论文部分内容阅读
提出一种新的星图中星获取算法——极值点法,利用最小二乘向量机(LSSVM)对原始星图的局部区域作灰度曲面最佳拟合,在拟合曲面上求解灰度极大值的像素点,获得星的中心点的初步位置。以初步位置为基础的星图像素聚类加速了星图中星的获取过程。以模拟星图中星的精确中心位置为参考,计算在不同噪声条件下测量位置与最近参考位置的距离平方倒数和的均值,优化 LSSVM 参数。为获得最佳星获取性能,卷积核为5×5像素的高斯LSSVM参数(σ2, γ)取(17,1.25)。该极值点法与矢量法相比,效率相当,但性能更好。
In this paper, a new star acquisition algorithm-extreme point method is proposed. The least squares vector machine (LSSVM) is used to best fit the local area of the original star map to the gray surface, and the gray Degrees of maximum pixel, get the initial position of the center of the star. The clustering of star images based on the initial position accelerates the star acquisition in the star image. Taking the accurate center position of the star in the simulated star map as a reference, the average of the square reciprocal sum of the distances between the measurement position and the nearest reference position under different noises conditions is calculated to optimize the LSSVM parameters. To obtain the best star acquisition performance, the Gaussian LSSVM parameters (σ2, γ) with a convolution kernel of 5 × 5 pixels are taken as 17,1.25. Compared with the vector method, the extremum point method is quite efficient, but the performance is better.