Research On The Algorithms Of Target Recognition And Tracking Based On Non-uniform Sampling Image

The algorithms of target recognition and tracking are widely used in military and civil fields. It is one of the highlights issue in scientific research such as image processing, pattern recognition and computer vision. When the target scales and rotates, the accuracy of conventional algorithms of target recognition will be affected. Therefore, it is meaningful to do the research on the algorithm of target recognition and tracking under the condition of scaling and rotation changes.In this paper, the software based on non-uniform mapping transformation has been proposed. Then, the research on the algorithms based on non-uniform sampling image target recognition and tracking has been taken, aiming at resolving the problems of target recognition and tracking when the target changes by scaling and rotation.In this thesis, major works are done as follows:(1) The analysis of human visual system has been done to complete the algorithm design of the log-polar transformation, including the direct algorithm and the inverse algorithm. The algorithm has implemented t he image mapping from Cartesian coordinates to log-polar coordinate. Simulation experiments have been carried out to verify its function.(2) The algorithm design of space target recognition based on the log-polar image has been completed. Based on the methods of projection and curve fitting realized the detection of target movement distance to realize the detection of the amount of scaling and rotation. Meanwhile, the validity of the algorithm has been verified by simulation experiment.(3) The frequency domain identification algorithm based on log-polar mapping image has been completed. The detection of the scaling and the rotation of the target based on the log-polar mapping image has been realized based on Fourier transform and phase correlation method. Meanwhile, the characteristics of different contexts and object are verified.(4) The software of the target recognition and tracking based on the non-uniform sampling image has been designed. This algorithm has realized the real-time video processing, as well as the recognition and tracking of the real-time dynamic target.The algorithm design and the recognition and tracking software programming of have been completed in this thesis, which laid a research foundation to achieve the target recognition and tracking with a high speed under the conditions of the scaling and the rotation of the target.