Satellite video target intelligent detection method based on multiple prior information constraints

Abstract

The invention discloses a satellite video target intelligent detection method based on multiple prior information constraints, and the method comprises the steps: 1, obtaining an image in a satellitevideo, and generating superpixels for the whole image based on a linear iterative algorithm; step 2, calculating boundary connectivity of each superpixel based on the human eye saliency model; 3, generating a motion heat map of the image by using a background modeling method; and step 4, constructing a cost function equation constrained by three priori information of the background, the foregroundand the motion heat map, performing optimization calculation on the cost function equation, and calculating a final saliency target map through a least square method. According to the method, multi-priori information constraints are utilized, the robustness of illumination changes and satellite staring angle subtle differences in satellite videos is greatly improved, and the calculation speed isgreatly increased while pixel-level interference is effectively resisted through calculation with superpixels as units.

Description

technical field [0001] The invention relates to the technical field of remote sensing, in particular to an intelligent detection method for satellite video targets based on multiple prior information constraints. Background technique [0002] In the early stage of remote sensing imaging development, the high-resolution aerospace optical remote sensing data obtained in my country were mainly static pictures. It is difficult to meet the effective observation of natural disasters such as forest fires, earthquakes and floods and the real-time monitoring of some sensitive areas. As a new type of earth observation satellite, video satellites continuously observe a certain area through "staring" imaging, and are especially suitable for the observation of dynamic targets. [0003] Compared with aerial imagery, vehicle pixels in satellite video sequences range from a few pixels to a dozen pixels, and also exhibit low contrast with the background. Therefore, no appearance or shape i...

AI Technical Summary

Problems solved by technology

However, due to the local differences in the dark current, quantum effect, and material structure in the CCD, there are many other noises, and the slight changes in the angle when the satellite "stares" also causes such things as illumination and angle changes in the process of shooting video by the satellite. Interference, prone to insufficient background modeling and suffers from parallax issues caused by camera motion

The saliency-based method is often subject to various interferences in satellite videos with complex backgrounds. In the case of insufficient prior information, roadblocks or conspicuous buildings on the road may be misdetected as targets.

Images