Image processing method for centimeter-level precise landing of UAV (unmanned aerial vehicle)

Abstract

The invention provides an image processing method for the centimeter-level precise landing of a UAV (unmanned aerial vehicle), and the method comprises the steps: making a target recognition pattern,enabling the UAV to fly to a place nearby a landing position, capturing an image, enabling a rotary PTZ camera to scan the ground surroundings if there is no target image recognized in a camera view at a current position, searching the target image, and adjusting the UAV to move above the landing point to start the form landing; recognizing the most apparent big rectangle at first in a landing process from a high place; recognizing three small rectangles again after the UAV descends to a certain height, enabling the UAV to land on the ground for recognizing a QR code at the center of the pattern best, determining the center position of the QR code after the target QR code is obtained, obtaining the horizontal distance at the centimeter-level precision, and achieving the centimeter-level precise landing. The method provided by the invention is accurate and reliable in recognition, is high in adaptability, has the centimeter-level precision, and solves the problems that a method for recognizing a QR image from a high place is limited in recognition height, is difficult to adapt to different brightness values and is weak in anti-interference capability.

Description

technical field [0001] The invention belongs to the UAV landing technology, in particular to an image recognition-based precise UAV landing method, in particular to an image processing method for centimeter-level precise UAV landing. Background technique [0002] The conventional method is to rely solely on GPS positioning to assist landing, but civilian GPS measurement errors range from 0.5 meters to several meters, and are easily affected by interference; further, using the current advanced RTK positioning technology to assist landing can provide the smallest centimeter-level positioning Accuracy, but the cost of equipment is expensive; therefore, the software method based on image recognition is valued, and the target image is collected through the standard camera of the drone, and image recognition is performed to provide information such as the position and angle of the target landing point. The following two patent documents are typical summaries based on current image...

AI Technical Summary

Problems solved by technology

This method reduces the computing power requirements for image recognition and can achieve accurate landing under normal lighting. However, due to the fuzzy details of high-level QR codes, the recognition height is limited; at the same time, it does not consider the application when the brightness is weak such as dark and dark. scene, lack of adaptability

Images