Abstract: Aerial photography of photovoltaic components often suffers from distortion due to shooting angle issues, which affects the measurement, recognition, and classification of elements like straight lines, angles, and proportions. To address this problem, this paper proposes an undistorted aerial photovoltaic component image correction method that requires no calibration. The approach first extracts four critical feature points from the photovoltaic component image using a feature point extraction algorithm and then matches them using a feature point matching algorithm to obtain their corresponding relationships before and after distortion. Secondly, the theoretical principles of computational geometry are utilized to compute the parameters T necessary for distortion correction through triangulation and coordinate transformation of the corresponding points. Finally, distortion correction is carried out on the image using parameter T to obtain the corrected photovoltaic component image. This algorithm not only improves the visual correction effect of the image but also increases the accuracy of data by 20%~55%.This method has a high application value in the foreign object recognition, operation and maintenance of the aerial photovoltaics module image.

Key words: photovoltaic module, aerial images, distortion correction, backward mapping