一种基于对数极坐标变换的快速目标识别算法

在目标识别的过程中,观察目标图像相对于基准目标图像会存在尺度、方向和位置的变化,使得识别速度和准确率降低. 针对这一问题,提出了一种快速对数极坐标变换算法,加快了从笛卡儿坐标转换到对数极坐标的过程. 通过采用双轴投影相似度分析算法对目标图像进行匹配,进一步加快了识别速度,同时保证了匹配的可靠性. 理论分析和试验结果表明,该算法在计算效率和目标识别正确率方面具有较好的性能.

A fast target recognition algorithm based on LPT

In the course of target recognition, recognition speed and accuracy decreases if the scale, orientation and position of the target image changeswith respect to the reference image. This paper p resents a fast log2polar trans2 formation (LPT) algorithm, which speeds up p rocessing from Cartesian coordinates to log2polar coordinates aswell as speeding up image matching by using a dual2axis p rojection algorithm. The new algorithm greatly accelerates tar2 get recognition and ensures the reliability ofmatches. Theoretical analysis and experiments showed that thismethod has better performance in both recognition speed and accuracy.