摘要
研究摄像机定位优化控制问题,摄像机镜头存在多种非线性畸变,对标定路径和图像质量产生一定的影响,难以采用精确数学模型来描述,针对传统方法的摄像机标定准确率低。为了提高摄像机的标定准确率,利用LSSVM较好的处理非线性预测能力,建立一种PSO-LSSVM的摄像机标定模型。将摄像头采集到的图像坐标作为输入,将世界坐标作为输出,通过采用LSSVM精确逼近输入与输出的复杂非线性关系,采用PSO寻找LSSVM最优参数,提高标定准确率。通过标定模型进行对比实验,实验结果表明,PSO-LSSVM不仅加快标定速度,且提高了摄像机标定的准确率。
In the camera calibration,camera has various nonlinear distortions,mathematical model can not describe that accurately,and the accuracy of traditional calibration methods is low.In order to improve the camera calibration accuracy,using LSSVM good nonlinear prediction ability,this paper proposed a camera calibration model based on PSO-LSSVM.The collect images coordinates were used in model inputs and the world coordinates were used in the output.Using LSSVM to approach the non-linear relationship between input and output,the LSSVM parameters were optimized by PSO LSSVM.The experimental results show that the PSO-LSSVM raised the camera calibration accuracy,speedd up the calibration speed,and has good real-time compared with other calibration modes.
出处
《计算机仿真》
CSCD
北大核心
2012年第3期280-283,共4页
Computer Simulation
关键词
摄像机标定
支持向量机
粒子群优化算法
Camera calibration
Support vector machine(SVM)
Particle swarm optimization(PSO)