摘要
对于生产流水线上输送较慢的大尺寸工件,利用线阵相机对工件进行检测的成本较高和相机的功效不能完全发挥的缺点,提出了一种基于面阵相机的在线图像拼接方法;该方法能在一定的速度范围内有效地对运动物体整体图像进行实时拼接;在图像匹配过程中,该方法运用基于匹配点对误差排序的方法,将正确的匹配提取出来;利用误差阈值内的角点组合来估计单应性矩阵;然后利用Levenberg-Marquart(L-M)非线性优化方法对求得的矩阵参数进行优化得到最终单应性矩阵参数;实验验证结果表明,该方法稳定性好。
According to the situation of using linear camera for big staff on the slowly running product line, the relative high cost and the camera＇ s full capability could not be applied. A image online masaic method based on aerial camera is presented. The method could masaic part images of the moving--slowly products in real--time at a speed range of the product line. When execute the images matching process, the method applies an image matching method based on error sorting, using the matching corner pairs whose error is under the error threshold to estimate the homograph matrix between two images. Subsequently by optimizing this initial parameter vector with Levenberg--Marquard algorithm, we can obtain finally accurate data. Experimental results have demonstrated the stability and validity of the method.
出处
《计算机测量与控制》
CSCD
2008年第8期1158-1160,共3页
Computer Measurement &Control
基金
天津市高等学校科技发展基金资助项目(20041112)
