerful the performance of pixels in the remote sensing image of the drone, the weaker the control capability of the remote sensing console for the image of the drone. In order to accurately correct the geometric distortion behavior of pixel points and enhance the control ability of remote sensing hosts over drone images, research has been conducted on the entire process control method for geometric distortion correction of drone remote sensing images. Extract the geometric descriptors of drone remote sensing images and define the scale space of distorted pixels to complete the registration process of distorted pixels. Implement resampling for geometric distortion pixels, solve the correction processing function by calculating distortion correction parameters, and achieve the correction of geometric distortion in unmanned aerial vehicle remote sensing images. By using distortion correction intersection points to determine the coverage range of geometric distortion areas in the image, and then jointly controlling object calibration conditions, the entire process of geometric distortion correction for unmanned aerial vehicle remote sensing images is completed. The experimental results show that the application of the control method can avoid significant changes in pixel values at geometric distortion nodes, and the pixel difference before and after distortion does not exceed 5.0pt. It can achieve accurate correction of pixel geometric distortion behavior in unmanned aerial vehicle remote sensing images, which meets practical application requirements.