基于惯性传感器和Kinect摄像机的OFCM图像匹配算法

OFCM Image Matching Algorithm Based on Inertia Sensor and Kinect Camera

摘要:

提出了一种基于图像三维信息及摄像机运动参数的图像匹配算法, 用于解决智能监控系统中图像匹配实时性差、鲁棒性低等问题。该算法分别利用惯性传感器和Kinect摄像机估计出摄像机的运动外参矩阵和图像的深度信息, 再根据射影几何原理, 结合当前帧图像的像素坐标计算出该像素点在下帧图像的像素坐标, 从而完成图像匹配。利用该算法对实际采集的图像序列进行了分析与处理, 并从配准精度、鲁棒性和实时性方面与经典匹配算法KLT进行对比。实验结果表明: 该算法极大地降低了计算量和计算时间, 不仅能满足智能监控系统对图像匹配精度和稳定性的要求, 更能满足系统实时性的要求。

Abstract:

In order to solve the problems of low real time and robustness in intelligent monitoring, an image matching algorithm which is based on the three dimensional information of the image and the motion parameters of the camera is put forward. The algorithm uses an Inertia sensor with magnetometer and a Kinect camera to estimate the camera’s motion parameters and the depth information of the image respectively. Then based on the equation of photography geometry, the new pixel coordinates in the next frame are estimated by using the pixel coordinates in the current frame. Thus image matching is realized. The experiments are performed based on actual image data, and the results of the image matching are compared with KLT algorithm widely applied in image procession, in the aspect of registration accuracy, robustness and real time. Conclusions indicate that the proposed algorithm, which reduces the computation time extremely, can meet the requirements of not only precision and stability but also real time of intelligent monitoring systems.