Detection And Speed Estimation Of Moving Vehicle From Remote Sensing Images

Target detection from optical remote sensing images is an important researchbranch of space information acquirement for earth observation. Especially theapproaches for vehicle detection based on spaceborne pushbroom optical remotesensing images are the emphasis of experts all over the world. It is widely used in thefield of traffic management, military monitoring etc.First of all, the commonly used image fusion algorithms are introduced in thispaper. Vehicles are extracted and velocities are calculated by using the PanSharp imageand Ghost of moving target. Then an automatic target detection method is proposedbased on satellite remote image by considering the feature of the large region andhigh-resolution pushbroom optical remote sensing images. Specifically, the acquiredpanchromatic and multispectral images are analyzed. Image segmentation, normalizeddifferential vegetation index, Morphological gray scale reconstruction are combined torealize high-speed background suppression, the detection of the target by using the highresolution characteristics of panchromatic image and rich spectral reflectance feature ofmultispectral image. Finally, according to the satellite sensor model and RPCs, theproper way to transform the images coordinates of vehicles into ground coordinatesprecisely is discussed, and moving parameters estimation of the targets are realized.