Image Matching of Low Altitude UAV Visual Navigation Methods: Current Situation and Prospect

The visual navigation technology of low-altitude UAV aims to match the real-time image taken by sensor with the existing image features in database, so as to determine the current position of the UAV and realize the control of the flight trajectory. Among visual navigation procedure, image matching is the most important step. Current image matching methods are divided into traditional methods and deep learning-based methods. This paper takes the classification of image matching techniques and methods as the breach to grasp the research status of image matching methods in low-altitude UAV visual navigation. First of all, the article analyzes steps of image matching and key problems of this field. Secondly, the paper introduces traditional methods of image matching and the method based on deep learning, summarizing the advantages and disadvantages of different methods. Then, the paper summarizes mainstream datasets used in low-altitude UAV visual navigation tasks and compares the performance indexes of different methods. Finally, the article generalizes the existing problems in advanced image matching methods to point out the possible development direction and trend, so as to provide reference for subsequent researchers to get further research.