星空背景下运动目标级联支持向量机(SVM)高精度检测

A High Precision Algorithm for Space Object Detection under Starry Background Based on Cascade Support Vector Machine(SVM)

为实现空间监视系统光学图像中目标高精度探测,提出了一种星空背景下高精度快速级联式支持向量机空间目标检测算法。通过提取空间目标不同尺度下目标二值规范化特征,训练前两级线性级联支持向量机分类器;继而提取目标的面积、周长、灰度、Hu矩特征作为组合特征,训练第三级支持向量机分类器。在目标检测过程中,采用前两级支持向量机分类器进行候选目标的窗口预测和评分,进而利用第三级支持向量机分类器进行目标确认而给出检测结果。仿真实验及结果分析表明,这种级联支持向量级目标检测方法的精度高、实时性强、适用于星空背景下的空间监视系统。

In order to achieve high precision for optical image object in space surveillance system,a high precision and fast space object detection algorithm under starry background based on cascade support vector machine(SVM)is put forward.By extracting binary normalization features of space object in different scales,the first two levels linear cascade SVM classifier are trained.And the area,perimeter,gray and Hu moment features of object as an combination of features is used to train the third level SVM classifier.During the process of object detection,the first two levels SVM classifiers are employed to carry on window prediction and score of candidate objects,and the third level SVM classifier is used to determine the final candidate window,so as to finish the object detection.Simulation experiments and the results analysis demonstrate that the proposed cascade SVM object detection algorithm possess the advantage of high precision and real-time,hence it can be applied to space surveillance system under the starry background.