基于PCA与不变矩的车标定位与识别

Method for Vehicle-Logo Location and Recognition Based on PCA and Invariant Moment

  • 摘要: 在车牌位置确定的情况下,利用车标边缘特点在车牌上方一定范围内检测车标。提出车标似真度的概念,将检测到的车标图像映入PCA生成的特征车标空间,得到的重构图像与原图像进行车标真实性检测,减少车标的误定位;然后利用不变矩的旋转、尺度及平移均不变的特性,定义不变矩的最小矩距离进行车标识别。通过实测车标图像的定位和识别实验表明,该方法是有效和可行的。

     

    Abstract: Vehicle-logo location can be executed using edges feature of the vehicle logo depending on the position relation between vehicle-logo and vehicle license plate,then the plausibility function of vehicle-logo can be proposed by the correlation function between the original vehicle-logo and the reconstruction image reflected into the eigen vehicle-logo space from the original vehicle-logo by PCA,the error of vehicle-logo location can be reduced.After the vehicle-logo location,the vehicle-logo can be recognized by invariant moments minima moment distance depending on the stability of invariant moments features for rotation,scale,translation.The experiments result of vehicle-logo location and recognition for virtual images shows that the method is effective and practical.

     

/

返回文章
返回