Target Text Location for Vehicle Inspection Based on Template Matching

Nowadays, vehicle inspection has an urgent demand for automatic target information recognition technology. Target text location is the prerequisite for target text recognition and has a great influence on recognition. Experimental results show that deep learning algorithms, such as faster RCNN, are not effective on such problems to some extent. Therefore, this paper adopted the method of template matching, took the vehicle license application form as the research subject, realized the target text location according to the relative location between the target texts in the image, and achieved satisfactory results. The application of this method on the driving license proves that it is a target text location technology that can be applied to document images with relatively fixed layout.