红外技术, 2018, 40 (3): 280, 网络出版: 2018-04-09  

基于改进 SIFT算法的弹载电视制导技术研究

Research on Missile-Borne TV Guidance Technology Based on Improved SIFT Algorithm
作者单位
1 陆军炮兵防空兵学院高过载弹药制导控制与信息感知实验室, 安徽 合肥 230031
2 陆军炮兵防空兵学院研究生管理大队, 安徽 合肥 230031
摘要
以图像匹配技术为代表的弹载电视制导技术具有信息直观的特点, 作为非常优秀的图像匹配技术, SIFT算法受到了广泛的关注和深入的研究。针对传统 SIFT算法实时性差的问题, 本文提出了一种改进的 SIFT算法。在提取特征点部分, 通过 Laplace算子找出图像边缘区域并进行 Laplace加权处理, 然后利用 FAST特征点检测算法提取区域特征点; 在生成特征点描述子部分, 将传统的 128维 SIFT算子降为 48维, 利用改进的 SIFT特征描述算子为特征点赋予方向和描述符使其具有旋转不变性; 在特征点匹配部分, 利用欧式距离提取匹配点对, 并采用 RANSAC算法提纯匹配点对, 得到最优矩阵。实验结果表明改进的 SIFT算法在目标旋转、尺度变化等条件下匹配效果良好, 与传统 SIFT算法相比具有很高的实时性, 可以很好地实现图像实时匹配。
Abstract
The technology of missile-borne television guided by image matching technology has the cha-racteristics of information visualization. As an effective image matching technology, the SIFT algorithm has received extensive research attention. Aiming at resolving the problem of the poor real-time performance of the traditional SIFT algorithm, an improved SIFT algorithm is proposed. In the extraction of feature points, the Laplace operator is used to find image edge regions and Laplace weighting is performed. In the genera-tion of feature point descriptors, the traditional 128 dimensional SIFT operator is reduced to 48 dimensions, and the improved SIFT operator is adopted to assign directions and descriptors to feature points with rota-tion invariance. In the matching of feature points, the matching points are extracted by Euclide distance and are refined by the RANSAC algorithm to obtain the optimal matrix. The results of experiments show that the improved SIFT algorithm provides good matching effect under the conditions of target rotation and scale change. It performs well in real-time and can realize real-time image matching in comparison with the tradi-tional SIFT algorithm.

刘桢, 任梦洁, 姜万里. 基于改进 SIFT算法的弹载电视制导技术研究[J]. 红外技术, 2018, 40(3): 280. LIU Zhen, REN Mengjie, JIANG Wanli. Research on Missile-Borne TV Guidance Technology Based on Improved SIFT Algorithm[J]. Infrared Technology, 2018, 40(3): 280.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!