| Photoelectric platform as a sophisticated aiming payload of modern warfare, iswidely used in satellite, airborne, land-based, carrier-based reconnaissance withintegration of combat in military and civilian fields. The complex environment of thecarrier aircraft movement will cause the sailing posture is changing, leading to theLOS of the airborne optoelectronic platform unstable, causing the image dithering, itis difficult to meet the high-precision LOS stablization demand as well as intelligencegathering and reconnaissance needs. In-depth study in the LOS of destabilizingfactors by constructing mathematical theory of image dithering caused the two-axisand multi-axis turrent, positioning and matching of image feature detection andmotion estimation; the use of image stabilization method to enhance the stability ofthe LOS to solve the moving base optoelectronic platform in intelligence gathering,reconnaissance in terms of the stability of demand.Firstly, the derivation of the optical imaging system characteristics and theestablishment of the camera motion model, build the relationship between themathematical model of camera movement and image movement mathematical model,eliminate image shake image stabilization solution for in-depth study of imagestabilization technology and theoretical foundation for the engineering of imagestabilization technology to build a basic turrent of practice.Translation-based two-axis turrent video sequences jitter, gray projection to theface of the dramatic impact of the external environment of low reliability, stability,error rate, of the image gray level information entropy image stabilization algorithm.Analysis of two-dimensional projection of the gray information integral to calculatethe inter-frame image of local mutual information for high-end extreme value, thelocal motion vectors; concentration to determine the algorithm to calculate the global motion compensation parameters, efficient and eliminate the pseudo-motion vectors,avoid the smoothness of the image output sequence. Algorithm works transplantationefficiency of the algorithm for720×576@25fps, the image sequence length500frames, using DSP320C6455the interframe stable reach10.2ms, stable error<1pixel/frame the stable turnover ratio <1%, relative gray projection stable algorithm,stable error rate reduces to10times.A comprehensive in-depth study on the multi-axis photoelectric turrent imagestabilization based on two-axis turrent image stabilization. Defects of theFourier-Mellion algorithm in the frequency domain analysis, frequency domain, suchas symmetry characteristics to build single-phase limit the actual number of Fourierimage stabilization algorithm, the rotation compensation for the time CFT is nearly70%increase in translation compensation70%increase in computation time. GPGPUtechnology combined with the characteristics of "cloud computing", designed tooverlap the movement of composite phase estimation structure to enhance thesynchronization performance of parallel threads, change the scale of parallel threads,adjust the graphics processor storage, improved concurrency of phase motionestimation can be high quality to estimate the parameters of image motion in themulti-axis framework, nearly36times, The image stable performance18ms/frame the1.6k×2.4k pixels of the image sequence, image sequences of512×512pixels, thealgorithm is compared to the CFT.The invariant feature image stabilization to combat a wide range of viewpointschange-depth study.By detecting the right scale, rotation, affine invariance porphyriticcharacteristic extreme points, robust feature improved fast algorithm to extract thelocal image region of feature points and their descriptions, design dynamicequilibrium KD tree for fast searching and matching, image stabilization performancegains four times; can withstand20%noise,100%of the repeat detection and locationcharacteristics, from the local motion, the impact of changes in light, able to estimatethe interframe affine and any form of global motion parameters, greatly enhancing theenvironmental adaptability. Anti-viewpoint change in the invariant features of imagestabilization extracted information than SIFT improved nearly20times, can handle upto±85°viewpoint changes caused by any movement in the form of imagestabilization.Concentration to determine the motion estimation and nonlinear differential trackermotion estimation strategy in order to reduce the influence of the pseudo-motion vectors of the visual axis stability. By analyzing the distribution of statisticalproperties of the local motion vector design histogram, convolution, filtering, dataprocessing process removed the pseudo motion vectors to reduce the pseudo-motionvectors of the overall stability of the image sequence, can be100%removedpseudo-sport is estimated to improve the algorithm the overall resistance to wrongestimation of the impact of image stabilization; computation time is only13%of theRANSAC algorithm, motion estimation performance of a substantial increase. Finally,intelligence gathering, reconnaissance and other non-visual axis image stabilizationand stability problems of orthogonal aerial exploration of the research proposedAHRS inertial measurement components and global positioning system for imagestabilization, and image mosaic integration of technologies such as cross, to expandinformation reconnaissance field of view.
|
PDF Full Text Request