A small target detection method based on an unmanned aerial vehicle

Abstract

The invention provides a small target detection method based on an unmanned aerial vehicle, and the method comprises the steps: enabling the unmanned aerial vehicle to fly autonomously to shoot a video; Sequentially reading two frames in the video sequence, calculating a Hessian matrix, and extracting key points in the image; Calculating the Euclidean distance between the key points, and matchingthe key points in the two images according to the Euclidean distance and the Hessian matrix; removing noise points in the key points, and calculating and calibrating a perspective transformation matrix between every two frames of images; And carrying out subtraction on the calibrated current frame and the reference frame to obtain a difference image, carrying out binaryzation, morphological operation, reconstruction and the like on the difference image in sequence, and extracting and marking a foreground target in the image. According to the method, small targets in the aerial video can be effectively extracted and detected through similarity of backgrounds, meanwhile, computing resources of the unmanned aerial vehicle are fully used, and the online target detection capability of the unmanned aerial vehicle is improved.

Description

technical field [0001] The invention relates to an unmanned aerial vehicle and target online detection technology, in particular to a small target detection method based on an unmanned aerial vehicle. Background technique [0002] Unmanned Aerial Vehicles (UAVs) are unmanned aircraft operated by radio remote control equipment and independent program control equipment, or completely or intermittently autonomously operated by an onboard computer. Compared with manned aircraft, drones have the advantages of small size, low cost, low environmental requirements, and strong survivability, and are often more suitable for tasks that are dangerous and harsh. With the rapid development of the UAV manufacturing industry, UAV systems are widely used in smart city management, intelligent traffic monitoring and other fields. Among them, the monitoring of moving targets is a basic but challenging functional requirement in UAV systems, which is closely related to applications such as infra...

AI Technical Summary

Problems solved by technology

However, in high-altitude environments, objects will have relatively small sizes, different types, and variable orientations, which make a priori features difficult to detect

[0005] Recently, the target detection and tracking method based on deep learning is a hot spot of research, including SSD, RCNN, YOLO, etc., but the multi-layer nonlinear processing for feature extraction and learning has high computing and hardware costs, which is not applicable. UAVs with limited computing resources

In addition, due to the free flight of the drone, how to solve the unstable shooting of the motion camera is also a difficult problem

In order to solve this problem, the common method is to use the optical flow method to compensate the camera motion, and then use the clustering algorithm to segment the target, however, the optical flow method is very sensitive to noise and illumination

In addition, compared with the UAV flight situation, the optical flow method has a higher computational complexity, which is more suitable for target detection when the UAV is hovering or flying at a low speed, and is not suitable for the online detection system based on the UAV.

Images