The rapid development of unmanned aerial vehicle (UAV) technology has opened up aerial transport channels for urban material transportation and high-rise emergency rescue. During UAV delivery missions, the distance between the UAV and the receiving platform directly determines whether goods can be safely delivered. However, in enclosed spaces such as balconies and windows between floors, conventional global positioning system (GPS) methods have errors of 1 m~3 m. Moreover, in complex environments characterized by low visibility, strong light interference, and complex terrain, traditional visual ranging methods struggle to achieve the required levels of high precision and high reliability. The monocular vision-based UAV beacon imaging and ranging method offers a novel approach for close-range UAV missions, owing to its low computational complexity, distinct visual features, and minimal environmental dependency. However, conventional beacons are susceptible to factors like illumination changes and occlusion in complex environments, making it difficult to ensure ranging accuracy. Fluorescent beacons have clear boundaries and shapes, with enhanced contrast in complex backgrounds, facilitating easier identification of beacon features. This paper uses nested fluorescent beacons as feature markers for UAV imaging ranging. By improving the clarity of beacon contour detection and the accuracy of beacon feature point identification, imaging ranging errors are reduced, thereby providing a theoretical reference for safe UAV operations in complex environments.

An imaging ranging method based on nested fluorescent beacon detection (NFBD) was developed. This method utilized the unique optical characteristics of nested fluorescent beacons to calculate the relative distance from the UAV to the target point. The specific process was as follows: for beacon edge detection, color space conversion and color threshold segmentation strategy, as well as multi-scale detection strategy, were incorporated into the conventional edge detection algorithm to improve the clarity of fluorescent beacon edge detection. For beacon feature point extraction, an improved scale-invariant feature transform (SIFT) detection algorithm was implemented, incorporating feature point screening and adaptive threshold segmentation based on local median. This enhancement improved the accuracy and illumination adaptability of fluorescent beacon feature point detection while reducing ranging errors. For imaging distance calculation, a multi-frame image data fusion approach employing weighted averaging was adopted to improve ranging accuracy and stability. Finally, by processing the nested fluorescent beacon detection images, distance measurement from the UAV to the target point was completed.

First, anti-interference detection was conducted for the nested fluorescent beacon detection algorithm. Simulations and experiments were performed to analyze the proposed NFBD algorithm in comparison with traditional beacon feature point detection algorithms. The simulation results (Fig.7) and experimental results (Fig.8) demonstrated that the detection accuracy of the proposed NFBD algorithm was higher than that of traditional methods. In practical detection, the NFBD algorithm demonstrated improvements in anti-interference performance for beacon detection by 4.5%, 18.2%, and 19.9% compared to the SIFT, speeded up robust features (SURF), and oriented fast and rotated brief (ORB) algorithms, respectively. Subsequently, experiments were carried out to evaluate the ranging accuracy of the UAV beacon imaging. Within the experimental distance range of 0.5 m to 3.0 m, the relative ranging error of the NFBD algorithm ranged from 0.2% to 11.1% (Fig.10). This corresponded to an improvement in beacon imaging ranging accuracy of 3.2%, 4.2%, and 4.6% over the SIFT, SURF, and ORB algorithms, respectively. Under varying illumination conditions, the relative error of the proposed algorithm across three testing scenarios remained consistently between 0.2% and 11.5%, and its fluctuation range was significantly smaller than those of the benchmark algorithms (Fig.13). When the beacon was occluded, at occlusion ratios of 25%, 50%, and 75%, the NFBD algorithm exhibited ranging accuracy improvements of 3.8%, 4.4%, and 6.1% over the SIFT algorithm, 4.2%, 5.0%, and 6.9% over the SURF algorithm, and 4.5%, 5.6%, and 10.3% over the ORB algorithm, respectively. These results verified the occlusion adaptability of the proposed algorithm (Fig.16).

This study proposes an imaging ranging method based on nested fluorescent beacon detection. The method enhances the clarity of beacon edge detection by introducing color space conversion, color threshold segmentation, and a multi-scale detection strategy. It improves beacon detection accuracy and illumination adaptability through feature point screening and adaptive threshold segmentation based on local median. Additionally, the method increases ranging accuracy and stability by employing a multi-frame image data fusion approach with weighted averaging. Compared with traditional SIFT, SURF, and ORB algorithms, the proposed algorithm exhibits improvements of 4.5%, 18.2%, and 19.9% in anti-interference performance in beacon detection, and 3.2%, 4.2%, and 4.6% in beacon imaging ranging accuracy, respectively. Furthermore, the proposed algorithm demonstrates superior illumination adaptability and occlusion adaptability. This method provides accurate distance information for safe UAV operations.