Adaptation Based On The Fractal Characteristics Of Scene Analysis

By using Digital Scene Matching Area Correlation (DSMAC), especially in the terminal phase, the guidance system of missile have many advantages, such as great anti-interference ability, great accuracy, etc. The selection of matching area is an important element of the performance of the scene matching system. This dissertation focuses on the study of scene navigability analysis.Some image features and some matching performance metric parameters commonly used are studied. Their characteristics and the role to the scene navigability analysis are discussed. This paper presents the method of scene navigability analysis: taking the fractal feature as image feature, taking the statistic threshold of co-correlation coefficients'peak value as matching performance metric parameter, and then establishing the relation between the fractal feature and the co-correlation coefficients.The stable and fast algorithm to estimate fractal dimension (D) and the role of the relation between fractal feature and co-correlation coefficients to scene navigability analysis are the innovation and the research emphasis of this paper.Algorithm to calculate fractal dimension is studied and a lot of experimental analysis is made. A modified Carpet Covered algorithm based on the Peleg Carpet Covered algorithm is proposed. The comparison of the experiments shows that the modified algorithm is more stable and has better anti-noise and texture(terrian)-differentiating ability.Considering that the modified Carpet Covered algorithm takes too much time to operate,a fast algorithm is presented based on the quad-tree theory and the characteristics of the modified Carpet Covered algorithm. The algorithm's validity to visible light image and synthetic aperture radar (SAR) image with speckle noise is proved by simulation experiments.On condition that the Hurst parameter (H, H=3-D) is homogeneous, the quantitative relationship between the Hurst parameter and the co-correlation coefficients of reference sub-map is deduced. And experiments show that the conclusion is valid. The consistency tendency of the Hurst parameter and the co-correlation coefficients is studied. On condition that the Hurst parameter is inhomogeneous, an approximate formula to calculate the co-correlation coefficients is proposed.The algorithm of scene navigability analysis based on the fractal feature and the co-correlation coefficients of reference sub-map is given. The experiment shows that this method can program the required regions successfully in certain conditions. And the calculation quantity of the algorithm can meet the requirement of the scene navigability analysis.