基于MSER的集装箱号低秩矫正研究

集装箱号的快速记录对于码头、港口业务至关重要．由于相机拍摄的图像常伴有严重的畸变，极大影响了字符识别算法的准确性，因此需要在识别之前对图像进行矫正．本文提出了一种快速有效地对存在畸变的集装箱号图像进行矫正的方法．首先应用基于MSER的图像分割算法对集装箱箱号区域进行检测，确定矫正的特征区域；然后根据低秩模型利用LADMAP*以及初始值热启动的方法来降低算法复杂度并矫正图像；最后，结合图像低秩和矫正前后的变换率优化算法的收敛条件，经过多次迭代后得到最终的矫正结果．实验结果表明，该方法不仅能对集装箱号进行较好的矫正，而且算法也具有稳健性．

Research on Container Number Correction with Low Rank Based on MSER

The fast record of container number is very important for the terminal and the port business. Since the image of the camera is often accompanied by the serious distortion, and greatly affect the accuracy of the character recognition algorithm, we need to correct the image before the recognition. In this paper, a fast and effective method to correct the distortion of the container image is proposed. Based on MSER element projection merging algorithm, the container area are detected and the correction feature region is determined. Then, according to low rank model with LADMAP* and initial value of warm start method, the method is able to reduce algorithm complexity and correct image. Finally, combining with the image of low rank and the convergence condition of optimization algorithm before and after the transformation, the final results are obtained. The experimental results show that the algorithm is not only good for container number correction but also robust.