图像制导的目标匹配算法与系统设计

Design and Algorithm for Image Matching Navigation and Guidance

根据航拍地形图像和目标图像的特点,提出Zoser图像匹配算法用于图像导航与制导。Zoser算法主要由高斯Zoser图像金字塔、24邻域极值点和SIFT128描述符等组成。Zoser算法使用整型数据,结构简单,二级算法独立性好,易于模块化设计;根据这些特点提出了基于FPGA和多DSP的系统设计方案,系统具有较好的实时性和图像匹配目标识别性能。Zoser算法对图像位移、旋转、仿射变换、尺度变化、噪声影响、轻微形变和照度变化有较好的鲁棒性,识别和匹配的准确性较好,速度比SIFT算法提高3倍左右,更适合实时系统。

The novel Zoser image matching algorithm fitting to the aerial images was presented for the image navigation and guidance. Zoser algorithm mainly consists of Gaussian Zoser （step） images Pyramid, 24-neighbor extreme, and 128-dimension SIFT descriptors, which uses integer data, simple image structures, independent sub-algorithms fitting to modularization design, so the design project is a FPGA and multbDSPs real-time system. Zoser algorithm has matc- hing invariance to image scale and rotation, and is shown to provide robust matching across a substantial range of affine distortion, change in 3D viewpoint, addition of noise, and change in illumination. Zoser has high matching precision, and its speed is 3 times quicker than SIFT.