弹上下视景象目标快速匹配的技术实现

弹上光学传感器拍摄的下视目标图像的配准是打击效果评估的前提.研究了一种稳定的SlFT算子,实现了大视场下视景象目标的快速匹配.其方法是:通过在尺度空间中计算局部极值,先确定匹配关键点的位置和尺度,同时去除低对比度的关键点不稳定边缘上的响应点;然后由关键点相邻像素的梯度方向直方图描绘出种子点及其特征向量,近而获得关键点的特征向量;最后使用距离函数作为相似性度量进行配准.实践证明,与传统的方法相比,实现的算法具有更高的实时性.由于在原有基础上增加了其特征向量的雏数,因而具有更好的鲁棒性.

Fast downward-looking target matching technology on missile in large field of view

Downward-looking target matching on missile is a prerequisite for hitting effect evaluation.A stable SIFT algorithm is proposed to implement fast downward-looking target matching in large field of view.The method initially identifis positions and scales of key points by calculating local maxima in scale space and eliminates the response points on the unstable edge of low-contrast key points.Then seed points and their eigenvectors are sketched based on the gradient direction histogram of key points neighboring pixels,further eigenvectors of key points are obtained.Finally,distance functions are used as similarity measure for image matching.Compared with several traditional methods,the proposed algorithm has higher real-time capability in experimental results.Furthermore,the improved algorithm has better robustness due to increasing of eigenvector dimensions on the basis of original one.