在复杂场景中多目标物的检测识别方法

本文提出了一种在复杂场景中实时检测、识别多个目标物体的方法。利用由矩阵码（matrix code）构成的 2D 矩阵来识别复杂场景中多个目标物。算法不仅能适应目标物体在旋转、缩放、平移变换下的不变性识别，而且能适应仿射及射影变换下的不变性识别。试验结果证明所提供的算法在识别和监视系统中有很好的鲁棒性和实时性。
关键词 ： 矩阵码; 目标识别； 拐角点提取； 透视变换； 图像匹配；
Abstract: An method for real-time detection and recognition of multi-targets in complex scenes is presented.This method utilizes 2D matrix markers, being composed of matrix code ,which can identify a large number of objects.The algorithm can adapt to object recognition，invariant not only under rotation、scaling、 translation , but
under affine and projective transformation. Experimental results are given to demonstrate the robust adaptation and real-time of the algorithm in recognition and surveillance system.
Key words: matrix code; object recognition; corner detection; perspective transformation; image matching.