Remote sensing image sequence moving target detection method based on improved ViBe background model

Abstract

The invention provides a remote sensing image sequence moving target detection method based on an improved ViBe background model. The method comprises the following steps: adding random noise in a setrange to a first frame image in a remote sensing image sequence to construct and initialize a ViBe background model; starting from a second frame image, carrying out moving target detection: based oncurrent frame and the background model difference, obtaining an initial foreground detection result; carrying out median filtering and Gaussian filtering on the initial foreground detection result toremove noise jamming; carrying out connected domain analysis on the filtering results, that is, converting pixel level processing into connected domain level processing, and carrying out preliminaryscreening on the detection result based on connected domain attributes; carrying out target level analysis: serving each connected domain in the result obtained after preliminary screening as a target, and carrying out further optimization on the detection result through ghost judgment and saliency detection; and carrying out threshold segmentation based on an Otsu method to obtain a final movingtarget detection result. The method takes detection result accuracy and calculating efficiency into both considerations, and can meet real-time processing requirements.

Description

technical field [0001] The invention belongs to the technical field of remote sensing image processing, in particular to a method for detecting moving targets in remote sensing image sequences based on an improved ViBe background model. Background technique [0002] Moving object detection is one of the most active research topics in computer vision and image processing. With the rapid development of satellite remote sensing technology, real-time observation of global hotspots through video satellites has become possible. How to effectively extract moving target information from image sequences is the basic task of video satellite remote sensing data processing, and also the basis for research on higher-level remote sensing applications (such as target tracking, object recognition and classification, etc.). Real-time monitoring and moving target extraction of hotspot areas can be used for air traffic control and abnormal target detection, which is of great significance to t...

AI Technical Summary

Problems solved by technology

The advantage of this update method is that the background model will not be affected by the error of a single image, but compared with the recursive update method, the cached image sequence occupies a large amount of memory and the amount of calculation is relatively large, but the detection effect is average.

[0006] Due to the large amount of data that needs to be processed for video satellite moving target detection and the complex scene content, in the traditional background difference method, the non-recursive update method with a large memory footprint and the recursive update method with a more complex model are not applicable; The simple recursive update method, such as the ViBe background model, although it meets the requirements of calculation speed, has the problem of poor detection results, which needs to be improved in a targeted manner

Images