Image stitching technology for police drones using an improved image registration method incorporating ORB algorithm

Abstract

Drone aerial photography technology can assist public security departments in conducting on-site investigations and evidence collection. The single-aerial image captured by drones has limited field of view. Due to the influence of reconnaissance scenes, there are various shooting angles and overlapping areas. An improved image registration algorithm was proposed based on ORB, which combined segmentation and denoising processing to evenly distribute ORB feature points throughout the entire drone image. On this basis, Laplace fusion algorithm was used to match the overlapping regions. The study compared the improvement of ORB with several other methods. Three sets of experiments were compared and analyzed with the other three different fusion algorithms. These experiments confirmed that compared with existing methods, the improved image registration method had significant improvements in accuracy and speed. The improved ORB registration algorithm was superior to traditional ORB in registration accuracy and registration speed, with a registration rate increase of 7.4% and a time reduction of 153.97 seconds. This method not only effectively improves the splicing effect, but also handles the details of the splicing seam well, providing a basis for police to analyze the scene.