Image moving target real-time detection method based on unmanned aerial vehicle platform

Abstract

The invention discloses an image moving target real-time detection method based on an unmanned aerial vehicle platform. The method comprises the following steps of S1, background estimation, S2, threshold segmentation of a moving target image, S3, morphological filtering of the moving target image, S4, region expansion, S5, target identification, S6, establishment of a target trajectory chain, S7,refreshing of target data, S8, target tracking and S9, target prediction. According to the image moving target real-time detection method based on the unmanned aerial vehicle platform, the moving target in the scene can be effectively detected in real time no matter whether the unmanned aerial vehicle is in a moving state or a static state, whether the lens of the camera is zoomed or not and whether the target suddenly moves or not.

Description

technical field [0001] The invention relates to a real-time detection method of an image moving target based on an unmanned aerial vehicle platform. Background technique [0002] There are three methods of moving target detection for UAVs and their advantages and disadvantages: [0003] Background Subtraction: Considering the current frame and the background for foreground segmentation. Primarily useful for static or pan zoom cameras or known backgrounds. This kind of method is to make the foreground and background have a reasonable boundary, and the eventful background of the scene is known or can be modeled, and there are few studies on multi-object segmentation. For example, structure-from-motion (SFM) methods estimate camera parameters, sparse 3D points, and depth maps. Although using this method can make the obtained foreground masks and moving object boundaries accurate, they are limited to scenes with large differences in foreground and background depths, so they a...

AI Technical Summary

Problems solved by technology

In addition, applying the mean shift algorithm to optimize the foreground segmentation can neither guarantee the real-time performance of the segmentation nor the consistency of the segmentation boundary

Other motion segmentation methods (such as Hierarchical Directed Acyclic Graph or Maximum Weight Clique), which are not necessarily designed to detect multiple independently moving objects, their performance will degrade when objects move suddenly

Images