基于版面理解的选票图像检测定位方法

Detection and location of ballot image based on layout understanding

针对目前选票上需印制同步道来辅助定位,存在印刷要求高、影响美观等问题,提出一种基于版面理解的选票图像检测定位方法。根据选票版面的几何结构和逻辑结构,首先,利用边缘检测、形态学操作等图像处理技术对选票进行预处理,初步定位填涂块;其次,利用支持向量机(SVM)模型对预处理得到的数据集进行分类预测,进一步识别定位填涂块;最后,利用填涂块坐标拟合直线来进行后处理,消除图像检测所带来的偏差,辅助定位。实验结果表明,该方法识别准确率达到99.9%,同时具备较好的时间效率,因此能有效去除同步道。

Nowadays the ballots should be printed with auxiliary synchronous tracks to locate filling blocks, which need high quality printing and lack beauty. A detecting and locating method of filling blocks based on layout understanding and Support Vector Machine （SVM） was proposed. According to the features of the ballots, firstly, image processing methods such as edge detection and morphological operation were introduced to locate the initial positions of filling blocks; then a pre-trained SVM classifier was used to classify the preprocessed dataset; at last, the bias of detected positions were corrected by fitting a straight line to assist location. The experimental results show that the method has high speed of location and high accuracy of recognition, the accuracy reaches 99.9%. So the method can be effectively applied to ballots processing without synchronous tracks.