Target Detection In Optical Remote Sensing Image Based On Complex Background Suppression

As an important research branch of space information acquirement for earth observation, target detection based on high-resolution optical remote sensing images has received a wide range of applications in the field of military surveillance, battlefield analysis and traffic management. Moving target detection in high-resolution remote sensing image has become one of the hotspots at home and abroad.Firstly, image segmentation methods commonly used in target detection are discussed for the characteristics of high-resolution optical remote sensing images of wide-area complex background, and the problem of a large number of background interference in target detection. Mean shift filter algorithm is used to overcome the defect of inadequate background suppression in segmentation, which helps to cluster and smooth the homogenous area, and reduce the gray difference among pixels.Secondly, this paper researches on panchromatic and multispectral images of single satellite, and combines mean shift filter with image segmentation effectively, which not only makes the edges between heterogeneous regions significantly, but also helps to cluster and merge homogeneous regions. And then, with utilization of panchromatic image attribute and multispectral channel imaging properties, a kind of remote sensing image background clutter suppression technology is proposed, which is based on different types of surface features normalized index, the region of interest extraction, morphological grayscale reconstruction and background subtraction. Vegetation, water and buildings clutter are suppressed and isolated, the detection area is reduced and the false alarm rate is decreased finally.Next, according to the target characteristics of remote sensing images, the diversity of scale, radiation characteristics of the images, and the complexity of background, reconstruction algorithm is applied to the panchromatic remote sensing images by using of different-direction structural elements. This algorithm is based on the results of complex background suppression. And morphological filter is applied to extract vehicle targets, then a priori information such as geometric characteristics is used to remove false alarm. Thetarget detection and recognition are realized and target detection accuracy is improved.Finally, based on common imaging sensor model, the moving target location is realized by considering the RPC coefficients and the relational model of target position coordinates in the image and space ground position coordinates.