Research on automatic quantitative quality inspection method for internal geometric accuracy of remote sensing images

• Published in: Journal of applied remote sensing Vol. 19; no. 1; p. 016511
• Main Authors: Ge, Huibin, Li, Ying, Wang, Boyu, Geng, Yu, Ba, Xiaojuan
• Format: Journal Article
• Language: English
• Published: Society of Photo-Optical Instrumentation Engineers 01.01.2025
• Subjects: triangulation; image internal accuracy; error visualization; automated quantitative quality inspection
• ISSN: 1931-3195; 1931-3195
• DOI: 10.1117/1.JRS.19.016511

The geometric accuracy of remote sensing images is very important to the mapping accuracy of land surveying and mapping industries, and its internal geometric accuracy affects the accuracy of length and area measurement. Therefore, it is of great significance to automatically screen out remote sensing images with acceptable accuracy. We propose a quantitative quality inspection algorithm for internal geometric accuracy based on automatic image matching to construct a triangulation network, which uses the automatic matching checkpoints to construct a triangulation network, characterizes the internal error between two points of the line segment by calculating the relative error of each line segment that composes each triangle, and finally counts the middle error of the relative accuracy of all line segments in the triangulation network to characterize the internal geometric accuracy of the image. Taking the remote sensing images of plains, hills, and mountains as experimental data, the experimental results show that the quality inspection results of the proposed method are consistent with those of the manual quality inspection method, and the deviation of the quality inspection accuracy is less than 0.3 pixels. The method indicates that it has nearly achieved the objectivity of manual quality inspection. With its characteristics of quantification, automation, and batch processing, this method holds significant practical importance for the internal geometric accuracy quality inspection of massive remote sensing images. In addition, the error results are rendered by color grading using segmentation thresholds, which can quickly and intuitively identify the internal geometric accuracy quality of different areas of the image.